Implantable Light-Emitting Diodes (A3036IL)

© 2019-2022 Kevan Hashemi, Open Source Instruments Inc.
© 2021 Alice Hashemi, Open Source Instruments Inc.


Optical Power
Fiber Tapering
Obsolete Stimulator


Note: The Implantable Light-Emitting Diode (A3036IL) is an active product. The Implantable Stimulator-Transponder (A3036) is discontinued, replaced by the Implantable Stimulator-Transponder (A3041).

[12-FEB-24] The Implantable Light-Emitting Diode (A3036IL) is a light-emitting diode (LED) designed for implantation in animals so as to provide optical stimulation for optogenetic experiments. They are accessories to our Implantable Stimulator-Transponders (ISTs), such as the A3041. The A3036IL uses a variety of LEDs: the blue and green EZ500, blue and green EZ290, and the deep red LXZ1-PA01. All A3036ILs are equipped with a thin-walled steel tube to allow the LED to be mounted during implantation. The tube is thinned near the base so it may be cut after the LED has been secured.

Figure: Fiber-Coupled LED (FCLED) For Depth Illumination in Mice. Fiber diameter 270 μm, length 4±0.3 mm. The tip emits 2.8 mW at 10 mA forward current.

The optical fibers we use with our implantable LEDs are polished at the base and tapered at the tip. We glue the polished base to the surface of an LED and the fiber carries LED light to the tapered tip. We call the combination of fiber and LED a Fiber-Coupled LEDs (FCLEDs). The fraction of optical power emitted by the LED that is delivered to the tapered tip of the fiber is the coupling efficiency of the fiber.

Figure: Tapered Light Guide. Made from a 300-μm diameter fiber. (Sam Orphanos)

We make FCLEDs out of bare LED chips wire-bonded to a printed circuit board. We have blue and green FCLEDs, but no red FCLEDs, because we have not yet found a red LED of adequate efficiency that we can purchase in die form.

Figure: Surface-Mount LED (SMLED, A3036IL-C) with Epoxy Dome For Surface Illumination. Near socket: orange L+ lead. Far socket: purple L− lead. A steel tube allows us to fix the ILED during surgery. The printed circuit board is 4.0 mm × 3.0 mm, and 0.8 mm thick.

With no optical fiber glued to the LED, the LED is instead covered with clear epoxy so that it may be placed on the surface of the tissue to be illuminated. We call these our Surface-Mount LEDs (SMLEDs). The surface of the epoxy may be flat or domed, but this makes little difference to the power delivered to the tissue surface. These epoxy-topped LEDs we call Surface-Mount LEDs (SMLEDs). We make red, blue, and green SMLEDs.

Figure: Implantable LED (ILED) Back Side. Near socket: L+ lead. Far socket: L− lead.

A steel mounting tube is glued to the back of each ILED and thinned just above the glue joint so it may be cut easily during implantation.


[03-AUG-22] The following versions of the Implantable Lamp (A3036IL) are in production. The gold-plated pins on the end of the stimulator's lamp leads mate with a pair of sockets on the A3036IL. Currently, we use wire-bonded EZ500 chips for our blue and green illuminators, both SMLED and FCLED. We use the LXZ1-PA01 for our red SMLEDs. For a comparison of a larger variety of illuminators, including those we have rejected for inferior performance, see Optical Power.

Version Type LED Light Guide Wavelength
Die Size
(μm × μm)
Optical Power
(mW at 10 mA)
A3036IL-A SMLED Blue EZ500 Epoxy Dome 460 480 × 480 10.3
A3036IL-B SMLED Green EZ500 Epoxy Dome 527 480 × 480 7.2
A3036IL-A270-4 FCLED Blue EZ500 270 μm Dia, 4 mm Len 460 480 × 480 2.8
A3036IL-B270-4 FCLED Green EZ500 270 μm Dia, 4 mm Len 527 480 × 480 1.8
A3036IL-A270-6 FCLED Blue EZ500 270 μm Dia, 6 mm Len 460 480 × 480 2.8
A3036IL-B270-6 FCLED Green EZ500 270 μm Dia, 6 mm Len 527 480 × 480 1.8
A3036IL-A450-8 FCLED Blue EZ500 450 μm Dia, 8 mm Len 460 480 × 480 3.3
A3036IL-B450-8 FCLED Green EZ500 450 μm Dia, 8 mm Len 527 480 × 480 2.4
A3036IL-E SMLED LXZ1-PA01 Epoxy Dome 650 1000 × 1000 8.0
A3036IL-X SMLED Any Epoxy Dome Any Any Test Lamp
Table: Versions of the Implantable Lamp (A3036IL). We quote expected optical power at the tip of the fiber or the surface of the LED for forward current 10 mA. Optical power is approximately linear with current.

We can make any taper-tipped, polished-base fiber 4 mm or longer, diameter 270 μm, 330 μm or 450 μm.

Figure: Three A3036IL-A270-6, Two A3036IL-A, and One Prototype A3036IL-A270 with 3-mm Polished End.

The EZ500 data sheet does not provide spectra for their light emission. But the Luxeon Z data sheet from Philips Lumileds does provide spectra, which we present below. We assume the EZ500 blue and green spectra are similar.

Figure: Spectra for Luxeon Z LEDs.

Based upon the above spectra, we estimate that the half-power width of the EZ500 460-nm blue LED is 20 nm, and of the 525-nm green is 50 nm.


[12-FEB-24] The ILED requires a printed circuit board onto which we mount the LED. In the case of the EZ290 and EZ500 LEDs, the LED's bottom-side cathode pad must be glued to the board with conductive epoxy, while it's top-side anode pad wire-bonded to the board. The EZ290 is smaller than the EZ500, which suggests that it may be a better choice for delivering light into a fiber. But the EZ290's anode pad is in the center of its emitting area, which forces us to keep the fiber at least 100 μm above the LED surface, while the EZ500's anode pad is on one edge of the emitting area, which allows us to place the fiber right onto the emitting surface. In the end, we obtain greater efficiency with the EZ500 despite the loss of light around its edges. The Luxeon Z LEDs are not suitable for FCLEDs because their area is so large. But they are suitable for SMLEDs. We mount the Luxeon Z LEDS by surface-mount reflow.

A303602A_Top: Rendering of top side of TR2227 Head.
A303602A_Bottom: Rendering of bottom side of TR2227 Head.
A303602B_Top: Rendering of top side of EZ500 Head.
A303602B_Bottom: Rendering of bottom side of EZ500 Head.
A303602C_Top: Rendering of top side of TR2227 Head.
A303602C_Bottom: Rendering of bottom side of TR2227 Head.
A303602D_Top: Rendering of top side of Wire-Bonded LED Head.
A303602D_Bottom: Rendering of bottom side of Wire-Bonded LED Head.
A303602E_Top: Rendering of top side of Luxeon Z Head.
A303602E_Bottom: Rendering of bottom side of Luxeon Z Head.
A303602E_Panel: Drawing of Luxeon Z Head Panel.
EZ500: Manufacturer's data sheet for EZ500 blue and green LEDs.
EZ290: Manufacturer's data sheet for EZ290 blue and green LEDs.
TR2227: Manufacturer's data sheet for TR2227 blue and green LEDs.
Luxeon Z: Manufacturer's data sheet for Luxeon Z family of LEDs.
Fiber Polishing and Tapering: Guide to polishing and tapering optical fibers.

The A3036 implantable stimulator is obsolete, but here are its design files.

S3030A_1: IST Version A Schematic.
S3030B_1: IST Version B Schematic.
Code: Logic Programs and Test Scripts.
LCMXO2-1200ZE: The programmable logic chip data sheet.
WLCSP-25: The 2.5 mm square BGA logic chip package. PCB for A3036A, Gerber files and drawing. Panel A303601A, Gerber files for assembly.
A303601A_Top: Rendering of top side of A303601A circuit board.
A303601A_Bottom: Rendering of bottom side of A303601A circuit board.
A3036AV1.ods: Bill of Materials and Pick and Place for A3036A. PCB for A3036B, Gerber files and drawing. Panel A303601B, Gerber files for assembly.
A303601B_Top: Rendering of top side of A303601B circuit board.
A303601B_Bottom: Rendering of bottom side of A303601B circuit board.
A3036BV1.ods: Bill of Materials and Pick and Place for A3036B.

Optical Power

[01-JUL-21] We calibrate our implantable LEDs by measuring their optical power output when the current through its LED is 10 mA. We measure the LED's power output with one of our calibration stands. We have one calibration stand for bare LEDs and another for fiber-coupled LEDs. The LED output power is approximately proportional to current, as shown in the plot below. We get roughly twice as much power at 20 mA and half as much at 5 mA.

Figure: Optical Power Output versus Current for Various of Bare LEDs. We are measuring output power with our Bare LED Calibration Stand (see below). An IST with battery voltage 3.7 V and lead resistance 56-Ω will deliver 14 mA to the LED with forward voltage drop 2.9 V.

Our Implantable LEDs (A3036IL) generate light with an LED mounted on a printed circuit board. Our favored LED is the EZ500, which comes in blue and green. The EZ500 light-emitting surface is 480 μm square, with a bond wire to one side, so that we can lower the base of an optical fiber right onto the emitting surface for greater coupling efficiency. For red light, we use the LXZ1-PA01 from the Luxeon Z family, which provides a 1-mm square emitting surface. The LXZ1-PA01 is an efficient source of 650-nm deep-red light for illuminating the surface of organ tissue. But its emitting surface is too large for efficient coupling of light into an optical fiber.

Figure: Optical Power Ouput versus Current for Two A3036IL-A450-8 Fiber-Coupled LEDs. We are measuring power emitted by the fiber tip using our Fiber-Coupled LED Calibration Stand.

Blue LEDs produce roughly 1 mW/mA (one milliwatt per milliamp forward current), green LEDs roughly 0.5 mW/mA forward current, and the deep-red LED roughly 0.8 mW/mA. We make surface illuminators by covering the LED with a clear epoxy dome. The A3036IL-A is a blue C460EZ500 with clear epoxy dome for surface illumination of brain tissue. We make depth illuminators by gluing a glass fiber with a tapered tip to the LED and inserting the tapered tip to the required depth. Roughly one third of the light emitted by the LED is captured by the fibre and transported to its tip, where it radiates in all directions. The A3036IL-A270-4 is a blue C460EZ500 with a 4-mm long, 270-μm diameter fiber light guide. The A3036IL-A270-4's average power output at the fiber tip is 2.8 mW for 10 mA forward current.

Figure: Optical Power Output at Fiber Tip for 10 mA Forward Current, Various Fiber-Coupled LEDs.

We measure optical power using an SD445 photodiode and an ammeter. We convert photocurrent into optical power using 0.18 mA/mW for 460-nm blue, 0.25 mA/mW for 527-nm green, and 0.40 mA/mW for 655-nm red light. We have two calibration stands that permit us to make consistent measurements of power for any combination of LED and fiber. The drawing below shows the arrangement of LED or FCLED and photodiode in each stand. For more information on the calibration and appearance of the test stands, see our development notes.

Figure: Bare LED Calibration Stand (Left) and Fiber-Coupled LED Calibration Stand (Right). On the left, 7.1% of the light emitted by the LED arrives at the detector. On the right, 39% of the light emitted by the fiber tip arrives at the detector.

When we cap an LED with epoxy, we make sure the surface is convex. We avoid loss by total internal reflection within the epoxy, and we direct the light more in the forward direction, which favors illumination of a surface directly in front of the ILED. But the epoxy dome is never uniform, so measuring the power emitted by the ILED after application of the epoxy dome is difficult. We assume the total power transmitted by the ILED is equal to the total power emitted by the LED die before we apply the dome.

We measure the power emitted by fiber-coupled LEDs directly with our FCLED calibration stand. The fraction of LED power reaching the fiber tip depends upon the area of the fiber and the LED, and upon the numerical aperture of the fiber itself. We make our light guides out of optical fiber with numerical aperture 0.86 (core index 1.72, cladding index 1.49). All light within ±60° of the fiber axis will be captured by the fiber and transported to the tip. For a typical LED, this ±60° accounts for 75% of the light emitted. A 450-μm diameter light guide, perfectly polished at the base, and perfectly positioned on a 480-μm square EZ500 LED will couple 51% of the emitted light to the fiber tip. The A3036IL-A8 uses a 450-μm fiber on a C460EZ500, and we obtain coupling efficiency of around 40%.

The Implantable Stimulator-Transponder (A3041B) boosts its battery voltage to a fixed 3.3 V. It also provides a current limit we can apply to the stimulus. When equipped with 45 mm long leads, each will have resistance 28 Ω, making a total of 56 Ω. Suppose we turn up the stimulus current to its maximum value of 10 mA. To determine the actual LED current we need to know the approximate forward voltage drop of the LED at 10 mA.

Figure: Current and Power versus Voltage for an A3036IL-A8 Implantable LED. Blue C460EZ500 LED coupled to 450-μm, 8-mm fiber, power measured at the tapered tip of fiber.

The A3036IL-A450-8 FCLED has forward voltage 2.73 V at current 10 mA. The voltage drop across 56 Ω at 10 mA is 0.56 V. Thus 3.29 V is enough to drive 10 mA through the leads and LED. At this current, the tip of the fiber produces 5 mW. Thus the maximum optical power from the FCLED when driven by the A3041B is 5 mW.

[03-AUG-22] The following table compares the optical power output of at 10 mA forward current for various combinations of fibers, domes, and LEDs.

Version Type LED Light Guide Wavelength
Die Size
(μm × μm)
(mW at 10 mA)
A3036IL-A SMLED Blue EZ500 Epoxy Dome 460 480 × 480 10.3
A3036IL-B SMLED Green EZ500 Epoxy Dome 527 480 × 480 7.2
A3036IL-A270-4 FCLED Blue EZ500 270 μm Dia, 4 mm Len 460 480 × 480 2.8
A3036IL-B270-4 FCLED Green EZ500 270 μm Dia, 4 mm Len 527 480 × 480 1.8
A3036IL-A270-6 FCLED Blue EZ500 270 μm Dia, 6 mm Len 460 480 × 480 2.8
A3036IL-B270-6 FCLED Green EZ500 270 μm Dia, 6 mm Len 527 480 × 480 1.8
A3036IL-A450-8 FCLED Blue EZ500 450 μm Dia, 8 mm Len 460 480 × 480 3.3
A3036IL-B450-8 FCLED Green EZ500 450 μm Dia, 8 mm Len 527 480 × 480 2.4
A3036IL-C SMLED Blue EZ290 Epoxy Dome 460 290 × 290 7.4
A3036IL-D SMLED Green EZ290 Epoxy Dome 527 290 × 290 3.9
A3036IL-C450-8 FCLED Blue EZ290 450 μm Dia, 6 mm Len 460 290 × 290 3.8
A3036IL-D450-8 FCLED Green EZ290 450 μm Dia, 6 mm Len 527 290 × 290 2.0
A3036IL-C270-6 FCLED Blue EZ290 270 μm Dia, 6 mm Len 460 290 × 290 1.8
A3036IL-D270-6 FCLED Green EZ290 270 μm Dia, 6 mm Len 527 290 × 290 1.0
A3036IL-E SMLED Red LuxeonZ Epoxy Dome 650 1000 × 1000 8.0
A3036IL-X SMLED Any Epoxy Dome Any Test Lamp >2.0
Table: Versions of the Implantable Lamp (A3036IL).

For all these devices, optical power is approximately linear with current. Notice that the EZ290 fiber-coupled LEDs are not as efficient as the EZ500, even though the surface area of the EZ290 is smaller. The central bond wire of the EZ290 interferes with the contact between the fiber and the emitting surface.

Fiber Tapering

[01-FEB-24] Our fiber tapering machine uses two motor controllers, two micrometer stages, a horizontal stand, and a heating coil. We mount a 100-mm length of fiber, polished flat at the right end, to mounting plates, one attached to each motorized stage. As we mount the fiber, we pass it through the heating coil. During the tapering process, the heating coil is red-hot and softens the glass. The two stages to the left, bringing the target taper location into the coil. The right stage stops while the left stage pulls away. The heat-softened fiber stretches into a thread above the coil. The right stage moves off to break the fiber and create the taper on the right section of fiber. The right section is the one we keep.

Figure: Tapermaker Tool on MacOS. These settings generate a 1-mm long, uniform taper on the end of a 270-μm to 450-μm diameter TD5 fiber.

We control the tapering machine with the Tapermaker Tool, available in LWDAQ 9.1.8+. The tapering procedure is as follows.

  1. Press Reset to configure the motor controllers and establish the home position. The motors move to their limit switches and then to the home position, which is the starting position for tapering.
  2. Adjust the settings if you need to. The default values work well for 450-μm and 270-μm TD5 fiber.
  3. Install a fiber in the stretcher by passing it through the heating coil and securing it with magnets.
  4. Turn on the heating coil.
  5. Press Taper.
  6. When both motors have been stationary for a few seconds, turn off the heating coil.
  7. Remove the tapered fiber and place in gel pack. Remove the scrap fiber and dispose of safely.
  8. Press Home to return to home position. You are now ready for next taper.
  9. When you are done tapering, press Off to turn off power to the motors.

At any time, you can press Stop to stop the motors. You will occasionally need to stop the motors when you forget to return to the home position before you press Taper. The stretcher will collide with the heating coil. The Off button turns off power to the motors, so you can rotate the shafts by hand. After you rotate the motors by hand, you must press Reset to re-establish the home position in the controller's memory. When you are done with the tapermaker for the day, turn the motors off so they don't stay hot all the time.

Video: The Tapermaker Creating a Tapered Light Guide. We taper a 300-μm diameter TD5 fiber. The base of the taper is 4 mm from the base of the fiber. The taper is 1 mm long.

The approach distance is how far the two stages move upwards together to bring the target point into the coil. The approach speed is how fast they move. The approach differential is ratio of the left stage speed to the lower stage speed during the approach. This differential should be slightly greater than one so that imperfections in the synchronous movement of the stages do not cause the fiber to buckle. With the stages separating very slightly, the fiber remains straight. The left stretch distance and speed define the movement of the stage once the target poin is in the heating coil. The right stretch delay is how long the lower stage waits before it starts to move down. The right stretch distance and speed define its movement downwards. More details of the tapering process are given in the comments of the Tapermaker.tcl file, which you will find in the Tools of the LWDAQ distribution.

Figure: Freshly-Tapered Fiber. Made from a 300-μm diameter fiber. (Sam Orphanos)

Once the tapermaker is done, the taper will be extended by a thread of glass. We remove this thread by sanding the tip of the fiber on hand-held 15-μm grit. Once sanded, the taper looks like this.

Obsolete Stimulator

[25-JUL-22] The table below lists the existing versions of the obsolete Implantable Stimulator-Transponder (A3036). These circuits are now obsolete, replaced by the A3041.

Version Battery Volume
(mm × mm × mm)
Length (mm)
Resistance (Ω)
A3036A LiPo PP031012AB 19 mA-hr 0.9 21 x 10 x 5 1.5 45 56 Prototype
A3036B LiPo PP031012AB 19 mA-hr 0.9 21 x 10 x 5 1.5 45 56 Synch signal 1024 SPS
A3036C LiPo PGEB201212 10 mA-hr 0.75 14 x 13 x 6 1.3 45 56 Synch signal 1024 SPS
A3036D LiPo GMB201021 25 mA-hr 1.2 22 x 11 x 5 1.8 45 56 Synch signal 1024 SPS
Table: Versions of the Implantable Stimulator-Transponder (A3036).

The Implantable Stimulator-Transponder (A3036) is a wireless electrical stimulator that receives commands and responds with acknowledgements, battery measurements, and synchronizing signals using a single antenna. The A3036 is obsolete, replaced by the Implantable Stimulator-Transponder (A3040. The displacement volume of the A3036A is only 0.90 ml. The electrical stimulus is delivered through two silicone-insulated helical wires terminated with miniature pins. A single radio-frequency command initiates an arbitrarily long stimulus. When combined with the A3036IL series of implantable LEDs, the A3036 provides optogenetic stimulus. When combined with a bipolar depth electrode, the A3036 provides direct electrical stimulus. Between implants, we can recharge the A3036's battery through its stimulus leads with the help of a Battery Charger (A3033D).

Figure: Implantable Stimulator-Transponder (A3036C) with Test LED (A3036IL-X). Volume 0.75 ml, extreme dimensions 14 mm × 13 mm × 6 mm. Orange L+ and purple L− stimulus leads are 45-mm long and terminated with gold pins. Their combined resistance is 56-Ω. Stimulus leads are also used for re-charging. Single loop antenna provides command reception and acknowledgement transmission.

In its standby state, the IST is ready to receive commands, but is doing nothing else. Its standby current consumption is only 7 μA, so the A3036A, with its 19 mA-hr battery, can remain in its standby state for 2700 hrs before its battery is exhausted. Commands arrive from a transmitter such as the 915-MHz Command Transmitter (A3029C), which we control with the same software we developed for the Implantable Sensor with Lamp (ISL). This software operates with our LWDAQ data acquisition hardware. The IST transmits its acknowledgements and metadata at 915 MHz also, for detection by a data receiver such as the Octal Data Receiver A3027E. The IST transmissions use the same protocol as our Subcutaneous Transmitter System (SCT). The IST command transmissions take tens of milliseconds, and during command transmission, SCT samples will be lost.

Warning: The IST cannot activate when its stimulus electrodes are touching. We ship all ISTs with a Test Lamp (A3036IL-X) attached. The test lamp is not intended for implantation.

All versions of the A3036 IST are equipped with a lithium-polymer battery. No other miniature battery can provide 20 mA for the lamp, nor provide sufficient current to start up its logic chip. The IST applies its battery voltage to the lamp through its lamp leads. The resistance of the lamp leads limits the LED current, and we make this resistance part of the design of the device. The average voltage of lithium-polymer battery during its lifetime is 3.7 V. The forward voltage of the blue C460EZ500 is 2.9 V at 20 mA. Suppose we choose the resistance of the leads to set the lamp current at 20 mA for battery voltage 3.7 V. The 19-mAhr battery of the A3036A will provide 57 minutes of continuous light output. If our stimuli consist of 10-ms pulses at 10 Hz, the A3036A can deliver a total of 570 minutes of stimulus before it exhausts its battery.

One application of the IST is to implanted it together with an SCT so that stimuli can be generated in response to EEG activity in real time, with the help of the Event Classifier running on the data acquisition computer. Earlier ISL device combined stimulator and sensor in the same device, powered by the same battery, and suffered from lamp artifact. When we separate the stimulator and the sensor into two circuits each powered by their own battery, we find that we eliminate lamp artifact in optogenetic applications. The IST's synchronizing transmission, when enabled, is recorded along with SCT signals, and allows us to obtain perfect superposition of the stimulus pulses and biometric signals.

Figure: Connections to an Optical Stimulator and Local Field Potential Sensor. We have L+ and L− from an Implantable Stimulator-Transponder (IST) connected to a fiber-coupled LED, and X+ and X− from a subcutaneous transmitter (SCT) connected to a depth electrode and a skull screw.

We first proposed the development of the IST in our IST Technical Proposal. Development of the IST began on 06-AUG-19 with a purchase order from UCL. Further development was funded by an SBIR grant from the NIH from April 2019 to April 2020. From April 2020 to present, development funded by OSI, with collaboration on imlantation at Cornell University and the Institute of Neurology at UCL.

[16-JUN-21] Each version of the IST has a nominal battery capacity. The A3036A battery, for example, has a nominal capacity 19 mA-hr, which means it can deliver 19 mA for an hour or 1 mA for nineteen hours. We discharged and charged the battery of device A215.10 five times, and charged once more with an A3033A. Once charged, it reported battery voltage 4.3 V when inactive. We attached a white LED, turned on lamp power continuously and measured 20.0 mA current with the battery reporting 4.1 V. The lead resistance is 56 Ω. The LED forward voltage drop is 2.98 V. We turned on a 50% stimulus ofd 2-ms pulses at 250 Hz and recorded battery voltage versus time.

Figure: Discharge of an A3033A Battery Delivering 50% Stimulus to a White LED. Total charge delivered is 17 mA-hr.

Part-way through, we measure 11 mA current with the lamp turned on continuously, and the battery is reporting 3.6 V, which is also consistent with LED forward voltage 2.98 V. Thus we are able to calculate the LED current for intermediate battery voltages and integrate to obtain the total charge delivered by the battery versus time. The current consumed from the battery while the lamp is on depends upon the stimulus leads and the battery voltage.

Figure: Battery Current with Lamp at Full Power versus Battery Voltage. The lamp is a C470EZ500 LED. We compare 120-mm long 0.8 mm diameter D-Leads (Blue) to 50-mm long 0.7-mm diameter B-Leads (Orange).

The table below gives the current consumption of all versions of the A3036 in various states and allows us to estimate the battery life of the A3036 in various use cases.

Inactive0.0075Logic powered off, crystal radio awaiting command.
Active0.15Logic powered on, lamp off, transmit off.
Transmit0.25Logic powered on, lamp off, transmit on.
Flash16Lamp on, 50-mm long 0.7-mm diameter B-Leads.
Flash22Lamp on, 120-mm long 0.8-mm diameter D-Leads.
Table: Current Consumption of the A3036 All Versions.

Suppose we implant a fully-charged A3036C in a mouse. Each day we turn on the lamp with 10-ms pulses at 10 Hz for 100 s three times to see if our mouse turns in circles. We have 50-mm B-Leads for the lamp connection. The lamp is on 10% of the time during each stimulus, so current consumption is 10% of 16 mA, or 1.6 mA. Five stimuli of 100 s is 800 mAs = 0.22 mAhr. For the remaining 85900 s of the day, the A3036C consumes 7.5 μA, or 0.18 mAhr. Each day we consume 0.40 mAhr. The A3036C's battery capacity is 10 mAhr, so we can run this experiment for 25 days.

Preparing an A3036B for encapsulation is similar to preparing a Subcutaneous Transmitter (A3028) for encapsulation, with the added steps of tuning the antenna matching network and calibrating the battery voltage sensor, and the added burden of a two-minute firmware compile time and thirty-second logic chip programming time. The first thing we do is prepare the board for tuning by configuring its logic chip.

  1. Attach an antenna to the circuit so we can tune its matching network. Solder a thin, 50-mm insulated wire to the antenna pad.
  2. We must program the A3036B so that it does not switch off the crystal radio while we are applying power to the antenna. Open the Lattice Diamond Programmer. Tell it to open the P3036B.xcf project. The project should specify a firmware file, something like P3036B_1_18_5_60.jed. Connect the programming cable to A3036B.
  3. We need to apply power to the circuit in order to program it, and we need a LiPo battery because the programming process draws up to 25 mA from the power supply. Connect a LiPo battery to the programming extension.
  4. We need 910 MHz power to turn on the A3036B. There is no convenient programming extension connection to allow us to turn the board on with a jumper link. There was supposed to be, but a design error stopped this feature from working. Use a Command Transmitter (A3029C) to apply +20 dBm of 910 MHz to a Loop Antenna (A3015C). With the Diagnostic Instrument transmit command 0x0081 to the A3029C to turn on its 910 MHz output. We don't need to connect boost power to the A3029C.
  5. Place the A3036B antenna next to the transmit antenna. Although we have not yet tuned the antenna matching network, the crystal radio will still be sensitive enough to turn on power to the logic chip. The board should now be powered up and ready to program, so program the logic chip. This will take two minutes. When complete, the board is now ready for tuning.

We tune the antenna matching network with a 870-960 MHz sweep generated by a Modulating Transmitter (A3014MT) and an amplifier. For our amplifier we are using a modified A3029C circuit board to produce a 16-dBm sweep. We have L11 removed, and we use J4 to bring in the −3 dBm sweep from the A3014MT, so that the A3029C output is our +16 dBm sweep. We take the sweep signal and split it in two. One part we mix with 910 MHz to produce a ±45 MHz intermediate frequency, which we view on an oscilloscope. The second part we apply to an A3015C loop antenna. We place the A3036B with battery on the loop antenna and measure VR at U5-3 with a ×1 probe. We must have the battery attached so the RF switch U4 is powered up, connecting the antenna to the matching network. We will see a sweep something like this. With favorable orientation of the antennas, the peak should be around 100 mV high, and somewhere within the range 900-930 MHz. If it is higher than 915 MHz, we increase C6 by 0.1 pF. If lower than 905 MHz, we decrease C6 by 0.1 pF. We wash and dry the board, then repeat our measurement and perform further adjustments until the peak response lies in the range 905-915 MHz.

The antenna matching network is tuned and the logic chip is configured. If the firmware file we uploaded to the logic chip was P3036B_1_18_5_60, then the circuit has been configured with the following constants, as presented in its VHDL source code, which will soon edit as part of our calibration and configuration procedure.

-- Configuration and Calibration Constants.
	constant device_id : integer := 1; -- The identifier number for this device
	constant fck_calib : integer := 18; -- Calibration of ring oscillator to give 10 MHz on FCK
	constant frequency_low : integer := 5; -- Low frequency code for data transmission.
	constant battery_calib : integer := 60; -- Offset to make battery measurement 200 at 3.70 V.
-- Stable Calibration Constants
	constant frequency_step : integer := 1; -- High minus low frequency code
	constant sample_rate : integer := 1024; -- How often we transmit diagnostic signal

Knowing the initial configuration values allows us to estimate the correct values after we measure the behavior of the circuit. To configure and calibrate the circuit, we follow the steps listed below. As we proceed, we record all measurements and configuration constants in a table on the back of the batch production record sheet.

  1. Set up a Command Transmitter (A3029C) and Loop Antenna (A3015C) as when we first programmed the A3036B. Turn off the A3029C output with the Sleep button in the Diagnostic Instrument.
  2. Connect a 3.7 V power supply to the programming extension through an ammeter. Measure the current consumption of the circuit. This measurement is the inactive current. Acceptable values are 6-9 μA.
  3. Open the Lattice Diamond Compiler and open the P3036 project. Open the main VHDL file and locate the above lines. These are the ones we will edit.
  4. Change the range of the ammeter to 20 mA or higher. A lower ammeter range will stop the short burst of 20 mA drawn from the battery when the circuit first powers up, and the circuit will not obey commands. Use the Stimulator Tool to turn on transmission of the A3036B synchronizing signal with the Xon button. We know the circuit's id because we set it with the initial program we used before tuning the matching network. Check that the synchronizing signal is being transmitted with a second Loop Antenna (A3015C) connected to an Octal Data Receiver (A3027). Measure the spectrum of the A3036B transmit signal using a third loop antenna connected to a Spectrometer (A3008C). The peak of the spectrum should lie in the range 913-918 MHz. If the peak is too high, reduce frequency_low in the VHDL file. If too low, increase frequency_low.
  5. We must calibrate the logic chip's ring oscillator. Put an oscilloscope probe on TP1, which is pin P3-1 on the programming extension, and presents us with the FCK (Fast ClocK) signal. We want FCK to be a 10-MHz square wave. If its period is greater than 110 ns, reduce fck_calib. If its period is less than 95 ns, increase fck_calib.
  6. Turn off the synchronizing signal with the Stimulator Tool. Use the Receiver Instrument to download continuously from the Octal Data Receiver. This continuous download provides acknowledgment and battery measurements to the Stimulator Tool. Use your oscilloscope to measure VB, which is pin P3-4. Obtain a battery measurement from the A3036B with the Battery button. If the measurement is more than 0.1V too high, increase battery_calib by 5. If more than 0.1V too low, decrease battery_calib by 5.
  7. Set the device_id to the value we want for this circuit. If we want the synchronizing signal to have some frequency other than the default 1024 SPS, we can set it so some other power of two samples per second now. Compile the code, connect the programming cable to the programming extension, and turn on command transmitter power with the 0x0081 command in the Diagnostic Instrument. The board should now be powered up, which you will observe in the ammeter, which will report current of 3-4 mA. Press the Program button in the Lattice Diamond Compiler and use the programing panel to program the logic chip with the new firmware.
  8. Repeat our measurements of transmit spectrum and transmit clock frequencies, adjust VHDL program, and re-program logic chip until both frequencies are correct.
  9. Use the Stimulator Tool Xon button to turn on synchronizing signal transmission. Measure current consumption. You will have to set the ammeter range to 20 mA or higher to turn on the synchronizing signal, then drop to a lower range to measure the current consumption in microamps. We call this measurement the transmit current. Acceptable values for 1024 SPS are 200-300 μA.
  10. Start a stimulus consisting of a large number of 10-ms flashes every 1000 ms with the Start button. Turn off signal transmission with Xoff. By starting the stimulus and then turning off transmission, you do not have to switch the ammeter to a higher range, because the board is powered up the entire time. The board needs 20 mA ony when it wakes up from its inactive state. Measure current consumption during the stimulus. This measurement is the active current. Acceptable values are 100-200 μA. No light will flash because we have no light attached to the circuit board.

Assuming the currents and frequencies are within their acceptable ranges, the board is now ready to for its batteries and leads, and after that we have quality assurance to check all the above measurements again.


[18-NOV-19] The A3036AV1 needs the following modifications.

  1. Connect D2-1 to U5-2 with wire link in order to correct a PCB error and complete the 0V net.
  2. Replace D1 SMS7621079LF with SMS7630079LF. The Pin-1 marker on the new diode is the anode, while on the previous diode it was the cathode.
  3. Add 0.5 pF in parallel with C6 to bring matching network resonant frequency close to 910 MHz.

[12-FEB-20] For the A3036BV1 we make the following enhancements to the circuit.

  1. Fix connection to 0V on ground plane at battery.
  2. Switch D1 to SMS7630040LF in P0402 package.
  3. Switch D2 to CDBQC0130L in P0402 package.
  4. Change C6 to 1.5 pF.



[23-AUG-19] Schematic complete. Arrange components on 10-mm × 10 mm circuit board, narrowed at the top to fit into the terminal end of our 19-mAhr LiPo battery. The A3036 crystal radio uses the SMS7630079LF detector diode in a 1.7-mm long SC-79 package. The threshold comparator is the same MCP6541 but in the smaller SC-70-5 package. The stimulus and battery-check switches are two N-channel enhancement mode mosfets provided by a DMG1024UV in a 1.7 mm × 1.0 mm SOT-563 package. The OR gate is is the same SN74AUP1G32, but in the much smaller UDFN-6 1.0 mm × 1.5 mm package, which we use on our A3028GV1 circuit boards. This package fails to load properly in roughly 1% of assemblies and is impractical to replace by hand, but it is small. The logic chip is the same LCMXO2-1200ZE, but in a 2.5 mm square WLCSP-25 package with balls on a 0.4-mm pitch. We are not yet sure how to make a footprint for so fine a ball pitch.

[04-SEP-19] The WLCSP-25 package has twenty-five balls on a 0.4-mm (15.7-mil) pitch. With 10-mil diameter pads the clearance between pads is only 6 mil, which is barely enough to run a 2-mil track, let alone the 5-mil tracks of our usual fabrication process. We consult with Epectec for a solution to the layout problem. They propose that we use 12-mil pads with 10-mil soldermask opening and a 4-mil laser-drilled micro-via from the top copper (L1) to the first middle copper layer (L2).

Figure: BGA Footprint on A303601A Printed Circuit Board. Shown are the top copper (L1), top soldermask, and the drill holes.

The second layer of copper (L2) has 10-mil pads to receive the 4-mil microvias, as shown below.

Figure: The Middle Copper (L2) Beneath the BGA. Shown are the middle copper tracks with 10-mil pads to receive the microvias, and the 4-mil microvia holes themselves.

There are no pads for the 4-mil holes on the remaining four copper layers: ground plane (L3), power plane (L4), middle copper (L5) and bottom copper (L6). But our drill file specifies 4-mil holes, and we see these rendered in the plot below.

Figure: The L3-L6 Copper Layers. There are no pads for the 4-mil holes, because the 4-mil holes are indented to pass only between the L1 and L2 layers. These holes are, nevertheless, shown in this plot, because our drill file cannot specify partial drill holes.

We could have routed tracks directly beneath the microvia holes, but our layout software is incapable of understanding drill holes that pass between only two layers. So we kept the bottom four layers clear of copper around the drill holes. In order to generate the gerber files for the bottom four layers, we deleted the 4-mil vias temporarily.

[06-SEP-19] Completed draft version of P3036A01 firmware, including battery voltage measurement by timing how long it takes to charge up C4. When we request a battery measurement, the firmware will assert BT, which closes the battery test switch U3-6 to U3-1. Capacitor C4 starts to charge through R1 and R2. These resistors present a voltage source 80.5% of VB through a resistance of 6.43 kΩ. Capacitor C4 is 1.0 μF. The charging time constant is 6.4 ms. The BTV signal connects to an input on U8, a 3.0-V logic input with logic threshold around 1.5 V. When VB = 3.6 V, voltage BTV will take around 3 ms to reach 1.5 V, or 100 cycles of our on-board 32.768 kHz clock. To the first approximation, time taken is inversely proportional to the battery voltage, so we expect 86 periods when VB = 4.2 V and 106 periods when VB = 3.4 V. Submit A303601A Rev 1 printed circuit board for fabrication on a ten-day turn.


[30-SEP-19] We have the A303601A printed circuit boards, quantity 100, in panels of 10 each. We solder a battery to a blank circuit board, and a charging connector. We succeed on the fourth battery, having figure out how to fold the battery tabs, solder them to the pads on the circuit board, and press the board into place. One of our 0-V connections is missing due to a bug in the way our PCB layout software implements relief connections to the ground plane. We find we can correct this error with a short wire link.

Figure: A 19-mAhr LiPo Battery Loaded On A303601A PCB.

Once the programming extension is clipped off, the battery and circuit will be contained within a 10 mm × 20 mm × 3 mm cuboid. Allowing for 0.5 mm of epoxy and silicone over all surfaces, final volume will be ≤0.92 ml. We are hoping for 0.80 ml. We charge the battery.

[03-OCT-19] We ship components and circuit boards to our assembly house. We have ordered 20 circuits made. We should receive first article in two weeks.

[04-OCT-19] We receive 1254 of C527TR2227 (green ≥10 mW @ 20 mA) and 1777 of C460TR2227 (blue ≥30 mW @ 20 mA). They come on plastic sheets. They are so thin they are transparent.

Figure: Twelve C527TR2227 On Plastic Sheet. Each is 270 μm × 220 μm.

We ship all the green ones to an assembly company we have contracted to load the dies onto the A303602A printed circuit board, which we submit for fabrication on a 25-day turn.

[11-OCT-19] We have our first three assembled A3036AV1 circuits, No1-No3. We connect D2-1 to U5-2 in order to correct a PCB error and complete the 0V net. When we connect power, current consumption is 7 μA. We see a 3.3-V 32.768-kHz square wave on both sides of R7. The A3036AV1 has no resistor between OND and its two destinations U6-2 and P3-3. If we tie P3-3 to P3-4, we connect a LO output of U8 to a Hi voltage. We remove U6 on No1 and use its pads to connect 3VB to 3VA. We set VB to 3.3 V. On our programming cable, we connect ispEN to VCC. When we plug the programmer into P1, the cable drives PEN HI so as to enable the JTAG interface. We program U8 repeatedly from our laptop. Once we program U8, we see a 2.8-V square wave on U7-2 and a 1.9-V square wave on RCK. No matter how we configure TP1 and TP2, we cannot change their state. TP1 remains Hi and TP2 remains Lo. We remove U6 on No2 and repeat the above steps, arriving at the same result.

Figure: Assembly of Prototype A3036IL-A450-8. A blue EZ500 LED is faintly illuminated to make the picture. Clear glue covers the base of a fiber light guide. The LED sits in a 3-mm square package, bonded to the interior pads.

[14-OCT-19] We use the Lattice Diamond Programmer V3.9, instead of the older ispVM 18.0, to program No1. Now RCK is a 3.0-V square wave, and we can program the behavior of TP1 and TP2 as we wish. We add output KEEPER on unused pin B1 and set equal to STBY in the firmware so as to stop the compiler eliminating the power control unit. We had deleted this output when we adapted the A3030 firmware for the A3036. We solder a white LED with 100-Ω resistor to the L+ and L− outputs of No3, and a steel antenna to A. The No3 circuit still retains U5. We place the antenna over a Loop Antenna (A3015C) driven with 10 dBm of 910 MHz. With the logic chip erased, XEN is Lo, the RF power causes RP to go Hi, and chip power turns on, after which the erased chip asserts OND, ONL, and XEN, keeping its own power turned on, turning on the lamp, and disconnecting the antenna from the tuner. We connect the programming cable. We remove from our source of 910 MHz. We program the device successfully. Standby current is 7.4 μA.

Figure: A3036AV1 Prototype Circuit with 3-mm White LED and Steel Antenna.

We look at the RF power signal on D1-1, which we call VR. We have the A3036AV1 over a source of 910 MHz. When VR reaches 10 mV, RP is asserted and both 1V2 and 3VB turn on. With a Command Transmitter (A3029C) we are able to flash the white LED. When the device receives a STOP command, it flashes its lamp. When 1V2 turns on, OND, ONL, and XEN all rise to about 1.5 V for 1 ms. We enable acknowledgments and the No3 circuit transmits an RF signal. The signal is not received by our data receiver. We have not yet calibrated the RF canter frequency or its modulation clock.

During stimulation, the current in No3 when the lamp is off is 220 μA. We solder antennas and lamps to No1 and No2. These are permanently on because of our removal of U6. Their quiescent current consumption is 150 and 250 μA respectively. This is what we call the "active" current of the device, and we expect it to be around 60 μA. Looking at the Lattice Diamond power calculator, this 200 μA quiescent current is consistent with a failure to turn off the band gap references. We try to calibrate the RF center frequency and the transmit clock, but we find that the ring oscillator is not working. It is being eliminated from the design. We believe we can fix these problems by attention to the firmware, so we will proceed with assembly of another 17 circuits.

We add routing priority for the ring oscillator bits, and the fck divisor bits, see P3036A.lpf. Now the compiler preserves all bits and we see the ring and divisor in the device view. We add frequency estimates for the four clocks. We note that we have SDM_PORT set to the default value DISABLE, which means that outputs DONE, INITN, and PROGRAMN are available for general-purpose I/O. We are using one of these, INITN, for our output XEN.

[16-OCT-19] We are able to persuade the compiler to retain our ring oscillator only if we disable the ring oscillator for some fraction of the time. So we turn it on and off with RCK and route it to TP1 directly on No1. The ring oscillator runs 50% of the time. Current consumption is 5.2 mA. The ring oscillator output has period 8.5 ns, frequency 118 MHz. Instead of running the ring oscillator directly to TP1 we run TCK to TP1 and RCK to TP2. We see TCK with period 201 ns. Current consumption 750 μA. Of this, 150 μA is present without the ring oscillator, so the oscillator consumes 600 μA for 50% of the time, the equivalent of 1200 μA for 118 MHz continuously, or 10 μA/MHz, which compares well to the 11 μA/MHz we observed with the A3030 oscillator.

[05-NOV-19] We calibrate the ring oscillator on No2, one of the circuits that is always on. We have not calibrated its RF center frequency. The transmit clock period (TCK period) is 190 ns, which is on the edge of being too low. But we receive acknowledgements from the device. We study the battery monitor, which consists of U3-2, R1, R2, and C4. When we assert BT, U3-2 turns on, charging C4 (1 μF) through R1 (8 kΩ), with R2 (33 kΩ) present to limit the voltage we attain at BTV, and to discharge C4 between measurements. Eventually BTV reaches the threshold of the Schmitt trigger input on U8-D5, which we have set to have "small" hysteresis. We measure the charging time in RCK periods, which are 30.5 μs. The charging time constant is 6.4 ms, or 200 periods. The A3036A returns the charge time in the data of an auxiliary message on channel 15. We vary VB and transmit battery commands to No2 and plot the charge time versus voltage.

Figure: Capacitor Charge Time versus Battery Voltage. The charge time is the number of 32.768 kHz periods it takes to charge C2 to the logic threshold of roughly 1.5 V. The device does not transmit for VB ≤ 2.5 V.

We prepare ISL Controller Tool 6.1, which supports the A3036A battery measurement. We are able to measure the battery voltage with better than ±0.5 V precision around the critical region of 3.6-3.8 V.

[08-NOV-19] We take No2 and remove the wire between 3VA and 3VB. We see 3VB at 0.2 V and 1V2 at 0.0 V. Current consumption from VB is 8.7 μA. We connect 3VB to VB with 1 kΩ using the pins of P1 and P3. We apply 3.3 V to VB. We see 2.3 mV across the 1 kΩ, implying that U8's I/O circuits, U9's shutdown current, and U2's enable input current are together 2.3 μA. We find we can program the board with the 1 kΩ VB to 3VB series resistor in place. When we turn on the lamp, TP2 is asserted and we see 360 mV across our 1 kΩ resistor, which is the current flowing through the 1 kΩ and R13 (8 kΩ). With the lamp off, the total current consumption from VB is 213 μA. The Lattice Diamond power calculators says the core current consumption of U8 when the band-gaps are turned off is 60 μA, which was what our A3030 circuits consumed with almost identical firmware.

We install the Lattice Diamond 3.11.0. We were previously using 2.2.0. In both cases, we select the LSE synthesis tool. We re-create the P3036 firmware for LD3.11 and compile. Total current consumption 220 μA, command reception and acknowledgement working. With fck_divisor = 11 we get TCK period 188 ns, implying ring oscillator 117 MHz. With fck_divisor = 12 we get 220 ns, implying ring frequency 109 MHz. We have some work to do stabilizing the ring oscillator with respect to fck_divisor.

We are using Lattice Diamond Programmer 3.9.0. We discover the TCK Divisor setting. Programming is unreliable with the divisor set to 1. With the divisor set to 2, scanning and programming works every time, but takes twice as long: 123 seconds. Current consumption of No2 remains 224 μA. In order to free up U8 pins A4, B4, A5, and C5 for general-purpose I/O, we have disabled the JTACK interface by default. We now enable the JTAG interface and program again. We remove R1 because BT (TDO) is HI. Current consumption 218 μA.

We turn on the RF oscillator continuously, at the LO frequency. With frequency_low = 7, the center frequency is 909.5 MHz, and with frequency_low = 8, center is at 915.5 MHz. We set frequency_step = 2 and frequency_low = 7 and obtain reception of acknowledgements.

We route the BTV logic signal to TP1. This signal is the voltage BTV at U3-1 passed through U8-D5's Schmitt trigger input. Let's call it BTVL. We look at BT, BTV, and BTVL during a battery test for No1. Battery voltage is 3.58 V. Acknowledgments are turned off, so the only response of the device is the battery message.

Figure:BTV (yellow, 500 mV/div), BT (blue, 2 V/div), and BTVL (red, 2 V/div) During Battery Test, No1. Charge time is 4.68 ms according to the cursors.

We are not receiving acknowledgements from No1. The TCK period is correct, but we have not checked the RF center frequency. We don't have its report of the charge time in RCK periods. But we expect from our earlier calibration that this will be around 120 periods, or 3.66 ms, which is not consistent with the above measurement. The above measurement was performed with this same circuit, but acknowledgments were enabled. We repeat for No2.

Figure:BTV (yellow, 500 mV/div), BT (blue, 2 V/div), and BTVL (red, 2 V/div) During Battery Test, No2. Charge time is 5.16 ms according to the cursors.

We do receive acknowledgements from No2, and it says 172 every time, or 5.25 ms, which is consistent with the the above observation.

[20-NOV-19] We check the response of the A3036AV1 to a −5 dBm sweep applied to an A3015C antenna, duplicating our earlier measurement of the A3030E antenna network. We see no response on VR in the A3036AV1. The A3036AV1 is loaded with D1 = SMS7621079LF, which turns out to be the wrong member of the family. The SMS7621 is designed to measure power from a 50-Ω source, and requires forward bias 300 mV for a current of 1 mA. The diode we are replacing is the now-obsolete HSMS285C, which requires only 150 mV for 100 μA and 250 mV for 1 mA forward current. Its video resistance is 8 kΩ at zero bias. Instead of the SMS7621, we should have used the SMS7630, a zero-bias detector diode with video resistance 5 kΩ, requiring 100 mV for 100 μA and 200 mV for 1 mA forward current. We order these. In the meantime, we supply 10 dBm of RF power to an A3015C loop antenna and place an A3036AV1 on top. Matching network resonance is at 1000 MHz. We add 0.5 pF to C6 and resonance drops to 920 MHz.

[25-NOV-19] We apply +10 dB to an A3015C. We place an un-tuned A3030E circuit board on the antenna and measure VR as we increase frequency from 900-1000 GHz. We see a maximum of 60 mV at 970 MHz. We repeat with A3036AV1 with SMS7621 diode and C6 = 1.5 pF. We see a maximum of 60 mV at 912 MHz. We apply −5 dBm sweep at 10 kHz and see 20 mV peak on the A3030E at around 975 MHz and <2 mV on A3036AV1 at around 915 MHz. Replace D1 with SMS7630, taking care to place the pin 1 marker on the other side of the footprint. We see 20 mV on A3036AV1 at around 915 MHz. The video resistance of the SMS7621 is so high that the diode cannot charge C7 = 100 pF fast enough to respond to the sweep with a 20-mV peak. That of the SMS7630 is 5 kΩ, giving a time constant of 0.5 μs with C7.

[26-NOV-19] We vary the value of the ring oscillator divisor and measure the transmit clock period. We want to see the period increasing with divisor, and we do from 7 to 14, but not at 15. We find that our ring oscillator is being optimized for speed, but not the fast clock divider. When we turn on optimization for the fast clock divider, the TCK period is unstable, so we are leaving the divider without optimization for now.

Figure: TCK Period versus FCK Divisor in P3036.

We connect a Modulating Transmitter (A3014MT) to a Loop Antenna (A3015C). We modulate the RF with a square wave so that it moves in and out of the center frequency of our A3036A with SMS7630 diode. Rise and fall times are around 1 μs.

Figure:Rise and Fall Time of VR with D1 = SMS7630. For the same response of the SMS7621 diode see here.

We select a circuit with D1 = SMS7621 and add 0.5 pF to C6. We repeat the above experiment with this new circuit and our rise and fall times are of order 4 μs. We sweep the A3014MT output frequency with a ramp. We use our SSG-6001 synthesizer and a ZAD-11 mixer to calibrate the sweep on our oscilloscope screen. We connect the sweep to an A3015C that is flat on our bench. We move the A3036A around on the loop until we maximize the peak in VR.

Figure:Response of VR to 10-kHz Sweep 840-940 MHz when D1 = SMS7630. For the same response of the SMS7621 diode see here. The left cursor is 890 MHz, the center is 910 MHz, and the right cursor is 930 MHz.

With the SMS7630 diode, we obtain a peak of 40 mV. But with the SMS7621 diode the peak is no more than 10 mV. We add 0.5 pF to C6 and switch D1 to SMS7630 on all remaining A3036 circuits. Peak of sweep response for all ten circuits in the range 903-913 MHz.

We switch the BTV signal input hysteresis from SMALL to LARGE. We set battery_calib in the P3036A03 firmware to 58 and now the battery counter is 200 for VB = 3.70 V. We do the same for No3. In order to calibrate No3 we send command 0081 to the Command Transmitter (A3029C) to turn on RF continuously, which allows No3 to remain on during programming. We have battery calibration 61, which results in battery counter 199 at 3.70 V, 184 at 4.00 V, and 218 at 3.40 V. We use −60 for the scaling factor in V/cnt and our reference will be a count of 200 for voltage 3.7 V.

[27-NOV-19] Inactive quiescent current of No3 and No4 is 7.3 and and 7.3 μA. We program and calibrate No4. We now have 3 A3036A calibrated for battery monitoring, their battery calibration values are 56-61. We attempt to program and calibrate more circuits, but have difficulty programming, and once programmed, they do not respond to commands, not even with an increase in current consumption.

[28-NOV-19] We devise an easier way to connect U5-3 to 0V, by connecting D2-1 to U9-3. This leaves D2-2 free for a wire to look at VR. We start with No6, and succeed in programming and calibrating. We modify No7 and program, but U7 is not producing RCK. We modify No5 and program, but VR is oscillating, spending 90% of the time at 10 mV and rising for 10% of the time to 100 mV, every 70-100 μs, the period varying as we move our finger around the circuit. This behavior is consistent with a matching oscillation on XEN, whereby U4 is connecting the antenna to the tuner only 10% of the time. WEe put a probe on XEN, but the oscillations stop and XEN stays HI, turning on U9, and disconnecting the tuner. Current consumption is 12 mA.

[02-DEC-19] Of 10 A3036AV1 circuits we have corrected with modifications, six are calibrated and ready for battery and encapsulation.

Figure: Calibration Constants and Current Consumption of A3036A. Circuit A3036AV1 with modifications. Active current is with stimulus with rare pulses, Xmit current is while transmitting synchronizing signal, no stimulus.

The active current consumption remains roughly 70 μA higher than the 60 μA suggested by the Lattice Diamond power calculator for this design. The same functionality loaded into the A3030E circuit yields active current consumption of around 70 μA.

[03-DEC-19] We have five A3036A programmed with firmware P3036A03 ready for encapsulation, with batteries loaded, 45-mm lamp leads, and 30-mm stranded antennas.

Figure: Batch of A3036A Ready for Encapsulation.

The leads are our 0.7-mm diameter silicone insulated leads, red for L+ and blue for L−. On one device, we measure lead resistance 27 Ω and 29 Ω. We expect 6.0 Ω/cm = 54 Ω. Assuming battery voltage 3.7 V, LED forward voltage drop 2.9 V and lead resistance 54 Ω, we expect LED current (3.7 − 2.9)/50 = 16 mA.

[04-DEC-19] We have 14 of 460-nm EZ500 LEDs mounted to A303602B printed circuit boards. We run 30 mA through each and measure optical power output. We get 25.3±1 mW. This compares well with the average 26.9 mW we obtained with LEDs from the same batch earlier.

[06-DEC-19] We have five encapsulated A3036A. Devices A215.3, A215.4, A215.8, and A215.10 respond to commands, transmit the synchronizing signal, and deliver power to LEDs. Device A215.9 does not respond to commands. We place the four working devices with lamps in our Faraday enclosure with data antenna 15 cm to one side and command antenna 15 cm to the other. We connect boost power to the command transmitter. Reception of commands is unreliable. We place the four devices in a beaker of water. Their bodies rest upon the glass bottom. Reception is now reliable. We turn on the lamps 90% of the time with a continuous stimulus. It takes ten to twenty minutes for the batteries to drop to 3.5 V according to the ISL Controller's battery voltage measurements. We connect 5.0 V to the lamp pins of A215.10 and see 9 mA flowing in.

Figure: Current versus Voltage for Charging Diode. We connect 200 Ω from VB to V0 and apply a voltage across VB and L−. We measure current flowing into VB and we measure the voltage between 0V and L−. The former is the diode current, the latter is the diode forward voltage drop.

The re-charge diode is the diode in parallel with the drain of U3-3, see schematic. The recharge current enters through L+ end exits through L− In order to exit through L−, it travels along the 0V net through the re-charge diode. The current through the battery is the total incoming current minus the circuit's inactive quiescent current, which is around 7.3 μA. We want to charge the battery to 4.2 V and then reduce the current slowly to zero while maintaining a battery voltage of 4.2 V. 5.0 V connected to the lamp leads, the voltage across the battery will be 4.2 V when the current flowing through the diode and the 54-Ω lamp leads is 4.0 V. When the current drops to 1 mA, however, the battery voltage will be 4.5 V, which could damage the battery. If we stop the re-charge when the current is 3 mA, we stop it when the battery voltage reaches 4.3 V, which may be okay. But if we set the charging voltage to 4.7 V, the battery voltage will be 4.0 V at 3 mA, 4.2 V at 1 mA, and 4.3 V at 0.1 mA.

We have been charging A215.10 for twenty minutes with 5.0 V and the charge current has dropped to 7.5 mA. We drop the charge voltage to 4.7 V and the current drops to 3.4 mA. We connect all four ISTs to the same 4.7 V source and leave to charge.

[09-DEC-19] We have two A3036IL-A8 prepared. No1 we moved the fiber before the epoxy was cured. No2 we allowed to cure without interference. Today we measure their output power with a photodiode and obtain 2.9 mW for No1 and 4.0 mW for No2. Average LED output at 30 mA is 25 mW, so let us assume 12.5 mW at 15 mA, giving us 23% and 32% coupling efficiency. In earlier work our best head fixture provided 41%, our worst provided 22%, and our median was 38%. Low coupling efficiency was a strong function of the quality of the taper base. These two tapers were both chipped at the base, but only in the cladding, so we trusted that they would work adequately.

[10-DEC-19] We push three of A3036A into a 10-ml graduated tube containing 5 ml of water. The water level goes up to 7.7 ml, making the individual A3036A volume 0.9 ml. We are testing lamps and batteries, and we notice that the lamp flashes when we perform a battery measurement, on both No8 and No10. We have not checked No3 and No4.

[11-DEC-19] We have A3036IL-A8 No3, having snapped the fiber on No2, leaving us with only No1 24% efficient. With 15 mA current into No3 we see photocurrent 0.33 mA, or 4.1 mW for 33% efficiency. We glue the fiber onto No4 and leave to cure. We use our prototype A3033A to charge A3036A No3, No4, No8, and No10 until charge current is less than 1 mA.

Figure: Green TR2227 Mounted on Circuit Board. From the Palomar Technologies.

[12-DEC-19] A3036IL-A8 No4 emits 3.7 mW at 15 mA for 30% efficiency. And then the fiber fell out of the glue joint. We throw away the glue cartridge, which is three or four years old, even though our glue samles have all set well. We take out a new cartridge that is less than a year old. We set up a new LED No5 and measure power output before applying fiber: 11 mW for 15 mA. We glue a new fiber.

[13-DEC-19] A3036IL-A8 emits 4.0 mW at 15 mA for 36% efficiency. We place an IST A3036A attached to an A3036IL-A in 200 ml of water and move the beaker around on an ALT platform, which is 32 cm by 16 cm, all parts, especially the corners. We try the IST lying flat on the bottom of the beaker, suspended against the side wall, and suspended in the center. We open and close the enclosure door. We send stimulus commands and observe by eye if the IST received the command. Of 100 stimuli, the IST received 96%. The lost 4% were all in one far corner of the platform with the door closed.

We have the IST Prototype system ready to ship: 3 of IST A3036A (A215.3, A215.4, A215.8), 8 of ILED A3036IL-A, 2 of FCLED A3036IL-A8, 1 of Command Transmitter A3029C, 1 BNC elbow, 1 Boost Power Supply for A3029C, 4 BNC feedthroughs, 4 short antenna cables, 1 of A3015C with cable, and 1 Battery Charger A3033A. We ship DHL to London.

We have A3036A A215.10. When inactive, battery measurement is 4.2 V. When active, 4.0. We start a 10-Hz, stimulus of 30-ms pulses. After 9 minutes, battery reports 3.9 V, after 17 minutes, battery reports 3.8 V. We stop stimulus, battery reports 4.1 V.

[16-DEC-19] Continue 30% stimulus with A215.10. We note that frequency of stimulation, as seen in Neuroarchiver spectrum of synchronizing signal, is 10.25 Hz, not the 10 Hz wer requested. After a total of 58 minutes since start of drain, battery reports 3.7 V. We stop stimulus, battery reports 3.8 V.

[17-DEC-19] Continue 30% stimulus with of A215.10. Before start, battery reports 4.0 V, after start, reports 3.7 V, but this rises to 3.8 V after a few minutes. After a total of 76 minutes since start of drain, 3.7 V. After a total of 126 minutes, 3.6 V. We stop, battery reports 3.8 V. Later, we continue for another 33 minutes until the battery voltage has dropped to 3.4 V, and then stop. Total time is now 159 minutes. We connect stimulus leads to 4.7 V.

[18-DEC-19] Disconnect A215.10 from 4.7 V charging voltage. It now reports 4.3 V battery when inactive, 4.1 V when transmitting its synchronizing signal, and 4.1 V when generating 30% stimulus. After 23 minutes, reporting 3.9 V. We now switch to a 1% stimulus: 1 ms pulses at 10 Hz and leave running. After 490 minuts of 1%, reporting 4.0 V.

[19-DEC-19] At 9:21 am lamp still flashing at 1%, battery reports 3.8 V. So far 22 min at 30% and 11 hrs at 1 %. Change to 5%, battery reports 3.8 V.

[20-DEC-19] At 9:13 am lamp not flashing. Cannot get a battery report or acknowledgement from the device. Connect to 4.7 V charger at 9:18 am, charge current 0.01 A. At 9:37 am charge current is 5.6 mA. Disconnect and obtain battery report 3.7 V. Reconnect to charger, 5.6 mA. At 10:33 am charge current is 4.6 mA. At 11:20 am battery reports 4.0 V and charge current is 4.0 mA. At this current, charging diode drops 0.55 V. Assuming leads are 56 Ω, we expect VB = 3.93 V. At 15:27 charge current 0.7 mA, reports 4.2 V. Start 50% stimulus, battery reports 4.0 V.


[23-DEC-19] Friday and today we get 129 minutes of 50% illumination from A215.10 at which point battery reports 3.4 V. We continue with 10% for 17 minute until battery reports 2.9 V. The lamp still flashes, but we perceive it to be dimmer. Connect to 4.7 V charger and see 11 mA. After 18 minutes 6.0 mA, 27 minutes 5.7 mA, 120 minutes 4.0 mA, 300 minutes 1.1 mA. Remove and connect LED. Battery report 4.2 V. Start 20% stimulation at 15:40.

[24-DEC-19] Our 20% stimulation ended some time before 10:30 am. We charge battery for half an hour, then run 250 Hz 1-ms pulses, after half an hour, battery reports 3.5 V. We attach to A3033A charger at 11:45 am. We have ten more A3036A circuits modified, cleaned, and dried. We are numbering these No4-No13 using the labels on circuit boards. Inactive current 7.47plusmn;0.1 μA for all but No9, which consumes 12 mA and we reject.

[26-DEC-19] We use the Tapermaker Tool to make tapered optical fibers using our new 270-μm diameter high-index glass fiber, with numerical aperture 0.86 and diameter tolerance ±30 μm. We have not bothered polishing the bases of the fibers. We are working on getting a symmetric, uniform taper at the tip.

Figure: Tapered 270-μm Diameter Fibers. The longest is 9.6 mm from tip to base, and the shortest is 6.5 mm.

We are confident we can make 6-mm long fibers with 1-mm tapers at the tip. We can control the length of the taper with the configuration of the tapering process.

[27-DEC-19] We have A215.10 recharged over-night, reporting 4.3 V when inactive. We start 10 ms pulses at 1 Hz for 1% stimulus at 12:30.

[30-DEC-19] Device A215.10 reports 3.5 V at 12:21. The active current consumption of A210.10 is 109 μA, combined with 1% of an average of 15 mA lamp current is 259 μA. So far the battery has provided 18.6 mA-hr.

[31-DEC-19] We enhance our ring oscillator implementation, separating it into its own VHDL file P3036A05_OSC. In P3036A05_Main we specify fck_calib to configure the ring oscillator. The ring oscillator output is FCK, which we route to TP1. We no longer route TCK to TP1 because we encounter timing and routing issues when we do so.

Figure: FCK Period versus FCK Calib. "A05" is for A215.19 with P3036A05 firmware.

We calibrate A215.17-19. We have enhanced the ISL Controller to accommodate higher channel numbers. For fck_calib = 17, we are making a ring oscillator containing nine gates, and dividing its frequency by six. The result is an FCK period of 99 ns. The single-gate ring oscillator would run at 545 MHz. A single gate delay is around 1 ns.

We try various alterations to the power controller firmware in the hope that we can reduce the standby current consumption to 60 μA, but our standby current consumption for 17-19 is 154, 143, and 116 μA with only slight variations regardless of firmware.

[03-JAN-19] We complete calibration of nine more A3036A taken from our second set of ten. There were no hardware issues with these devices except for 12 mA inactive current in one circuit. Calibration constants and current consumption shown below.

Figure: Calibration Constants and Current Consumption of A3036A Batch 2. Circuit A3036AV1 with modifications. Active current is with stimulus with rare pulses, Xmit current is while transmitting synchronizing signal, no stimulus. Table updated as we damage and repair circuits.

The average difference between the Xmit and Active current ius 93 μA with standard deviation 4 μA. The slope of current versus sample rate, dI/dS, is 0.091 μA/SPS, which is less than the 0.11 μA/SPS we observe in the A3028GV1 circuits. Our concern remains the average active current, which is 160 μA with standard deviation 30 μA when we expect 70 μA with standard deviation 3 μA.

[06-JAN-20] Take A215.25, place on RF power. Current 179 μA. Connect P1-4 to P1-7, making sure PEN is unasserted. Current 180 μA. VB = 3.7 V, 3VA = 3.0 V, 1V2 = 1.2 V. Immerse entire circuit in acetone. Place in vacuum chamber, evacuate, allow to boil for ten seconds, remove, blow dry. Current 169 μA. Start stimulus with a flash every three seconds. Cool U8 with freezer spray. Current drops to 25 μA and lamp is still flashing. We cool again, then attach a thermocouple and record current with temperature as the device warms up.

Figure: Active Current versus Temperature. We compare the A3036A equipped with LCMXO2-1200 in WLCSP-25 and A3030E equipped with same device in TQFP-100.

The Lattice Diamond Power Calculator tells us that the typical current consumption of the chip at 25°C is 67 μA, while the worst case is 238 μA. At 37°C the typical is 96 μA and the worst case is 353 μA. All our chips are within this range of typical to worst case.

[07-JAN-20] We shatter the corner of U8 on A215.24. We heat U8 from above with iron at 400°C. After twenty seconds, it comes loose. We load another WLCSP-25 package by hand, heating from above with iron at 400°C. After ten seconds, the package attaches. We heat for another ten seconds. We wash and dry and the new U8 programs first time, and operates with the same calibration constants except we must drop battery_calib from 72 to 57. Active current consumption 150 μA.

The U8 on A215.24 may have been damaged by heat when we loaded it at 400°C. But it was not damaged by handling. We held it with tweezers. We measure the variation in current consumption with temperature and add to the above plot. The U8 on A215.25 may have been damaged by machine placement. But it was not damage by temperature. The reflow temperature profile is here, and reaches only 247°C for four seconds.

Summary of failures for B88120, 20 of A3036AV1, 4 faulty circuits. 2 with short or open circuit under U6. 1 with U7 not producing RCK. 1 with open circuit under U8. We replace U8 on this last one and produce fully-functional A215.26. We measure active current versus temperature for an A3030E, E157.1, and plot.

[09-JAN-20] We have 5 of A3036IL-C blue TR2227 and 5 of A3036IL-D green TR2227 back from Palomar Technologies. We connect 5 V through 200 Ω expecting 10 mA to flow into each LED and see of order 15 mW of light from blue LEDs and 6 mW from green LEDs. Some emit no light, some emit a flickering dim light, some emit a steady dim light. We pick a blue one and measure current and power output versus applied voltage. At each measurement, we connect power, make our measurements quickly, and disconnect power so as to allow the LED to cool down.

Figure: Current (mA) and Optical Power (mW) versus Forward Voltage (V) for Prototype Blue TR2227 LED of A3036IL-C.

The plot below is taken from the TR2227 data sheet. The optical power output is, to the first approximation, proportional to the current. Our plot shows the optical power increasing with voltage as suggested by the data sheet. But the forward current does not agree with the data sheet. At 3.5 V, for example, the forward current in our prototype is 120 mA, while the data sheet suggests 30 mA. The data sheet suggests 45 mW of blue light for forward voltage 3.5 V, but we see only 12 mW.

Figure: Current (mA) versus Forward Voltage (V) for Blue TR2227 LED of A3036IL-C, Manufacturer's Data Sheet.

We try cleaning one board with hot water and a brush, but the LED comes off. We knock another off by accident. When we apply 7.0 V through 200 Ω, all eight remaining LEDs emit light, some dim, some bright. Forward current is 75 mA for the six bright ones and 90 mA for the two dim ones. We measure the resistance between the input terminals. One of the bright green LEDs has resistance 2 kΩ in both directions. One of the bright blue LEDs has resistance 300 kΩ in both directions. Among the rest, the resistance varies 10-100 kΩ, the same in both directions, with no correlation between brightness and resistance.

[10-JAN-20] We equip A215.26 with two 120-mm long leads, OD 0.8 mm, 6-mil wire, MDC26398 spring. Lead resistance 19.6 and 19.8 Ω. We connect to an A3036IL-A and power from bench-top supply through ammeter. We measure VB at the circuit board. We turn lamp on to full power continuously. We measure total current versus VB, see plots, where each plot gives lead length, lead wire diameter, and LED version. The inverse slope of the graph is 44.5 Ω. We repeat with A215.25 with two 50-mm long leads, OD 0.7 mm, 4-mil wire, MDC13867A. Lead resistance 28.2 and 27.0 Ω. The inverse slope of the graph is 63.9 Ω.

[14-JAN-20] We hear from Lattice Semiconductor in response to our question about combining their Mico8 microprocessor with our existing VHDL code in the LCXO2-1200. "You just need to have a top design connecting to the MICO8. Just add the .msb generated by LMS to your Diamond Design. Just follow the Development Guide of LMS (MICO8/32) and check if how you can connect the .msb file to your Diamond Design."

[21-JAN-20] We started an A3030E and A3036A flashing their lamps at 1 Hz with 10 ms pulses yesterday at 11 am. Today, the A3036A flash is occurring 100 ms before the A3030E flash, suggesting that the difference in their clocks is around 1 ppm.

We have No25 at room temperature. We connect 3.7 V and activate with continuous RF power on antenna beneath the circuit. Current consumption is 177 μA. Re-program with same code again, freshly compiled. Immediately after programming see 198 μA, falling to 178 μA. We enable the JTAG interface and disable mux_configuration_ports, recompile and re-program. Standby current now 375 μA includes current caused on by TDO = BT being HI, and BTV being neither HI nor LO. Restore code, 185 μA. Try with No1, 175 μA before. Disable JTAG and see 400 μA. Remove R1, R2, and R3 and R13. Now current is 151 μA. Disable JTAG again and get 158 μA after a couple of minutes to cool down.

[24-JAN-20] We create standby-only firmware STDBY_WLCSP25 for the A3036A and STDBY_TQFP100 for the A3030E so as to make it easier to measure the standby current consumption of LCMXO2 chips. The VHDL source code is the same for both, but the pin numbers assigned to RCK and STDBY are different, to suit the WLCSP-25 and TQFP-100 packages, and we generate a larger number of unused inputs in the for the TQFP-100. We enable JTAG port and program A3030E E157.1, get standby current 65.3 μA. With E157.2, 65.7 μA. We connect U10-82, which is JTAGEN, to P1-4 and disable JTAG. We program E157.2 and get 65.6 μA. Reprogram with JTAG enabled again, get 65.8 μA. We program A3036A No1 with STDBY code with JTAG disabled and get 154 μA.

We have another 40 of LCMXO2-1200ZE in WLCSP-25 from another distributor, date code 1810. We load one onto our A3036A No1. We program with our STDBY code, current is 138 μA.

[28-JAN-20] We removing parts from A3036A No1 in order to isolate the component that is consuming an extra 70 μA. We remove the lamp switch, battery test switch, and DAC resistors. We have already removed U6, and we connect 3VB to VB. We still see 140 μA total consumption. We remove U8. Consumption is 70 μA. So U8, in standby mode, was consuming only 70 μA, which is correct. We connect 3VB to 0V. We remove U2 and C3. No change. We remove U4, current drops to 6 μA. Consulting the PE4239 data sheet, we see that its quiescent current is 250 nA so long as CTRL1 is HI.

Figure: Table from PE4239 Data Sheet. Current consumption is low provided U4-4, the CTRL input, is HI.

When the A3036A is inactive, quiescent current is 7.5 μA, and CTRL of U4 is LO. So the current consumption of U4 with CTRL LO must be less than a microamp also. When CTRL is floating, however, the current consumption is unspecified. We take No26, activate with RF power, heat U4 with soldering iron. Quiescent current rises from 110 μA to 200 μA in three seconds. We do the same with U8 and see an even more rapid increase. We remove U4 and U9 and the 110 μA current persists. We program with our STDBY code. Current persists. We remove U9, current drops to 9 μA. Load a new U9 and program with STDBY and see 140 μA.

[29-JAN-20] Rebuild No26 with new U4 and U9, re-program, re-attache thick-wire leads. Works fine. Active current 120 μA.

[30-JAN-20] We receive 3 green and 5 blue TR2227 mounted to printed circuit board, having passed 5 V through 1 kΩ 2-mA current and bright light test at Palomar Technologies. All eight emit bright light when we run current through them. We measure upward optical power with an SD445 photodiode. We observe 6.2 mW at 10 mA from one blue LED, and 0.67 mW from another at 1.0 mA. This compares well to 7.0 mW at 10 mA we observed with the blue EZ500. We damage two blue LEDs by connecting 5 V to them without a resistor. Another two suffer damage when we leave them with 10 mA running through them for thirty seconds. We damage a green LED with 10 mA as well. The remaining two green LEDs emit 0.44 mW at 1.0 mA, compared to 6.0 mW at 10 mA for the green EZ500.

[31-JAN-20] We left a green LED with 1 mA for eight hours and it had failed by the time we returned. We separate failures into two types. One type is open-circuit, where the LED appears to have broken from its epoxy bonds. This occurred in two LEDs that we over-heated with a direct application of 5 V with no resistor. The other failure is similar to the one for which we plotted current and light output versus voltage above. The LEDs are producing close to the nominal output power versus forward voltage. This suggests another component in parallel with the LED. But this component is not a resistor, because its current is 2 mA at 1 V, 40 mA at 2 V, and 80 mA at 28 V.

Figure: Damaged Green TR2227 Mounted on A3036C Circuit, Closeups. The LED is 270 μm long. We see a signs of a dark film on the bottom side of the LED in left and right images. Note that the 3-mil gab between the gold-plated pads is almost too wide for the pads on the LED.

We place a working LED in water to see if it will remain cool while illuminated. The LED durns off and acts as if damaged. We remove and blow dry, but the LED is now acting like the rest of the damaged LEDs. We examine another damaged LED with a 20× loop. We see a dark stain on the underside of the LED. We take the above photographs with our microscope, in which the stain is not clear, but suggested. We now suspect that metal migration is creating a parallel diode junction in the gap under the diode. This will happen after 8 hr at 1 mA in air, 30 s at 20 mA current in air, and 1 s at 10 mA in water.

We cover a fully-functional green LED with epoxy, place in our vacuum chamber to remove bubbles, and drive epoxy beneath the LED. Later in the day, the epoxy is not yet cured, but it's tacky. We run 10 mA through the LED and see 1.4 mW of green light with 2.98 V forward voltage. We increase to 20 mA. Light is flickering. Current is changing too quickly for us to make self-consistent measurement of power, voltage, and current. After a few minutes we return to 10 mA and see 0.88 mW. The current is 0.3 mA at 1 V, 2 mA at 2 V, and 7.5 mA at 2.8 V. At this final current, power output is 0.64 mW, ten times less than the data sheet suggests.

[04-FEB-20] We have ordered the A303602C printed circuit board for TR2227 with 10-mil long mounting pads separated by 3 mil. We take one of our existing A3036IL that draws around 100 mA for a forward voltage of 3.0 V, producing roughly the correct amount of green light, and we push the TR2227 off its pads and over-turn it. We obtain this view of the epoxy and underside of the LED chip.

Figure: TR2227 Removed and Overturned Near Epoxy Mounting.

Now that we have removed the LED, the current is 0 mA for forward voltage 3.0 V. We have eight remaining A3036IL-C/D circuits that produce light. With forward voltage 3.0 V they shine brightly, but the current they consume varies greatly, from 70 mA to 150 mA.

We take A3036A No22, active current 213 μ and transmit current 300 μA at time of calibration. Active current 208 μA today. We replace U8 and re-program. Active current 143 μA, transmit current 227 μA. We apply 7.5 V as the battery voltage for the A3036A through 200 Ω. No harm comes to the circuit. We load batteries onto six A3036A with the help of water-soluble flux. We check immediately after soldering, and all six boards work fine. We triple wash in hot water after and three of them have ONL asserted, which we observe on P1-3. We check PEN on P1-4 and find it LO, but as soon as we connect a probe to PEN, ONL is unasserted. We re-program these three boards using their original calibration constants and they all work fine after. We assemble two more, the first works after washing, the second we re-program. We now have eight working boards with batteries loaded. To load the battery, we bend the battery tab back upon themselves twice, so as to bring to solderable surfaces face-up. One of the tabs, the one with the spot-welded extension, we can solder only on one side of the extension. We place a lump of solder on each tab end, add lumps to the circuit board battery pads, add flux, and lower the pads onto the tabs. We melt solder from the side of one joint then the other, alternating as we push the circuit board down onto the tabs. The resulting circuit plus battery is 19 mm long.

[05-FEB-20] We have eight A3036A with batteries loaded yesterday A215.18-25. We measure battery voltage with a voltmeter on the programming extension and get 3.8=7-3.95 V. We turn on and off data transmission. We check the battery voltage measurement and get 3.9, 4.1, 3.9, 4.0, 3.7, 3.9, 4.0, 3.9 V respectively. The lowest reading must be a poor calibration. The existing value of batter_calib we used when re-programming this device after washing is 62, but we find that 52 gives an accurate battery measurement. Now A215.22 says 3.9V and our multimeter agrees.

[11-FEB-20] We have eight encapsulated A3036A circuits, numbers 18-25. All respond to commands, flash their lamps, transmit synchronizing signal, and report battery voltage2 3.8-3.9 V when inactive. We have 13 of A3036IL-A. We measure optical power for 10 mA current and see 4.4-6.0 mW, with the exception of one with an air bubble over the LED, which produces 4.2 mW. The average power of 5 mW is roughly half what we expect from our earlier measurements of the same LED with no epoxy cap. We have two EZ500 without epoxy cover, they emit 9.8 and 9.5 mW for 10 mA. We add one more drop of epoxy to all thirteen caps, so that the caps are domed. We now measure 6.7-7.8 mW at 10 mA.

[14-FEB-20] We receive 5 blue and 5 green mounted TR2227 LEDs from Palomar. Of these, all turn on brightly with 1 mA forward current for one second, except for one blue, which turns on brightly with 70 mA and then flickers and fails. We cover the nine working LEDs each with a drop of clear DP270 epoxy and place in our vacuum chamber at −25 psi for five minutes, then drop pressure over five minutes. We leave to cure in their gel pack.

[15-FEB-20] At 4 pm yesterday we left No10, 24, and 25 flashing an A3036IL for 100 ms at 1 Hz. Today No10 has stopped. We connect to charger. No24 and No25 flashing 100 ms apart. We measure this separation by looking at their synchronizing signals. Battery reports are 7.0 V and 3.4 V. We stop stimulus and battery reports are 2.9 V and 3.7 V. After half an hour of charging, No10 reports 3.9 V.

We number our nine new epoxy-encapsulated TR2227 LEDs. No1 is green. We run 2.0 mA through it and observe 0.76 mA light output. After ten minutes, forward voltage is 2.80 V for 2.03 mA and power output is 0.72 mW upwards. We flip the circuit board over and measure 0.16 mW coming through the printed circuit board. So the total power output is 0.88 mW. While we are handling the circuit board, the LED flickers and goes dim. We take out No6, which is blue. At 2.0 mA its forward voltage is 2.8 V and power upwards is 1.0 mW. After one minute, it flickers and goes dim. We take out No2, green, and run 10 mA through the LED. It shines brightly for ten seconds, then goes dim. We take No7, blue, and run 0.5 mA through it, seeing 2.5 mW. But it starts to flicker after twenty seconds.

[17-FEB-20] We consider how best to mount the TR2227. Right now, we are mounting the chip by its two bond pads with silver epoxy to a gold-plated pad, as shown below.

Figure: Drawing of Our Silver-Epoxy TR2227 Mounting Scheme.

We consult the TR2227 manual, which contains the following drawings of the device package. We assume these sketches are to scale.

Figure: TR2227 Package Drawings from Data Sheet.

According to the data sheet, we are mounting the die with the top side down. The data sheet states in its opening paragraph, "The design is optimally suited for industry standard sideview packages as it is die attachable with clear epoxy and has two top contacts, consistent with industry standard packaging." We see that the LED junction comes within 10 μm of the cathode pad. We suspect that our silver epoxy is spreading from the cathode pad to the junction and creating another current path in parallel with the LED. We propose to try the following mounting arrangement of the TR2227 on our A303602C printed circuit board.

Figure: Drawing of Our Wire-Bonding TR2227 Mounting Scheme.

In the new mounting scheme, we attach the die to the substrate with clear epoxy and then wire bond to the nearby gold-plated pads. We are not sure what fraction of light emitted will end up propagating upwards past the bond pads, but we want to know if this scheme produces a reliable LED.

The A303602D printed circuit board provides a generic two-pad wire-bonding footprint for an LED die. The top side provides the footprint with masked vias leading to the bottom side. The top side will of course be the one that is facing down when implanted. The bottom side provides two pads for sockets, each anchored with its own via, as well as a 1.4-mm diameter clearing in the solder mask directly opposite the center of the LED. This clearance allows us to glue a steel tube directly to the FR4 substrate.

Figure: Rendering of the Top (Left) and Bottom (Right) Sides of the Wire-Bonded LED Head. The vias on the top side are masked. The boards will be made with ENIG finish to allow wire-bonding. The anode pad is the larger square, the cathode pad the smaller square.

This wire-bonding footprint is compatible with the TR2227, EZ290, and EZ500. The TR2227 will mount with clear epoxy to the center of the anode pad, and then be wire-bonded to the anode pad and cathode pad. The gold-plated pad beneath the LED will reflect downward-going light back upwards, increasing the efficiency of coupling into our fiber. The EZ290 and EZ500 are both square, 170-μm thick chips with bottom-side anode and top-side cathode. These will be glued with conducting epoxy to the center of the anode pad and wire-bonded from the top-side to the cathode pad. We receive the following drawing from Cree showing how they intend the TR2227 to be mounted in an epoxy package.

Figure: Drawing from Cree Showing How TR2227 Should be Mounted.

[19-FEB-20] Yesterday we recharged three A3036A, but something went wrong with one of them. Afterwards, its lamp power was stuck on, and it would not respond to commands. We left it in front of a video camera to exhaust its battery. After 100 minutes, the light is half as bright. After 120 minutes, the light is out. We connect to 4.7 V and see 11 mA flowing in. After five minutes, current is 8.5 mA. We disconnect and are able to turn on data transmission with a command. We return to charger. The other two devices we leave flashing at 1 Hz and sixteen hours later they are flashing 30 ms apart.

[21-FEB-20] We charged No25 and No25 overnight. We set them to flashing A3036IL-A for 300 ms at 1 Hz, as well as transmitting synchronizing signal. At first, their battery voltages are 4.0 V and 3.9 V respectively during stimulus. After 190 min their voltages are 3.3 V and 3.4 V respectively. Turn off all functions and batteries report 3.6 V and 3.7 V. We charge No10 for five hours. Battery reports 4.2 V. Start 300 ms, 1-Hz continuous stimulus. Battery reports 3.9 V. After 210 minutes, battery reports 3.4 V. Turn off stimulus, reports 3.6 V. After 210 minutes of charging, No24 and N25 report 3.9 V and 4.0 V.

[24-FEB-20] We have No24 and No25 recently charged. We connect to A3036IL-A and A3036IL-B and set to 300 ms pulses at 1 Hz average, but random spacing. The result is an engaging, never-repeating display. We begin at 09:30. One stops flashing at 12:24 and the other at 12:39.

[25-FEB-20] We turn three A3036A on at the same instant, set to random 10-ms flashes every second. They flash randomly, but synchronously. The random number sequence is the same for all three.

[28-FEB-20] Today something went wrong while re-charging one of our three A3036As, No24, where the lamp is stuck on, we are letting it run its battery down. This same problem arose on 18-FEB-20, but we did not record which stimulator is was that exhibited the problem. Both times, however, one of our staff was working next to the three recharging devices and moving them around.

[02-MAR-20] Charging No24. Connect No26 to power supply with ammeter in series. When we issue an Xon command, we look for current to jump up to about 150 μA for half a second to indicate that RP has been asserted, and for current to remain at 150 μA to indicate that data is being transmitted, and we look at reception from No26 to see if data transmission is effective. The minimum value of VB for which we can turn on data transmission is 2.6 V. Reception is 100%. Once data transmission is turned on, VB = 2.4 V is the minimu battery voltage that keeps it on. At VB = 2.3 V, transmission turns off. With the device inactive, the minimum voltage for which RP is asserted is 2.4 V. We insert 50 Ω in series with VB to simulate a battery's source impedance. The minimum VB to turn on data transmission is 3.4 V. Once turned on, it remains on all the way down to 2.3 V. The minimum voltage for which RP is asserted is 2.4 V.

[03-MAR-20] No24 recharged, connect lamp, it is stuck on. It does not respond to commands. Leave with lamp shining from 9:35 am to 2:30 pm, when it is dark. Connect to 4.7 V charger and see only 5 mA flowing in. This is the same behavior as No9, which failed after encapsulation. We put No24 and No9 in our physical sample bag.

We deliver 4.0 V through 50 Ω to A3036A No26 and observe VB when we issue an Xon command to activate the device. The power-up takes 2 ms and peak current is 42 mA. We cannot deliver such a current surge with an MnLi battery. We need a LiPo battery.

Figure: Power-Up Current from Inactive to Active. Yellow: VB 500 mV/div. Blue: 4V0 power supply 500 mV/div. We deliver the 4.0-V power supply through 50 Ω to VB. Instantaneous current is displacement of VB below 4V0 at 10 mA/div.

We resolve to try the ADP5301 micropower buck converter. It can deliver 50 mA at 1.2 V, which is sufficient. When delivering 200 μA to the sensor circuit from a 3.7-V LiPo battery, the ADP5301 is 87% efficient, implying a battery current of 75 μA. The ADP5301 comes in a BGA-9 with 0.5-mm pitch, and the entire buck converter with inductor, capacitors, and control resistor occupies 3.6 mm × 3.0 mm.

[10-MAR-20] We receive form Micro Precision Technology the first 10 C460EZ500 on Wire-Bonded LED Head (A303602D) printed circuit board.

Figure: EZ500 LED Die Bonded to Anode and Cathode Pads of A303602D.

The cathode bond wire is close to the edge of the cathode pad so as to allow us to put more optical fiber on the LED surface. The anode is glued with conducting epoxy. We damage one of the cathode wire bonds during handling. We apply 10 mA to the remaining nine. Average power 10.1 mW with standard deviation 0.5 mW. According to the manufacturer calibration of these LEDs, we expect 155 mW total power output for 150 mA. The EZ500 data sheet suggests that should see something like 11 mW at 10 mA. Given that we are losing light by unwanted reflection off the edges of our photodiode, and by imperfect reflection off the conducting epoxy below the LED, we are well-satisfied with 10.5 mW at 10 mA. We also have an EZ500 in 3-mm package coupled to a 450-μm fiber polished entirely by hand, without using a puck. For this one, we get 2.8 mW at the fiber tip for 10 mA forward current.

[17-MAR-20] We have a dozen A3036IL-A assembled. The first eleven with the EZ500 in a 3-mm SMT package, the last one with the EZ500 mounted directly onto an A303602D wire-bonding board. We apply 10 mA forward current and measure output power at the taper tip using an angled photodiode.

Figure: Fiber Tip Optical Power. A 450-μm fiber on EZ500 LED, 10 mA forward current.

When the output power is less than 3.5 mW, as for 9-12, misalignment of the fiber base with respect to the LED surface is obvious when viewed with a loupe.

[19-MAR-20] We have our first set of ten blue TR2227 glued and wire-bonded to our A303602D wire-bonding board. Nine out of ten emit 3-4 mW of light for 10 mA forward current. One draws current but emits no light. These LEDs have been glued with conducting epoxy rather than clear, non-conducting epoxy. We suspect the conducting epoxy has wicked up the side of one of the LEDs and shorted the anode pad to the cathode junction. We expect 10 mW for 10 mA. We suspect the conducting epoxy is absorbing the downward light. We leave one LED with 20 mA forward current, emitting 6 mW.

[20-MAR-20] Today our blue TR2227 is still shining, now emitting 3 mW at same current. We receive 240 of C460EZ500 mounted to circuit boards. We apply 10 mA and measure output power for 25 LEDs, get an average of 9.4 mW. Manufacturer's calibration of these diodes says an average of 156 mW @ 150 mA.


[23-MAR-20] We have 250 C527EZZ500 wire-bonded to A303602D boards. We apply 10 mA and measure output power with photodiode. Average is 6.0 mW. Manufacturer's calibration of these diodes says an average of 71 mW @ 150 mA.

[30-MAR-20] We construct a power calibration stand for bare LEDs mounted on our A303602D wire bonding circuit board. We mount an SD445 photodiode, 10 mm square, in a vertical plane 10 mm from the LED die, with the die at the height of the lower edge of the photodiode, and centered upon it.

Figure: Lateral Photodiode Power Meter for Wire-Bonded LEDs. The bottom edge of the photodiode is at the same height as the top surface of the printed circuit board. The center of the LED die is 10 mm from the surface of the photodiode. The photodiode receives 7.1% of the light emitted by the LED.

We use a Monte-Carlo simulation, PCV1.tcl, of rays of light originating at LED die randomly distributed over a quarter-sphere to determine the fraction of of photons the photodiode would receive from a point source. We modify this calculation to determine the fraction of photons received from a planar source, where the intensity is proportional to the sin of the angle between the ray direction and the plane of the source (Lambertian distribution). The result of our simulation is shown below.

Figure: Power Fraction for Increasing Range of Lateral Photodiode. Fraction of hemisphere covered by photodiode (Blue), fraction of emitted rays striking photodiode (yellow), and ratio of photodiode area to hemispherical area (orange).

At small ranges, the fraction of photons striking the photodiode is equal to the fractional coverage of the photodiode, or 0.5. At large ranges, the fractional coverage of the photodiode approaches the ratio of the photodiode area to the hemispherical area. But the fraction of photons striking the photodiode decreases more quickly, because the photodiode sees a smaller emitting surface from its lateral position. At range 10 mm, the simulation predicts that 7.1% of photons emitted by the LED will strike the photodiode. The simulation itself produces a map of its simulated rays. The map below shows the rays generated for range 10 mm with ten thousand iterations, Lambertian distribution.

Figure: Hits (red) and Misses (green) For Photons Generated by PCV1 Simulation (y_range = 10 mm, z_max = 10 mm, z_min = 0 mm, x_min = −5 mm, x_max = +5 mm). We assume a Lambertian distribution of photons across the hemisphere above the LED, place the 10 mm square photodiode in a vertical plane 10 mm from the die, with lower edge of photodiode at same height as LED.

We place a C460EZ500 in our power meter. With 10 mA forward current, we see 148 μA photocurrent. Using 7.1% we get 148 μA ÷ 7.1% ÷ 0.18 mA/mW = 11.6 mW. We place the LED 1 mm from the surface of the photodiode, facing the center, and get 1.6 mA photocurrent, suggesting 8.9 mW is received by the photodiode. We consider the reflection of light for large angles of incidence. The SD445 data sheet says its field of view is ±60°. If we assume all light with incidence >60° will be reflected, the fraction lost will be 25%. In which case, our 8.9 mW suggests total power output of 11.8 mW, which is close to our measurement with 10-mm lateral photodiode.

We repeat with C460TR2227 and see 82 μA for 10 mA forward current, 6.2 mW. With LED up against photodiode, face-on, 720 μA, or 0.72 mA ÷ 0.75 mW/mW ÷ 0.18 mW/mA = 5.3 mW. The TR2227 radiation pattern deviates from Lambertian: at 68° from the optical axis, intensity is 70% of the value of the optical axis. The EZ500 radiation pattern is close to Lambertian: 33% at 68°, while cosine 68° is 0.37.

[01-APR-20] A green C527EZ500 in our calibration stand produces a photocurrent of 127 μA at 10 mA, suggesting power 0.127 mA ÷ 7.1% ÷ 0.25 mW/mA = 7.1 mW.

[02-APR-20] We construct a power meter for fiber-coupled LEDs, where we want to know the total optical power emitted by the fiber tip. Our objective is to provide a meter that does not damage the surface of the photodiode with the hard fiber tip, and provides a consistent measurement from one fiber to the next.

Figure: Lateral Photodiode Power Meter for Fiber-Coupled LEDs. The fiber tip is 1.3 mm from the reflecting surface of the silicon photodiode, and approximately centered on the photodiode area. Installed is an A3036IL-A8. Note the aluminum foil baffle along the bottom edge of the photodiode.

The aluminum-tape baffle at the base of the photodiode reflects light from the LED away from the photodiode. When we cover the base of the fiber with a separate piece of foil, we reduce the photodiode current by less than 5%. We use our PCV1.tcl ray simulator to estimate the fraction of light emitted by the fiber tip that is received by the photodiode. We set Lambertian to disabled, z_max = +5.0 mm, z_in = −5.0 mm, and y_range = 1.3 mm.

Figure: Hits (red) and Misses (green) For Photons Generated by PCV1 Simulation (y_range = 1.3 mm, z_max = 5 mm, z_min = −5 mm, x_min = −5 mm, x_max = +5 mm). We assume a uniform distribution of photons across the sphere around the fiber tip.

The simulation suggests that 39% of the light will be received. This results in our measuring 3.7 mW for this fiber tip at 10 mA. We get 3.5 mW with our existing angled-photodiode measurement.

[15-APR-20] We receive 5 Blue EZ290 and 5 Green EZ290 from our wire-bonding company. We measure power output at 10 mA forward current in our lateral-photodiode calibration stand.

Figure: Optical Power Output of C460EZ290 (Blue) and C527EZ290 (Green).

In our original tests with the EZ290, we measured an average of 7.2 mW at 10 mA for the C460EZ290 and 4.0 mW at 10 mA for the C527EZ290. Those measurements we performed with a photodiode held over a 3-mm plastic package in which the LED was mounted.

We apply a −6 dBm sweep to an A3015C loop antenna. We load 30-mm wire antennas onto our three A3036BV1 circuits. We set the circuit down on the antenna and measure VR. We must apply power to the circuit to see antenna resonance. We solder a battery directly to the circuit, or we plug one into the programming extension. In the latter case, we must have the battery leads running away from the A3015C antenna or else we see additional peaks in the VR response. The directly-soldered battery gives us a reliable peak. All three circuits show resonance at 885±5 MHz.

Figure: VR versus Frequency with C6 = 1.5 pF. Green: VR 2 mV/div. Blue: Sweep signal downshifted by 915 MHz and low-pass filtered to ±21.4 MHz. Yellow: ramp applied to A3014MT voltage-controlled oscillator. Sweep power is approximately −6 dBm.

We try 40-mm and 20-mm antennas and resonant frequency remains 885±5 MHz. These A3036BV1 boards come loaded with C6 = 1.5±0.1 pF. We replace with 1.0±0.1 pF and resonance is at 940 MHz. We load 0.2 pF in parallel with 1.0 pF to make 1.2 pF and see the following sharp resonance at 910 MHz.

Figure: VR versus Frequency with C6 = 1.2 pF. Green: VR 5 mV/div. Blue: Sweep signal downshifted by 915 MHz and low-pass filtered to ±21.4 MHz.

Once tuned, we obtain robust reception of commands at range 1 m from our command antenna. We have two bare A3036AV1 circuits, and we find they, too, have resonant frequency 885 MHz for C6 = 1.5 pF when we connect the battery. We drop C6 to 1.2 pF and resonance for both is at 910 MHz. During our work, one A3036AV1 and one A3036BV1 required re-programming.

[08-JUN-20] Although we have 25 each of C460TR2227 and C527TR2227, these are not proving suitable for fiber-coupled LEDs. They emit more power sideways, and the result is poor capture efficiency in the fiber. For 270-μm fibers we see average capture efficiency 20%, 30%, and 24% for TR2227, EZ290, and EZ500 LEDs respectively. for 450-μm fibers we see 30% and 39% for EZ290 and EZ500 respectively. We will include the TR2227 in our survey of LEDs, but we abandon plans to build FCLEDs with them.

Figure: EZ500 Coupled to 270-μm Diameter Fiber.

We solder five LXZ1-PA01 to A303602D circuit boards. Three have damaged silicone covers. The other two we place in our power calibration stand with 10 mA forward current and measure 7.7 mW and 9.1 mW, using 0.40 mA/mW as the photodiode sensitivity.

[18-JUN-20] We are tuning the split-capacitor matching network C5/C6/L1 of our A3036Bs. We are applying a sweep 860-970 MHz of about +10dBm (A3014MT sweep through first amplifier stage of A3029C) to an A3015C loop antenna. We place the A3036B on the loop. We leave L1=10nH. We have one circuit with C5=1.0pF, C6=1.5pF, 0033. With a wire antenna soldered and no battery, peak 915 MHz. With a clip instead of a wire, 915 MHz. With a battery connected to the programming extension on 60-mm leads, 880 MHz. With a battery soldered to the circuit directly, 880 MHz. Return to no battery and wire antenna, 915 MHz. We repeat these experiments with boards 0001 and 0025 and see the same ≈30 MHz drop in peak frequency when we load the battery. The half-amplitude width of the peak is around 30 MHz, so by tuning without a battery and loading the battery later, we risk our encapsulated device being half as power-sensitive. Our 0001 device has C5=1.0pF, C6=1.2pF, and with a battery loaded, its peak is 900 MHz. Without battery loaded, peak is 930 MHz.

[19-JUN-20] We connect a power supply to VB on A3036BV1 0002. The circuit has a 50-mm antenna. We place the 50-mm antenna over an A3015C loop antenna to which we apply a 10-dBm 860-970 MHz sweep. We set VB=0V. The peak in VR is 40 mV at 924 MHz. We turn up VB to 3.7V. The peak in VR is 100 mV at 897 MHz.


[23-JUN-20] We have 26 of A3036BV1 programmed with default firmware with their antenna matching networks all tuned in the range 900-920 MHz. Average peak frequency is 908 MHz, standard deviation 6 MHz, maximum 920 MHz, minimum 900 MHz. All circuits have C5 = 1.0 pF and L1 = 10 nH. Almost all circuits have C6 = 1.2 pF.

Figure: Typical A3036B Matching Network Response. Sweep 860-970 with 910 at center, 890-930 marked with purple cloud. Yellow trace is VR with ×1 probe. Transmit antenna A3015C driven with +10 dBm. Battery power connected, C5=1.0pF, C6=1.2pF, L1=10nH.

[13-JUL-20] We measure the peak frequency of our antenna matching network as we increase C6, with and without a battery. With battery loaded, the antenna switch is powered up and connecting the antenna to the matching network. With the battery disconnected, the antenna switch behavior is unspecified.

Figure: Tune Frequency vs. Value of C6, With and Without Battery.

It appears that 0.1-pF changes in C6 can bring the peak frequency into the range 905-915 MHz.

[03-AUG-20] We replace the pins on A215.10 and A215.25 and ship to Cornell. These we have already recharged and discharged ten times. We have a batch of ten A3036B with batteries loaded. The A3036B has orange and purple for L+ and L−. We charged them with A3033A through leads, and also directly into battery with bench-top power supply. Battery voltages vary from 3.8 to 4.5 V at time of calibration. The 4.5 V is after four days charging with one particular A3033A. Of ten, nine work well, one has no RF output.


[12-JUL-20] When illuminating cardiac tissue for optogenetic stimulation, deep-red light penetrates farther than green or blue light. We are currently building a test fixture to measure the penetration of light through animal tissue. So far as we can tell from a survey of the literature, the absorption length of red light in heart tissue is of order 5 mm, and of blue light in heart tissue is 2 mm.

[25-SEP-20] We have three A3036B that we charged with A3033D. We attach one each of green, blue, white LEDs. Over the past three weeks we initiated eleven stimuli, each of ten thousand 5-ms pulses at 10 Hz. All three devices report battery voltage 4.0 V at the end. Total lamp on-time is 550 s. Estimated LED current is (4.0 V - 3.0 V) ÷ 56 Ω = 18 mA. Total charge consumed 2.8 mA-hr. Reception of commands is immediate in all 33 cases.

We receive back from Edinburgh A215.18, A215.19, and A215.20, after hearing report of poor reception while implanted. We place each one on a petri dish and move throughout FE2F enclosure, 10 different locations, and measure reception, then repeat while in water. For A215.18 we get command reception 70% in air and 70% in water, VBAT = 3.9 V. For A215.19 we get 100% in air and 100% in water, VBAT = 4.1 V. For A215.19 we get 60% in air and 90% in water, VBAT = 4.1 V. We try a newer B215.37, which has been through our imrproved antenna tuning process, and get 100% in air and 100% in water.

[28-SEP-20] Devices A215.18 and A215.20 we soaked in acetone, then clean acetone, then rinsed in alcohol. All dental cement is gone. We pull the lamp pins out of the lamp sockets with no difficulty. A215.20 flashes its lamp, when we reconnect the lamp. A215.18 does not. A215.20 is bloated. We squeeze it and acetone squirts out of a tiny hole in one battery corner. We squeeze as much acetone out as we can. We set both devices to charging with A3033D. The two epoxy-dome implantable lamps we soaked no longer work. Their epoxy is shattered and flaking. We have a new A3036IL-C450-6, Blue EZ290 with 450-μm diameter fiber 6 mm long from base to tip. With 10 mA forward current we see 3.8 mW at the tip.

[06-OCT-20] We have a selection of A3036IL with epoxy dome. We measure photodiode current with epoxy dome 1 mm in front of photodiode center, with 10 mA forward current. We see ±10% variation in photocurrent for any given type. For Blue EZ500 we get 6.3 mW, Green EZ500 3.4 mW, Blue EZ290 4.7 mW, Green EZ290 2.6 mW.


[20-JAN-21] We prepare two 270-μm diameter 6-mm long fiber-coupled LEDs. One is an A3036B270-6 that produces 1.8 mW of green light at the tip of the fiber for 10 mA, the other an A3036A270-6 that produces 2.9 mW of blue light. We begin with the full circuit board with mounting extension, the fiber glued to the LED on one side, and the mounting tube on the other side. We set our iron to 500 F. We apply flux to the socket pads and apply solder. We solder the two sockets in place. Rinse in hot water and blow dry. Rinse and dry. Rinse and dry. Clip mounting extension beyond the extend of the glue that holds the fiber. Hold between thumb and index finger and grind away excess circuit board and glue until close to the white line. Rinse in hot water and blow dry.

[21-JAN-21] We have run out of the 19-mAhr batteries we used with the A3036A/B. We assemble six A3036 with PGEB201212 10-mAhr batteries and GMB201021 25-mAhr batteries. We clip the tabs of the smaller battery short, then solder directly to the circuit board and fold over to make a 12-mm square 4-mm thick A3036C. We solder wires to the circuit board and run them to the tabs of the larger battery, so that we can align the base of the antenna and leads with one end of the battery, and make a 20-mm long A3036D that is 4 mm thick at one end and 2 mm thick at the other.

Figure: Assembled A3036C (smaller battery) and A3036D (larger battery) Ready for Encapsulation.

To hold the batteries to the circuit boards, we tied a thread around both with a granny knot, pull tight, then add a knot and trim.

[17-MAR-21] Of five A3036C we have on the shelf, four are not responding to commands. We recharge them all. Two are now functioning once more, and two are not. No50 and No57 flash their lights once when we transmit a command, or when we first turn on 910 MHz continuous power. The flash lasts several milliseconds. We remove silicone from battery tabs and find both batteries at 4.2 V. We short one battery for one second to reset all internal circuits on No57, but this makes no change to the circuit's behavior. We do the same on No50 and afterward the circuit does not respond to commands at all, despite battery voltage 4.0 V. We name these symptoms "unexplained logic failure" (ULF). We are left with three fully-charged A3036C. We ship No44 and No49 to ION along with an A3033D charger and three A3036IL-A implantable lamps.


[24-MAR-21] We charged No53 yesterday. Today it has ULF. We construct five new A3036C, but do not encapsulate. They are No1, No46, No51, No52, and No55. We are not confident that No1's tuner has been calibrated, but the other four are calibrated. We will exercise these circuits in the hope that we will observe the unexplained logic failure, then be able to examine the circuit to determine what is going wrong.

[25-MAR-21] Rob Wykes at ION/UCL implanted No49 a few days ago and today he tries to flash the lamp, but No49 is exhibiting the symptoms of ULF. The partner device No44 flashes easily. No49 flashes briefly whenever we start a command.

[26-MAR-21] We have two devices here, No50 and No53, that have ULF. We dissect both. We peel off the silicone, the battery, and burn away epoxy so we can obtain the following trace as we issue a stimulus start command for the device.

Figure: Signals from Example ULF Device, No53. Blue: L−. Yellow: U5-1, the source of RP. Green: 3VB.

Sometimes 3VB stays on until we disconnect power, in which state both circuits consume 4.7 mA, which is the current consumption of the A3036 before it has been programmed. Sometimes 3VB stays on for 20 ms, 40 ms, or 100 ms and then turns off. The circuit's current consumption when inactive is 7.5 μA. After soldering a wire to C3 so as to get to 1V2, we observe 1V2 turning on and off with 3VB. No53 always stays on after a command, consuming 4.7 mA until we disconnect power. We file the stub of the programming extension and make sure there are no connecting tracks on the cut surface. Now No53 always turns itself off after a command transmission.

[09-APR-21] We have C215_54 exhibiting ULF. We take this picture of its cut, silicone-encapsulated programming extension.

Figure: Cut Programming Extension, Covered With Silicone. The neck has been cut too far from the stimulator circuit, so that the clearings in the ground and power planes have been missed, and we see the power and ground planes in cross-section. Note also the specks of copper in silicone.

We cut the programming extension off after epoxy encapsulation, sand down the cut edge, then dip in silicone. During this process, the stimulator is inactive. The central two layers should be clear of copper except on the far left and right sides. But in the case of No54, we cut too far from the stimulator circuit and we see the complete ground and power planes. We have VB, which is the battery voltage, on a track that could be shorted to 3VB by our cutters, turning on the logic chip. But no other shorts between copper signals could caused an immediate problem so far as we can see. If we fail to remove the copper fragments generated by sanding, however, we have fragments in the silicone and along the surface of the cut circuit board. The PEN signal has a "weak pull-down" resistor inside the LCMXO2-1200ZE, but the value of this resistor is not specified. We connect external power to C215_54 and see 30 μA flowing in for five minutes, then 6.2 μA. The lamp flashes when we deliver 915 MHz, but only for a few milliseconds. After a few more trials, we see the current latch up at 4.7 mA and the lamp will no longer flash. Disconnect current, reconnect and current is 30 μA but latch-up occurs with any RF power and lamp flashes only with first RF pulse. We have ULF in No53, where the neck was cut in the correct place. The power and ground planes are reduced to 20-mil tracks on the left and right side only. So whatever causes ULF does not require the ground and power planes in the cut.

We take an unencapsulated circuit and connect TP2 (P1-8) to PEN (P1-4). The circuit works perfectly for multiple stimuli and multiple commands to other stimulators. We do the same for TP1 (P1-6) and PEN (P1-4). The connection has no effect upon the circuit function. For 3VB (P1-1) and PEN (P1-4) we see the light turn on so long as the logic chip is active, but the chip is otherwise unchanged. We erase the logic chip program on No1 and the lamp turns on continuously. Reprogram and it's back to normal. Earlier stimulators failed with the light stuck on, which is consistent with a logic chip erase, but ULF is not consistent with an erase. We have two more C215.56 and C215.83 still functioning. One has neck cut in correct location, the other too far from stimulator. We have been exercising them half a dozen times a day, and recharging them as needed.

[19-APR-21] We have No45 and No56 left of 5 A3036B/C we kept at OSI. These we have run down and recharged in the past week. No further ULF. Our five un-encapsulated A3036 boards have run down their batteries and been recharged. No ULF. At Edinburgh, we have two A3036B implanted in two CD1 mice (30% larger than normal strains) four weeks after AAV injection. Just before implantation, we connected both A3036B to an A3033D charger. At time of implantation, both devices flashed their lamps readily. Now they do not flash. We calculate that the initial charge with which the two were shipped on 11-DEC-20 was sufficient for 2500 hrs of inactive operation, which would take them to 01-APR-21, two weeks after implantation. We find that the A3033D battery charger at Edinburgh shorts the ±15 V power supplies of their LWDAQ Driver, and so will not provide recharge.

[12-MAY-21] We have discharged and charged the batteries of No46, 51, 52, and 55 five times with several sets of stimuli every day. We recharge through their lamp leads only. All continue to function perfectly. We set these aside for epoxy encapsulation. We still have No1 unencapsulated, but its receiver needs tuning.

[21-MAY-21] We have No46, 51, 52, and 55 encapsulated in epoxy, A3036C. We have clipped and sanded the programming neck. All have been working well for the past four days, and continue to work well. We have No33, No37, and No38 back from Edinburgh. No33 was never implanted, but we did connect to the faulty battery charger. We connect 4.6 V to lamp leads and see a few milliamps flowing in. After a few hours, however, No33 will not flash a lamp. We dissect and find the battery is healthy but the circuit will not obey commands. Connect external 3.6 V. Inactive 7.5 μA. Upon command transmision, circuit powers up briefly, but does not obey command. No37 and No38 have dental cement head fixtures. We soak in acetone for 24 hours, rinse in acetone, rinse in propanol. We have the two lamps and the lamp pins free of cement and clean. Both lamps still work. We connect both to 4.6 V charging voltage with 10 mA limit, and see the volage rising as 10 mA flows in. When the batteries are charged, we disconnect, leave for 48 hrs, and test. No37 flashes its lamp and transmits synchronizing signal. No38 does not respond. We dissect. The battery provides no voltage. Inactive 8.0 μA, active behavior normal. We flash a lamp and watch synchronizing signal. The broken A3033D battery charger, meanwhile, contained a faulty capacitor that was shorting +15V to 0V. We conclude: the faulty charger damaged No33's circuit but not its battery. No38's battery was too deeply drained to hold a charge for more than 24 hours. No37's battery was drained.

[28-MAY-21] We are continuing to discharge and re-charge our collection of ISTs. We find that No37's battery runs down faster than the others. No further failures.

[09-JUN-21] We add No1 to our collection of encapsulated A3036C. We have five fresh circuits with batteries and leads loaded, ready for epoxy. We are upgrading two older circuits for A3036C, to be loaded with the last two 10-mAhr batteries we have in stock. All existing devices working well.

[14-JUN-21] We implanted an IST June 3rd in collaboration with Brandeis University, Nelson Laboratory. We used a 34d old wild type male mouse for the experiment. The surgery consisted of three major parts: AAV injection, IST implantation, and surface A3036IL-A blue implantation (surface illuminator).

AAV injection: We injected AAV-CaMKIIa-hChR2(H134R)-EYFP into the right hemisphere of the M2 of the mouse. This should lead to expression of CHR2 in excitatory neurons. We will not be able to confirm expression until we image the brain of the mouse, but our last test of AAV injection on a previous mouse showed successful opsin expression.

Craniotomy/Surface illuminator: A craniotomy was conducted using the coordinates of the M2 cortex of the mouse. A hole was drilled where the lamp was intended to be placed. The surface illuminator was then held directly on the hole, and secured to the skull with dental cement. The metal tube on the back of the surface illuminator was snipped and the remains were covered with dental cement to prevent the sharp spike from injuring the mouse.

IST Implantation: The IST battery was checked prior to surgery and was found to be 4.2 V when turned off and 4.0V when on. The stimulator tool was used to confirm that the IST was receiving commands and the LED was responding to them. We implanted the A3036-C IST under the skin of the back of the mouse. We then fed the leads through under the skin and up to the head, where we connected the leads to the surface illuminator via pins and sockets.

We closed up the mouse and applied post-operative ointments and then left it to recover for the night. We then continued post-op care for three days. The wounds on the mouse healed nicely, with no infection or reopening. The mouse was very active one day after surgery and maintained its ability to groom itself, which is a sign of health.

One week after implantation, on Thursday 6/10, we tested the equipment in the faraday enclosure up at Brandeis. The IST was still fully charged and the A3036IL- A blue light flashed as intended. The mouse did not have any noticeable reaction to the flashing of the light at 10Hz for 2 seconds. It could be that the AAV is not sufficiently expressed yet in the brain. We will test the animal again with longer periods of light flashing on Tuesday 6/15 and Friday 6/18.

[15-JUN-21] We ran the light flashing experiment again today at 11am in the Brandeis Faraday enclosure. Today marked 12 days post-surgery and AAV injection. The light was flashed with 10-ms pulses, a period of 100 ms, and the number of periods was 1200 (2 minutes). The light flashed and the IST reported battery voltage 3.9V. The mouse demonstrated no detectable behavioral reaction to the flashing stimulus. We repeated it 3 times, with no reaction. The working theory is that either the AAV is not expressed in the intended are of M2 or it is not fully expressed yet. We will try again on Friday 6/18 and we will trying varying pulse lengths and frequencies.

The A3036C is equipped with a 10-mAhr LiPo battery. Its inactive current consumption is 7.5 μA. Its current consumption during a stimulus in which the lamp is on for 10% of the time is approximately 1.6 mA, see Battery Capacity. With 5 of 100-s stimuli per day we expect the batter to last for 25 days. With 5 of 100-s stimuli every three days, the battery will last for 36 days. Our plan is to check and test the animal every three days or so, which should allow us five weeks.

[17-JUN-21] We examine the seven A3036B we assembled with batteries ready for encapsulation. They are touching one another in their tray. Two have flat batteries, No2 and No28. We recharge directly through battery connector until charge voltage is 4.1 V. Currents remain around 8 mA. We place in enclosure and at 12:30 pm we initiate stimulus 10% duty cycle with test lamps. No28 is equipped with a lime-green LED with forward voltage 2.5 V, so we expect lamp current of 25 mA and average current of 2.5 mA. We expect No28 to flash for 4 hours. But No2 is equipped with a royal blue LED with forward voltage 2.9 V, so we expect lamp current current 13 mA, average stimulus current around 1.5 mA. We expect No2 to flash for 6 hours. Four hours later, No28 stops flashing, No2 still going, battery voltage measured directly is 3.8V. We stop the stimulus, intending to continue tomorrow.

[18-JUN-21] Re-start stimulus in No2 at 7:48 am. We recharged No28 overnight with an A3033D and it now reports battery voltage 4.0 V. We measure 3.89 V. Still flashing at 9:00 am, but we forget to turn it off when we leave.

[21-JUN-21] We re-charge No28. Both No2 and No28 report VBAT=4.2V. We begin flashing 10 ms at 10 Hz at 13:15. No2 is equipped with a red Luxeon Z with VF = 1.8 V, No28 with a lime green Luxeon Z with FV = 2.5 V. A couple of hours later we measure VBAT 3.9 V and 4.0 V with the lamps off. Lamp current with VBAT = 3.9V will be 38 mA and 25 mA respectively. We expect No2 to run for 2.6 hr and No28 to run for 4.0 hrs. After 2.5 hrs, No2 stops. Its battery voltage is 3.5 V. After roughly 4 hours, No28 stops. After five and a half hours, we check its battery and find the voltage is 1.1 V. Recharging both with A3033D.

[18-JUN-21] We ran the light flashing experiment again today at 10 am in the Brandeis Faraday enclosure. Today marked 19 days post-surgery and AAV injection. We flashed the light was flashed with the following parameters:

  1. 10ms pulses, a period of 100ms, and the number of periods was 1200(2 minutes).
  2. 10ms pulses, a period of 50ms, and for 2,400 periods.
  3. >10ms pulses, a period of 200ms, and for 600 periods.

The light successfully flashed during each trial and the IST reported battery voltage 3.7 V. The mouse demonstrated no detectable behavioral reaction to the flashing stimulus of trials 2 and 3. During the first trial, however, with a 100ms, the mouse did stand in once place and rear on its hind legs repeatedly. This was abnormal behavior but not significant enough to conclude it is due to the light.


[02-JUL-21] A week ago we encapsulated No2 and No28 with epoxy. After encapsulation, No28's battery was flat. We recharged the battery, executed several 100-s stimuli and left for 48 hours. The battery is now flat. No2 reports battery voltage 4.2 V. With a hot iron we remove the battery from No28 and find a cavity beneath.

Figure: Cavity Revealed After Removing Faulty Battery. Resistors appear to be covered with a think coat of epoxy.

We load a new battery onto the circuit, check that it works, and dip in epoxy.

We ran the light flashing experiment again July 2nd at 10am in the Brandeis Faraday enclosure. July 2nd marked 4 weeks post-surgery and AAV injection. We flashed the light with the following parameters:

  1. 10ms pulses, a period of 100ms, and the number of periods was 1200(2 minutes).
  2. 10ms pulses, a period of 50ms, and for 2,400 periods.
  3. 10ms pulses, a period of 200ms, and for 600 periods.

The light successfully flashed during each trial. One noticeable issue was that the light only began flashing after I clicked START and STOP once or twice and then clicked START again. It seems the IST is not receiving every message send from the computer. The device number light up green every time I clicked START, which suggested that the message was being sent to the IST, but was not being received 100% of the time. The IST reported battery voltage of 3.7V. The mouse demonstrated no detectable behavioral reaction to the flashing stimulus during any of the trials. The repeated rearing on hind legs that I had seen last week was not noticeable this week.

[07-JUL-21] We ran the light flashing experiment again today, July 7th, at 10am in the Brandeis Faraday enclosure. July 7th marks 5 weeks post-surgery and AAV injection. We flashed the light with the following parameters three times. We waited 2 minutes in-between each trial to give the mouse time to recover: 10ms pulses, a period of 100ms, and the number of periods was 1200 (2 minutes).

The issue of the IST not responding to every stimulus sent by the computer was still occurring this week. I had to click the START and STOP buttons multiple times before the light began to flash. When the light did begin to flash, it flashed for the entire time it was told to (1200 periods). The IST reported battery voltage of 3.7V. The mouse demonstrated no detectable behavioral reaction to the flashing stimulus during any of the trials. The mouse was still in good condition despite living with the IST and ISL under its skin. It has continued to groom itself and be active throughout the experiments.

[12-JUL-21] Have No2 encapsulated and No28 freshly charged. All ISTs flashing except No11, which we set to recharge. We have nine A3036C in-house, two A3036D, and one A3036B. Of the nine A3036C, one drains its battery every day. The other eight we believe are fine, but we are going to be watching their batteries more closely this week and next. Another A3036C has been implanted at Brandeis for four weeks. We set No28 to flashing 10 ms, 10 Hz, red LED, random at 13:00. Flashing continues until at least 16:00, but is done by 18:00. At 18:00 we disconnect No11 from the charger. It reports battery voltage 4.1 V. We connect No28 to the charger.

[13-JUL-21] No28 flashing. All others flashing.

[14-JUL-21] No28 still flashing. All others flashing except No17, which we attach to charger. In late afternoon, load lamp onto No17. All ISTs flashing. At 18:30, 36 hours after we last charged it, No28 is still flashing. All others flashing.

[15-JUL-21] We have No46 back from Brandeis after five weeks implanted. We soak in acetone overnight and are able to break up the white dental cement the next morning and recover lamp pins. We connect to charger. All our other ISTs flash except No6, which we connect to charger. We note that No28 is still flashing. Battery reports are 3.6 V for No7 and No45, 4.0 V for No17, and 3.7-3.9 V for all the rest, including 3.8 V for No28.

[16-JUL-21] All ISTs flashing except for No28, which does not respond. No7, No45, and No56 report VB = 3.6 V, No6 and No17 say 4.0-4.1 V. All others 3.7-3.9 V. No46, the five-week implant, says 3.9 V. After two hours we connect No28 to lamp and flash it.

[17-JUL-21] All ISTs flashing. All respond to battery instruction first time. We have No7 and No45 reporting 3.5 V. No1 and No56 report 3.6 V. All others 3.7-4.1 V. We put No7 and No45 on the charger.

[18-JUL-21] Remove No7 and No45 from charger. All ISTs flashing except No28. Put No28 on charger. We have No7 reporting 4.1 V, No45 reporting 4.2 V. All others report 3.8-4.2 V. We run another stimulus.

[19-JUL-21] Take No28 off charger. All ISTs flashing. Battery voltages 3.7-4.2 V. Two stimuli.

[20-JUL-21] All ISTs flashing. Battery voltages 3.8-4.2 V except No56 is 3.6 V. Put No55 on charger by mistake.

[21-JUL-21] Remove No55 from charger. All ISTs flashing except No28. Battery voltages 3.8-4.2 V except No56 is 3.5 V. Put No56 on charger.

[22-JUL-21] Take No28 and No56 off charger. All nine A3036 and all four A3037 are flashing. Battery voltages for the nine A3036 are 4.0 to 4.3 V.

[27-JUL-21] Report from Alice Hashemi at ION/UCL.

SURGERY: On Monday July 26th 2021, Rob Wykes and I conducted a surgery on a mouse that had been injected with AAV-CaMKIIa-hChR2(H134R)-EYFP. The AAV was injected into the middle of the M2 cortex 5 weeks ago in order to allow for optimal opsin expression once we started the surgeries. Rob injected the AAV using two different holes in the skull and injected 1mm deep.

The main components of the surgery included implanting the IST (A3036C), DC-40 Hz transmitter (A3028WZ-AAA-B45-B), 3 X-Electrodes, and a 460-nm Blue 270um x 4mm fiber-coupled LED (FCLED, A3036IL-A270-4). The IST and DC-40 Hz transmitter were implanted into the lower back of the mouse, with one on either side of the back. We then cut away the skin from the surface of the skull and drilled four holes through the skull to the surface of the cortex. Rob was able to drill a hole in-between the two holes he had used 5 weeks ago to inject the AAV. This allowed us to be confident that we were placing the FCLED in the M2 cortex. We drilled a hole near the FCLED on the anterior end of the brain for the first X-electrode. We inserted the fiber of the FCLED 1 mm deep into the M2. We fed the purple and orange leads of the IST under the skin of the back up through the hole we cut in the scalp. We inserted the pins of the IST leads into the sockets on the FCLED. We inserted an x-electrode and connected it to one of the EEG leads of the DC transmitter (NOT the reference lead, which is the blue one). The leads were again fed up through the cavity between the skin and the back of the mouse, out through to the exposed hole in the head. We used the crimping technique for all three electrodes in this surgery. Once the electrode was lowered about 1 mm into the surface of the cortex, we started to secure the FCLED and electrode in place. We used silicone to surround the remaining parts of the fiber that was exposed above the skull. We completely covered the electrode and the ISL with dental cement. In order to help secure the dental cement we had inserted a screw into the anterior part of the skull, out of the way of the electrode and FCLED.

Once the dental cement dried on the top half of the skull, we drilled two more holes in the lower half of the brain to insert the remaining two electrodes. We repeated the insertion, crimping, and dental cement procedure as previously described. Once the dental cement cured, we went back over any exposed metal on the mouse brain and covered it with more dental cements. We sutured the incision on the back and gave the mouse pain medicine and saline for recovery.

POST SURGERY: On July 27th, we conducted the first experimental protocol on the mouse. We placed the cage in a table top Faraday enclosure and turned on the IST and DC transmitter. We turned on the recorder instrument in LWDAQ and detected the synchronizing signal from the IST as well as the signal from both channels of the transmitter. The IST reported a battery voltage of 3.8 V. The ISL light was then flashed using the stimulator tool using the following parameters: pulse_ms = 10, period_ms = 100, num_pulses = 900. The mouse began having a seizure, which we could detect in the EEG recordings as well as visually in the mouses behavior. The 30-second seizure was followed by the mouse circling around its cage counterclockwise for about 30 seconds. When the light stopped flashing, the mouse stopped moving immediately after and sat in the corner of its cage.

[28-JUL-21] All stimulators flash except No28 and A3037A No69. When our eight A3036s have stopped flashing, they report battery voltages 4.0-4.2 V.

[30-JUL-21] All ISTs flashing except No28. Put it on charger. All others report battery 4.0-4.2 V. Ship our final five functional A3036C to ION/UCL leaving us with only No28.

[01-AUG-21] Report from Alice Hashemi at ION/UCL.

Circling Experiment protocol: Pulse_ms:10 Period_ms: 100 Num_pulses:900. Trial 1: The mouse was asleep but got up as soon as the light started flashing, just as we had observed on Tuesday. The mouse then entered what looked like a seizure for the reminder of the trial. We stopped the light after 75 seconds. The mouse showed no signs of movement outside of the seizure like twitching, and the seizure had not subsided, so I did not want to put it through any unnecessary suffering. Trial 2: The protocol was repeated again with the parameters previously stated. When the light began flashing, the mouse walked around for a few seconds, then stood on its hind legs and had another seizure. The seizure lasted a minute, and then the mouse began to circle once or twice around the cage clockwise. The light stopped flashing before it began properly circling like we had seen Tuesday.

Figure: Recording M1627563858 of Optogenetically Induced Seizures. The two 90-s stimuli occur during the two downward shifts in the two LFP recordings from X-Electrodes, as seen by the 0.0-40 Hz response of the WZ transmitter. Channel X is shown black, Y is shown in gray. Channel numbers 35 and 36 with colors altered for clarity.

The recording M1627563858.ndf contains the synchronizing signal from the A3036C as well, and the stimulus lines up with the two LFP downward shifts in the plot above. During each stimulus, we see ictal activity in both channels.

Figure: Optogenetically Induced Seizure, Eight-Second Interval with Spectrum. Voltage versus time of X on left, spectrum of X on right, 2 μV/div, 5 Hz/div.

In the example above, we see ictal pulses at 3.3 Hz with clear harmonics at 6.6 Hz and 10.3 Hz. The stimulus pulses are at 10-ms at 10.3 Hz. We see no sign of lamp artifact in the LFP signals.

[04-AUG-21] At ION/UCL, No6 has exhausted its battery after six days implanted. We believe the current drain was caused by accidentally turning on the command transmitter's 910 MHz power and leaving it on, so that No6 was actively expecting a command for thirty hours. We find that seizures are the result of 10-ms flashes at 10 Hz, but with 1-ms flashes we see immediate circling that ends when the flashes end. We have two-channel EEG with synchronizing signal and video of repeated circling and also a CSD provoked by turning on the lamp for 20 s.

[07-AUG-21] All IST/ISS flashing except No28. Put No28 back on charger. At ION, the first two A3036Cs implanted have failed after six days. We suspect their synchronizing signals were left running after some problems with turning on and off the command transmitter. They have four animals implanted, and are able to cause all to circle, even the one with surface illuminator. Circling with 1-ms flashes at 10 Hz, shaking in place for 10-ms flashes.

[09-AUG-21] All IST/ISS flashing including No28. When we open the Faraday enclosure we notice that No28 flashes its lamp once, which seemed odd at the time, but we are going to watch for such flashes in future, following Alice's reports of random flashes at ION.

[10-AUG-21] At ION, animal No3 with IST 215.7 and the surface illuminator. Reports battery voltage 3.9 V. Trial_1: 10 ms pulse, 2 Hz (500 ms period), 90 s (180 flashes), some scratching/digging, no circling, no seizure. Trial_2: 10 ms pulse, 5 Hz (200 ms period), 90 s (450 flashes), no circling. Trial_3: 10 ms pulse, 5 Hz (200 ms period), 90s (450 flashes), no circling. Trial_4: 10 ms pulse, 10 Hz (100 ms period), 90 s (900 flashes), minor circling but slowly, stopped circling after abbot 40 s. Trial_5: 10 ms pulse, 15 Hz (66.7 ms period), 90 s (1349 flashes), one small circle, then mouse seemed tp fidget. Was not a clear seizure or clear circling. Trial_6: 10 ms pulse, 20 Hz (50 ms period), 90 s (1800 flashes), circled a few times in both directions. Trial_7: 10 pulse, 50 Hz (20 ms pulses), 90 s (4,500 flashes), seemed irritated. Could be CSD, mouse was very still. Trial_8: 10 ms pulse, 100 Hz (10 ms period), 90s (9,000 flashes), looked like a CSD reaction. No circling, no seizure.

Animal No4 (IST 215.11), IST not responding. Implanted on August 4th. We had near perfect reception in the enclosure and there IST in animal No3 worked well. The mice are not kept in a Faraday enclosure. They are kept in a room where the ISTs flash occasionally at random. In the surgery room, when we hold an IST in our hand, it flashes often.

Previous experiments on Animal No4, IST 215.11. August 5th. Trial_1: 1 ms pulse at 10 Hz for 90 s, no circling, seizure like behavior for entirety of time. Trial_2: 2ms pulse at 10 Hz for 90 s, had seizure. Trial_3: 10 s continuous for CSD, unclear what happened, see direct artifact and then U-shape afterwards, 60 s baseline before, and 7 minutes after. August 6th. Trial_1: 1 ms pulse at 10 Hz for 90 s, circled once, went into seizure right after, and seizure continued entire time. Trial_2: 0.1 ms pulse at 10 Hz for 90 s, no behavioral changes. Trial_3: 0.5 ms pulse at 10 Hz for 90 s. circled once, then went into seizure. Trial_4: 0.3 ms pulse at 10 Hz for 90 s, circled about 10 times towards left, no seizure. Trial_5: 0.5 ms pulse at 2 Hz for 90 s flashes, no behavioral response, light on at 110.0 s, ends at 200.0 s. Trial_6: 0.5 ms 20 Hz for 90 s, had a seizure.

[16-AUG-21] Take No28 off charger. All four ISTs flashing and three of ISS, but No83 not responding, connect to charger.

[17-AUG-21] No28 won't flash. No83 recharged. Others flash except ISS No68, put on charger.

[18-AUG-21] Put No28 on charger. No68 now recharged. All flashing.

[22-AUG-21] Remove No28 from charger. All flashing except No37. We last re-charged No37 on 21-MAY-21, 13 weeks ago. Even without stimuli, of which there have been many, this amounts to 16.5 mAhr of inactive battery use. The No37 battery is 19 mAh LiPo. We put No37 on the charger.

[23-AUG-21] Take No37 off charger. All flashing (Nos 28, 37, 45, 56, 68, 69, 83, and 84).

[01-AUG-21] Take No28 off charger but it won't flash. Tired of charging this one every other day, so toss it out. We sent No37 to Cornell where it provoked circling behavior. All others flashing (Nos 45, 56, 68, 69, 83, and 84).


[06-OCT-21] We ship A3036D numbers D215.45 and D215.56, along with four A3036IL-A270-4 FCLEDs and an A3033D battery charger to Edinburgh University.

[28-OCT-21] We have twelve A3036C back from ION. We begin testing, recharging, and dissecting them. With seven dissected, we have five that show the 4.6-mA drain after activation by RF power that is characteristic of our Unexplained Logic Failure. At ION, we observed implanted ISTs to be flashing their lamps occasionally and randomly when being carried around. We did not observe this random flashing while the animals were in a Faraday enclosure. We can sometimes cause an IST lamp to flash with proximity of a mobile phone. We suspect interference causing the ISTs at ION to power up randomly. We hypothesise that short power-up pulses are corrupting the configuration memory of the logic chip through aborted readout. Two ISTs are in perfect working order, one of them is No46, which spent five weeks implanted at Brandeis. We generate random pulses of RF power 2-20 ms long and leave No46 flashing in response to them. After an hour it is unharmed.

We consult the A3030 manual, which shows how the LCMXO2-1200 powers up in response to RP, see 07-AUG-14. The chip configures itself in only 620 μs. Our 2 ms pulses were too long to cause any damage. In our experience, the 7-μs pulses of an implanted transmitter are too weak and to short to turn on the power supplies, let alone initiate configuration of the logic chip. We should try pulses of 10-1000 μs.

[01-NOV-21] We set up an A3014MT, raising U4-4, the VCO shutdown input, and connecting to our pulse generator. We find that 420-μs pulses of 915 MHz is the minimum required to get a flash from the LED of No46. We apply pulses 10-3000 μs at 10-100 Hz. No46 continues to function. There follows a list of returned and ISTs with diagnosis for each one. We use "ULF" for Unexplained Logic Failure.

The Unexplained Logic Failure (ULF) has three constant features: 4.6 mA after command, flash on command, and 33 μA sleep current. This last one is new, and suggests that the problem is not with the logic chip.

[03-NOV-21] Remove battery from C215.6. Burn off epoxy around R7. Apply external 3.7 V. Sleep current 33 μA. No RCK on R7. Issue RF command, current 4.7 mA. Remove U7, clean and dry, current 19 μA and falling. Connect external 3-V 32.768 kHz. On RF command, device powers up briefly. We look at RP on R6 and it looks fine during commands. Despite RP being asserted, however, 3VB is not turning on, as seen on C9. We remove U6, but the UDFN-6 resists our efforts and we end up damaging the top layer of the circuit board. Applying external 3.0 V to VB and 3VB draws 300 mA.

Remove battery from C215.7 and connect 3.7 V. 30 μA sleep current, flash on any RF command, powers down after. Remove epoxy from around R7. Sleep current 7.5 μA then dropping to 5.4 μA. No RCK on R7. Remove U7, sleep 6 μA. Connect external RCK and look at RP. See robust bit reception on RP. But 3VB does not budge from 0V during command reception despite RB being applied to U6-1 and stable 3VA on U6-6. Remove U6, but compromise the tuner, see negative pulses on VR. Connect +3.7V to 3VB and see 300 mA flowing in.

Remove battery from C215.55. Connect 3.7 V. 30 μA sleep and 4.7 mA after command, with lamp flash upon command. Expose R6 and see no RCK. Expose decoupling capacitors. See steady 3VA but RCK is stuck HI. Sleep current 31 μA. On RF command, 3VA remains stable. Sleep current is now 8 μA and RCK is a random series of transitions of about 10 Hz average.

[04-NOV-21] Connect power to C215.55, sleep current 30 μA. After ten minutes, drops to 6.4 μA and we see erratic transitions on RCK. Lamp flashes on RF command and current returns to 6.3 μA. Solder leads to RCK and 3VB. Now sleep current is 33 μA and RCK is stuck LO. On RF command, lamp flashes and current is 4.6 mA. Disconnect power, 33 μA. We watch 3VB during RF command and see it turn on and stay on while RCK remains LO. Connect external 32.768 kHz to RCK. Now see stimulus on command to No55, sleep current 33 μA as before. With XON see synchronizing signal transmission. Initiate stimulus of 1000 pulses, all goes well.

Diagnosis of Unexplained Logic Failure: The 32.7680kHz oscillator, U7 ASTMTXK, fails. It does not produce RCK. It draws 33 μA instead of 1 μA. When the IST receives an RF command, the logic chip U8 powers up. Because RCK is not running, the power controller process does not move into its standby state, leaving U8 consuming 4.7 mA. The state of OND is uncertain. If OND is HI after power-up, the IST remains powered up, consuming 4.6 mA. If it's LO, 3VB will turn off as the command proceeds, causing further indeterminate behavior of OND. If at any time OND goes HI, it will stay HI, bringing us back to the 4.6-mA ULF current. We have observed the failure of the ASTMTXK in the past, caused by cracking its ceramic package during assembly, see here, here, and here. Our assembly house recommended that they place these parts by hand so as to avoid cracking. Hand-placement has proved reliable in a thousand SCTs so far. In these A3036BV1 assemblies, we observe the failure of U7 after some weeks or months in our laboratory, but more quickly when implanted in animals, and in almost every case. Our assembly house checked their record of the A3036BV1 build, and they are certain they placed U7 by hand. Accelerated failure in warm, humid air is a characteristic of cracked ceramic objects. The crack propagates with corrosion until it reaches a critical part of the component, and the component fails. We plan to stop using the ASTMTXK and switch to another part in a more robust package, such as the SiT1533.

[10-NOV-21] Receive 9 A3036 and 1 A3037 from Cornell. Soak in acetone, rinse in alcohol to remove dental cement and other residue. Clip damaged lead and expose wire at lead ends. Start charging and testing. Many of the batteries are puffy and a few ISTs smell sickly sweet.

We take a CR1025 coin cell and connect 100 Ω across its terminals. We watch the evolution of the voltage across the resistor with time. We see 25 mA for a few microseconds, then 23 mA decreasing to 22 mA over the next 20 ms. The CR1025 can start up the LCMXO2-1200 and provides 30 mAhr capacity. We need only 1.5 mA lamp current for 10 ms flashes to provoke optogenetic response, so we could power the entire A3036 off one such battery, do away with the puffy battery issue, forget about recharging, and increase logevity.

[11-NOV-21] We connect CR1025 to A215.21 and an A3036IL-X. Device responds to commands and flashes its light. Battery voltage is 3.0 V during stimulus with synchronizing transmission, and drops to 2.9 V during lamp flashes, see below.

Figure: Battery Voltage (Yellow) and Lamp Power (Blue). Battery voltage AC-coupled 50 mV/div. Lamp power DC-coupled, 1 V/div. Timebase 5 ms/div. Stimulus 10-ms flashes at 10 Hz.

We commence 10-ms flashes at 10-Hz at 09:15, along with synchronizing signal transmission. We are receiving an average of 1020 SPS on channel No21. At 10:15 flashing stops. Battery voltage while inactive is now 2.9 V, but during command reception drops to 2.0 V in jagged steps like this. With 3.0-V LED drive and 50-Ω lamp leads, lamp current can't be much more than 1 mA for 10% of the time. Average current may be 300 μA for this stimulus and transmission, so we have used 0.3 mAhr from a battery with capacity 30 mAhr. Detach battery, connect to 100 Ω to battery terminals. See 21 mA supplied to resistor for 100 ms. With no load, battery returns to 2.9 V. Disconnect lamp. Battery voltage drops to 2.0 V upon command reception. The A3036 uses a single TPS70930 to provide 3.0 V to the crystal radio and the logic I/O. When the logic chip starts to power up, a few hundred microseconds after RP is asserted, VBAT drops to 1.9 V, at which point, the TPS70930 will turn off its output.

Figure: TPS70930 Power Down. This plot of VIN and VOUT is for the TPS70933, 3.3-V micropower regulator. We are using the TPS70930 3.0-V regulator.

The repeated saw teeth we see in VBAT during command reception appear to be the TPS70930 turning off and on as its input drops below its minimum, then rises again after power turns off, and so on, which continues so long as RF power is detected by the crystal radio. Disconnect battery, put 10 Ω in series with 0V and connect 3.7 V power, measure voltage across resistor on power-up.

Figure: Power-Up Current with C9 = 10 μF. Scale 20 mA/div, timebase 100 μs/div. For C9 = 1.0 μF see here. The IST won't start up with C9 = 0.1 μF.

We divide the power-up current into three regimes: a "front porch" of 24 mA for 200 μs, a "pulse" of 140 mA for 100 μs, and a "back porch" of 22 mA for 1 ms. The front porch is consistent with charging C9 to 3.0 V. The pulse is consistent with charging C3 to 1.2 V, and the back porch appears to be the start-up of U8. We reduce C3 to 4.7 μF. Peak of the pulse is 120 mA, but duration of pulse is only 50 μs. Reduce C3 to 1.0 μF and see the power-up shown below.

Figure: Power-Up Current with C3 = C9 = 1.0 μF. Scale 10 mA/div, timebase 100 μs/div. For C3 = 4.7 μF with scale 20 mA/div see here.

Connect our CR1025 battery again. With C3 = C9 = 1.0 μF the battery is sufficient to power up the circuit. Connect lamp and see stimulus flashes. After a few minutes, however, we are once again unable to control the IST and VBAT shows saw tooth when we transmit command.

[03-AUG-22] We note that CREE has a new LED die out, the EZ400-p. The light-emitting surface is 380 μm square with the anode bond wire in one corner. With this LED, we should see a significant increase in light coupled into 270-μm fibers. Meanwhile, the EZ500 has been replaced by the EZ500-p with a corner anode pad. This die would be more convenient for 450-μm fibers also.

[15-AUG-22] We use the SD445 to measure the power of an infrared emitter with and without reverse bias. Saturation of the photodiode is evident at a photodiode current of 4 mA with no reverse biase.

Figure: Saturation of SD445 at High Optical Power.


[11-JAN-25] After a long break, we have revived our tapermaker, arranging it horizontally, with the product fiber on the right and the cast-off fiber on the left. We make the following taper out of 290-μm diameter TD5 fiber.

Figure: First Taper with New Tapermaker Location. Fiber diameter 290 μm.

[01-FEB-24] After removing a broken limit switch, we now have our tapermaker working well. We make a 6-mm long light guide, 4 mm of fiber and 2 mm of taper. We take this movie.