Head-Mounting Transmitter (A3040)

© 2021-2023, Kevan Hashemi, Open Source Instruments Inc. © 2022, Calvin Dahlberg, Open Source Instruments Inc.

Contents

Description
Warranty
Versions
Connector
Analog Inputs
Battery Life
Wrapping
Antenna
Design
Modifications
Development

Description

[12-DEC-23] The Head-Mounting Transmitter (HMT, A3040) is a telemetry sensor that connects to an Electrode Interface Fixture (EIF) cemented permanently to the skull of a subject animal. The HMT provides high-fidelity amplification and recording of up to four biopotentials. The EIF provides leads, electrodes, and a connector that mates with the HMT. Once cemented in place, the EIF provides not only electrical connection to the HMT, but is also the means by which the HMT is secured to the subject animal. The HMT transmits its signals using its flex-circuit antenna, which hovers over the back of the animal. The HMT uses the same wireless communication system as our Subcutaneous Transmitters (SCT).


Figure: Head-Mounting Transmitter (A3040D) Wrapped in Teflon Tape. Battery is CR1225, 2000 μA-dy, total mass 2.2 g. A flexible antenna hovers over animal's back.

The HMT needs to be splash-proof and dirt-resistant, but need not be water-proof or corrosion-resistant. We wrap the circuit in teflon tape during use, so as to cover the electronics, but allow the connector and antenna to protrude. When we deploy an HMT, we hold the subject animal securely, press the HMT's connector onto the EIF, and secure the two connectors together with a drop of silicone sealant such as KwikCast. The sealant prevents the animal from removing the HMT. When we ourselves want to remove the HMT, we hold the animal securely, peel off the sealant with forceps, and slowly work the two connectors apart. Once the HMT has been detached, we remove its wrapping, push the battery from its retainer, load a new battery, and wrap in fresh tape. It is now ready for re-deployment. The HMT will provide continuous recordings of indefinite length, so long as we can tolerate five-minute interruptions for battery replacement.

PropertyValue
Volume of Transmitter Body1.6 ml
Mass of Transmitter Body with Teflon Wrap2.2 g
Body Dimensions14 mm × 14 mm × 8 mm
Total Height with Connector11 mm
Maximum Operating Current235 μA
Battery TypeCR1225 Coin Cell
Battery Capacity2000 μA-days
On-Off Switchload or remove battery
Number of Inputs4
Input Impedance10 MΩ || 2 pF
Sample Rate (Each Input)256 SPS
Sample Resolution16-bit
Input Dynamic Range27 mV
Input Bandwidth0.3-80 Hz
Input Noise≤8 μV rms
Total Harmonic Distortion<0.1%
Absolute Maximum Input Voltage±5 V
Minimum Operating Life14 days
Table: Specifications of the A3040D2 Four-Channel 0.3-80 Hz Head Mounting Transmitter.

The HMT has no on-off switch. When we slide the battery into the retainer, it starts transmitting. It keeps transmitting until the battery runs down or we remove the battery. We can deduce the state of the battery from the average value of teh telemetry signals, so we will always have a day or two warning before we need to replace the battery if we want to maintain a continuous recording. The HMT provides four amplifiers with a common reference potential. It is designed for recording EEG from one to four locations on the brain, sharing a common reference potential over the cerabellum.


Figure: The A3040AV1 Circuit with Battery Loaded. Visible are the flexible antenna on the left, electrode interface connector in the center, and the battery holder on the right. This battery is a CR1225.

Current consumption is linear with total sample rate. We use intecept ≤25 μA and slope ≤0.12 μA/SPS to calculate the maximum current consumption of an HMT, and divide the nominal battery capacity by this current to obtain our minimum operating life.

Warranty

[02-MAY-22] The HMT comes with a one-year warranty against unfortunate events. And unfortunate event is any event during the course of an experiment that damages the HMT but which appeared at the start of the experiment to be unlikely. In the case of an unfortunate event, return the HMT to us. We will either repair your HMT and return it to you promptly, or replace your HMT and send the replacement promptly. The replacement will be covered by the warranty of the original HMT.


Figure: An Unfortunate Event. Connector broke off prototype HMT while removing from animal. We avoid this problem by epoxying the connector to the HMT.

The HMT is designed to be used repeatedly. We replace the battery and wrap it again. Unlike our implantable transmitters, the HMT is not encapsulated in epoxy, and is not hidden away beneath the skin of an animal. The HMT is vulnerable to scratching, abrasion, and possible accidents involving water and urine. If an HMT becomes detached from an animal during an experiment, it might be chewed and destroyed while it is loose upon the floor of the animal cage. While cutting off our wrapping, we might inadvertently cut the antenna, or we could crack one of the HMT's tiny, ceramic components.


Figure: Amputated Antenna. This HMT was pulled off its electrode interface fixture by its host mouse. The mouse then chewed through the antenna. The transmitter circuit still functions, but without its antenna, reception is poor.

The vulnerability of the HMT to physical damage during its multi-use lifespan introduces a financial risk into any experiment we plan with the device. In order to relieve our customers of this financial risk, and instead take this risk upon ourselves, we warranty the HMT for one year after shipping against inadvertent damage as well as manufacturing defect.


Figure: Chewed Connector. This HMT was pulled off its electrode interface fixture by its host mouse. The mouse chewed the wrapping and bit the connector.

If, despite taking reasonable precautions, you find that six of your ten HMTs have been dismounted and chewed within a year of your purchasing them, send them back to us. We will confirm the nature of the damage, repair them if possible, and replace them otherwise.

Versions

[12-MAY-23] All versions of the A3040 have the same profile: 14 mm × 14 mm × 8 mm with 14-mm rear flex antenna. But they may be programmed to enable one to four input channels at various sample rates and bandwidths. We can equip them with CR1025, CR1220, or CR1225 batteries, but we doubt we will ever see the device deployed with anything other than the largest battery, the CR1225. The CR1225 provides 1.6× the capacity of the CR1025, at the cost of an increase in mass of only 10%. The table below gives example versions with their signal bandwidths and operating life.

Version W X Y Z Battery
Capacity
(μA-dy)
Mass
(g)
Life
(dy)
A3040C2 X1: 0.3-160 Hz, 512 SPS, 27 mV X2: 0.3-160 Hz, 512 SPS, 27 mV X4: 0.3-160 Hz, 512 SPS, 27 mV Disabled 2000 (CR1225) 2.2 9
A3040D1 X1: 0.3-40 Hz, 128 SPS, 27 mV X2: 0.3-40 Hz, 128 SPS, 27 mV X3: 0.3-40 Hz, 128 SPS, 27 mV X4: 0.3-40 Hz, 128 SPS, 27 mV 2000 (CR1225) 2.2 23
A3040D2 X1: 0.3-80 Hz, 256 SPS, 27 mV X2: 0.3-80 Hz, 256 SPS, 27 mV X3: 0.3-80 Hz, 256 SPS, 27 mV X4: 0.3-80 Hz, 256 SPS, 27 mV 2000 (CR1225) 2.2 14
A3040D2Z X1: 0.0-80 Hz, 256 SPS, 270 mV X2: 0.0-80 Hz, 256 SPS, 270 mV X3: 0.0-80 Hz, 256 SPS, 270 mV X4: 0.0-80 Hz, 256 SPS, 270 mV 2000 (CR1225) 2.2 14
A3040D3 X1: 0.3-160 Hz, 512 SPS, 27 mV X2: 0.3-160 Hz, 512 SPS, 27 mV X3: 0.3-160 Hz, 512 SPS, 27 mV X4: 0.3-160 Hz, 512 SPS, 27 mV 2000 (CR1225) 2.2 7.4
A3040D3Z X1: 0.0-160 Hz, 512 SPS, 270 mV X2: 0.0-160 Hz, 512 SPS, 270 mV X3: 0.0-160 Hz, 512 SPS, 270 mV X4: 0.0-160 Hz, 512 SPS, 270 mV 2000 (CR1225) 2.2 7.4
A3040D4 X1: 0.3-320 Hz, 1024 SPS, 27 mV X2: 0.3-320 Hz, 1024 SPS, 27 mV X3: 0.3-320 Hz, 1024 SPS, 27 mV X4: 0.3-320 Hz, 1024 SPS, 27 mV 2000 (CR1225) 2.2 3.8
Table: Version of the A3040 Head-Mounting Transmitters. Minimum operating life at 25°C in days with Renata or Panasonic batteries.

In the table above, we specify mass with Teflon. A silicone wrap is more rugged, but adds another 0.3 g to the mass.

Connector

[08-FEB-23] The A3040A provides an eight-way, dual-row, 0.025" pitch, hermaphroditic surface-mount connector that mates with a connector of the same type on the subject animal's head. We use Omnetics PZN-08-VV (A79612) on the circuit board, where we solder the connector's gull-wing leads to its footprint. We use PZN-08-DD (A79614) on the Electrode Interface Fixture (EIF), where we bend the connector's through-hole pins inwards, cut them short, and solder our leads directly to the cut ends. In the HMT, we use the pin numbering given by the manufacturer, see S3040B_1.

PinEIF8 Lead
Color
Function
1blueGND
2blueGND
3greenX3
4yellowX2
5nonenone
6nonenone
7redX1
8salmonX4
Table: Color Codes, Pin Numbers, and Functions for Wired Connector.

In the HMT circuit diagram, we refer to the reference potential as "VC", the "common voltage". Once we connect this potential to an animal body, we say the HMT is "grounded". Our assumption is the one of the two GND pins on the EIF will be used for a low-impedance connection to the animal body. On the Electrode Interface Fixture (EIF), the numbers we assign to the pins do not match the manufacturer's data sheet, but they do match the pin numbers on the HMT circuit.


Figure: Electrode Interface Fixture (EIF) Connector, Top View. Connector is PZN-08-DD. Pinout matches that on the HMT circuit diagram.

The EIF pin numbers in the sketch match the pin numbers of the mating connector on the HMT circuit. So we have Pin 1 and 2 are GND (VC in schematic, the reference voltages for the amplifiers). We have X1, X2, X3, and X4 on pins 7, 4, 3, and 8 respectively. When we construct the EIF, we will cut off all unused pins. We bend the remaining pins and solder up to five leads. One will be GND. The others can be X1-X4. Each electrode lead can be terminated with bare wire, a pin, or a depth electrode.


Figure: Electrode Interface Fixture (EIF) Connector, Bottom View. Connector is PZN-08-DD. Pinout matches that on the HMT circuit diagram. This is the view we have when assembling an EIF.

The Electrode Interface Fixture may be equipped with up to five leads. One lead must always be present: the low-impedance ground, which will be connected to Pin 1 or Pin 2 or both. For historical reasons, the ground lead is always blue. The leads for X1, X2, X3 and X4 are optional, but they will be red, yellow, green, and cream-colored if present. The leads are 20 mm long by default. They are 0.5 mm in diameter, silicone-insulated springs. On the far end of each lead, we may solder a pin, a depth electrode, or leave a bare wire for securing in place with a screw, as requested by our customer.

Analog Inputs

[02-MAY-23] The A3040D3 is equipped with four amplifiers with gain ×100 and frequency range 0.3-160 Hz. The A3040D3 samples all four inputs with sample rate 512 SPS. Each sample is a sixteen-bit number 0..65535, where the number zero corresponds to the bottom of the input dynamic range, and 65535 corresponds to the top of the range. The plot below shows their response to a 15-mVpp sinusoidal input with 10-MΩ source resistance as we increase the sinusoidal frequency from 0.25 Hz to 1 kHz. Because the input resistance of the amplifiers is 10 MΩ, we see 7.5 mVpp on each input.


Figure: Frequency Response of A3040D3 and A3040D2 Transmitters. The D2 corner frequency is 80 Hz for all four channels, and the D3 has corner frequency 160 Hz for all four channels.

The dynamic range of the inputs is equal to the battery voltage divided by the amplifier gain. For the A3040D3 at 10 Hz, the gain is ×100, so the dynamic range is 27 mV. All inputs are referred to a single reference voltage, VC = 1.8 V. The A3040D3 provides a high-pass filter at its inputs. The average value of any of its input is zero. The amplifiers use VC = 1.8 V to represent an input of zero. With battery voltage 2.7 V, an input of zero will be converted to 1.8 V / 2.7 V × 65535 = 43690 cnt (ADC counts). Count zero corresponds to −18 mV at the input, and count 65535 corresponds to +9 mV. The A3040 amplifiers may be configured to provide low-pass cut-off frequencies 40 Hz, 80 Hz, 160 Hz, 320 Hz, and 740 Hz. These cut-off frequencies accompany sample rates 128 SPS, 256 SPS, 512 SPS, 1024 SPS, and 2048 SPS respectively. The low-pass filter is a three-pole Chebyshev filter with some ripple in the pass-band to allow for a sharper cut-off at the top of the pass band. The default high-pass filter is a single-pole RC filter with cut-off at 0.3 Hz. We can remove the high-pass filter to provide response down to 0.0 Hz.


Figure: Spectrum of A3040D3 Input Noise, All Four Channels. Vertical 0.41 μV/div. Horizontal 10 Hz/div.

Noise on the A3040D3 inputs in a Faraday enclosure with leads ends in water is 20 cnt rms. Full scale is 27 mV or 65536 counts, so each count is 0.41 μV. We multiply 20 cnt rms by 0.41 μV/cnt to get 8 μV rms. The plot below shows the spectrum of this noise.

Battery Life

The HMT current consumption from a 3-V Lithium Primary cell will be no greater than:

Ia = 25 μA + (R × 0.12 μA/SPS)

We divide the nominal battery capacity by the maximum active current to obtain our minimum operating life. The typical operating life is 10% higher, because the typical active current is 10% lower than that given above.


Figure: Active Current Consumption of A3040AV1 versus Total Sample Rate. Straight line fit gives intercept 17 μA and slope 0.110 μA/SPS.

Wrapping

[30-MAR-22] The A3040A has no permanent epoxy or silicone encapsulation. Before loading the transmitter on the animal, we fold the circuit as shown in our sketch. We wrap the circuit in tefloon tape, building up two or three layers, keeping the connector free, and finally adhering the tape to itself. When we replace the battery, we unwrap the tape with our fingers. We do not use scissors because these might make contact with the circuit inside and cause peramanent damage. We remove all the tape and remove the battery by pushing it out of its holder with a wooden applicator. We now we have the opportunity to wash the circuit if we like. We rinse in hot water while scrubbing gently with a tooth brush. We blow dry with compressed air. Now we soak the circuit and its next wrapper in ethanol for a few minutes, blow dry, load battery, and wrap with sterile gloves.


Figure: Top View of A3040P Wrapped in Parafilm Tape. We prefer plumber's teflon tape to parafilm.

Any water-proof tape that adheres well to itself is a likely candidate for wrapping the HMT during use. Self-fusing silicone tape is rugged and water-proof, but adds 0.5 g to the transmitter mass, see here. Plumber's teflon tape is thinner, and adds only 0.2 g to the mass. The British variety of plastic wrap, which they call clingfilm adds only 0.2 g, see here. Parafilm adds 0.2 g see here. Gardener's tape adds only 0.2 g as well, see here.

Antenna

[08-FEB-23] The antenna on the A3040 is a flexible circuit with a rounded end. Counting the neck at its base, the antenna is 14 mm long. It is 4 mm wide. It carries a zig-zag conductor of total length 70 mm, which is slightly less than one quarter of the wavelength of our telemetry transmissions. We transmit in 902-928 MHz, for which the wavelength is around 330 mm.


Figure: Transmit Antenna. Ruler marks are millimeters.

The A3040AV1 provides footprints for a series-parallel matching network between the circuit's radio-frequency (RF) oscillator and the antenna. The AV1 does not attempt to match the antenna to the oscillator: the two are coupled with a single 1 nF capacitor. But the AV2 loads 27 nH for L1 and 0.2 pF for C26 to increase power output by a factor of ten.

Design

[28-AUG-23] Here is the HMT conceptual sketch from our technical proposal.


Figure: Head-Mounting Transmitter Sketch. The transmit antenna protrudes behind the body of the device. The mouse's nose will be to the left in the drawing, and its tail to the right. The connector axis is parallel to the mouse body axis.

Below are schematics, data sheets, and design files.

S3040A_1.gif: 4×0.3-160 Hz, BGA-64, flex antenna, 9-way socket.
A304001A: Gerber files for A3040A PCB, Rev 3.
A304001A_Top: Top view of A304001A.
A304001A_Bottom: Bottom view of A304001A.
A3040AV1_Top.gif: Top side component map of A3040AV1 assembly.
A3040AV1_Bottom.gif: Bottom side component map of A3040AV1 assembly.
A3040AV1.ods: Bill of materials for A3040AV1 assembly.
A3040AV2.ods: Bill of materials for A3040AV2 assembly.
S3040B_1.gif: 4×0.3-160 Hz, BGA-64, flex antenna, 8-way socket.
A304001B: Gerber files for A3040A PCB, Rev 1.
A3040BV1_Top.gif: Top side component map of A3040BV1 assembly.
A3040BV1_Bottom.gif: Bottom side component map of A3040BV1 assembly.
A3040BV1.ods: Bill of materials for A3040BV1 assembly.
S3040C_1.gif: 4×0.24-160 Hz, rail-to-rail amplifiers, as in A3040BV2.
A3040BV2.ods: Bill of materials for A3040BV2 assembly.
Code: Logic chip firmware library.
CR1225: Diameter 12 mm battery electrical data sheet.
A79612: PZN-08-VV, 8-way connector, vertical, surface mount.
A79614: PZN-08-AA, 8-way connector, vertical, through-hole.
A78914: PZN-08-WC, 8-way connector, stranded 316SS leads.
A78967: PZN-08-WC, 8-way connector, solid silver leads.
A79617: 10-way hermaphroditic connector, vertical surface mount.
A78538: 9-way socket with guide post for HMT, vertical surface mount.
A78682: 9-way plug with guide post for EIB, vertical surface mount.
BC-2009: Battery retainer for CR1025, CR1220, and CR122, showing cuts we make before loading.
EIB-8: Eight-Pin Electrode Interface Board, by Neuralynx.

The A304001A is a rigid-flex circuit board that provides two rigid areas for transmitter components, one rigid area for programming and calibration, which we cut off once the transmitter is configured correctly, and a flexible antenna.


Figure: Rigid-Flex Printed Circuit Board for Prototype HMT, Top Side. Connector is on the bottom side.

The A304001A's battery holder supports the CR1220, CR1225, and CR1025 coin cells.

Modifications

[11-JAN-22] To make the A3040AV1 with the A304001A PCB:

  1. SCK Reroute: Remove R7, connect U3-7 to D0, a logic output on Bank 0. This output is accessible by wire modification, but provides only 1.8-V logic. The 1.8-V HI level proves sufficient to drive the ADC when powered by a 2.7-V coin cell.
  2. Antenna Detune: Remove L1 and replace with 1.0 nF. Remove C26.

[11-JAN-23] When updating A304001A to A304001B we should perform the following modifications.

  1. Replace existing J1 footprint for PZN-08-VV, which has only eight pads and the pads rows are farther apart.
  2. Remove J1 from solder paste stencil so it is flat gold.
  3. Add polarity to U5 silk screen and permanent footprint silk screen.
  4. Connect SCK through A4 to B5, an output in Bank 1 that provides 2.7-V logic output.
  5. Need a ground pad on the programming extension.
  6. Need a silk screen white area on programming extension.
  7. Add ground pad to top layer next to analog multiplexer.
  8. Add pull-down resistor on TCK on programming extension.

[04-APR-23] To make the A304001C out of the A304001B we need the following modifications.

  1. Connect the extension ground pad to 0V.
  2. Move cross-over of tracks in flex region to mid-point of flex region.

Development

[19-DEC-21] Sketch of prototype here.

[03-JAN-22] Finish first draft of A304001A rigid-flex circuit board, equipped with four amplifiers. We are using the ADG804 four-to-one analog multiplexer to select one of four signals for the ADC. We are using the LTC1865L in SOP-8 package because we want to conserve our stock of MSOP-10 packages for our smaller transmitters. We find that the op-amp and passives for each amplifier weigh 34 mg.

[06-JAN-22] We receive battery retainers and modify them by cutting off end flanges and solder to an existing 12.5-mm square transmitter board. The corners of the retainer protrude by 0.5 mm. We prefer to bend the retaining spring so that it pushes down on the center of the battery. After these modifications, the battery is secure in the retainer, with its negative tab pressed on the center of the circuit board, where we expect to have our gold-plated negative battery pad.


Figure: CR1225 Mounted in Battery Retainer.

[03-FEB-22] Change pinout of our Omnetics socket connector, J1, to match the manufacturer's A78538 drawing, modify all sketches, schematic, and printed circuit board files.

[16-FEB-22] We have A304001A rigid-flex boards in panels of five. Load Omnetics connector we burn off an EIB-8 onto the footprint of J1. Trim battery holder, bend down its spring, insert CR1225. Wrap in blue, self-adhesive silicone. Mass is 2.3 g.

[28-FEB-22] Kit for 15 of A30304A on 10-day turn is ready to ship.

[16-MAR-22] We have fifteen A3040AV1 assembled circuits. We have assigned an input-only pin on the logic chip to the SCK output. We remove R7 and use the D0 tuning bit for SCK, connect the R7 pad to U3-7 with a wire link. The D0 bit is U4-G1, on Bank 0 of the logic chip, which runs on 1.8 V, so the HI level will be only 1.8 V. We program as D3: all channels enabled, 512 SPS. Active current is 231 μA, less than the 270 μA we predict with our current calculation formula.


Figure: Top Side of A3040AV1. Red wire corrects design error.

We load a battery holder and insert CR1225 battery. Obtain robust reception in FE3A enclosure with one pick-up antenna in a few locations, no systematic test of reception. Right now, the antenna matching network is our first guess: 10 nH series inductor with 1 pF parallel capacitor.


Figure: Bottom Side of A3040AV1. Connector J1 not loaded.

Our nine-way Omnetics connectors with guide pin, A78538, are due to ship to us on 21-MAR-22. We touch the pads of the connector and see voltage induced in all four input channels on our Receiver Instrument display. The ADC appears to be working despite the fact that SCK is being driven by only a 1.8 V logic signal when the ADC expects 2.7-V logic.

[20-MAR-22] We clip the programming extension of our working, programmed A3040D3, No224. We wrap in silicone and trim. We are still missing J1. Mass is 2.51 g. Of this 1.10 g is the A3040AV1 assembly, 0.87 g is the CR1225 battery, and 0.54 g is the silicone wrap. When we load a CR1025 battery, the mass of the circuit without wrap is 1.76 g, the mass of the battery is 0.65 g. Apply wrap, total mass 2.30 g. Running off the CR1025, we record from all four channels and see VBAT = 3.0 V (fresh battery). We prepare two more AV1s and program as D1..D4, measuring current consumption, which we find to be linear with sample rate, 17 μA + 0.110 μ/SPS. Our A3040A02 firmware brings the transmit channel numbers and input channel numbers into agreement. If the base transmit channel number is b, the channel numbers for X1, X2, X3, and X4 inputs will be b, b + 1, b + 2, and b + 3.

[22-MAR-23] We solder silicone leads to our No224 AV1 and connect these to a 15-mVpp, 10-MΩ signal source so as to measure gain versus frequency. With 10-MΩ input resistance, we expect to see 7.5 mVpp at X1..X4. Assuming VB = 2.7 V and gain ×100 in the middle of the amplifier pass-band, we expect amplitude roughly 6.5 kcnt rms, with cut-off frequency 160 Hz, and this is indeed what we see.

[23-MAR-22] We experiment with our antenna matching network, L1 and C26 in schematic. We program an A3040AV1 to turn on its transmitter (U6) continuously. We connect an external tuning ramp 0-5 V to the TUNE input of U6. We place the antenna of the A3040AV1 near the coil antenna of an A3038DM-C detetector module in which we have replaced the 900-930 MHz bandpass filter with a wire link. We look at the logarithmic power measurement, P, on the detector module. The peak power received is a strong function of the location of the transmitter. The variations in power received are a weak function of location. We place the transmitter in a petri dish with the antenna sticking up in the air, about 10 cm from the detector antenna.


Figure: Three Tuning Networks. Blue: tuning ramp. Yellow: logarithmic power measurement, 500 mV/div = 16 dB/div. Cursors mark 900-930 MHz band. Left: L1 = 10 nH, C26 = 1.0 pF. Center: L1 = 10 nH, C26 = 2.0 pF. Right: L1 = 0 nH, C26 = 0 pF.

Our original tuning newtork was 10 nH with 1.0 pF. This network produces a 20 dB reduction in transmit power in the 902-930 MHz band when compared to no network at all (0 nH and 0 pF). When we increase the tuning capacitor to 2.0 pF, we see a slight increase in power emitted in our 900-930 MHz band compared to no network. We equip one A3040D3 with 10nH/2pF and another with 0nH/0pF. Both have had their programming extensions removed. We fold up and wrap in silicone. We place on a petri dish in FE2F enclosure and move around, watching reception from a single A3015C damped loop antenna connected to an Octal Data Receiver. We try different arrangements of the antenna and transmitters. In each arrangement, we have both transmitters oriented the same way and close to one another. With transmit antennas vertical, we obtain robust reception from both transmitters in ten arrangements. With antennas horizontal, we obtain robust reception from the 10nH/2pF network in 18 of 20 arrangements, and from the 0nH/0pF network in 10 of 20 arrangements. We damage No5. We load 10nH/2pF onto No224 and No1. We clip No1, fold both, and compare. We obtain robust reception in only half the arrangements, and equally for both. We try 100pF/10nH (100 pF in place of L1, 10 nH in place of C26) on No1. Reception is far inferior to No224 with 10nH/2pF. We try 100pF/0pF on No1. Reception from No1 100pF/0pF and No224 10nH/2pF robust with antennas vertical. With antennas horizontal, we see robust reception from 100pF/0pF in two thirds of arrangements, and robust reception from 10nH/2pF in only one third of locations. Load No224 with 1nF/0pF. Reception from No1 100pF/0pF and No224 1nF/0pF is equally robust with antenna horizontal.

[24-MAR-22] We receive 15 Omnetics connectors. Solder one to No1, takes about five minutes. Wrap with 40-mm tape and CR1225 battery, mass 2.6 g. The connector weighs 57 mg. We solder leads to an EIB-8 and plug our A3040D3 into the EIB-8. We connect the EIB-8 leads to a four-way SCT-BNC interface, and by this means we apply a frequency sweep to all four inputs and measure frequency response. At first, amplitude is low and varies from channel to channel. We heat up and blow dry, now obtain uniform response.

[28-MAR-22] We wrap No1 in Saran Wrap. The wrap does not stick to itself well enough to provide a lasting seal, but mass with three layers of Saran Wrap is 2.1 g. Some residual wrap material adheres to the tubes of socket J1.

[30-MAR-22] We have five A3040D3 ready to ship to UCL. Frequency response D1_1.gif. We ship with ten CR1225, ten CR1025 batteries, and one meter of silicone wrapping tape. Calibration constants below.


Figure: Calibration Constants and Active Current for Batch D1_1. Six A3040D3 HMTs.

During calibration, we measure transmit spectrum of several devices before and after we replace the A3040AV1's 10nH/1pF antenna matching network with 1nF/0pF. In all cases, we first perform the SCK/R7 modification. The transmit center frequency is stable to ±1 MHz, and power stable to ±3 dB as we change the matching network and re-measure the spectrum.

The prototype A3040A provides a nine-way, dual-row, 0.025" pitch, surface-mount socket with a single guide post for connection to the electrode interface board (EIB-8). We use custom Omnetics NSD-09-VV-G, which we order with part number A78538-001.


Figure: The NSD-09-VV-G Socket Mounted on A3040AV1. Note guide pin and nine socket tubes. Pin one is front-left.

We use the pin numbering given by the manufacturer. The figure below is an exerpt from the S3040A_1 schematic, showing pin numbering as seen looking down on a soldered socket. We add pin "G" for "guide post" so that we have a label for the tenth pad on the printed circuit board. This pad is not used by the connetor, because there is no connection pin below the guide post.


Figure: Pinout of NSD-09-VV-G Socket Manufacture's Pin Numbering. View as seen when holding the A3040A and looking down on the connector.

The NSD-09-VV-G socket mates with any NSD-09-G plug, provided the plug's guide post hole is in the G position. We designed the A3040A to mate with Neualynx's EIB-8 electrode interface board. In this figure, we see the top of the EIB-8's nine-way plug, with the guide post hole in position G on the top-right. The EIB-8 provides ten connection pads. With J1 on the A3040AV1 circuit loaded in its original (0°) orientation, the two GND pads on the EIB-8 will connect to VC, the reference voltage of the four amplifiers. The other pads connect as shown below. We also show how the pads on the EIB-8 will connect to the circuit if we reverse J1 on the A3040AV1 footprint. In the original orientation, the A3040AV1 antenna extends to the right of the figure. In the alternate (180°) orientation, the antenna extends to the left. The latter orientation turns out to be the one that we prefer, because the EIB-8 is designed to have the GND pads towards the nose of our host animal.


Figure: EIB-8 Connections for HMT Inputs (X1-X4) and Reference Potential (GND). Magenta: With J1 in the original (0°) orientation on the A3040AV1. Blue: With J1 in the alternate (180°) orientation.

Within the HMT circuit, we refer to the reference potential as "VC", the "common voltage". Once we connect this potential to an animal body, we say it is "grounded". Our assumption is the the GND pads on the EIB-8 will be used for a low-impedance connection to the animal body.

[08-APR-22] We wrap an HMT in plumber's teflon tape, and are well-satisfied with the result. Mass with CR1225 battery is only 2.2 g.

[11-APR-22] We receive this picture from ION of an A3040D3 wrapped in British clingfilm.

[05-MAY-22] We have four hours of recording from ION of No17 and No37 mounted on mice. In the first hour we nave No37 running. In the second hour we have No17 and No37 running.


Figure: Reception from Mounted HMTs in IVC Rack in London. Average is 95% from No37 and 89% from No17.

[06-MAY-22] We receive another six hours of recording from the IVC rack at ION. There are half a dozen A3028C transmitters running along with two A3040D3 HMTs. We measure reception from HMT channels No17 and No37, and compare to SCT channels 201, 203, 205, and 210.


Figure: Reception from Transmitters in IVC Rack in London. A3040D3 HMTs are 17 and 37. The others are A3028C SCTs.

Reception from all transmitters drops dramatically from 2.2-2.5 hr. We assume the enclosure door was opened for access. From time 2.5-6.0 hr, reception from each channel is as follows.

Channel   17   37   201  203  205  210
Average 95.2 97.6  99.4 99.1 99.2 97.0

In the period 0.0-2.2 hr, reception from the two HMTs is inferior to reception from the SCTs. After 2.5 hr, reception from HMTs is slightly inferior on average (97% compared to 99%). When we implant an SCT, the presence of water around the device slows down the electric waves propagating back and forth along the antenna. We find that antennas of thirty to forty millimeters are ideal for subcutaneous operation. The HMT antenna is operating in air, where the resonant length of a straight quarter-wave antenna is 80 mm. The HMT antenna is straight, but it is only 14 mm long. The antenna conductor is 80 mm of zig-zag track, but this zig-zag does not behave the same as a straight wire. It was our hope that the 14-mm zig-zag antenna would work fine without any matching network at its base. But it now appears that we are going to have to spend a few days figuring out what inductor and capacitor value we should load for L1 and C26 in order to get the antenna to resonate.

[17-MAY-22] Our collaborators request that we reduce the distance between the mouse's head and the HMT circuit. We cannot modify the existing Omnetics connector, but we order a selection of 0.4-mm and 0.35-mm pitch board-to-board connectors that will reduce the separation to less than 2 mm.


Figure: The HMT Prototype on a Mouse. Marked is the 8.5-mm dimension our collaborators wish to reduce.

[20-MAY-22] Two A3040D3 have failed when the socket on the HMT broke off at the solder joints. The connector on the EIB is glued in place. Our collaborators do the same for the connectors on their HMTs. They use UV curing epoxy.


Figure: Connector Glued to HMT to Hold It Firmly in Place.

[25-MAY-22] We try alternatives to the 9-Way Omnetics connector. We have a 6-Way 0.35-mm pitch 5861-series board-to-board plug and socket from Kyocera and a similar plug and socket from Molex in their Slimstack series. We glue these new connectors to circuit boards so that we can press them together and pull them apart.


Figure: Slimstack Board-to-Board Socket.

The separation of the circuit boards is less than a millimeter. It is difficult to get the plug and socket to mate when we cannot see either connector. We would need two alignment pins on the electrode interface board to align the HMT with the connector when mounting on the animal.


Figure: Connector Comparison. (Photo and measurements by Calvin Dahlberg.)

The connection force provided by these new connectors is of order 0.1 N compared to 3 N provided by the Omnetics connector. If we add threaded ends to our two alignment pins, these could pass through holes in both layers of the HMT circuit. But the pins would eventually encounter the battery an be unable to emerge from the HMT. If we separate the pins sufficiently to avoid the battery, the electrode interface board will have to be 14 mm long. Even if we figure out how to get the pins past the battery, we still have to figure out how and when to wrap the circuit in teflon.

[26-MAY-22] Our collaborators inform us that the orientation of J1 on the A3040A is incorrect for use with the EIB-8. The orientation we chose places the hole in the EIB-8 at the front end of the mouse's head, rather than at the back end. We now consider what will happen if we rotate J1 with respect to its original orientation. We refer to the EIB-8 manual for the conections between its pads and its Omnetics plug.


Figure: EIB-8 Pads, Top View. Black circle pin is guide pin aperture. Original (0°) orientation of J1 on A3040AV1 places the antenna on the right side of the figure. Alternate (180°) orientation places the antenna on the left side.

In the following table, we map out how the A3040AV1 circuit's four inputs and ground connections will be mapped to the EIB-8 if we rotate J1 on its footprint.

Connector
Orientation
Footprint
Pin
Connector
Pin
EIB-8
Pad
Electrical
Function
11GNDGND
22A1X2
33A2NC
44A3X4
55A4NC
66A8X3
77A7NC
88A6X1
99A5NC
GGGNDGND
180°19A5GND
180°28A6X2
180°37A7NC
180°46A8X4
180°5GGNDNC
180°64A3X3
180°73A2NC
180°82A1X1
180°91GNDNC
180°G5A4GND
Table: Electrical Function of EIB-8 Pads for Original (0°) and Alternate (180°) Orientation of J1 on the A3040AV1 Circuit.

If we rotate the connector on the A3040AV1, we will have the EIB-8 pads connected to the A3040AV1 circuit in the following manner: A5=GND, A1=X1, A3=X3, A6=X2, and A8=X4, see diagram.

[19-JUN-22] We complete a study of how we might fasten the HMT to an animal's head with magnets, while making the electrode connection with a smaller, lower-profile connector.


Figure: Magnetic Clamping Force, With Omnetics Connector Comparison (By Calvin Dahlberg).

Calvin makes the following comments, "smaller connector tends to rock back and forth. Aligning magnets works so-so. Smaller connector may be shorter than other components on the board. Magnets are hard to work with, although hot glue works okay. Two and three magnets on both sides would definitely go over 1g weight limit." The rocking of the two boards with respect to one another, while the plug and socket rotate, is of particular concern to us: we may see movement artifact introduced into our EEG. The additional weight is the greatest cost.

[11-JAN-23] We are working on a new electrode interface fixture (EIF). We abandon the NSD/NPD connector in favor of a new design by the same manufacturer, Omnetics Corporation. The PZN connector is a hermaphroditic, intrinsically-polarized connector. We issue a new version of the schematic, S3040B_1, in which we use the manufacturer's pin numbering for the PZN-08-VV on the HMT. In the EIF, we modify the pin numbering to match the HMT, and use the same connector but with through-hole pins, PZN-08-DD.

[12-JAN-23] Study D1.17 returned last summer from ION. Consumes 150 mA from external power supply. Replace MAX2623, U6, and now it works fine. We re-tune by loading 1 kΩ in parallel with R3. Current consumption 245 μA, we leave it running with a battery. We are applying modifications to six more circuits. We D41 in fine shape. We have a prototype D242 that may be fine. We have D33 for antenna tuning only. We have another board that is scrap for spare parts. Our objective is to prepare 4 for shipping with the PZN-08-VV connector.

[13-JAN-23] We prepare an EIF prototype with five 0.5-mm diameter leads soldered to its pins. The 0.5-mm diameter leads are far more flexible than our 0.7-mm leads, and the spring itself is smaller, so it takes up less space in a lap solder joint like those shown below. We experimented with sliding the 0.7-mm lead's spring over the connector pins and soldering, but we find we must twist the spring onto the pin, which can weaken the pin.


Figure: Electrode Interface Fixture. Left: Bottom, showing solder joints and epoxy reinforcement. Right: Top, showing connector pins and sockets.

We apply a little black DP270 encapsulation epoxy to the bottom of the connector to reinforce the bends in the leads.

[19-JAN-23] The black epoxy did not adhere well to the leads shown above, but instead collected elsewhere. We continue making prototypes, trying to get the pins and joints smaller and shorter. We assemble the following with five leads, then apply epoxy. With the bottom facing up, the black DP270 creeps off the solder joints and starts making its way down the outer walls of the connector. So we flip the connector right way up, supporting it over an aperture with a weight on its leads. The result, photographed the next day, shown on the right belwo.


Figure: Electrode Interface Fixture, Shorter Joints. Left: Before covering with epoxy. Right: After covering with epoxy.

There are several metal points and one solder surface where the epoxy covering is so thin as to be transparent. In our experience, coatings this thin break off or wear through quickly. We will try rotating the EIF while the epoxy is curing, as we do with SCTs, so as to neutralize gravity and leave the epoxy distribution to be dominated by surface tension.

[20-JAN-23] We solder two X-Electrodes to our latest prototype, producing the EIF8-XAAX shown below. We construct four clips with magnets that mount on our rotation


Figure: First EIF8-XAAX. The X-Electrodes are soldered to X1 and X4. We have bare wires for X2 and X3, and a bare wire for GND.

[30-JAN-23] We have 5 of A3040A3 with PZN-08-AA loaded. Current consumption is 235-246 μA.

[08-FEB-23] By loading 27 nH for L1 and 0.2 pF for C26 we find that antenna output power increases by a factor of ten. Full report to follow. Reception in air in Faraday canopy with four pick-up antennas is 98%. We receive PZN-08-DD, connector with wire contacts, shown below.


Figure: PZN-08-WD for Connection to A3040 Amplifiers.

Consulting our connector pinout, we construct the following table of colors, pin numbers, and functions.

PinColorFunction
1brownGND
2blackGND
3orangeX3
4redX2
5greennone
6yellownone
7purpleX1
8blueX4
Table: Color Codes, Pin Numbers, and Functions for Wired Connector.

We record frequency response of our five A3040D3, all twenty channels, see D1_17.gif. When one device fails, we examine the circuit and note a missing capacitor. We load the capacitor, wash, dry, and test again: it passes.

[13-FEB-23] We have 8 of EIF8-XAAX ready to ship. We have four A3040D3 ready to ship. A fifth shows low gain in one channel. While performing quality control we discover a problem with the digitization. When we apply a 15-mVpp sinusoid to X1, we see 400 μVpp crosstalk in X2. When we apply the sinusoid to X2 we see the crosstalk in X3, X3 talks to X4, and X4 talks to X1.


Figure: Feedthrough Crosstalk. Here we have 400 μVpp crosstalk on X3 when we apply 15 mVpp to X2 in D1.17 with firmware A02.

Examination of the firmware reveals that we are changing the analog multiplexer address lines (A1 and A0 in the schematic) on the rising edge of ACTIVE, which is only 10 μs before we initiate conversion in our ADC. In the four-channel A3047, we don't see this crosstalk. In that firmware, we are changing the address on the rising edge of End Clock (ECK), which is at least 70 μs before conversion. We correct our P3040 firmware, starting a new version P3040A03.abl, in which we change the address on the rising edge of ECK. The feedthrough crosstalk disappears.


Figure: Feedthrough Crosstalk Eliminated. Here we have 60 μV rms noise on X3 when we apply 15 mVpp to X2 in D1.37 with firmware A03.

[15-FEB-23] We increase the output power of the A3040D3 by a factor of twenty with a new antenna matching network. We try several inductors in place of L1 and vary the capacitor we load for C26 until we arrive at 27 nH and 0.2 pF as the most effective network.


Figure: Relative Antenna Power Output for Various Matching Networks. We use 0.01 pF as a stand-in for 0 pF in our log plot.

To measure relative power, we place the HMT, folded over with battery loaded, in the same orientation on a petri dish in an FE3A 20 cm from an A3015C loop antenna connected to an A3008 spectrometer. We used the peak power of the spectrum as our measurement. Peak power is 13 dB higher with the new network than with our 0 nH and 0 pF.

[24-FEB-23] We test JBWeld structural epoxy for circuit corrosion by applying it to a circuit board then poaching at 60°C for a week. We see no sign of corrosion. The glue is non-conducting and water-proof. Our DP270 epoxy has been crawling all over our PZN-08-DD connector but failing to secure the leads and pins. The JBWeld is more viscous, and we obtain the superb result shown below.


Figure: EIF8 Made with JBWeld Epoxy.

The epoxy has wicked into the lead coils and around the joints. But it has not progressed around the connector. To obtain the above result, we held the connector in our rotator while the epoxy cured. We obtain almost exactly the same result with DP460-NS (the no-sag version of DP460).

[27-FEB-23] We have poached DP460, clear DP270, JPWeld, and dental cement on four circuits for a week. We see no sign of adhesive-induced corrosion in any of the samples. We are satisfied that any of these may be used for our electrode interface fixtures. We complete layout A304001B with modificaionts listed. The board is made for the PZN-08-VV connector, corrects an error in the original layout, adds a 0V pad to the programming extension and a signal ground pad near the analog switch.

[08-MAR-23] We measure power emitted by the HMT antenna when it is immersed in water to within a millimeter of its base. The HMT body remains out of water, but folded over in its usual fasion. With the antenna in water, we see more power with an inductor from the antenna base to 0V. So we try various values of base inductor and series capacitance. We compare power emitted in air and water for each matching network.


Figure: Relative Antenna Power Output for Various Matching Networks, Water and Air. We use 100 nH as a stand-in for L1 open-circuit.

[14-APR-23] Settle on the A3040BV1 bill of materials: we will give all four amplifiers frequency response 0.3-80 Hz. We are shipping A304001BR1 circuit boards and kit to NPi Technologies today.

[02-MAY-23] We receive 45 of A3040BV1 assemblies. Update firmware and program for D2, four channels 256 SPS. Current consumption is 139 μA compared to our expected 148 μA. We load the PZN-08-VV connector, but we do not epoxy. We cut flanges off the battery clip and load. We connect the ground pad on the programming extension to P1-7, because the pad ground connection is missing on the A304001BR1. We apply 10-MΩ sweep to all inputs and see a nice 0.3-80 Hz response. We are now able to increase the center frequency of the transmission in steps of one 5-bit dac count. The A3040A allowed only steps of two because of the way we corrected a layout error.

[12-MAY-23] We are learning how best to load the PZN-08-AA onto our A3040BV1 assemblies. While programming and testing A3040B1 we are struck by the power of the emitted radio frequency signal, and the extent of our operating range. We forgot to remove the battery from one HMT, placed it in our library five meters away, and still obtained better than 80% reception on our desk.

[09-JUN-23] We have 34 of A3040D2 ready to ship. We lost one circuit because we glued the connector before performing QC1, and one contact was faulty. Other than that our yield was 100%.

[28-AUG-23] We have five of A78914. These are PZN-08-WC with teflon-insulated, stranded, stainless-steel wires loaded (wires are A-M Systems 793200). We hope to use these connectors as a starting point for the assembly of electrode interface baords.


Figure: PZN-08-WC with Stranded, Stainless Steel, Teflon-Insulated Wires.

[30-AUG-23] We receive quotation from Omnetics Corporation for PZN-08-WC with A-M Systems 78600 teflon-insulated solid silver wires loaded, part number A78967, $72 each, 12 weeks.

[31-AUG-23] At UCL our collaborators have deployed two HMTs with EIF8-XAAX head fixtures. They report, "The recordings from one surgery were exceptional, with a spontaneous seizure recorded in unprecedented detail. Changing batteries equally has thus far been successful and we have been able to record of over a month. We have however run into a problem with the Omnetics 8 pin Nano connector working itself loose due to animal activity. Three times the mouse has been able to completely remove the HMT in the days following connection, resulting in a minor mauling of the transmitter by the mouse. Thus far the transmitters have resisted destruction, but some damage to the antenna can be seen."

[08-SEP-23] Mouse removed HMT and chewed through antenna completely. Collaborators having some success with Kwik-Cast to hold the connectors together. They also note that it is the mouse pulling on the wrapper with its claws that allows the mouse to pull off the HMT. They have not been using plumbers tape, but rather some stronger tape. They try clingfilm and the mouse tears the clingfilm but fails to remove the HMT.

[29-SEP-23] Working on P3040B firmware. We try uniform sampling by adding a conversion at the end of each sample interval, in addition to the scattered conversion during transmission. Current consumption of the A3040D2 increases from 132 μA to 148 μA as the LT1865L makes the extra conversion. We don't think the benefit of uniform sampling in the HMT is worth the 10% decrease in battery life. In the A3049, which uses the ADS8860, the increase is negligible.

We prepare firmware for the A3040C2. We use a PZN-08-WD wired connector to deliver ground and signals to the device. Wires are black GND, purple X1, blue X4, orange X3, red X2. We program No69. We connect 20 mVpp to X1 and see 6.5 mV rms on W, 12 μV on X, and 8 μV on Y. The Z-channel is disabled. With 20 mVpp on X4 we see 10 μV on W, 6.5 mV rms on X, and 9 μV on Y. With 20 mVpp on X2 we see 8 μV rms on W, 6.5 mV on X, and 10 μV on Y. In this latter case, we look at the spectrum of W and Y and see sharp peak at 10 Hz, amplitude 10 μV. Crosstalk is roughly −56 dB. Current consumption 189 μA. We program No21 with the same and see 182 μA. Using 25 + 0.12 × SPS we get 179 μA.

[02-OCT-23] We adjust A3034C2 so it provides three channels at 512 SPS. Reprogram No21 and No69. Current increases from 182 μA to 196 μA and from 189 μA to 209 μA respectively. Operating life drops from 10.5 days to 9.6 days. Using 25 + 0.12 × SPS we get 209 μA.

[03-OCT-23] We prepare two A3040D3Z using two of the original A3040AV1 circuits with PZN-08-VV connectors loaded. We replace C6, C11, C16, and C21 with 50 kΩ and C7, C12, C17, C22 with 620 kΩ.

[10-OCT-23] Our A3040D3Z are noisy. We update P3040A with the code present in P3040B that reduces sampling noise. Gain of amplifiers is ×10, dynamic range 270 mV, noise 20 μV rms on all channels, spectrum below.


Figure: Noise Spectrum of A3040D3Z. Amplitude scale 8 μV/div. Frequency scale 25.6 Hz/div.

[14-DEC-23] We have three HMTs back from ION. On device 0x6170 one joint on the BC-2009 battery clip failed, see below. In future, we must make sure we solder both sides of the clip. In this case, the nickel plating on the clip peeled off.


Figure: Battery Clip Joint Failure.

Devices 0x74D0 and 0x552F draw over 100 mA. We replace U6, MAX2623 and current consumption returns to normal. Diagnosis: static damage to U6, which we have observed and even caused deliberately with a plasma ball in our A3048 and A3049 circuits. Solution: encapsulate U6 and antenna network in epoxy. We check gain versus frequency and find 0xFF2F and 0x6710 are perfect, but channels No70 and No72 on 0x74D0 are -6dB at all frequencies for 10-MΩ sweep, but correct gain for 100-kΩ sweep. We place in the oven to bake. After twenty minutes, remove and frequency response is correct for 10-MΩ sweep.