Figure: Previous (left) and Current (right) RF Amplifier Layouts. The left one is stable, the right one oscillates.
The Demodulating Receiver (A3017) contains a 902−930 MHz 30-dB amplifier, a downshifer, a 60-dB 50-MHz limiting IF amplifier, a frequency discriminator, and an amplitude demodulator. The A301701A circuit board also includes, just for good measure, more copies of the A3016SO SAW Oscillator and the A3008C Spectrometer.
If you can't understand our terminology (RF, IF, LO, dB, dBm, etc), please consult our Terminology page.
S3017_1: RF Amplifier, Downshifter, IF Amplifier, and Discriminator.
S3017_2: Demodulator and Power Supplies.
Below you see the A3017A RF Amplifier layout next to its predecessor, the A3016DR RF Amplifier. The former amplifier worked well: 27-dB of stable gain from the RF input to the IF output after the mixer and IF low-pass filter. Consult the A3016DR manual for measurements of the RF passband.
The new, unmodified amplifier, as shown below, oscillates.
It's not the missing capacitors around L201 that stopped oscillation in the previous amplifier. As we describe here, we replaced these capacitors with solder lumps in the previous amplifier, and it was stable.
We applied copper tape, cut tracks, changed resistors, and soldered wire links. So far as we can tell, the problem is the loop closed by our +15 V power tracks. In the previous layout, +15 V came up through the ground plane in two separate places for the two amplifier stages. In this layout, +15 V comes up through the same ground plane penetration. Our final solution is in the figure below, and we find it to be stable regardless of where we touch it with our fingers.
The printed circuit board traces that connect the amplifiers are 30 mils wide (0.75 mm). The thickness of the circuit board is 62 mils (1.6 mm). The circuit board contains four layers of copper traces. The top layer is the one we see in the figure, with the 30-mil traces. The second layer is a continuous plain connected to 0 V, which we call the ground plain. A 30-mil trace 20 mils above a ground plane makes a 50-Ω transmission line. We want the traces to look like 50-Ω transmission lines so that they do not reflect RF power (see Terminology). But we subsequently discovered that the ground plain was only 10 mils below the top layer. The correct width for a 50-Ω trace 10 mils above a ground plain is 15 mils. According to this calculator, our 30-mil traces 10 mils above a ground plane, separated by FR-4 fiberglass, act like 30-Ω transmission lines. These 30-Ω traces will reflect 10% of the power delivered by a 50-Ω source. Another 10% will reflect back into the 30-Ω trace where it ends in a 50-Ω load. So each 30-Ω traces introduces a loss of roughly 1 dB. This 1 dB is small compared to our gain of 20 dB at each stage. Although our traces are twice as thick as they should be, we don't expect any dramatic drop in performance as a result.
We connected the A3017A RF input to the output of one of our Modulating Transmitters (A3014MT). The A3014MT was tuned to roughly 910 MHz. Its output is −4 dBm, and we passed it through 36 dB of attenuition, so the power deliverd to the RF input was −40 dBm. For LO we used our 864-MHz A3014SO, with its +9 dBm output. The A3017A's IF output was −14 dBm, so its RF to IF gain is 26 dB. According the manufacturer's data sheets, we can expect 20 dB of gain from each amplifier, 3 dB of loss in the SAW filter, 7 dB loss in the mixer, and 1 dB loss in the IF low-pass filter. Our expected gain is 29 dB.
We replaced the A3017A with our A3016DR. Its IF output with the same input was −12 dBm, for an RF to IF gain of 28 dB, only 1 dB lower than our expected gain. We switched the amplifiers, and used high-frequency capacitors insted of general purpose, but something about the A3017A layout causes a 2-dB loss of gain with respect to the A3016DR.
[15-OCT-08] We built two more A3017s with serial numbers P0194 and P0193. We found that U201 and U202 were failing in one of the new receivers after running for a few days. We changed R203 and R206 to 150 Ω. Now the amplifier quiescent current is 35 mA instead of 40 mA. We also found that water stuck under the SAW filter could reduce the gain by 6 dB. With the new resistors we obtain a −7 dBm output for a −36 dBm input (−4 dBm from the A2014MT, −30 dB attenuator, and −2 dB in a 36" cable). The total gain of the RF amplifier is now 29 dB. The IF power of receiver Q0127 with the same input was −10 dBm. We replaced U201, U202, R203, and R206 in receiver Q0127 and cleaned the circuit board. The IF power rose to −7 dBm. Receiver Q0127 failed in London: after running for a while, it would stop receiving. We suspect that long-term damage to U201 and U202 caused by the 40-mA quiescent current are to blame. Receiver Q0126, which has been running for two years in Boston, now shows a gain of +29 dB, without the modifications. We changed R203 and R204 and the gain remained unchanged.
With −4 dBm on the RF input, the IF output of the A3017A saturates at −2 dBm, which suggests U202 is producing around +10 dBm.
The figure below shows the IF output with RF input frequency for a −34 dB supplied by one of our Modulating Transmitters (A3014MT). The pass-band is defined by both the SAW filter (L201, a B3588) and the IF low-pass filter (L203, an SCLF-65) acting together. For a picture of the frequency response without the SAW filter, see here. (The linked image shows the response of the A3016DR without the SAW filter, and the A3016DR is almost identical to the A3017A). The cut-off at the low end of the passband is due to the SAW filter. The variations in response within the passband are also due to the SAW filter.
If we define the signal pass-band as lying betwen two −3 dB frequencies, then we see that the A3017A pass-band extends from 900 MHz to 930 MHz, which encloses the 902−928 MHz ISM band very nicely.
But we can also define the interference acceptance-band as lying between two −20 dB power frequencies. Mobile phone interference in the UK, for example, is ten times more powerful than our transmitter signals in the 930−950 MHz band (see here). If our rejection is only 20 dB in this band, our transmitter power will be only 10 dB greater than the interference, even with the antennas oriented favorably. The −20 dB pass-band of our RF amplifier is not easy to see in the figure above, because −20 dB is a 90% drop in amplitude. But we can measure the −20 dB pass-band with the help of different attenuators, and we find it to be 890−940 MHz. Without the IF low-pass filter, the −20 dB pass-band extends up to 950 MHz, and perhaps even as hight as 960 MHz.
In other words: the IF low-pass filter is essential for removing 940−950 MHz interference from mobile phones in the UK. Our earlier Demodulating Receiver (A3005C) did not have this same low-pass filter, and was therefore more vulnerable to interference in the UK, while performing well in the US.
The A3017A schematic specifies the B3588 SAW filter. Its pass-band is nominally 915±13 MHz. We tried the AFS915S3 instead. Its pass-band is nominally 915±4 MHz. Below are the RF frequencies responses for both filters, obtained in the same way as we did here, and using the same apparatus we describe below.
The AFS915S3 pass band is around 30 MHz, far wider than we suspected after reading the data sheet. The insertion loss of the AFS915S3 appears to be at least 3 dB higher than that of the B3588.
We added the three OPA699 limiting op-amps that make up the A3017A's IF amplifier. Together, these amplifiers provide 60-dB of gain, with amplitude limited to ±700 mV. The amplifiers worked well. With no input, the IFL output is a near-random saturated signal with some steady frequency content. When we applied −64 dBm, or 400 pW, of steady 910-MHz to our RF input, the IFL signal you can see in the schematic saturated cleanly. Thermal noise in our 40-MHz SAW bandwidth is around 1 pW. The ERA-3SM noise figure is just under 3 dB. So our effective RF input noise is 2 pW (see Noise and Interference). When we applied −82 dBm, or 7 pW, the IFL output was dominated by our 46-MHz IF signal, but noise was clearly visible.
The noise in our receiver is the thermal and amplifier noise at the antenna input. The interference is RF power in the receiver's pass-band arriving at the antenna from transmitters other than those the receiver is intended to receive.
We concern ourselves with noise and interference only at the antenna input. At all other points in the receiver, our signal is at least ten times more powerful. Noise and interference at the antenna input has at least ten times the effect as it does anywhere else.
The antenna input connects through a decoupling capacitor to the input of amplifier U201, which is a ERA-3SM. The capacitor generates no noise. But the 50-Ω input impedance of the amplifier generates thermal noise. Thermal noise is electronic noise that arises from the random, thermal movement of electrons. A circuit with bandwidth B suffers from noise power 4kTB, where k is the Boltzmann Constant with value 1.38×10−23 J/K, and T is absolute temperature. Our pass-band is roughly 900-930 MHz, which is a bandwidth of 30 MHz. This gives us a thermal noise power of around 0.5 pW at the input of our RF amplifier. The ERA-3SM noise figure is 2.6 dB, which means that the effective noise power at its input is 2.6 dB higher (twice as great) as the thermal noise. Our effective RF input noise should be around 1 pW, or −90 dBm.
We disconnected the antenna and terminated the RF input with 50 Ω. We measured the rms amplitude of the signal at IFL by eye on the oscilloscope screen, and decided it was close to 200 mV rms. This 200 mV rms is well within IFL's ±700 mV saturation. The gain of the IF amplifier is 113 ≈ 64 dB. Our RF gain we measured to be 26 dB. The total gain from the RF input to IFL is 90 dB. With −90 dBm of noise at the RF input, we expect IFL to be 200 mV rms, which is what we observe.
We connected the antenna, and found that IFL saturates. The output of U204 is 200 mV rms. The gain from the RF input to the output of U204 is 68 dB. The interference power at our RF input must be around −68 dBm.
Our antenna noise power is −90 dBm, just as we expect. Our laboratory interference is −68 dBm, which is 22 dB more powerful than the noise. Our sensitivity to transmissions is limited by interference. This means that if we put our antenna in series with a 12-dB attenuator, we should see very little loss in performance, because the attenuator attenuates both the interference and our signal, leaving the interference still 12 dB more powerful than the noise. Certainly, a cable with attenuition 2 dB, like a 240-cm RG-58C/U cable we use with the A3015A will not degrade our sensitivity to transmissions.
We tested the A3017A in our OSI office, which is on the second floor of a large brick building on Main Street, Waltham. There are seven or eight 2.4 GHz wireless networks overlapping the office, and offices upstairs and on either side. Downstairs is a grocer. We saw no bad messages from interference. The amplitude at the output of U204 was a repeating 100-mV rms 50 MHz waveform, with bursts of 150 mV rms amplitude several times a second. If we take the bursts as our interference power, we arrive at −70 dBm.
We added the remaining parts to our A3017A, which completed the Discriminator and Demodulator. We applied a sweeping TUNE input to our Modulating Transmitter (A3014MT), and looked at the demodulator output versus frequency for various values of attenuition in the RF supply cable. We performed our measurements with the circuits on a wooden table, with no enclosure, as we detail below, we obtained much bette results when we eclosed them in a metal box
We touched the SAW filter in the RF amplifier, and the sweep was effective down to −72 dBm. We wrapped foil around the entire A3017A, and the sweep was again effective down to −72 dBm. When we combined the A3017 with other circuits in a metal enclosure to make the Data Receiver (A3018), we obtained much better performance. When we put the lid on with −76 dBm, the trace looked like the −64dBm trace above. There was no trace of the sweep when we disconnected the cable carrying the sweep RF power.
In theory, our sweep input is competing with the thermal noise at our antenna input. Our noise power is −90 dBm (see above). Our sweep is effective down to −76 dBm. This suggests that our sweep fails when it is less than 14 dB above the noise power. Compare this to our observed failure of message reception when the message power is less than 12 dB above the interference power.
We disconnected our source of RF power and inserted an 80-mm quarter-wave antenna into its RF output socket. We connected an A3015A loop antenna to the RF input of our A3017A via a 2.4-m (96") RG58 coaxial cable. We obtained the following trace.
According to our antenna measurements, we expect a minimum 32 dB loss as our RF signal travels through a quarter-wave antenna, across 1 m, to a loop antenna and through a 240-cm cable. We expect a maximum loss 17 dB greater, or 49 dB. Across 2 m, we expect a mininum loss of 38 dB, and a maximum loss of 55 dB. Given that our RF source was −4 dBm, we expect our RF power at the antenna input to be between −59 dBm and −42 dBm. But our interference power, after passing through the 1-dB loss of the antenna cable, is −69 dBm. Our signal power is at most 27 dB greater than the interference, at least 10 dB greater, and almost everywhere it is 12 dB greater. This is what we observe: when we hold a transmitter at 2 m, we get reliable reception in almost every orientatioin.
With both antennas in favorable directions at 2 m, we should get 27 dB more signal than interference. Ignoring reflections, at 6 m we would get 12 dBm more than interference. So we expect our transmitters to work up to 6 m in the absense of reflections. In fact, we find that they work at up to 14 m in our lab.
When we interposed our body between the two quarter-wave and loop antennas, we observed occasional sharp holes in the sweep respons. These are reception dead-spots, caused by destructive interference of the RF signal at the receiving antenna (see here and here). With a quarter-wave and dipole antenna, we were able to orient the two at range 2 m to obtain a reception dead spot without the use of our body. This allowed us to take a photograph of the dead spot as it appears in the response of antenna-to-antenna transmission power versus frequency. You can see clearly the hole caused by destructive interference at the receiving antenna.
To record messages from a transmitter, we must to connect the A3017 to a Data Recorder (A3007). We present the results of our work with these two circuits combined in our Subcutaneous Transmitter report.
