The S-Pixie uses uses an LM386 as its audio amplifier (TI datasheet, note that I could not determine the manufacturer of the chip included in my kits).
The design closely follows the design shown in Figure 12 of the datasheet. The 10 uF capacitor between pins 1 and 8 (shown as the two GAIN pins in the schematic above) sets the amplifier gain to 200 from the default LM386 gain value of 20.
While the LM386 can operate at Vcc, the 1k resistor (Vcc to pin 6) allows the transmitter circuit to disable the audio amplifier during transmission. Note that the output coupling/filter capacitors values are different from those specified in the datasheet. I’ll test the specified values to see if they make a noticeable difference in output quality.
Testing the Audio Amplifier
Audio Amplifier Noise – Minimum RF Signal
Similar to my tests of the S-Pixie mixer, I tested the audio amplifier by feeding an RF test signal into the S-Pixie antenna input port. I first looked at what input level was detectable on my oscilloscope. Without an RF input signal, I measured a noise level of about 140 mV P-P at the audio amplifier input. With a 50 mV, 7.027 MHz RF test signal, I was just able to detect a slight sinusoidal wobble above the noise level in the input signal to the amplifier. This was with my probe on 1X attenuation. At 10X, this signal was still indistinguishable from noise.
Much below this level the input signal was indistinguishable from noise on the oscilloscope. This signal however, was clearly audible on at the audio amplifier output on both headphones and an 8 ohm speaker as well on the oscilloscope (magenta trace below).
So what signal level was barely detectable? On the oscilloscope this occurred with an RF signal at about 10 mV.
Below this level the audio amplifier output was indistinguishable on the oscilloscope as with no RF signal at all. Through headphones though, the audio amplifier output was detectable with an RF signal at much lower levels, at times down to 500 uV, depending on the frequency of the RF signal.
I found that the minimum detectable RF signal level depended on frequency, because the noise level varied by frequency. Here is the frequency spectrum of the audio amplifier output with a 50 mV, 7.027 MHz RF signal.
This signal produced an output tone between 4 and 5 kHz, clearly visible above the noise level on the plot above. As I reduced the RF signal level though, the output faded into the noise. Here is the frequency spectrum with the RF signal at 750 uV.
The 4-5 kHz output tone is just above the noise level. At 500 uV, it’s at the noise level and just barely detectable on the headphones.
The audible noise level at lower frequencies is higher though, increasing the required RF signal level needed to obtain an audible tone from the audio amplifier. Here is the frequency spectrum of the audio amplifier output with a 5 mV, 7.024 MHz RF signal.
The output tone at just under 2 kHz is just at the noise level in this frequency range but was still detectable on the headphones. With a 900 uV RF signal at 7.024 MHz, I could just barely detect an output from the audio amplifier.
We expect the LM386 to have a gain of 200 with a bypass capacitor between pins 1 and 8. I measured the gain using a 200 mV, 7.027 MHz RF signal. At an RF signal much above this level, the output of the audio amplifier started clipping.
The RF signal produced a very noisy signal at the input to the audio amplifier. The output signal, however, was very clean.
While the amplifier output voltage of 4.28 V is easy to determine from the oscilloscope trace, I puzzled a bit for what input voltage to use for the gain calculation. The oscilloscope shows the P-P input voltage of 148 mV, but that includes a substantial amount of noise. The actual sinusoidal signal is about 20 mV. This gives a gain of 214. With this, I assume it is correct to ignore the noise in the input signal for the gain calculation. I admit to kind of backing into this. Is that cheating?
Audio Output Quality
I found the S-Pixie audio output to be somewhat harsh with no RF signal and with annoying 2 kHz tones. This wasn’t particularly important as long as a strong RF signal was present. In that case the headphone volume could be reduced to where the noise wasn’t audible, but the CW related tone was still clearly audible.
To see if the audio quality could be improved, I tested the recommended coupling/bypass capacitors from the datasheet, or more exactly, I used a 47 nF bypass capacitor and a 220 uF coupling capacitor. I couldn’t tell much difference between the two, either audibly or on the oscilloscope. I’m guessing the S-Pixie designers saw the same and decided to just use capacitor values already used in the kit to decrease build complexity/confusion.
Source of Audio Amplifier Noise – Replacing the Local Oscillator
I tested the alternate local oscillator design discussed in this post. I found that while this local oscillator redesign produces a better sinusoidal output it didn’t produce better audio output from the S-Pixie. This is because the mixer loads the local oscillator circuit, producing a signal very similar to the original design.
I found that coupling the local oscillator and mixer with a 100 pF capacitor instead of the 10 nF capacitor solved this problem and somewhat decreased audible noise at the audio output. It also made the noise less harsh.
Looking at the audio amplifier input signal, I found the noise level with the alternate design (with the 100 pF capacitor) to be about 40% of the original S-Pixie design.
With the 10 nF coupling capacitor, the noise level was a bit higher at 70 mV.
I also replaced the local oscillator with an externally generated 7.023 MHz signal. This has a more dramatic effect, both audibly on the headphones, on the frequency spectrum, and on the oscilloscope.
With a clean, external local oscillator signal, the noise at the output to the audio amplifier was less harsh, with no annoying 2 kHz tones. The noise level at the audio amplifier input was substantially less, about 80%.
The cleaner and lower noise level is clearly visible on the frequency spectrum.
The external local oscillator voltage level had an effect on the audio amplifier output level. A 5 volt LO signal gave only a 1.35 volt P-P audio amplifier output, 2 volt LO signal gave a 1.64 volt P-P output, and a 1.235 volt LO signal gave a 2.5 volt P-P output. LO signals below this rapidly decreased the audio amplifier signal.
Comparison with the PCB Build
I haven’t done a lot of comparisons with the PCB build yet, in part because of the difficulty in separating the various circuits on the PCB build. Now that the receiver circuit is completed on the breadboard build, I can do some comparisons.
Once again I used my Analog Discovery 2 to generate an RF signal to apply with a coax cable to the S-Pixie antenna jack. Right off I noticed that the zero beat on PCB build was 7.024 MHz, rather than 7.023 MHz on the breadboard build. Unfortunately I didn’t measure the resonant frequency of the PCB crystal before I installed it.
The quality of the audio amplifier output was similar to the breadboard build, a bit harsh and not particularly pleasant. However, with enough frequency separation, I found I could detect RF signals down to 50 uV on the PCB build (the limit of the AD2 waveform generator). With this ability, the quality of the noise is not material as you can simply reduce the volume until the noise disappears. The audible CW tone is still clear, at least to the level of my test equipment.
This difference may not be between the two build, but one based on my testing method. With the PCB build, the RF signal was connected directly to the S-Pixie with a coax cable. I used an oscilloscope probe with jumper wires for the breadboard build. That alone could account for the difference. I’ll have to do some more testing.
But enough for now. It’s time to move on to the transmitter.
Update: Testing the Breadboard Build with a BNC Connector
I added a BNC connector to the breadboard build for the RF signal, similar to the PCB build. The result was dramatic. I was able to clearly detect a CW tone at the audio amplifier output for RF signals down to 50 uV, the same as on my PCB build. Adding the connector had no effect on the noise at the input to the audio amplifier. That’s as expected. That noise is coming from the local oscillator.
So lesson learned again during this testing. If you want accurate tests and measurements, pay attention to how your test equipment is connected to your radio, especially when dealing with high frequencies.