So I received my v0.11 boards back from the fab house and soldered one up last night to begin testing. I managed to squash a few bugs in the initial design, as well as identify a couple of new ones (which were relatively easy to bodge for now). Luckily the only minor issues I found are really more cosmetic than functional – and they led to design considerations for the next version.
The assembly was quite straightforward, however, some of the surface mount stuff can be a challenge for some builders. The keys to soldering SMT components include:
- cleaning the board prior to soldering (isopropyl alcohol)
- use plenty of high quality flux
- a small beveled soldering iron tip works best
- pick a small diameter electronics solder (with flux) – I prefer Sn60/Pb40 (leaded)
- use a small solder “tack” for opposing corners to keep the component in place
- whenever possible, use magnification as you may find it helps steady your hand
Perhaps the most tricky component to solder up on this board is the FT232RL USB to serial chip since it is in a 28-SSOP (small shrink outline package) and the pins can appear to be quite small. Using a minimum of solder, it is actually quite straightforward to solder up – you just have to be patient for the fist little while as you master SMT soldering techniques.
Once everything was soldered up in stages (testing each stage as it was built), I decided to test the fidelity of the audio signal being passed by the isolating transformers.
To accomplish this task, I fired up my Rigol DG1022 function generator and dialed in a 1kHz tone at 5Vpp.
Recall that the audio in a single-side band transmission is limited to between 300Hz and 3kHz so my choice of 1kHz falls well within the audio pass-range.
I then probed the output of the transformer using a 10x probe as to avoid “loading down” the circuit under test. I was rather shocked and dismayed to see the result.
The waveform being displayed showed a very peculiar characteristic which I was not expecting to see!
Notice the gradual (truncated) rising edge of the waveform and the rather sudden drop-off on the falling edge? This may give a clue as to what is happening to the core of the transformer at 5Vpp input.
I was rather stumped about this phenomenon until I realized today that I must be driving the transformer much harder than its design parameters. When I back off on the input voltage, I observe the following waveform output.
The signal comes out at ~39mVpp which is a tad too low, risking too much noise injection from nearby incidental radiators!
Out of interest’s sake, I bodge wired two 47µF electrolytic capacitors in place of the right-hand isolation transformer and probed the output once again.
The output waveform which was passed by the capacitors was by far better looking! There was no distortion noted and the output voltage was only 0.05mV less than the input!
I tried one of my other transformers and realized results very similar to those of the capacitors. Despite being orders of magnitude cheaper than transformers, the capacitors lack the break-down voltage margin that the transformers provide. It is extremely tempting to re-tool the board to use caps however – given the cost savings and all ..
I’d love to hear your input: which form of isolation would you propose? Fire off a comment and I’ll be happy to consider all opinions!
I think an isolation transformer would be the safest bet if you don’t trust your radio’s isolation capability. Otherwise, decoupling capacitors would be fine I guess. What about using a 2:1 transformer instead?