Calibrating output power, TX audio input, and relationship of CW output power to SSB output power
Rob French (KC4UPR)
So I'm having great fun and success with a little Teensy 3.2-based I/O module that I've added to my uBITX, including a Teensy Audio Adaptor, which is allowing me to do a bunch of different things (transmit and receive DSP, integrated USB audio input/output, various other I/O control). Another thing I've added is a G4HUP IF buffer at the 45 MHz IF, which outputs to an RTL-SDR dongle... I've been using that as a panadaptor. As a side benefit, although the buffer switches off in transmit, I still see the rig's TX spectrum at the 45 MHz IF when I transmit... which leads to the following.
I had to do some tweaking of the output levels out of the Teensy Audio Adaptor, as I noticed that I was easily exceeding my expected output power of 8 W on 40 meters (that's what I get in CW), and hitting 14 watts. Looked great on the power meter, but I could see on the 45 MHz IF spectrum that I was getting all sorts of splatter. Initially, I was thinking I would just turn down the audio level from the Teensy until I didn't see the splatter in the 45 MHz IF spectrum. Easy enough, but I noticed that even after I cleaned up the spectrum (for example, turned down the level so that I just saw one clean spike in WSJT-X tune mode), I was still getting nearly 14 watts.
So that leads to my question, what should I expect in terms of output power relationship between CW and SSB? I know for sure that I need to turn down the audio level to get rid of the splatter... but what if that power level is still higher than what I get in CW? Is that ok? (I realize the ultimate answer is looking for splatter and harmonics at the final output.)
I did implement in the Teensy software a two-tone 700/1900 Hz signal generator, so it's very easy for my to push that into my transmit audio, if that will be helpful for calibration purposes.
One thing I added in the Teensywas a two tone test generator.
There are others on the thread that are more experienced in RF engineering than I, though I will offer what I think.
Since base signals for CW and SSB differ in how they are generated, as you are surmising, there would be differences in characteristics in how they are applied to the final PA stages. There can be differences in levels that would explain what you are seeing. There is also a known dip in RF output for 40 meters as compared to 80 and 20 meters that has not been explained, and I have been searching for the root cause. You may have uncovered a direction for me to look into that anomaly.
The only way to tell if the output of the device meets requirements is to measure the actual output. How you do that will depend on the test equipment that you have at your disposal. A spectrum analyzer would be the bast, in my opinion, though it can be done as hams have for many years through other means.
Linearity can be tested with a two-tone signal and an oscilloscope.
Harmonics could be tested with another receiver, though I do not know how to measure that to ensure that the power levels are below the required -40db below the fundamental (I have read differing values for the reduction requirement, and it may be -43db). I use an SDR Play RSP1A (approx. $120 delivered in the US) with good results. There is similar software for the RTL-SDR dongle, though at HF frequencies it seems to be limited. The challenge that I found was to ensure that the level measured as the base signal was accurately related to the harmonic measurements. When I tried the RTL-SDR the software could not measure the full range in one scan, you had to measure at the fundamental, then change the RTL-SDR scan frequency to measure the harmonics. There is also a scaling issue with the RTL-SDR. It has only 8 bits of resolution, the RSP1A has 14. There are ways around these limitations that include switchable DB attenuators that I have not tried.
The last measurement that I have done is look for spurs outside of the base signal band. Here again, it can be done with a second receiver. The measurement accuracy comes into play with that approach. The RSP1A fills that measurement need for me.
For all of the above testing, there is the need for a good "tap" that can be used to measure the output signal. I built this one:
I would also recommend a step attenuator as well. This is needed to adjust the signal level out of the tap, as well as to verify that you are measuring the true signal, and not one of the internal noise products of the measurement device.
All of the above is just background for the main point that I made earlier in the message: the only way to be sure is to measure.
I hope this helps, and any corrections or additions are welcome. I am always looking to learn new things.