Playing with the electrons most decidedly is like a British mystery novel or movie. Not unlike Murder on the Orient Express -- they all did it!
A couple of huge clues were totally missed by me in the resolution of a problem, and this also links back to the need to document everything. But 1st a bit of stage setting.
The above cool looking commercial SSB Transceiver was built by a company in Canada by the name of Spilsbury Tindal. It is all solid-state with 4 crystal channels operating in the 2 -10 MHz range.
Inside is a Collins 455kHz Mechanical Filter and a Motorola RF Transistor good for about 15 watts. This was a 1970's built radio. It was single conversion with a 455kHz IF, thus even making it to 10MHz involved a very large pair of cojones! [With such a low IF and single conversion as you go higher in frequencies images ARE a problem.]
The four crystal channels could be simplex (same frequency for Transmit and Receive) or entirely different. Internal to the set were a battery box and even a wire antenna. The primary market was the Canadian logging and mining industries. These often show up on eBay but command a pretty hefty price.
Of course, getting one would limit you to 80, 60 or 40M. But with a bit of Juliano Magic, the crystal approach could be changed over to the Arduino and Si-5351. Were you anchored down to keeping it stock then I would opt for four channels on 60M. Documentation and schematics exist on the Internet. So, the other option is to scratch build one. I did although not pure Spilsbury Tindal -- not even close!
This unit was built about two years ago but had some issues. The 1st notable point is the use of the ILI9341 2.2-inch display. This same code will work with the 4-inch display which would consume the whole front panel. If you are thinking Color TFT - get the ILI9341. A series of voltage dropping resistors takes care of the level shifting to the display.
Now to the mystery -- it just didn't sound proper on receive or transmit. The first test was to switch back and forth between LSB and USB. I added the two-sideband selection in the event I would make this into a 20/40M rig. A switch and a couple of added BPF and LPF filters and some relays would make that happen.
With any properly operating transceiver, switching sidebands should not make a huge (if any) difference in background noise. Such a state means you have the proper Carrier Oscillator frequencies, and these are placed on the filter slope at the right location. A non-symmetrical filter might result in frequencies that are not equally spaced from the filter center frequency i.e. not +/- 1500 Hz.
My replica uses a 9 MHz commercial filter and when you switched sidebands there was a huge difference. Initially, I had the LO at 23MHz and with a 9 MHz IF nets 14 MHz. LSB sounded stronger than USB and USB sounded pinched. I checked the code and the BFO frequencies were correct for that filter. So, my jump to the gun thought was a bad filter.
The next step was to install a new filter -- two wires and two nuts. Same problem. Then I tried a test where the LO was now at 5 MHz and thus what was USB was now LSB and the reverse. A most interesting outcome -- USB was now strong and LSB sounded pinched.
Then a small light in the tunnel led me to another possibility and one often is the last thought or never looked at all. I cruised up to 15.0 MHz to listen to WWV. I was receiving WWV at 14.998.8 MHz. The frequency was off by 1.2 kHz. The Carrier Boards for the Si-5351 have either a 25MHz or 27MHz reference crystal that must be calibrated to get the proper frequencies generated. You dial in 15MHz, and you receive (or transmit) on 15MHz when all is calibrated.
Failure to calibrate the Si5351 with a proper correction factor for the on-board crystal means all three clocks will be generating improper frequencies. The code may say one thing but what results will actually be different without a proper correction factor.
There are many references on the internet as to how to properly calibrate including a software routine. In thinking back some two years ago, I seem to remember I had smoked the Si5351 board which had been calibrated but the replacement was not.
My calibration process is to use my SDR radios where I sample CLK0 and adjust the correction factor in the sketch so what frequency that is being called up matches what I see on the SDR -- that can get me to about 1Hz. Sometimes on purpose I leave it at 20Hz off -- just so the FLEX guys can tell me I am 20Hz low. A few trial runs will get you the right calibration.
So, a calibration of the replaced Si5351 resulted in that switching sidebands has no change in background noise and when you tune 15MHz WWV --the rig reads 15MHz. Mystery solved with a culprit not 1st on the list to check.
TYGNYB!
73's
Pete N6QW
PS. When I searched the Internet for a calibration routine, I found a website that had the process. When I went to copy the link --it was on one of my own websites. Scary is a good word.