Homebrew SBITX - Tx Modules PA, LPF and Mic

Much progress since the last post. After resolving my ground-bounce and hallucinations with wisdom, I moved on to the transmitter which consisted of a 5 watt power amplifier, straight from the UBITX 40 module.  Thanks to Bill, N2CQR for the recommendation and his excellent Manhattan layout.

The three stage PA starts with a 2n3904 pre-driver followed by a 2N2219A driver with a heatsink cap and the fnal is an RD0HHF1 biased at around 6.5V - which puts it squarely in class A.  Bill's build and other bias more for class AB which I may experiment with later.

Next I needed a microphone and microphone amplifier.  I built the microphone amplifier straight from the SBITX schematic and built a homebrew electret mic.

The homebrew electret capsule in a 3D printed case  The PTT is a SPDT microswitch.  When I first hooked it up the mic did not seem very sensitive - almost had to shout to see output on the scope. A little googling and noodling and I figured it out.   In brief - when I finalized the mic amp build I mistakenly powered it from the 5V rail instead of the 12V rail, that accounted for most of the low output. By then I had read through the long groups.io thread from last year on electret bias and the recent thread on SSB audio quality and had ordered a ag of the -25dB capsules that Gordon recommended.  They accept VCC of 2-10 V through a 2.2K resistor.  So replaced the R21 10K biase resistor with 2.2K to the 5V rail.  With that setup the mic is working great.

With the mic amp and PA done I began work on the LPF module. The module consists of 4 filters covering the 80 meter to 10 meter bands.  In the original SBITX it is a diode switched module. I spent some time understanding the diode switching but decided instead to use relays - it takes away a lot of the complexity of the module.

Next lesson learned: when I first built the LPF board I had the relays only on the input and had tied all of the outputs together.  This created all kinds of problems with the filters interferring with each other even with their input switched out of the circuit.  Solution and best practice is to switch both the inputs and outputs and ground the unselected filters.  Problem solved.  But not out of the woods - still too much loss in the filter module.  I put that asisde for now and moved on to fixing the frequency alignment. 

I had previously tried to align the frequency (so that the displayed frequency matches the actual Tx or Rx freqency without success.  The alignment is done by creating a compensation offset for the 25MHz clock of the SI5351.   Its usually straight forward. Inject a known 10MHz signal into the receiver, zero beat and note the frequency display.  The delta between 10MHz and the frequency display can be used to derive the actual crystal value.  Pop that into the code and the the transceiver should be aligned.  The problem was it was aligned at 10MHz but off at 5 and 20.   After much testing and head scratching and with an assist from Farhan himself we discovered the problem was that the MIKROE WM8731 prototype board uses a different clock rate than the Linux driver in the Raspberry Pi.  In brief, the driver expects the codec to use the standard USB clock rate of 12MHz but the MIKROE board uses the more standard audio processing frequency of 12.288MHz.   The fix is to replace the 12.288MHz crystal on the MIKROE board wiht a 12MHz crystal.  Like so:

The surface mount crystal did not come off the board cleanly - in fact it lifed one of the tracks clean off the board.  So I put in a field expedient patch - the red enamled wired to replace the damaged trace.  To my delight it worked straight away.  I used blue painters tape to mask off the adjacent pins and that made the sodering to the chip much easier.

The results is very satisfying -  here is the first loggable QSO I had after completing the rig:

I'm very pleased with the way the rig is working.  I have a few kinks to iron out but we are appoaching the finish line.

73 from Geat Falls

Dean, KK4DAS

Homebrew SBITX Receiver - Ground Bounce, Hallucinations and Wisdom

 

After a long hiatus I am back at the blog.  I have a number of projects that I have been working on that I will share going forward, but today I want to talk about my latest project - a homebrew version of the SBITX transceiver designed by Ashhar Farhan.  The SBITX is a hybrid analogue superhet transceiver / software defined radio.   The analogue portion of my build is based on the Furlough 40 / SimpleSSB that I built in 2020.  The SDR software runs on a Raspberry Pi 3 or 4 with software written by Farhan.    I'll explain the title of the post before we are done today but first lets take a look at the KK4DAS SBITX.

Here is a demonstration made shortly after I completed the receiver:

For a quick overview of how it works, lets look at the block diagram.

Beginning with the antenna on the upper right let's follow the received signal path.  First we pass through a single 30 MHz low pass filter which passes the entire HF band.  We amplify the incoming signal with a broad band RF amplifier and then pass it through and ADE-1 mixer to mix the signal up the the 40MHz IF.  The LO and BFO clocks are provided by an SI-5351 PLL controlled by the Raspberry Pi.  The homebrew 40MHz crystal ladder filter is 25KHz wide which controls how much of the spectrum you can see on the waterfall display at any one time. The bidirectional IF board I am using is the first board I built for the SimpleSSB at the beginning of 2020.  I have replaced the 9MHz commercial filter with my homebrew 40MHz filter.  The second mixer then drops the signal to a 24Khz IF which is well within the range of the ADC in the codec board. From the second mixer we go through a low noise amplifier to boost the signal and pass it in to  the left line-input channel of the codec. The 24KHz signal is digitized in the codec and passed on to the Raspberry Pi where further signal conditioning and filtering occurs, the waterfall display is generated and the digital audio is extracted.  The digital audio is sent back to the codec where the digital to analog conversion occurs and the analog signal is sent out the left line output to headphones or to an amplified speaker.   The SBITX software also supports FT8, RTTY and CW decoding natively - no additional software or computer is needed.  For a detailed description of the SBITX you should read Farhan's SBITX description linked above.  Transmit will work much the same but in reverse.  I'll cover that when I get the transmitter implemented.

I'm very happy with how the receiver is performing. Its fun to listen to and sounds great.  But getting to this point has not been without a few stumbles and sidetracks.  I was honored to be included as a guest on the SolderSmoke Podcast Episode #250 with N2CQR, Bill Meara and N6QW, Pete Juliano where I shared my tales of woe - a few of which I will describe here in more detail and a others which I will save for another day. 

Ground Bounce - shortly after completing the receiver I made the unsettling discovery that signals that were being transmitted on 20 meters were being received on 20 meters but also at exactly half the frequency on 40 meters. This was not good - it seemed that it had to be strange mixing products in the first mixer, but I had tested the entire IF before hooking it up to the digital board - and this very same IF board was pulled from a working receiver. I looked at the output of both mixers and I couldn't see how the the 20 meter signal was leaking in on 40.  After thinking about it for a bit I decided to look at the SI5351 outputs on my TinySA Ultra spectrum analyzer and instead of seeing one clean signal on each of the LO and BFO clocks I saw both signals on both clocks.  This was clearly the source of my problem.  Skipping over a day or two of troubleshooting I sent a note to Farhan and he immediately identified the problem.  It was "ground bounce.  Apparently if the clock outputs are not properly grounded it causes current to rise internal the SI-5351 and signals to bleed between the clocks.   In following a separate piece of advice from Farhan on buildiing the digital board I had very carefully insured that there was one and only one ground connection in the digital board and that was directly back to the main DC input.  I had installed the SI-5351 directly onto the digital board and it shared that common ground.  But that meant that I couldn't also ground both ends of the coax shield between the SI-5351 and the mixers.  That was the cause of the ground bounce.  The solution was to remove the SI-5351 from the digital circuit and put it on the analog circuit - with the only connection between the SI-5351 and the Raspberry Pi were the two I2C control lines.  And also to ground the coax connecting the SI-5351 and the mixers at both ends.  That fixed it - the ground is no longer bouncing!

Hallucination - after curing the ground bounce I spent an evening listening to the rig enjoying the glow you get after fixing a thorny problem.  But my enjoyment was short-lived.  I noticed that from time-to-time that the waterfall display would go a little crazy,.  It appeared as if the the receive signal was being duplicated all up and down the band somewhere internal to the SBITX.   It looked like this:

The signal at the center is the received signal - all of the mirror images are false.  Those are the hallucinations.  Farhan identified that fairly quickly and let me know about a software fix in the SBITX 3.2  which led me to:

Wisdom - I don’t have a complete understanding, but the hallucinations are artifacts created during the Fast Fourier Transform of the received signal under certain circumstances.  The SBITX uses the open-source FFTW (Fastest Fourier Transform in the West) library.  There is an extension to the FFTW library called FFTW-Wisdom that is used tune the FFT algorithm the first time it is used.   The tuning  parameters are saved in what is known as an FFT “Wisdom” file .  The Wisdom file, which only has to be computed one time, contains saved information about how to optimally compute Fourier transforms of various sizes. The FFTW Wisdom File man page has more details.  That was what was implemented in SBITX V3.2 which eliminated the hallucinations.

I'm still chasing a few problems in the receiver.  Top of my list is a tuning problem. When I zero beat WWV on exactly 10 MHz, the displayed frequency on the SBITX is a few hundred Hz off of 10MHz, and when I tune to 15MHz WWV the display is off by a different amount. So, the delta between the displayed frequency and the frequency the radio is receiving changes with frequency – but not in any linear way.  I’ve tried several different ways to align the radio but have not yet been successful  The last thing I did was disconnect the analog receiver entirely from the SDR and used a signal generator to put a fixed 24KHz signal into the audio codec which should result in a signal displayed dead center on the waterfall – but it did not – it is a few hundred Hz off.  the current suspicion is that the crystal on my WM8731 protottype board is out of spec.  Farhan has offered to send me one of the codec boards he produced for the early SBITX prototype.  When that arrives, I will replace my audio codec with the one he sends.  That should resolve this last issue but it still doesn’t explain why the delta moves with HF frequency.  That’s what was puzzling me and what I was referring to on the podcast. 

That's it for now,

73 from Great Falls

Dean

KK4DAS