Transceiver Architecture 2.01
Please note the DifX is not a singular transceiver (like the Big Kahuna) but instead is a concept to demonstrate that successful transceiver projects can be achieved with something other than the Bitx20 footprint. The Bitx is a long standing successful design; but is not the only approach to homebrewing a rig. We are now at a point where with the aid of low cost technology we can build in many new features from the outset. The DifX series of radios will provide some insight into the "how to do it."
As promised in the previous post, I will explore homebrew transceiver architectures that are different than a Bitx. I can think of no better place to start than with my KWM-4 design which began in late 2012 and resulted in a completed transceiver in early 2013. This project was published in an 2013 article in QRP Quarterly.
What is significant about this project was that it is a dual conversion transceiver and covered six amateur bands and for my good friend N2CQR had a digital VFO. Just announced and mentioned in the SolderSmoke Podcast 194 and just appeared in Hackaday is the Micro-Bitx or uBitx from VU2ESE. The uBitx is a dual conversion transceiver and has created quite a stir in the homebrew community. While I can't comment on the design details of the uBitx as I simply don't know, I can share with you my concerns about a dual conversion architecture and what I had to "noodle my way through" on the KWM-4 rig.
Let us start with a high level block diagram of the KWM-4 and identify some of the key areas of noodling. This is shown below and highlighted in green shading.
Let us first start with the two blocks identified as BLA --yes Bilateral Amplifiers. But these amps are based on a design by Ron Taylor, G4GXO and appeared in the GQRP SPRAT #128. Essentially the amplifying devices can be either a Dual Gate MOSFET such as a BF991 or two J310's connected in a cascode circuit (source of one connected to drain of the second). The stage gain is 17 dB and the signals are diode steered. Z in/out is matched to 50 Ohms. This BLA is not the Bitx BLA! So Ok you want to see it. Noteworthy I first used this BLA in a 2007, 17M/40M Transceiver and later in the 2009 Tri-Band transceiver that used the HW-101 frequency scheme.
This circuit was selected as the key element over the Bitx or Plessey BLA for one specific reason, that being Gain Control. This circuit has the capability for either manual or Automatic Gain Control at the IF stage. Yes the W7ZOI Hycas AGC circuit provided the gain control in the KWM-4. The Bitx does not have that as a direct capability. There was a second reason for this approach which is shown later in a detailed block diagram and that is CW. In the KWM-4 the output/input of the 1st BLA ahead of the Collins Mechanical filter is relay switched so that in SSB it is always connected to the Collins Filter. But on CW transmit that connection is switched over to a keyed buffer amp that is supplied a signal from a third BFO frequency centered on 455 KHz.
Thus CW does not go through the Collins Filter on Transmit. For CW receive USB is used. But had I used an additional panel mounted switch the operator could have chosen either LSB or USB for receive --CW reverse. This is not found in the Bitx. Optionally if you had a Collins CW filter --then the ne plus ultra for the CW enthusiast -- selectable receiver filter bandwidths. W7ZOI in the SSDRA shows how to diode steer two filters and even provided an extra gain stage for the CW filter.
Now to the dual conversion part of the KWM-4. This project started innocently with my acquiring a 455 KHz Mechanical Filter as used in the KWM-2. I also had managed to scrounge up the USB and LSB crystals as well as a crystal smack on 455 KHz. There was a siren's call to "use me in a transceiver". But if used only with a single conversion (and dual conversion too as you will see) becomes problematic because of image and frequency mixing issues as you go higher in frequency.
Here is an example. I generate a SSB signal (we'll use 455 kHz as it will make the math simple) at 455 KHz and I mix that with 20.845 MHz LO. Normally the sum frequency would be 21.300 -- a really great place to be on 15 Meters! But the difference frequency is 20.845 - .455 = 20.390 -- That is really close to the 15 Meter band and unless you had some really brick wall Band Pass Filter --difficult to knock down. As you go even higher in frequency the spread becomes even less. So LOW IF frequencies tend to need some way of distancing themselves. Thus dual conversion.
In the three shaded blocks that show the LO, and 1st and 2nd mixers, I will now detail how I resolved that matter. Touring through the Mouser catalog, I found that they carried crystal filters at 10.7 MHz and these came in two bandwidths --15 KHz and 7.5 KHz. I also discovered that one of the stock computer crystals was 10.245 MHz and the light bulb went on. Initially I bought the 15 KHz filter but then changed that out for the 7.5 KHz filter and one crystal at 10.245 MHz. So here is why the light went on. If you add .455 MHz to 10.245 MHz, the sum is 10.7 MHz (hold that number) Now if you subtract the 0.455 MHz from 10.245 MHz the sum is 9.79 MHz. The 7.5 kHz bandwidth of the 10.7 MHz crystal filter will reject the difference frequency. Thus anything coming through the 10.7 MHz filter would only be he SUM frequency. That eliminated the undesired frequency product.
The K5BCQ frequency generator thus converted the 10.7 MHz SSB/CW signals to the appropriate ham bands. The only negative to this scheme is the 30 Meter band is too close to this 10.7 MHz conversion frequency and thus was not included -- not a problem for me but a problem for some I am sure.
The K5BCQ frequency generator has some 900 channels (900 VFO's) and I used that for assuring that what the dial says is the real transmit frequency. Thus the LO frequency is offset (switched to a different channel and different offset frequency) depending on the mode. There is about a 3 kHz spread in BFO frequencies between USB/LSB, thus by changing the LO injection frequency by that amount accounts for the mode. In the old days of analog VFO's this was done with a varactor diode somewhere in the VFO box that would shift the VFO frequency for the same analog dial setting.
Today using an Arduino with a Si5351 would eliminate the three BFO frequencies and the need to switch channels -- all done in software --so 4+ years does make a difference in the technology.
Now for the detailed block diagram of the KWM-4.
One other noodle problem is how to change the band pass and low pass filters at the same time you change bands? Luckily the K5BCQ frequency generator also had built in a 3 digit binary code that could be linked to the band change. I simply decoded that information and that provided the appropriate set of Band Pass and Low Pass filters for the band in use. I had to learn how to decode the information and steer the signals --more on that in a subsequent posting. This is a DifX!