Alberto made an interesting comment about the DDS scheme.
Well divide by 4 has quite a few reasons I know of. One thing though, and simple, the use of a divide by 4 section in the local oscillator frequency chain reduces the LO IN oscillator jitter by 4, and hence the phase noise by 4. And so, the idea of a digitally stabilized VFO then divided by 4 is something to think about. Hence a crystal oscillator divided by 4 is something to ponder in regards to extreme frequency stability. And perhaps way more stability than the radio we might place a QSD board into. You know such as SR40 in a ham rig. A VFO Divided By 4: A home brew VFO divided by 4 would help to improve the stability of a VFO as a LO IN source. At least improved 4 times over its undivided fundamental frequency. Something like a FM broadcast band VFO board from Fair Radio Sales would be good divided down 4 times for HF use. A 74HC74 chip type, for a divide by 4 section or a CMOS Shift Register IC, rated up as high as they can make them, usually to 80 MHz might still work fairly well at 120 MHz. I forget, there might be a 74HC74 rated higher in frequency? You would have to look it up to find out. So following a FM Broadcast band VFO or tuner, one can divide that VFO down. But you would also need a vernier drive reduction for fine tuning it. Inter Modulation Products: Someone made a comment in this thread and it is true that square waves are full of spurious inter modulation and harmonic content. And so, these products seem to me to be a matter for the 3rd order intercept point views? You know inter modulation products. Some argument has existed however that square waves with their content can help to better add energy in the passband to mix with the spectrum of lets say 48 kHz wide. However sine waves do the same. So that argument is sort of limbo. DRM demodulators use allot of analog LO IN sources. It is a single demodulator and the I/Q QSD is two demodulators. Digital versus analog? Well that battle will go onwards in heated arguments. People should go off in both directions and then emerge in the end together again, and compare notes. Analog LO IN: If a real serious VFO is used, it has to be a Huff & Puff VFO idea or a divided by 4 idea or the combination of both. And the VFO has to be stable to 50 Hz before digital control is added. Start off with a good VFO circuit and then add the digital stabilization. Hans Summers tells us that the H&P will not make a poor VFO better. Of course some people may use a VFO without stabilization and just divide it by 4 to improve its drift (divide its drift by 4). Allot of people look at these ideas for the reason that they can not afford something that cost hundreds of dollars. The USA is headed into poverty thanks to people's bad choices at the election poles and they are doing it again. We got a bunch of shady characters up for president. We are gonna be reduced down to a third world world nation, so we all are on a budget here. Money is super tight and gas prices went up and never came down over these post 9 11 years. Transportation cost went up and now the price at the stores reflects increased transportation cost. We are not buying things anymore here. So things are moving slow in production terms. Comparator Loop Filter: In the Huff and Puff experiments you can do something you can not do with a Phase Lock Loop filter. In a comparator loop filter I used 2 ~ 1000 uF caps and had a time constant of 20 seconds. The charge pump built up slowly and just nudged the frequency around as needed. So here is where the low noise digitally stabilized VFO comes in. Can you imagine using 2 ~ 1000 uF caps for a loop filter in a PLL circuit? I doubt it. See how the low oscillator phase noise thing goes now? No spurs either at least not in this stage of things. And this filter does not require some complex PLL equation to solve either. Those 1000 uF caps make a heavy noise filter. Now the thing to do is to move the VFO of this circuit up into the 100 to 200 MHz range and then divide this by 4. For instance if the incremental tuning you can achieve at 1 to 3 MHz ceases to be as incremental at high VFO frequencies then division of that frequency will restore the small incremental steps. I added 100 to 1000 uF caps to the V+ pins of the IC's and added ceramic disc across them. Directly onto each IC's on a proto board layout. Each chip was thus isolated from noise at it V+ supply pin. And across the main supply line to this circuit section, I added a big old 10,000 uF cap and so I got some good low noise out of that and I was amazed at how stable it was also. Which I think comes from ridding the circuit of ripple voltages, coupled with the digital stabilization. The long time constant allows the charge pump to slowly make corrections, real smooth like. Yet if it charges down to make a correction it responds fast enough. In slight corrections the voltage changes are small in amplitude and in sudden large corrections the voltage changes are in a much larger amplitudes, so sudden corrections work well in fast response terms. They call the loop filter the integrator section here. There is actually another loop filter cap ahead of this section, which with a resistor forms a low pass filter. Incremental tuning is a matter of tricking the circuit to pass a correction signal only on the leading rising edge of the waveform. And thus not pass all of the waveform. So the output now looks more like saw waves than square waves. The other or first capacitor in the low pass filter (loop filter) of the integrator helps to do this. It loads the 74HC74 output such that it only has time to respond on the rising leading edge. One day you will be able to buy the completed DDS boards in volume sales for $2 to $5 due to mass production when demand and market competition comes into view. The mass production prices will drop. I bought a Western Digital 160 GB hard drive at Wally World here in KY for $70. 00. I could not have bought that last year at that price. I could not have bought a 80 GB at that price. So in time we will have these things at cheap prices. There is yet allot of work and exploration ahead into the future uses of the Huff and Puff VFO, it is not over with. The problem that people have with the Huff & Puff is in using models that have large locking steps and it unlocking on them. It does not move much when used in incremental mode. Unlocking occurs when a AC line transient comes through the power supply and and hits the circuit with a transient pulse, then it unlocks. Rid the circuit of noise and transient pulses and this will not occur.
