[time-nuts] WWVB SDR discussion

[jimlux]

On 8/10/20 11:45 AM, paul swed wrote:
> Hello to the group.
> Have been looking forward to seeing how the STM32 SDR project might be
> going.
> SDR is a weak spot for me. So been reading. And believe the answer is that
> a SDR solution may work for AM code and even BPSK code to an extent. But
> doesn't the sampling destroy the quality of the incoming signal for
> establishing a locked reference?
>

Not really.  Consider you want to know the phase of a sinewave, and 
you're sampling at 3.5321 samples/cycle (i.e. it doesn't have to be 
integer samples/cycle).  You can solve for the phase of the sinewave to 
any level of precision, limited essentially by the SNR of the samples 
(including any quantization effects).


You can do this repeatedly - and implement all you need in terms of a 
phase locked loop - entirely in the digital domain.

As it happens, some things are *easier* if you, for instance sample at 4 
times the frequency of the input signal - various clock rate noises 
alias to places that are not near the input signal.

So, sample with enough bits, digitally filter, run your PLL, and you can 
conceivably steer your processor's VCXO to be in a fixed frequency 
relationship with the input.(and maybe phase).


This is essentially what we do with modern deep space transponders - we 
digitize the received uplink signal (actually, at the IF), run that 
through a PLL tracking loop using the frequency error term to drive a 
DDS which generates the downlink signal. So we can keep the transmit 
frequency exactly at 880/749*receive frequency (For X-band, anyway).

There is some arithmetic involved to make sure that all the ratios come 
out right - and some design involved to make sure that errors (i.e. 
noise on the reference oscillator, which is fixed frequency) cancel 
appropriately.

The previous generation of transponders did a quasi digital PLL - the IF 
was digitized and processed to generate an error signal, which was then 
used to push the VCTCXO around - so the reference frequency was always a 
fixed fraction of the receive frequency (usually around 78-80MHz, for 
historical reasons). Then the transmit side just multiplied the 
oscillator up.