[time-nuts] wtd: WWVB info
turner at ussc.com
Fri Aug 7 18:30:06 EDT 2015
The use of the PIC for WWVB carrier/data detection was only ever
intended for use with a visual clock, thus uncertainty (e.g. lag, delay
or whatever you want to call it) was par for the course in the
implementation that I described.
On 8/7/2015 3:51 AM, time-nuts-request at febo.com wrote:
> The gotcha with under sampling is the need for tight bandpass filters in front of the sampler. Narrow bandwidth always
> equates to long delay. If the filters are analog (rather than digital) that delay will have drift and temperature sensitivity.
> Both of those things are to be avoided (if possible) in a receiver intended for high accuracy use.
Neil, as for the link below, unfortunately that's not it. The project
in question used the PIC's A/D converter to directly process the
signal. This would rule out the PIC16F84 used in the link, below, as
that has no A/D capability. I've looked some more and have still been
unable to find it: I'm sure that it's on the Wayback Machine somewhere,
but things can be tricky to find if you don't already have a URL!
> Is this the design you are looking for?
I did see a mention of a "Tayloe" detector (or "QSD" - Quadrature
Sampling Detector) that might also be used to advantage in a project
like this. As with A/D converters, they, too may be undersampled with
reasonable effect - Some of the readily-available SDR receiver kits do
this - so it should be very practical to do something like the following:
- Produce an audio/sine wave DDS in software using the PWM hardware in
the processor (PIC, Arduino) at 4x the desired frequency using outboard
- Slice it using the processor's onboard comparator or an outboard: Many
PICs have comparators with outputs that may be made external.
- Apply this sliced signal to a divide-by-four system or counter to
produce the quadrature signal, or use the interrupt from the comparator
have the processor produce a count on a pair of pins for a multi-channel
- Use a QSD (a.k.a. Tayloe) to yield "baseband" at/around DC.
- Apply said baseband quadrature output to a pair of A/D inputs. If the
A/D's are sampled in quick succession compared to the detection
bandwidth, reasonable balance could be maintained.
Again, the QSD could be operated at a fraction of the desired frequency
using undersampling techniques provided that the input was adequately
bandpass-filtered - but this would seem like overkill since
undersampling using the A/D converter could accomplish practically the
same thing and the quadrature channels (or Costas) be done in software.
* * *
Taking a different approach, one could feed the sine output (at audio
frequencies) to a plain-old 4046 VCO/PLL and multiply the audio
frequency to 4x the receive frequency (240 kHz for WWVB, 310 kHz for
DCF77, etc.) and then produce the quadrature clocks for a direct
conversion at-frequency, the advantage being that there would need not
be any particular bandpass filtering in front of the QSD - just standard
low-pass filtering - to produce the baseband/quadrature outputs. The
phase/jitter incurred by the squaring/frequency multiplication would be
largely irrelevant in the long-term detection windows involved.
An audio-frequency DDS synthesizer with 32 bit accumulator resolution is
very easy to produce in software and with microHertz tuning resolution,
very fine phase control may be achieved in the long term: I've used
PIC-based audio DDS generators referenced from stabilized oscillators to
produce references to synthesize VHF frequencies as well as discipline
VHF/UHF oscillators with excellent results - with special steps taken to
mitigate phase modulation issues - so such should be practical at 60 kHz
with trivial hardware. (See links below for information on using audio
DDS techniques with respect to VHF oscillators.)
What would produce delay/uncertainty would be the necessary lowpass
filtering on the output of the QSD needed to limit the detection
bandwidth, but some of this could be mitigated with multiple windowed
detectors (in software), stable analog components and appropriate
characterization of the circuits involved.
It is probably fair to say that given the limited detection bandwidth
and, more importantly, the rather limited processing resources of a
low-end processor one will never quite achieve the same timing accuracy
that one might get with long-term correlation techniques to determine
the phase reversal of the original carrier down to the half-cycle -
minus propagational uncertainties, of course!
(One would have to be nuts to want to do all of this, but that's half of
the name of this group!)
References for using PIC-generated DDS audio signals as references for
- http://utaharc.org/rptr/synchronous_62.html - using the same DDS
techniques to discipline VCXOs.
More information about the time-nuts