[time-nuts] Phase modulation detection/NIST plan

Wed Jul 11 19:48:52 UTC 2012

 David
Read your comments and have been traveling. So finally a chance to email.

I read the document also and walked away with what I shared.
In your reading would you believe the following.
Its an absolute phase and that when it switches to 0 there is 1 transition
at the beginning of the second to 180 degrees staying that way to the next
bit or flipping again to 0 degrees if its a 1 at the 1 sec tic???
Is there a way to sense from the document that there is a bias towards 0
lets say.
I could not figure that out.
Regards
Paul.

On 7/9/2012 1:53 AM, David I. Emery wrote:

On Sun, Jul 08, 2012 at 09:02:53PM -0400, Bob Camp wrote:

 Hi

  The gotcha is that they may change the sync word based on test data.
They may also tweak other vague points in the spec based on the troubles
they run into in their tests or with their silicon.

 	I finally read the wwvb.pdf paper (yes, do so before opening
mouth)...

	I think I read the "Binary Phase Shift Keying Modulation"
paragraph on page 10 to indicate they are using ABSOLUTE, not
differential BPSK.

	They refer to the "baseband waveforms s0(t) and s1(t)".   To me
this is the absolute I vector... and this clearly says  that a 0 is
always upward (or by convention in phase), and a one is always downward
(180 out)...   They clearly say the phase shift is 180 degrees...

	I would think this clearly could be phrased better...

	It appears the data format they propose is quite well defined in
the paper, though they clearly indicate that a proposed extension is
changing the barker code sync word for frames every so often so as to
indicate a different frame type that might contain highly entropic (eg
volatile and unpredictable) information of  undefined character
including a possible mechanism for sending arbitrary and completely
apriori unpredictable bitstreams, though doubtless constrained by the
hamming codes used for FEC/error detection and the barker code sync
word.

	On a quick read it appears the complete 60 second time frame
format is defined unambiguously.   There are somewhat unpredictable DST
bits and leap second bits in there... but in practice those change VERY
infrequently from 60 second frame to frame or even from week to week
or year to year. (Yes Congress likes to muck with DST every decade or
so...).

	I am still reading more carefully, but I think this means that
the entire phase and amplitude sequence of the signal is defined for the
current initial version if you know the time of day and date and the
current leap second and DST settings (which change VERY infrequently).
And I *THINK* I understand this means the absolute phase sequence
relative to the 60 KHz going into the modulator at the transmitter....

	Thus the initial signal phase modulation could be removed by
some comparatively simple itty bitty microp software driving a balanced
modulator  BUT future signal extensions might not have that property.

	As for acquiring bit sync with the signal, both the amplitude
and phase information should allow a micro to do this easily and
relatively quickly if the I vector were provided to the micro somehow.
This would presumably be possible by either sampling the 60 KHz directly
with an A/D (at 240 Ksample/sec) or by using an external balanced mixer
driven by local synthesized 60 KHz.   Even just an envelope detector
would work with strong signals because of the AM  component, and this
might be enough to acquire adequate bit sync for some purposes.

	Software PLLs at 1 second rate are duck soup for even a SLOW
micro... and frequency errors are tiny so tracking can be tight. And
acquisition for these is also very fast given reasonable SNR. Only takes
forever if SNR is so low it takes that many seconds correlation to see a
reliable tick.

	I admit as I think about this that if one synthesized the clock
for a itty bitty simple micro from say a local DUT 10 MHz whose phase
relative to WWVB one is monitoring one could do much of the entire job
by using programmable timers on the micro and its internal A/D.   This
includes phase error versus WWVB output and of course TOD output.

	One would almost certainly want to either use external balanced
mixers (FET switches ?) and produce an analog I and Q (low pass
filtered) for processing by a really slow micro or use a fast enough one
to take a stream of actual real 60 KHz input samples at 240 KHz and
compute filtered I and Q)  (and LP filter/decimate it) (yes, with
accurate A/D clocking from suitable microp output pin interval timers
you might well be able to subsample by a lot and not actually ever deal
with even any close to a  240 KHz sample stream with the micro).

	This would of course allow computation of the vector positions
of the WWVB signal modulation in I and Q space relative to the 10 MHz
clock from the DUT.   And from that one should be able to compute
the various moments of 10 MHz DUT clock drift and do a decent job
of compensating for it (better and better as the DUT clock gets more
stable/predictable) and ride out fairly long fades and outages without
losing a pretty  good idea of the expected WWVB phase.

	Presumably most standards whose phase one is tracking with such
setups are very stable, thus the holdover should be considerable
if one uses a good error and drift estimate to adjust ones local
idea of WWVB phase relative to local clock derived from the standard
to compensate.   And guess what, determining a local error and drift
estimate is precisely what such a system is doing...

	And yes one could also drive a balanced modulator with the micro
to generate a de-psked signal for legacy receivers in the museum, but
in fact the micro would already know all that those things typically
provided by legacy equipment (phase of local DUT standard versus WWVB
and time of day and possibly SNR and or signal strength).

	It has occurred to me that it might be necessary to either
squelch (eg turn off) the de-psked output to the legacy receivers on
signal fades or perhaps BPSK modulate it with a several times the 1
second bit rate pseudo random bit stream to ensure that the legacy
receivers never saw anything they thought was good lock until the time
of day lock allowed determination of absolute WWVB signal phase reliably
(and the new code with FEC seems to make this VERY reliable).

	Enough drunken ramblings late at night...

 Bob

On Jul 8, 2012, at 8:07 PM, Magnus Danielson wrote:

 On 07/09/2012 12:46 AM, paul wrote:

 Peter indeed there could be
But it should not need to be decoded to undo the psk.
Plus documentation lacks some of the details I think to actually do it.
But that would be a significant project since the formats not been
settled completely yet.

 I have looked at the PTTI 2011 paper (wwvb.pdf) and much of a format
is being shown. Has anyone established the 14 bit sync-word and
verified the format? It seems that aligning up with the normal AM
broadcast should be possible.

Can someone record it as it has been reduced to say 2 kHz and analyze
the produced audio file? Recoding with 48 kHz sampling rate should
allow almost trivial 2 kHz I-Q demodulation to illustrate phase swaps.

Cheers,
Magnus

_______________________________________________
time-nuts mailing list -- time-nuts at febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.