Hi:

So how do we make a receiver to decode the PM format?

Have Fun,

Brooke Clarke
http://www.PRC68.com
http://www.end2partygovernment.com/2012Issues.html

J. Forster wrote:

File an FOIA request?

-John

==========

Hi

Given the low frequency, it's not to hard or expensive to do a DSP radio. A true FPGA (not a CPLD) would do it pretty easily. You could probably do it with a reasonably fast micro controller.  Normally the ADC would be a significant chunk of the cost. At 60 KHz … not so much.

Bob

On Sep 25, 2012, at 5:18 PM, Brooke Clarke brooke@pacific.net wrote:

Brooke
There is indeed additional detail in the spec that we had not seen. So I
will do some reading tonight.
However I am still working on the d-psk-r and at least at this point may
have a mod that works. It does on my local home brew bpsk generator even
when I add a great deal of noise. Though the noise added is really a flat
noise nothing like the ugly stuff on 60 KHz with impulse and such. But what
I have is not a very generalized approach.
Whats very interesting and others have mentioned there can be some very
creative ways to approach the BPSK if all you are interested in is getting
rid of the BPSK. You can down convert the signal and sample on a sound card
using tools like Spectrum Lab. Talk about a sledge hammer approach. Its
pretty interesting.
But like you I am curious is there a chip set that does the magic available
from DigiKey for $3.95? Would hope that what ever comes out might have a
phase control signal. Or does this cost an arm and a leg?
Regards
Paul
WB8TSL

On Tue, Sep 25, 2012 at 5:18 PM, Brooke Clarke brooke@pacific.net wrote:

I'm actually not joking about an FOIA. It would probably be worth asking
John Lowe for the design and code first though.

YMMV,

-John

=========

Bob way faster then the low on the analog side I believe.
I was using the gps tick to try to catch it and never did. Maybe a digital
oscope would have done the job. Also tried the gps tic as a cheatn compare
point for checking phase. That wasn't useful.
Regards
Paul.

On Tue, Sep 25, 2012 at 5:13 PM, Bob Camp lists@rtty.us wrote:

On Tue, Sep 25, 2012 at 2:18 PM, Brooke Clarke brooke@pacific.net wrote:

With a transmit freq so low you could directly sample the RF with a
not-very-fast A/D converter.    However I'd want some selectivity in the
front end before sampling.

Chris Albertson
Redondo Beach, California

On Tue, Sep 25, 2012 at 06:15:06PM -0400, paul swed wrote:

Paul,

http://www.xtendwave.com/ is a little bare.  The only mention

they have of the product is:

http://www.xtendwave.com/EverSet.pdf

It does not appear that they have a shipping product today.

I'm not really clear on why it's necessary to change the

modulation before the chipset is even on the market, but that seems to
be the way it is.

Various patent searches haven't turned up much, either, but I

could be missing something.

As for an FOIA request, if someone's going to file one, I'd be 

happy to chip in towards any duplicating fees.  I can also scan
documents if needed.

--msa

Hi

As long as you deal with straight overload, an ADC based receiver doesn't need as much front end selectivity as you might think. A fairly simple L/C filter should be plenty. A tuned antenna probably is going to provide all the selectivity you'd need.

Bob

On Sep 25, 2012, at 6:44 PM, Chris Albertson albertson.chris@gmail.com wrote:

In message BC97E391-7ABE-413D-868A-1878A2E3A653@rtty.us, Bob Camp writes:

It's called "A PC with a 192kHz soundcard..." and it's ridiculously cheap
and has amazing computing power :-)

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Hi,

From: "Bob Camp" lists@rtty.us

Another possibility is to modify your old circuit like that:
http://www.maxmcarter.com/rubidium/2012_mod/index.html

Bye,
Jean-Louis

Have you actually tried it and gotten it working, except possibly in a
very strong signal area?

-John

================

Also curious on the rcvrs doubling to 120 KHz another multiplier or did the
rcvr happen to have a way to change that frequency? The note says that the
rcvr local carrier doubled.
Regards
Paul

On Wed, Sep 26, 2012 at 9:18 AM, J. Forster jfor@quikus.com wrote:

As I read it, the feedback divider was reprogrammed.

David

On 9/26/12 9:49 AM, paul swed wrote:

Thats what I was reading. But have to say I have never seen a commercial
divider chain that could be changed. The spectracoms divide to 20 Khz and
mult by 3. So would like to see that circuit.
Regards
Paul

On Wed, Sep 26, 2012 at 10:00 AM, David McGaw n1hac@alum.dartmouth.orgwrote:

Hi

The issue is dithering the input to the ADC. If you filter "to much" you get
to the point that you are not dithering very much. That actually reduces the
linearity of many ADC's (ie spurs go up). It very much impacts your ability
to pull out low level signals.

Done that? yes lots of times. Got it working? yup. It's a very standard text
book approach that's been around for many decades.

Can you carry dithering too far? Indeed you can. You do not want to overload
the ADC. Once it's clipping, you aren't getting anything useful out of it.

Things like broadcast AM and switching power supplies are both threats to
your receiver. If your loop antenna has a Q of 100 or so, it's going to have
a bandwidth of about 600 Hz. That eliminates the AM band pretty well.
Switchers can be anywhere, yes you might have one at exactly 60.0001 KHz.
Call up Javad for a brick wall filter in that case.

Your 600 Hz wide antenna has taken out a lot of noise. More than likely,
your ADC isn't getting much into it. A 6KHz wide antenna would still kill
the AM band. The lower Q antenna also may give you less trouble over
temperature. Where between Q=10 and Q=100 you get reasonable dithering and
nuke your local crud and get a reasonably stable system.... TBD.

Bob

-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of J. Forster
Sent: Wednesday, September 26, 2012 9:18 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] WWVB PM Receiver

Have you actually tried it and gotten it working, except possibly in a
very strong signal area?

-John

================

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

You have not answered my question. Yes or no, do you have it working on
the new WWVB format?

-John

=================

This is precisely the issue.  Squaring the WWVB signal results in a
significant SNR penalty.  At high SNR it doesn't matter that much; at
low SNR you are in a world of hurt.

I had suggested to John Lowe that they consider retaining some carrier
in the signal, which would be trivial to do---instead of a modulation
index of 180 degrees for BPSK, make it 120 degrees or 150 degrees, so
that there remains a little bit of pure carrier at 60 kHz that's
trackable with PLL receivers.  He said something about being receptive
to the idea, but apparently it was not adopted, since the latest
document still says antipodal BPSK.

Ironically, GPS is heading in the opposite direction.  The legacy C/A
and P(Y) signals have no unmodulated pilot spreading codes, while the
newer ones at L2 and L5 have strong pilots that allow much better
tracking.

Cheers,
Peter

ps: what happened to the public comment process?  Did I miss the window?

On 9/26/12 9:11 PM, Peter Monta wrote:

I'm not sure about residual carrier aiding the tracking process.  A
Costas loop recovers the carrier pretty well, and a symbol aided loop
(where the I channel has a hard limiter, for instance) does even better.

After all, the energy is still the same.

Yes, these work (and a soft tanh() limiter improves on the hard
limiter a little bit), but I think they don't work as well as a PLL
with a pure carrier, where performance is measured as the variance of
the phase estimate at a given SNR.

True, but information has been lost as a result of introducing these
unknown phase transitions.  Now if the phase transitions are known,
one can certainly wipe them off by multiplying by a noiseless replica
of the known phase modulation, and then you're back to pure carrier.
But if you don't know the transitions ahead of time, you need the
Costas loop to find them for you, and that costs SNR.

In WWVB's case, many of these phase transitions probably can be
predicted.  But the point is not so much that good timing receivers
for the new signal are problematic.  On the contrary, they're no
problem at all with a little DSP.  But for the sake of backward
compatibility, putting 5 or 10 percent of the signal power into a
carrier seems a small price to pay.

Using a Costas-loop preprocessor to a legacy phase receiver is almost
to the point where you're better off tossing the legacy receiver and
just using the preprocessor.

I don't want to sound too negative here.  I'm glad WWVB is getting
these improvements, and the clarification from John Lowe earlier today
about the openness of the signal is helpful.  But backward
compatibility would have been so easy to put in.

Cheers,
Peter

I am feeling a bit slow here.
There is a carrier always. Thats how the AM works. So somehow we are
speaking about a semi non coherent carrier perhaps??
So whats the nickle solution and it is not squaring in a low s/n
environment. Been there done that. Very bad results on the east coast.
Regards
Paul
WB8TSL

On Thu, Sep 27, 2012 at 1:15 AM, Peter Monta pmonta@gmail.com wrote:

On 9/26/12 10:15 PM, Peter Monta wrote:

Or, you can use an acausal processing scheme.. demodulate the bits, then
go back and remove them from the input signal.  If what you're doing is
recovering timing to a gnat's eyelash, that would probably work.

Moin,

On Thu, 27 Sep 2012 08:53:03 -0400
paul swed paulswedb@gmail.com wrote:

Yes, there is still a carrier left. A BPSK spektrum looks like a
sin(x)/x signal which is centered on the carrier. Yes, you can
sync on this signal without much difficult... if you use the right method.
And that's where the problem lies. Most old devices use a PLL to sync
to the signal, which tracks the phase. Now if the phase jumps 180°
every second, the PLL will jump back and forth every second (or actually
on every 0-1 and 1-0 transition). Ie you have a jump of the PLL error
signal every second, possibly unlocking the PLL until the PLL has locked
in again. This jumping will disturb any frequency normal.

A way to mitigate this problem is to use

1. a smaller phase step than 180°

2. a higher modulation frequency

3. a modulation that has an average of 0, even on the short term
   (as in shorter than the time constant of a "standard" PLL)

4. and 2) allows the PLL filters to filter out the phase steps and 3) ensures
   that the modulation pattern doesnt change the frequency, even in the short
   term.

If you find yourself reminded of something... Yes, this is exactly what
DCF77 has done for over 20 years....[1]

I really wonder why they didn't copy the modulation scheme of DCF77 for
WWVB or at least take ideas from it...

		Attila Kinali

[1] http://remco.tk/handig/DCFp.pdf

--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson

Why not DCF. NIH. Just joking.
Good comments that are detailed. The phase does not change every second
though it can.
There are many times when the phase sits for 3 or more seconds. The PLL TC
is about 2-3 seconds on the older receivers so they tend to jump around
never really locking or only for a few seconds.
Now NIST has provided more detail then the earlier documents and that is
that for at least 2 seconds at the top of the minute the carrier is at 0
phase. (It may actually be 3 seconds. There is a comment about the last
second of the minute).

So what I have ended up doing along with discussions with JohnFor of
time-nuts fame is use a inverting and non inverting path with switch ahead
of the spectracom phase detectors
Detect the phase shift and through a simple microproc and very simple code
flip the incoming signal in about 2-5 cycles of 60 Khz. (This could also be
done in simple logic.) The value of the proc is it lets me easily tinker
with the decision to flip or not. Noise and lock out ignore etc.
But its dependent on the likes of a spectracom 8163 kindly given to me from
another time-nut for the cost of shipping.
So its not a general purpose solution.

Given the new information released a few days ago by NIST there are several
more possibilities that essentially look at the top of the minute. Gather
the phase of a local non locked source slightly adjust it per minute and
then through the minute at every 1 sec + 100 ms check the phase and flip
teh path.(Long fades are still a issue) This approach actually
re-establishes a 0 phase coherent solution and though its a pll its a
really loose pll without the need for various filters and feed back. Its a
feed forward approach.

An interesting item of note in a bandwidth limited channel such as used in
most of these rcvrs that have a xtal filter. The phase flip also causes a 0
carrier (Or hugely reduced)for a small amount of time. May be useful.

I am actually looking for the next test to see if the approach really does
work as written here. I have had enough false starts, used lots of opamps,
and solder over the last 6 months.
Regards
Paul
WB8TSL

On Thu, Sep 27, 2012 at 9:34 AM, Attila Kinali attila@kinali.ch wrote:

Jim,

What you are suggesting is essentially a spread spectrum system where the
chip pattern is time varient.

IMO, this is an incredible kludge. And, there is no gurantee that the
algorythm for generating the chip pattern will not change down the road.

YMMV,

-John

==================

On 9/27/12 7:23 AM, J. Forster wrote:

I think I poorly explained what I was thinking.

Store the raw samples
Run the samples through a demodulator to recover the bits using whatever
technique works best: for instance, you can make your symbol transition
decisions based on many bits at once, as opposed to only those you have
already seen.

Then, take those decoded bits and use them in a second pass through the
data to remove the bits (sort of like the inphase arm in a Costas loop)
so you can get a "carrier only" version of the input signal (with some
noise at the symbol boundaries, most likely).
Excise the transitions where the SNR is lower.
Then, do your carrier frequency and phase recovery on what's left over.

There's probably some elegant approach to deciding what to excise and
what not to.

But, in any case, no a priori knowledge of the bits is needed.

(We did something like this at JPL to recover telemetry bits from
Phoenix coming out of the plasma on EDL.  Recover the carrier and symbol
timing when you're farther down and then run the demodulator backwards
in time).  It's always easier to track than to acquire, after all, so
why not acquire later when the signal is strong, and track backwards to
where the signal is weak.