[ntp:questions] Re: WWV audio driver, FreeBSD, Can't seem to get it going

David L. Mills mills at udel.edu
Tue Jun 7 18:04:56 UTC 2005


Welcome to Real Radio. The WWV staff told me some years ago that, while 
the published specification of subcarrier level was 10 dB down from the 
carrier, in 1980 they changed it to 6 dB down from the carrier because 
the Heathkit GC-1000 WWV Most Accurate Clock needed a couple extra dB to 
be well.

I was a consultant on the Precision Time Inc./Traconex WWV/H receiver 
design, I argued for an approach just like the WWV/H audio driver, but 
they elected a more conventional design as you and I find. I have two 
PSTI/Traconex receiver here, which I confiscated when they went bankrupt 
and didn't pay my fee. They work just fine, even if you have to replace 
the RS232 chips every few years. You might find the audio demodulation 
routines frustrating at times, but they are demonstratively the best 
according to theory.

Dave, w3hcf

wa6zvp at gmail.com wrote:
> Hi all.
> I haven't done an update in a week or two, so thought I would update
> everyone on whats going on.
> I bought an SB card, but the snd_driver load didn't even see it, so
> I put the 'white box' special back in.  After restarting, I saw lots
> more action.  Started getting wwv3 and 4 messages, but it still would
> not quite get happy.  After an hour or so, it even lost minute sync
> and did a full restart.  Never again did it get any better.
> On a hunch, opening up the case and voila.  Hot.  Bad fan.
> I replace (er, kludge) some fans.  Things get better, for a while,
> and even set the clock once, before getting unhappy.  Even days of
> running with s9+ WWV signals would not fully sync.
> So, I start wrapping my head around the refclock_wwv driver.  (Actually
> it
> was more like it was wrapping itself around _my_ head, but I digress.
> At least I started understanding the sequence of the debug messages,
> and
> what to watch for.  I even added one myself to watch the watchcat
> variable.
> The WWV docs say the 100Hz 1bps data is at 50% modulation.  This does
> not correspond to the source code, so I changed the value of DGAIN from
> 10 to 2.  I varified the levels with a scope.  It is 50% of the tick
> levels.
> Seems better, but still not there yet.
> Today, in frustration, I find another scrapbin audio card.  Another one
> that
> that the snd_driver load wouldn't find. :(
> So, I put the semi-working on back in, rebooted, and I'll be damned.
> It progressed non-stop from Minute sync (took 2 minutes) directly to
> second sync.  3 minutes later it had all the digits.  After it had seen
> all the digits with high certainty, it set the clock and stayed there.
> Total elapsed time about 20 minutes.  It then proceeded to ramp up the
> audio clock freq updates to 1024 and is happy as a lark!!
> I'm writing this some 5 or 6 hours after the lock.  I'm still seeing
> clockstat output like:
> 53528 14520.008  0 2005 158 04:02:00  D -6 0 127 WV15 21 5
> 88.0 1024
> 53528 14522.016 wwv5 001f 175 4924 3207/34.4 5678/18.2
> WV15 fe91 37 3514/40.0 WH15 0002 19 1895/33.3
> Which is very good.  It is starting to take some data hits though.
> It is starting to fall off compared to a few hours ago.  This is
> expected.
> Propagation on 15 Mhz is starting to drop off to CO.  10 would be
> better.
> The Traconex receiver is not very agressive at checking better freqs.
> It
> has to really get weak.
> Anyway, I'd like to tune the fudgetime value now to match the GPSs, but
> it probably won't relock.  Wish I knew what made it do it today.
>           Roger
> wa6zvp at gmail.com wrote:
>> I suspect that interest in drivers other than GPS and PPS are waning,
>>in fact I have 3 GPS based servers as well, but thought I'd try
>>something else.
>>I have a Traconex 1020 WWV receiver used with driver 3 which works
>>rather well, so I thought I would add the audio driver (36) to the mix.
>>System is an older Dell workstation with FreeBSD 5.3.  The NTP version
>>is 4.2.0a-20050303, although I've tried others, and even BSD 5.4
>>The sound 'card' is on the motherboard.
>>I did:
>>kldload snd_driver
>>which resulted in:
>>May 19 19:57:01 Fuzzball kernel: pcm0: <Intel ICH2 (82801BA)> port
>>0xdc40-0xdc7f,0xd800-0xd8ff irq 10 at device 31.5 on pci0
>>May 19 19:57:01 Fuzzball kernel: pcm0: <Analog Devices AD1885 AC97
>>May 19 19:57:01 Fuzzball kernel: pcm0: measured ac97 link rate at 55945
>>May 19 20:08:03 Fuzzball ntpd[970]: ntpd 4.2.0a at 1.1191-r Wed May 18
>>20:59:11 PM
>>and it created /dev/audio0.0 to which I symlinked ntp.audio.0
>>ntp.conf contains:
>>fudge flag2 1 flag3 1 flag4 1
>>Here is the last 3 samples from clockstats:
>>53510 59149.914 wwv8 2  91 WV15 0   226  10.1 62698  4880
>>53510 59149.914 wwv8 2  91 WH15 0   248  12.5 443431  4880
>>53510 59208.430 wwv8 2 111 WV15 0   361  13.2 24318  2480
>>53510 59208.430 wwv8 2 111 WH15 0   264  15.7 35736  2480
>>53510 59266.946 wwv8 2  91 WV15 0   298   9.5 28666    80
>>53510 59266.946 wwv8 2  91 WH15 0   254  12.0 239884    80
>>The AGC does show action and responds to removing/inserting audio.
>>However, the tone filters never seem to get a grasp on things, like
>>perhaps the output of the A/D is garbage?
>>I know that Dave Mills has said that this driver is no longer
>>maintainable, at least by him, but I found references from last year
>>that it did work under FreeBSD.
>>Does anyone have any ideas on where I should go from here?
>>        Roger

More information about the questions mailing list