source: trunk/third/xntp/html/driver36.htm @ 17260

Revision 17260, 38.6 KB checked in by zacheiss, 23 years ago (diff)
This commit was generated by cvs2svn to compensate for changes in r17259, which included commits to RCS files with non-trunk default branches.
Line 
1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN">
2<html>
3<head>
4<meta name="generator" content="HTML Tidy, see www.w3.org">
5<title>Radio WWV/H Audio Demodulator/Decoder</title>
6</head>
7<body>
8<h3>Radio WWV/H Audio Demodulator/Decoder</h3>
9
10<hr>
11<h4>Synopsis</h4>
12
13Address: 127.127.36.<i>u</i> <br>
14Reference ID: <tt>WWV</tt> or <tt>WWVH</tt> <br>
15Driver ID: <tt>WWV_AUDIO</tt> <br>
16Autotune Port: <tt>/dev/icom</tt>; 1200/9600 baud, 8-bits, no
17parity <br>
18Audio Device: <tt>/dev/audio</tt> and <tt>/dev/audioctl</tt>
19
20<h4>Description</h4>
21
22This driver synchronizes the computer time using data encoded in
23shortwave radio transmissions from NIST time/frequency stations WWV
24in Ft. Collins, CO, and WWVH in Kauai, HI. Transmissions are made
25continuously on 2.5, 5, 10, 15 and 20 MHz. An ordinary shortwave
26receiver can be tuned manually to one of these frequencies or, in
27the case of ICOM receivers, the receiver can be tuned automatically
28by the driver as propagation conditions change throughout the day
29and night. The performance of this driver when tracking one of the
30stations is ordinarily better than 1 ms in time with frequency
31drift less than 0.5 PPM when not tracking either station.
32
33<p>The demodulation and decoding algorithms used by this driver are
34based on a machine language program developed for the TAPR DSP93
35DSP unit, which uses the TI 320C25 DSP chip. The analysis, design
36and performance of the program running on this unit is described
37in: Mills, D.L. A precision radio clock for WWV transmissions.
38Electrical Engineering Report 97-8-1, University of Delaware,
39August 1997, 25 pp. Available from <a href=
40"http://www.eecis.udel.edu/~mills/reports.htm">
41www.eecis.udel.edu/~mills/reports.htm</a>. For use in this driver,
42the original program was rebuilt in the C language and adapted to
43the NTP driver interface. The algorithms have been modified
44somewhat to improve performance under weak signal conditions and to
45provide an automatic station identification feature.</p>
46
47<p>This driver incorporates several features in common with other
48audio drivers such as described in the <a href="driver7.htm">Radio
49CHU Audio Demodulator/Decoder</a> and the <a href="driver6.htm">
50IRIG Audio Decoder</a> pages. They include automatic gain control
51(AGC), selectable audio codec port and signal monitoring
52capabilities. For a discussion of these common features, as well as
53a guide to hookup, debugging and monitoring, see the <a href=
54"audio.htm">Reference Clock Audio Drivers</a> page.</p>
55
56<p>The WWV signal format is described in NIST Special Publication
57432 (Revised 1990). It consists of three elements, a 5-ms, 1000-Hz
58pulse, which occurs at the beginning of each second, a 800-ms,
591000-Hz pulse, which occurs at the beginning of each minute, and a
60pulse-width modulated 100-Hz subcarrier for the data bits, one bit
61per second. The WWVH format is identical, except that the 1000-Hz
62pulses are sent at 1200 Hz. Each minute encodes nine BCD digits for
63the time of century plus seven bits for the daylight savings time
64(DST) indicator, leap warning indicator and DUT1 correction.</p>
65
66<h4>Program Architecture</h4>
67
68<p>As in the original program, the clock discipline is modelled as
69a Markov process, with probabilistic state transitions
70corresponding to a conventional clock and the probabilities of
71received decimal digits. The result is a performance level which
72results in very high accuracy and reliability, even under
73conditions when the minute beep of the signal, normally its most
74prominent feature, can barely be detected by ear with a shortwave
75receiver.</p>
76
77<p>The analog audio signal from the shortwave radio is sampled at
788000 Hz and converted to digital representation. The 1000/1200-Hz
79pulses and 100-Hz subcarrier are first separated using two IIR
80filters, a 600-Hz bandpass filter centered on 1100 Hz and a 150-Hz
81lowpass filter. The minute sync pulse is extracted using a 800-ms
82synchronous matched filter and pulse grooming logic which
83discriminates between WWV and WWVH signals and noise. The second
84sync pulse is extracted using a 5-ms FIR matched filter and
858000-stage comb filter.</p>
86
87<p>The phase of the 100-Hz subcarrier relative to the second sync
88pulse is fixed at the transmitter; however, the audio highpass
89filter in most radios affects the phase response at 100 Hz in
90unpredictable ways. The driver adjusts for each radio using two
91170-ms synchronous matched filters. The I (in-phase) filter is used
92to demodulate the subcarrier envelope, while the Q
93(quadrature-phase) filter is used in a tracking loop to discipline
94the codec sample clock and thus the demodulator phase.</p>
95
96<p>The data bit probabilities are determined from the subcarrier
97envelope using a threshold-corrected slicer. The averaged envelope
98amplitude 30 ms from the beginning of the second establishes the
99minimum (noise floor) value, while the amplitude 200 ms from the
100beginning establishes the maximum (signal peak) value. The slice
101level is midway between these two values. The negative-going
102envelope transition at the slice level establishes the length of
103the data pulse, which in turn establish probabilities for binary
104zero (P0) or binary one (P1). The values are established by linear
105interpolation between the pulse lengths for P0 (300 ms) and P1 (500
106ms) so that the sum is equal to one. If the driver has not
107synchronized to the minute pulse, or if the data bit amplitude,
108signal/noise ratio (SNR) or length are below thresholds, the bit is
109considered invalid and all three probabilities are set to zero.</p>
110
111<p>The difference between the P1 and P0 probabilities, or
112likelihood, for each data bit is exponentially averaged in a set of
11360 accumulators, one for each second, to determine the semi-static
114miscellaneous bits, such as DST indicator, leap second warning and
115DUT1 correction. In this design, an average value larger than a
116positive threshold is interpreted as a hit on one and a value
117smaller than a negative threshold as a hit on zero. Values between
118the two thresholds, which can occur due to signal fades or loss of
119signal, are interpreted as a miss, and result in no change of
120indication.</p>
121
122<p>The BCD digit in each digit position of the timecode is
123represented as four data bits, all of which must be valid for the
124digit itself to be considered valid. If so, the bits are correlated
125with the bits corresponding to each of the valid decimal digits in
126this position. If the digit is invalid, the correlated value for
127all digits in this position is assumed zero. In either case, the
128values for all digits are exponentially averaged in a likelihood
129vector associated with this position. The digit associated with the
130maximum over all of the averaged values then becomes the maximum
131likelihood selection for this position and the ratio of the maximum
132over the next lower value becomes the likelihood ratio.</p>
133
134<p>The decoding matrix contains nine row vectors, one for each
135digit position. Each row vector includes the maximum likelihood
136digit, likelihood vector and other related data. The maximum
137likelihood digit for each of the nine digit positions becomes the
138maximum likelihood time of the century. A built-in transition
139function implements a conventional clock with decimal digits that
140count the minutes, hours, days and years, as corrected for leap
141seconds and leap years. The counting operation also rotates the
142likelihood vector corresponding to each digit as it advances. Thus,
143once the clock is set, each clock digit should correspond to the
144maximum likelihood digit as transmitted.</p>
145
146<p>Each row of the decoding matrix also includes a compare counter
147and the difference (modulo the radix) between the current clock
148digit and most recently determined maximum likelihood digit. If a
149digit likelihood exceeds the decision level and the difference is
150constant for a number of successive minutes in any row, the maximum
151likelihood digit replaces the clock digit in that row. When this
152condition is true for all rows and the second epoch has been
153reliably determined, the clock is set (or verified if it has
154already been set) and delivers correct time to the integral second.
155The fraction within the second is derived from the logical master
156clock, which runs at 8000 Hz and drives all system timing
157functions.</p>
158
159<p>The logical master clock is derived from the audio codec clock.
160Its frequency is disciplined by a frequency-lock loop (FLL) which
161operates independently of the data recovery functions. At averaging
162intervals determined by the measured jitter, the frequency error is
163calculated as the difference between the most recent and the
164current second epoch divided by the interval. The sample clock
165frequency is then corrected by this amount using an exponential
166average. When first started, the frequency averaging interval is
167eight seconds, in order to compensate for intrinsic codec clock
168frequency offsets up to 125 PPM. Under most conditions, the
169averaging interval doubles in stages from the initial value to over
1701000 seconds, which results in an ultimate frequency precision of
1710.125 PPM, or about 11 ms/day.</p>
172
173<p>It is important that the logical clock frequency is stable and
174accurately determined, since in most applications the shortwave
175radio will be tuned to a fixed frequency where WWV or WWVH signals
176are not available throughout the day. In addition, in some parts of
177the US, especially on the west coast, signals from either or both
178WWV and WWVH may be available at different times or even at the
179same time. Since the propagation times from either station are
180almost always different, each station must be reliably identified
181before attempting to set the clock.</p>
182
183<p>Station identification uses the 800-ms minute pulse transmitted
184by each station. In the acquisition phase the entire minute is
185searched using both the WWV and WWVH using matched filters and a
186pulse gate discriminator similar to that found in radar acquisition
187and tracking receivers. The peak amplitude found determines a range
188gate and window where the next pulse is expected to be found. The
189minute is scanned again to verify the peak is indeed in the window
190and with acceptable amplitude, SNR and jitter. At this point the
191receiver begins to track the second sync pulse and operate as above
192until the clock is set.</p>
193
194<p>Once the minute is synchronized, the range gate is fixed and
195only energy within the window is considered for the minute sync
196pulse. A compare counter increments by one if the minute pulse has
197acceptable amplitude, SNR and jitter and decrements otherwise. This
198is used as a quality indicator and reported in the timecode and
199also for the autotune function described below.</p>
200
201<h4>Performance</h4>
202
203<p>It is the intent of the design that the accuracy and stability
204of the indicated time be limited only by the characteristics of the
205propagation medium. Conventional wisdom is that synchronization via
206the HF medium is good only to a millisecond under the best
207propagation conditions. The performance of the NTP daemon
208disciplined by the driver is clearly better than this, even under
209marginal conditions. Ordinarily, with marginal to good signals and
210a frequency averaging interval of 1024 s, the frequency is
211stabilized within 0.1 PPM and the time within 125 <font face=
212"Symbol">m</font>s. The frequency stability characteristic is
213highly important, since the clock may have to free-run for several
214hours before reacquiring the WWV/H signal.</p>
215
216<p>The expected accuracy over a typical day was determined using
217the DSP93 and an oscilloscope and cesium oscillator calibrated with
218a GPS receiver. With marginal signals and allowing 15 minutes for
219initial synchronization and frequency compensation, the time
220accuracy determined from the WWV/H second sync pulse was reliably
221within 125 <font face="Symbol">m</font>s. In the particular DSP-93
222used for program development, the uncorrected CPU clock frequency
223offset was 45.8&plusmn;0.1 PPM. Over the first hour after initial
224synchronization, the clock frequency drifted about 1 PPM as the
225frequency averaging interval increased to the maximum 1024 s. Once
226reaching the maximum, the frequency wandered over the day up to 1
227PPM, but it is not clear whether this is due to the stability of
228the DSP-93 clock oscillator or the changing height of the
229ionosphere. Once the frequency had stabilized and after loss of the
230WWV/H signal, the frequency drift was less than 0.5 PPM, which is
231equivalent to 1.8 ms/h or 43 ms/d. This resulted in a step phase
232correction up to several milliseconds when the signal returned.</p>
233
234<p>The measured propagation delay from the WWV transmitter at
235Boulder, CO, to the receiver at Newark, DE, is 23.5&plusmn;0.1 ms.
236This is measured to the peak of the pulse after the second sync
237comb filter and includes components due to the ionospheric
238propagation delay, nominally 8.9 ms, communications receiver delay
239and program delay. The propagation delay can be expected to change
240about 0.2 ms over the day, as the result of changing ionosphere
241height. The DSP93 program delay was measured at 5.5 ms, most of
242which is due to the 400-Hz bandpass filter and 5-ms matched filter.
243Similar delays can be expected of this driver.</p>
244
245<h4>Program Operation</h4>
246
247The driver begins operation immediately upon startup. It first
248searches for one or both of the stations WWV and WWVH and attempts
249to acquire minute sync. This may take some fits and starts, as the
250driver expects to see three consecutive minutes with good signals
251and low jitter. If the autotune function is active, the driver will
252rotate over all five frequencies and both WWV and WWVH stations
253until three good minutes are found.
254
255<p>The driver then acquires second sync, which can take up to
256several minutes, depending on signal quality. At the same time the
257driver accumulates likelihood values for each of the nine digits of
258the clock, plus the seven miscellaneous bits included in the WWV/H
259transmission format. The minute units digit is decoded first and,
260when five repetitions have compared correctly, the remaining eight
261digits are decoded. When five repetitions of all nine digits have
262decoded correctly, which normally takes 15 minutes with good
263signals and up to an hour when buried in noise, and the second sync
264alarm has not been raised for two minutes, the clock is set (or
265verified) and is selectable to discipline the system clock.</p>
266
267<p>As long as the clock is set or verified, the system clock
268offsets are provided once each second to the reference clock
269interface, where they are saved in a buffer. At the end of each
270minute, the buffer samples are groomed by the median filter and
271trimmed-mean averaging functions. Using these functions, the system
272clock can in principle be disciplined to a much finer resolution
273than the 125-<font face="Symbol">m</font>s sample interval would
274suggest, although the ultimate accuracy is probably limited by
275propagation delay variations as the ionspheric height varies
276throughout the day and night.</p>
277
278<p>As long as signals are available, the clock frequency is
279disciplined for use during times when the signals are unavailable.
280The algorithm refines the frequency offset using increasingly
281longer averaging intervals to 1024 s, where the precision is about
2820.1 PPM. With good signals, it takes well over two hours to reach
283this degree of precision; however, it can take many more hours than
284this in case of marginal signals. Once reaching the limit, the
285algorithm will follow frequency variations due to temperature
286fluctuations and ionospheric height variations.</p>
287
288<p>It may happen as the hours progress around the clock that WWV
289and WWVH signals may appear alone, together or not at all. When the
290driver is first started, the NTP reference identifier appears as
291<tt>NONE</tt>. When the driver has acquired one or both stations
292and mitigated which one is best, it sets the station identifier in
293the timecode as described below. In addition, the NTP reference
294identifier is set to the station callsign. If the propagation
295delays has been properly set with the <tt>fudge time1</tt> (WWV)
296and <tt>fudge time2</tt> (WWVH) commands in the configuration file,
297handover from one station to the other will be seamless.</p>
298
299<p>Once the clock has been set for the first time, it will appear
300reachable and selectable to discipline the system clock, even if
301the broadcast signal fades to obscurity. A consequence of this
302design is that, once the clock is set, the time and frequency are
303disciplined only by the second sync pulse and the clock digits
304themselves are driven by the clock state machine and ordinarily
305never changed. However, as long as the clock is set correctly, it
306will continue to read correctly after a period of signal loss, as
307long as it does not drift more than 500 ms from the correct time.
308Assuming the clock frequency can be disciplined within 1 PPM, the
309clock could coast without signals for some 5.8 days without
310exceeding that limit. If for some reason this did happen, the clock
311would be in the wrong second and would never resynchronize. To
312protect against this most unlikely situation, if after four days
313with no signals, the clock is considered unset and resumes the
314synchronization procedure from the beginning.</p>
315
316<p>To work well, the driver needs a communications receiver with
317good audio response at 100 Hz. Most shortwave and communications
318receivers roll off the audio response below 250 Hz, so this can be
319a problem, especially with receivers using DSP technology, since
320DSP filters can have very fast rolloff outside the passband. Some
321DSP transceivers, in particular the ICOM 775, have a programmable
322low frequency cutoff which can be set as low as 80 Hz. However,
323this particular radio has a strong low frequency buzz at about 10
324Hz which appears in the audio output and can affect data recovery
325under marginal conditions. Although not tested, it would seem very
326likely that a cheap shortwave receiver could function just as well
327as an expensive communications receiver.</p>
328
329<h4>Autotune</h4>
330
331<p>The driver includes provisions to automatically tune the radio
332in response to changing radio propagation conditions throughout the
333day and night. The radio interface is compatible with the ICOM CI-V
334standard, which is a bidirectional serial bus operating at TTL
335levels. The bus can be connected to a serial port using a level
336converter such as the CT-17. The serial port speed is presently
337compiled in the program, but can be changed in the driver source
338file.</p>
339
340<p>Each ICOM radio is assigned a unique 8-bit ID select code,
341usually expressed in hex format. To activate the CI-V interface,
342the <tt>mode</tt> keyword of the <tt>server</tt> configuration
343command specifies a nonzero select code in decimal format. A table
344of ID select codes for the known ICOM radios is given below. Since
345all ICOM select codes are less than 128, the high order bit of the
346code is used by the driver to specify the baud rate. If this bit is
347not set, the rate is 9600 bps for the newer radios; if set, the
348rate is 1200 bps for the older radios. A missing <tt>mode</tt>
349keyword or a zero argument leaves the interface disabled.</p>
350
351<p>If specified, the driver will attempt to open the device <tt>
352/dev/icom</tt> and, if successful will activate the autotune
353function and tune the radio to each operating frequency in turn
354while attempting to acquire minute sync from either WWV or WWVH.
355However, the driver is liberal in what it assumes of the
356configuration. If the <tt>/dev/icom</tt> link is not present or the
357open fails or the CI-V bus or radio is inoperative, the driver
358quietly gives up with no harm done.</p>
359
360<p>Once acquiring minute sync, the driver operates as described
361above to set the clock. However, during seconds 59, 0 and 1 of each
362minute it tunes the radio to one of the five broadcast frequencies
363to measure the sync pulse and data pulse amplitudes and SNR and
364update the compare counter. Each of the five frequencies are probed
365in a five-minute rotation to build a database of current
366propagation conditions for all signals that can be heard at the
367time. At the end of each rotation, a mitigation procedure scans the
368database and retunes the radio to the best frequency and station
369found. For this to work well, the radio should be set for a fast
370AGC recovery time. This is most important while tracking a strong
371signal, which is normally the case, and then probing another
372frequency, which may have much weaker signals.</p>
373
374<p>Reception conditions for each frequency and station are
375evaluated according to a metric which considers the minute sync
376pulse amplitude, SNR and jitter, as well as, the data pulse
377amplitude and SNR. The minute pulse is evaluated at second 0, while
378the data pulses are evaluated at seconds 59 and 1. The results are
379summarized in a scoreboard of three bits</p>
380
381<dl>
382<dt><tt>0x0001</tt></dt>
383
384<dd>Jitter exceeded. The difference in epoches between the last
385minute sync pulse and the current one exceeds 50 ms (400
386samples).</dd>
387
388<dt><tt>0x0002</tt></dt>
389
390<dd>Minute pulse error. For the minute sync pulse in second 0,
391either the amplitude or SNR is below threshold (2000 and 20 dB,
392respectively).</dd>
393
394<dt><tt>0x0004</tt></dt>
395
396<dd>Minute pulse error. For both of the data pulses in seocnds 59
397and 1, either the amplitude or SNR is below threshold (1000 and 10
398dB, respectively).</dd>
399</dl>
400
401<p>If none of the scoreboard bits are set, the compare counter is
402increased by one to a maximum of six. If any bits are set, the
403counter is decreased by one to a minimum of zero. At the end of
404each minute, the frequency and station with the maximum compare
405count is chosen, with ties going to the highest frequency.</p>
406
407<h4>Diagnostics</h4>
408
409<p>The autotune process produces diagnostic information along with
410the timecode. This is very useful for evaluating the performance of
411the algorithm, as well as radio propagation conditions in general.
412The message is produced once each minute for each frequency in turn
413after minute sync has been acquired.</p>
414
415<p><tt>wwv5 port agc wwv wwvh</tt></p>
416
417<p>where <tt>port</tt> and <tt>agc</tt> are the audio port and
418gain, respectively, for this frequency and <tt>wwv</tt> and <tt>
419wwvh</tt> are two sets of fields, one each for WWV and WWVH. Each
420of the two fields has the format</p>
421
422<p><tt>ident score comp sync/snr/jitr</tt></p>
423
424<p>where <tt>ident</tt>encodes the station (<tt>C</tt> for WWV,
425<tt>H</tt> for WWVH) and frequency (2, 5, 10, 15 and 20), <tt>
426score</tt> is the scoreboard described above, <tt>comp</tt> is the
427compare counter, <tt>sync</tt> is the minute sync pulse amplitude,
428<tt>snr</tt> the SNR of the pulse and <tt>jitr</tt> is the sample
429difference between the current epoch and the last epoch. An example
430is:</p>
431
432<p><tt>wwv5 2 111 C20 0100 6 8348/30.0/-3 H20 0203 0
43322/-12.4/8846</tt></p>
434
435<p>Here the radio is tuned to 20 MHz and the line-in port AGC is
436currently 111 at that frequency. The message contains a report for
437WWV (<tt>C20</tt>) and WWVH (<tt>H20</tt>). The WWV report
438scoreboard is 0100 and the compare count is 6, which suggests very
439good reception conditions, and the minute sync amplitude and SNR
440are well above thresholds (2000 and 20 dB, respectively). Probably
441the most sensitive indicator of reception quality is the jitter, -3
442samples, which is well below threshold (50 ms or 400 samples).
443While the message shows solid reception conditions from WWV, this
444is not the case for WWVH. Both the minute sync amplitude and SNR
445are below thresholds and the jitter is above threshold.</p>
446
447<p>A sequence of five messages, one for each minute, might appear
448as follows:</p>
449
450<pre>
451wwv5 2  95 C2 0107 0 164/7.2/8100  H2 0207 0 80/-5.5/7754
452wwv5 2  99 C5 0104 0 3995/21.8/395  H5 0207 0 27/-9.3/18826
453wwv5 2 239 C10 0105 0 9994/30.0/2663 H10 0207 0 54/-16.1/-529
454wwv5 2 155 C15 0103 3 3300/17.8/-1962 H15 0203 0 236/17.0/4873
455wwv5 2 111 C20 0100 6 8348/30.0/-3 H20 0203 0 22/-12.4/8846
456</pre>
457
458<p>Clearly, the only frequencies that are available are 15 MHz and
45920 MHz and propagation may be failing for 15 MHz. However, minute
460sync pulses are being heard on 5 and 10 MHz, even though the data
461pulses are not. This is typical of late afternoon when the maximum
462usable frequency (MUF) is falling and the ionospheric loss at the
463lower frequencies is beginning to decrease.</p>
464
465<h4>Debugging Aids</h4>
466
467<p>The most convenient way to track the driver status is using the
468<tt>ntpq</tt> program and the <tt>clockvar</tt> command. This
469displays the last determined timecode and related status and error
470counters, even when the driver is not discipline the system clock.
471If the debugging trace feature (<tt>-d</tt> on the <tt>ntpd</tt>
472command line)is enabled, the driver produces detailed status
473messages as it operates. If the <tt>fudge flag 4</tt> is set, these
474messages are written to the <tt>clockstats</tt> file. All messages
475produced by this driver have the prefix <tt>chu</tt> for convenient
476filtering with the Unix <tt>grep</tt> command.</p>
477
478<p>In the following descriptions the units of amplitude, phase,
479probability and likelihood are normalized to the range 0-6000 for
480convenience. In addition, the signal/noise ratio (SNR) and
481likelihood ratio are measured in decibels and the words with bit
482fields are in hex. Most messages begin with a leader in the
483following format:</p>
484
485<p><tt>wwvn ss stat sigl</tt></p>
486
487<p>where <tt>wwvn</tt> is the message code, <tt>ss</tt> the second
488of minute, <tt>stat</tt> the driver status word and <tt>sigl</tt>
489the second sync pulse amplitude. A full explanation of the status
490bits is contained in the driver source listing; however, the
491following are the most useful for debugging.</p>
492
493<dl>
494<dt><tt>0x0001</tt></dt>
495
496<dd>Minute sync. Set when the decoder has identified a station and
497acquired the minute sync pulse.</dd>
498
499<dt><tt>0x0002</tt></dt>
500
501<dd>Second sync. Set when the decoder has acquired the second sync
502pulse and within 125 <font face="Symbol">m</font>s of the correct
503phase.</dd>
504
505<dt><tt>0x0004</tt></dt>
506
507<dd>Minute unit sync. Set when the decoder has reliably determined
508the unit digit of the minute.</dd>
509
510<dt><tt>0x0008</tt></dt>
511
512<dd>Clock set. Set when the decoder has reliably determined all
513nine digits of the timecode and is selectable to discipline the
514system clock.</dd>
515</dl>
516
517<p>With debugging enabled the driver produces messages in the
518following formats:</p>
519
520<p>Format <tt>wwv8</tt> messages are produced once per minute by
521the WWV and WWVH station processes before minute sync has been
522acquired. They show the progress of identifying and tracking the
523minute pulse of each station.</p>
524
525<p><tt>wwv8 port agc ident comp ampl snr epoch jitr offs</tt></p>
526
527<p>where <tt>port</tt> and <tt>agc</tt> are the audio port and
528gain, respectively. The <tt>ident</tt>encodes the station
529(<tt>C</tt> for WWV, <tt>H</tt> for WWVH) and frequency (2, 5, 10,
53015 and 20). For the encoded frequency, <tt>comp</tt> is the compare
531counter, <tt>ampl</tt> the pulse amplitude, <tt>snr</tt> the SNR,
532<tt>epoch</tt> the sample number of the minute pulse in the minute,
533<tt>jitr</tt> the change since the last <tt>epoch</tt> and <tt>
534offs</tt> the minute pulse offset relative to the second pulse. An
535example is:</p>
536
537<p><tt>wwv8 2 127 C15 2 9247 30.0 18843 -1 1</tt><br>
538<tt>wwv8 2 127 H15 0 134 -2.9 19016 193 174</tt></p>
539
540<p>Here the radio is tuned to 15 MHz and the line-in port AGC is
541currently 127 at that frequency. The driver has not yet acquired
542minute sync, WWV has been heard for at least two minutes, and WWVH
543is in the noise. The WWV minute pulse amplitude and SNR are well
544above the threshold (2000 and 6 dB, respectively) and the minute
545epoch has been determined -1 sample relative to the last one and 1
546sample relative to the second sync pulse. The compare counter has
547incrmented to two; when it gets to three, minute sync has been
548acquired.</p>
549
550<p>Format <tt>wwv3</tt> messages are produced after minute sync has
551been acquired and until the seconds unit digit is determined. They
552show the results of decoding each bit of the transmitted
553timecode.</p>
554
555<p><tt>wwv3 ss stat sigl ampl phas snr prob like</tt></p>
556
557<p>where <tt>ss</tt>, <tt>stat</tt> and <tt>sigl</tt> are as above,
558<tt>ampl</tt> is the subcarrier amplitude, <tt>phas</tt> the
559subcarrier phase, <tt>snr</tt> the subcarrier SNR, <tt>prob</tt>
560the bit probability and <tt>like</tt> the bit likelihood. An
561example is:</p>
562
563<p><tt>wwv3 28 0123 4122 4286 0 24.8 -5545 -1735</tt></p>
564
565<p>Here the driver has acquired minute and second sync, but has not
566yet determined the seconds unit digit. However, it has just decoded
567bit 28 of the minute. The results show the second sync pulse
568amplitude well over the threshold (500), subcarrier amplitude well
569above the threshold (1000), good subcarrier tracking phase and SNR
570well above the threshold (10 dB). The bit is almost certainly a
571zero and the likelihood of a zero in this second is very high.</p>
572
573<p>Format <tt>wwv4</tt> messages are produced for each of the nine
574BCD timecode digits until the clock has been set or verified. They
575show the results of decoding each digit of the transmitted
576timecode.</p>
577
578<p><tt>wwv4 ss stat sigl radx ckdig mldig diff cnt like
579snr</tt></p>
580
581<p>where <tt>ss</tt>, <tt>stat</tt> and <tt>sigl</tt> are as above,
582<tt>radx</tt> is the digit radix (3, 4, 6, 10), <tt>ckdig</tt> the
583current clock digit, <tt>mldig</tt> the maximum likelihood digit,
584<tt>diff</tt> the difference between these two digits modulo the
585radix, <tt>cnt</tt> the compare counter, <tt>like</tt> the digit
586likelihood and <tt>snr</tt> the likelihood ratio. An example
587is:</p>
588
589<p><tt>wwv4 8 010f 5772 10 9 9 0 6 4615 6.1</tt></p>
590
591<p>Here the driver has previousl set or verified the clock. It has
592just decoded the digit preceding second 8 of the minute. The digit
593radix is 10, the current clock and maximum likelihood digits are
594both 9, the likelihood is well above the threshold (1000) and the
595likelihood function well above threshold (3.0 dB). Short of a
596hugely unlikely probability conspiracy, the clock digit is most
597certainly a 9.</p>
598
599<p>Format <tt>wwv2</tt> messages are produced at each master
600oscillator frequency update, which starts at 8 s, but eventually
601climbs to 1024 s. They show the progress of the algorithm as it
602refines the frequency measurement to a precision of 0.1 PPM.</p>
603
604<p><tt>wwv2 ss stat sigl avint avcnt avinc jitr delt freq</tt></p>
605
606<p>where <tt>ss</tt>, <tt>stat</tt> and <tt>sigl</tt> are as above,
607<tt>avint</tt> is the averaging interval, <tt>avcnt</tt> the
608averaging interval counter, <tt>avinc</tt> the interval increment,
609<tt>jitr</tt> the sample change between the beginning and end of
610the interval, <tt>delt</tt> the computed frequency change and <tt>
611freq</tt> the current frequency (PPM). An example is:</p>
612
613<p><tt>wwv2 22 030f 5795 256 256 4 0 0.0 66.7</tt></p>
614
615<p>Here the driver has acquired minute and second sync and set the
616clock. The averaging interval has increased to 256 s on the way to
6171024 s, has stayed at that interval for 4 averaging intervals, has
618measured no change in frequency and the current frequency is 66.7
619PPM.</p>
620
621<p>If the CI-V interface for ICOM radios is active, a debug level
622greater than 1 will produce a trace of the CI-V command and
623response messages. Interpretation of these messages requires
624knowledge of the CI-V protocol, which is beyond the scope of this
625document.</p>
626
627<h4>Monitor Data</h4>
628
629When enabled by the <tt>filegen</tt> facility, every received
630timecode is written to the <tt>clockstats</tt> file in the
631following format:
632
633<pre>
634        sq yy ddd hh:mm:ss.fff ld du lset agc stn rfrq errs freq cons
635
636        s       sync indicator
637        q       quality character
638        yyyy    Gregorian year
639        ddd     day of year
640        hh      hour of day
641        mm      minute of hour
642        fff     millisecond of second
643        l       leap second warning
644        d       DST state
645        dut     DUT sign and magnitude
646        lset    minutes since last set
647        agc     audio gain
648        ident   station identifier and frequency
649        comp    minute sync compare counter
650        errs    bit error counter
651        freq    frequency offset
652        avgt    averaging time
653</pre>
654
655The fields beginning with <tt>year</tt> and extending through <tt>
656dut</tt> are decoded from the received data and are in fixed-length
657format. The <tt>agc</tt> and <tt>lset</tt> fields, as well as the
658following driver-dependent fields, are in variable-length format.
659
660<dl>
661<dt><tt>s</tt></dt>
662
663<dd>The sync indicator is initially <tt>?</tt> before the clock is
664set, but turns to space when all nine digits of the timecode are
665correctly set.</dd>
666
667<dt><tt>q</tt></dt>
668
669<dd>The quality character is a four-bit hexadecimal code showing
670which alarms have been raised. Each bit is associated with a
671specific alarm condition according to the following:
672
673<dl>
674<dt><tt>0x8</tt></dt>
675
676<dd>Sync alarm. The decoder may not be in correct second or minute
677phase relative to the transmitter.</dd>
678
679<dt><tt>0x4</tt></dt>
680
681<dd>Error alarm. More than 30 data bit errors occurred in the last
682minute.</dd>
683
684<dt><tt>0x2</tt></dt>
685
686<dd>Symbol alarm. The probability of correct decoding for a digit
687or miscellaneous bit has fallen below the threshold.</dd>
688
689<dt><tt>0x1</tt></dt>
690
691<dd>Decoding alarm. A maximum likelihood digit fails to agree with
692the current associated clock digit.</dd>
693</dl>
694
695It is important to note that one or more of the above alarms does
696not necessarily indicate a clock error, but only that the decoder
697has detected a condition that may in future result in an
698error.</dd>
699
700<dt><tt>yyyy ddd hh:mm:ss.fff</tt></dt>
701
702<dd>The timecode format itself is self explanatory. Since the
703driver latches the on-time epoch directly from the second sync
704pulse, the fraction <tt>fff</tt>is always zero. Although the
705transmitted timecode includes only the year of century, the
706Gregorian year is augmented 2000 if the indicated year is less than
70772 and 1900 otherwise.</dd>
708
709<dt><tt>l</tt></dt>
710
711<dd>The leap second warning is normally space, but changes to <tt>
712L</tt> if a leap second is to occur at the end of the month of June
713or December.</dd>
714
715<dt><tt>d</tt></dt>
716
717<dd>The DST state is <tt>S</tt> or <tt>D</tt> when standard time or
718daylight time is in effect, respectively. The state is <tt>I</tt>
719or <tt>O</tt> when daylight time is about to go into effect or out
720of effect, respectively.</dd>
721
722<dt><tt>dut</tt></dt>
723
724<dd>The DUT sign and magnitude shows the current UT1 offset
725relative to the displayed UTC time, in deciseconds.</dd>
726
727<dt><tt>lset</tt></dt>
728
729<dd>Before the clock is set, the interval since last set is the
730number of minutes since the driver was started; after the clock is
731set, this is number of minutes since the time was last verified
732relative to the broadcast signal.</dd>
733
734<dt><tt>agc</tt></dt>
735
736<dd>The audio gain shows the current codec gain setting in the
737range 0 to 255. Ordinarily, the receiver audio gain control or IRIG
738level control should be set for a value midway in this range.</dd>
739
740<dt><tt>ident</tt></dt>
741
742<dd>The station identifier shows the station, <tt>C</tt> for WWV or
743<tt>H</tt> for WWVH, and frequency being tracked. If neither
744station is heard on any frequency, the station identifier shows
745<tt>X</tt>.</dd>
746
747<dt><tt>comp</tt></dt>
748
749<dd>The minute sync compare counter is useful to determine the
750quality of the minute sync signal and can range from 0 (no signal)
751to 5 (best).</dd>
752
753<dt><tt>errs</tt></dt>
754
755<dd>The bit error counter is useful to determine the quality of the
756data signal received in the most recent minute. It is normal to
757drop a couple of data bits under good signal conditions and
758increasing numbers as conditions worsen. While the decoder performs
759moderately well even with half the bits are in error in any minute,
760usually by that point the sync signals are lost and the decoder
761reverts to free-run anyway.</dd>
762
763<dt><tt>freq</tt></dt>
764
765<dd>The frequency offset is the current estimate of the codec
766frequency offset to within 0.1 PPM. This may wander a bit over the
767day due to local temperature fluctuations and propagation
768conditions.</dd>
769
770<dt><tt>avgt</tt></dt>
771
772<dd>The averaging time is the interval between frequency updates in
773powers of two to a maximum of 1024 s. Attainment of the maximum
774indicates the driver is operating at the best possible resolution
775in time and frequency.</dd>
776</dl>
777
778<p>An example timecode is:</p>
779
780<p><tt>0 2000 006 22:36:00.000 S +3 1 115 C20 6 5 66.4
7811024</tt></p>
782
783<p>Here the clock has been set and no alarms are raised. The year,
784day and time are displayed along with no leap warning, standard
785time and DUT +0.3 s. The clock was set on the last minute, the AGC
786is safely in the middle ot the range 0-255, and the receiver is
787tracking WWV on 20 MHz. Excellent reeiving conditions prevail, as
788indicated by the compare count 6 and 5 bit errors during the last
789minute. The current frequency is 66.4 PPM and the averaging
790interval is 1024 s, indicating the maximum precision available.</p>
791
792<h4>Modes</h4>
793
794<p>The <tt>mode</tt> keyword of the <tt>server</tt> configuration
795command specifies the ICOM ID select code. A missing or zero
796argument disables the CI-V interface. Following are the ID select
797codes for the known radios.</p>
798
799<table cols="6" width="100%">
800<tr>
801<td>Radio</td>
802<td>Hex</td>
803<td>Decimal</td>
804<td>Radio</td>
805<td>Hex</td>
806<td>Decimal</td>
807</tr>
808
809<tr>
810<td>IC725</td>
811<td>0x28</td>
812<td>40</td>
813<td>IC781</td>
814<td>0x26</td>
815<td>38</td>
816</tr>
817
818<tr>
819<td>IC726</td>
820<td>0x30</td>
821<td>48</td>
822<td>R7000</td>
823<td>0x08</td>
824<td>8</td>
825</tr>
826
827<tr>
828<td>IC735</td>
829<td>0x04</td>
830<td>4</td>
831<td>R71</td>
832<td>0x1A</td>
833<td>26</td>
834</tr>
835
836<tr>
837<td>IC751</td>
838<td>0x1c</td>
839<td>28</td>
840<td>R7100</td>
841<td>0x34</td>
842<td>52</td>
843</tr>
844
845<tr>
846<td>IC761</td>
847<td>0x1e</td>
848<td>30</td>
849<td>R72</td>
850<td>0x32</td>
851<td>50</td>
852</tr>
853
854<tr>
855<td>IC765</td>
856<td>0x2c</td>
857<td>44</td>
858<td>R8500</td>
859<td>0x4a</td>
860<td>74</td>
861</tr>
862
863<tr>
864<td>IC775</td>
865<td>0x46</td>
866<td>68</td>
867<td>R9000</td>
868<td>0x2a</td>
869<td>42</td>
870</tr>
871</table>
872
873<h4>Fudge Factors</h4>
874
875<dl>
876<dt><tt>time1 <i>time</i></tt></dt>
877
878<dd>Specifies the propagation delay for WWV (40:40:49.0N
879105:02:27.0W), in seconds and fraction, with default 0.0.</dd>
880
881<dt><tt>time2 <i>time</i></tt></dt>
882
883<dd>Specifies the propagation delay for WWVH (21:59:26.0N
884159:46:00.0W), in seconds and fraction, with default 0.0.</dd>
885
886<dt><tt>stratum <i>number</i></tt></dt>
887
888<dd>Specifies the driver stratum, in decimal from 0 to 15, with
889default 0.</dd>
890
891<dt><tt>refid <i>string</i></tt></dt>
892
893<dd>Ordinarily, this field specifies the driver reference
894identifier; however, the driver sets the reference identifier
895automatically as described above.</dd>
896
897<dt><tt>flag1 0 | 1</tt></dt>
898
899<dd>Not used by this driver.</dd>
900
901<dt><tt>flag2 0 | 1</tt></dt>
902
903<dd>Specifies the microphone port if set to zero or the line-in
904port if set to one. It does not seem useful to specify the compact
905disc player port.</dd>
906
907<dt><tt>flag3 0 | 1</tt></dt>
908
909<dd>Enables audio monitoring of the input signal. For this purpose,
910the speaker volume must be set before the driver is started.</dd>
911
912<dt><tt>flag4 0 | 1</tt></dt>
913
914<dd>Enable verbose <tt>clockstats</tt> recording if set.</dd>
915</dl>
916
917<h4>Additional Information</h4>
918
919<a href="refclock.htm">Reference Clock Drivers</a> <br>
920<a href="audio.htm">Reference Clock Audio Drivers</a> 
921
922<hr>
923<a href="index.htm"><img align="left" src="pic/home.gif" alt=
924"gif"></a>
925
926<address><a href="mailto:mills@udel.edu">David L. Mills
927&lt;mills@udel.edu&gt;</a></address>
928</body>
929</html>
930
Note: See TracBrowser for help on using the repository browser.