Chris, I think I've figured it out. Here is the key phrase, I think: "The EFC corrects the frequency to give the proper 6800 HEX over the following samples, but the phase shift from the previous error during the correction period remains."
Let's say that the frequency drifts off less than one cycle per second. Let's further say that it is a phase angle of 30 degrees. If we correct the frequency so that subsequently the frequency is spot on, the phase of our clock is still 30 degrees off. Let's say that that happens 11 more times. We are now a total of 360 degrees off in phase, but our frequency is exactly correct, and has been within our error tolerance through all 12 total error cycles. Without a phase correction our clock will wander around in phase, and thus in frequency, over the long term, though it will stay within tolerance over the short and medium term. If our clock was slightly fast this cycle, then we need to slow it down by just that much, i.e. that phase angle, on the next timing cycle. At least that's how I understand it. I need to think on it some more. Bob ----- Original Message ----- > From: Chris Howard <ch...@elfpen.com> > To: mccor...@ptialaska.net; Discussion of precise time and frequency > measurement <time-nuts@febo.com> > Cc: Bob Stewart <b...@evoria.net> > Sent: Thursday, June 13, 2013 8:58 PM > Subject: Re: [time-nuts] Fw: New to list and GPSDO questions > > > I'm with Bob in that I don't really understand this description. > > And what are we trying to be in phase with? the 1PPS? > Is it possible to reproduce the actual phase of the > clock in the GPS birds once it has been through my GPS > and come out as 1PPS? > > I picture the EFC steering the car straight down the road. > And you want to adjust not just the drift to the right but > steer back to the center of the lane. I think I get > that part. I don't get who is setting the lane > and why I should care. > > Do the statistics (ADEV, etc.) show that this is an improvement? > If I'm doing my slow zigzag down the center line or the > right shoulder, again, do I care? > > Chris > w0ep > > > > > > On 6/13/2013 5:51 PM, Richard H McCorkle wrote: >> Hi Bob, >> >> The VE2ZAZ controller is a frequency locked loop that keeps the source >> on frequency but over time the phase drifts, as past corrections are >> not compensated for. When the system is stable every 16 seconds the >> 16-bit counter rolls over 2441 times with an extra 26624 counts (6800 >> HEX). The previous count is subtracted from the current count, the >> difference from 6800 HEX is used to update the EFC to correct the >> frequency, and the current count is stored as the previous count for >> the next sample. When the current difference is not exactly 6800 HEX >> then a phase shift of 6.25ns has occurred over the sample period for >> each count of the difference from 6800 HEX. The EFC corrects the >> frequency to give the proper 6800 HEX over the following samples, but >> the phase shift from the previous error during the correction period >> remains. >> If you add a little code to provide a correction history you can >> add phase correction to the VE2ZAZ controller as well. Once lock is >> established store the sample count as a phase reference count. Add >> 6800 HEX to the reference count every update and store the result as >> the next reference value. Use the difference between the current and >> previous samples from 6800 HEX to coarse correct the frequency as >> currently done, but add the difference between the current count and >> reference count to the EFC correction as a fine phase correction. If >> hold or unlock occurs disable the phase correction routine until lock >> is re-established. Then store a new reference and restart the phase >> correction process. By updating the reference every update it provides >> a phase history so the accumulated count error over time can be >> removed and the phase of the source can be stabilized. >> >> Richard >> >>> >>> >>> Atilla, >>> >>> Isn't the VE2ZAZ circuit functionally equivalent to your example > 3? Granted, he's >>> not picking the 10 millionth transition and checking its phase >>> difference to the reference, but I've only got a 1PPS reference > with a >>> 1uS or so jitter from pulse to pulse. Bert is averaging over 16 >>> seconds, and creating a PWM signal to drive an integrator (simulating a > DAC), which >>> will drive a Trimble 34310-T. And like I mentioned >>> earlier, I just like the way Bert did it. It has a kind of elegance >>> that appeals to my inner hacker. =) >>> >>> >>> Bob - AE6RV >>> >>> >>> >>> ----- Original Message ----- >>>> From: Attila Kinali <att...@kinali.ch> >>>> To: Bob Stewart <b...@evoria.net>; Discussion of precise time > and frequency >>>> measurement <time-nuts@febo.com> >>>> Cc: >>>> Sent: Thursday, June 13, 2013 3:39 PM >>>> Subject: Re: [time-nuts] New to list and GPSDO questions >>> snip >>>> [3] Describes how to use a clock >>> synchronizer to build a GPSDO. Probably >>>> not the easiest and not the cheapest way, but definitly one with a > very >>>> low parts count. >>> snip >>>> [3] The AD9548 as a GPS Disciplined Stratum 2 Clock, by Gentile, > 2009 >>>> > http://www.analog.com/static/imported-files/application_notes/AN-1002.pdf >>> _______________________________________________ >>> 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. >>> >> >> >> _______________________________________________ >> 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. >> >> > _______________________________________________ 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.