Re: [time-nuts] Thermal Compensation: Digital vs Analog
On 26/02/14 17:09, Bob Stewart wrote: I've been experimenting with digital thermal compensation on my GPSDO. The results have been favorable for a 14 bit dithered PWM-based DAC, but leaves a bit to be desired in the big picture. And it takes up a lot of program bytes on my PIC.. What's the general consensus on this? Should thermal compensation be completely analog? Considering that time-constants can be fairly long, there is a benefit in having at least that part of the integrator in digital domain. There are benefits of both approaches. Resolution of input and output can be a problem for the digital solution. Cheers, Magnus ___ 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] Thermal Compensation: Digital vs Analog
I've been experimenting with digital thermal compensation on my GPSDO. The results have been favorable for a 14 bit dithered PWM-based DAC, but leaves a bit to be desired in the big picture. And it takes up a lot of program bytes on my PIC.. What's the general consensus on this? Should thermal compensation be completely analog? Bob - AE6RV ___ 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.
Re: [time-nuts] Thermal Compensation: Digital vs Analog
On Wed, 26 Feb 2014 08:09:44 -0800 (PST) Bob Stewart b...@evoria.net wrote: I've been experimenting with digital thermal compensation on my GPSDO. The results have been favorable for a 14 bit dithered PWM-based DAC, but leaves a bit to be desired in the big picture. And it takes up a lot of program bytes on my PIC.. What's the general consensus on this? Should thermal compensation be completely analog? Out of pure interest. Could you elaborate what results you got? Ie. what does your GPSDO look like? How do you compensate for the temperature coefficient? How much did that improve performance compared to non-compensated operation? Did you try any other approaches? Why? Why not? Yes, i'm a curious mind :-) Thanks in advance Attila Kinali -- The trouble with you, Shev, is you don't say anything until you've saved up a whole truckload of damned heavy brick arguments and then you dump them all out and never look at the bleeding body mangled beneath the heap -- Tirin, The Dispossessed, U. Le Guin ___ 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.
Re: [time-nuts] Thermal Compensation: Digital vs Analog
Hi Atilla, The GPSDO is VE2ZAZ's circuit with new code. It detects phase crossings to change the DAC, so it has a phase crossing for every update. I'm working on a TIC design but haven't started on the hardware. In the interim, I hooked up an LM34 thermistor and have been playing with that. In the 8 hour plot below, there are no frequency updates, only temperature updates. I've tried a rolling average, but it doesn't smooth it enough, so I'll have to try hysteresis next. The orange/green/blue line is the DAC. The red line is the thermistor. The cyan smear is the phase plot of 1PPS from my Adafruit (MT3339) against the OCXO (Trimble 34310-T). The units on the right correspond to the temperature - 100 degrees at the EFC divider directly beneath the OCXO. Also, they correspond to the wrapped phase, where 0-20 is 0-360 degrees. The OCXO is limited to a swing of about +/- 1.1Hz at the moment. http://www.evoria.net/AE6RV/TempComp/GPSDO.png In the plot below is the ADEV. Hopefully it's self explanatory. The phase has varied a bit more than 180 degrees during the test. http://www.evoria.net/AE6RV/TempComp/ADEV.png The biggest influence on temperature seems to be the low quality divider resistors in the EFC divider chain. I have new low TempCo resistors but I couldn't resist playing with these first. Without temperature compensation, phase would vary through about one cycle every change to the red (thermistor) line. Bob - AE6RV From: Attila Kinali att...@kinali.ch To: Bob Stewart b...@evoria.net; Discussion of precise time and frequency measurement time-nuts@febo.com Sent: Wednesday, February 26, 2014 10:23 AM Subject: Re: [time-nuts] Thermal Compensation: Digital vs Analog On Wed, 26 Feb 2014 08:09:44 -0800 (PST) Bob Stewart b...@evoria.net wrote: I've been experimenting with digital thermal compensation on my GPSDO. The results have been favorable for a 14 bit dithered PWM-based DAC, but leaves a bit to be desired in the big picture. And it takes up a lot of program bytes on my PIC.. What's the general consensus on this? Should thermal compensation be completely analog? Out of pure interest. Could you elaborate what results you got? Ie. what does your GPSDO look like? How do you compensate for the temperature coefficient? How much did that improve performance compared to non-compensated operation? Did you try any other approaches? Why? Why not? Yes, i'm a curious mind :-) Thanks in advance Attila Kinali -- The trouble with you, Shev, is you don't say anything until you've saved up a whole truckload of damned heavy brick arguments and then you dump them all out and never look at the bleeding body mangled beneath the heap -- Tirin, The Dispossessed, U. Le Guin ___ 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.
Re: [time-nuts] Thermal Compensation: Digital vs Analog
Hi Bob, better use an FIR. Your rolling average didn't smooth out enough because it doesn't have a cutoff low enough. Hysteresis is not going to help here that I know of. Cheers, D. On Wed, Feb 26, 2014 at 6:05 PM, Bob Stewart b...@evoria.net wrote: Hi Atilla, The GPSDO is VE2ZAZ's circuit with new code. It detects phase crossings to change the DAC, so it has a phase crossing for every update. I'm working on a TIC design but haven't started on the hardware. In the interim, I hooked up an LM34 thermistor and have been playing with that. In the 8 hour plot below, there are no frequency updates, only temperature updates. I've tried a rolling average, but it doesn't smooth it enough, so I'll have to try hysteresis next. The orange/green/blue line is the DAC. The red line is the thermistor. The cyan smear is the phase plot of 1PPS from my Adafruit (MT3339) against the OCXO (Trimble 34310-T). The units on the right correspond to the temperature - 100 degrees at the EFC divider directly beneath the OCXO. Also, they correspond to the wrapped phase, where 0-20 is 0-360 degrees. The OCXO is limited to a swing of about +/- 1.1Hz at the moment. http://www.evoria.net/AE6RV/TempComp/GPSDO.png In the plot below is the ADEV. Hopefully it's self explanatory. The phase has varied a bit more than 180 degrees during the test. http://www.evoria.net/AE6RV/TempComp/ADEV.png The biggest influence on temperature seems to be the low quality divider resistors in the EFC divider chain. I have new low TempCo resistors but I couldn't resist playing with these first. Without temperature compensation, phase would vary through about one cycle every change to the red (thermistor) line. Bob - AE6RV From: Attila Kinali att...@kinali.ch To: Bob Stewart b...@evoria.net; Discussion of precise time and frequency measurement time-nuts@febo.com Sent: Wednesday, February 26, 2014 10:23 AM Subject: Re: [time-nuts] Thermal Compensation: Digital vs Analog On Wed, 26 Feb 2014 08:09:44 -0800 (PST) Bob Stewart b...@evoria.net wrote: I've been experimenting with digital thermal compensation on my GPSDO. The results have been favorable for a 14 bit dithered PWM-based DAC, but leaves a bit to be desired in the big picture. And it takes up a lot of program bytes on my PIC.. What's the general consensus on this? Should thermal compensation be completely analog? Out of pure interest. Could you elaborate what results you got? Ie. what does your GPSDO look like? How do you compensate for the temperature coefficient? How much did that improve performance compared to non-compensated operation? Did you try any other approaches? Why? Why not? Yes, i'm a curious mind :-) Thanks in advance Attila Kinali -- The trouble with you, Shev, is you don't say anything until you've saved up a whole truckload of damned heavy brick arguments and then you dump them all out and never look at the bleeding body mangled beneath the heap -- Tirin, The Dispossessed, U. Le Guin ___ 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.
Re: [time-nuts] Thermal Compensation: Digital vs Analog
If you can understand the temperature effects and can model them accurately and you can measure temperatures and your DAC steps are small enough, then digital compensation can be perfect. But you are unlikely to meet all those conditions. In theory if the problem is that the voltage diver's ratio is a function of temperature then you can epoxy the LM34 to the divider and adjust the output of the DAC based on the current divider ratio. But in the real word you don't know the exact function or temperature and the DAC might have larger steps. I think the best plan is to reduce the source of the error, (the analog fix) Then if there is still any error source you can measure and model to that too. On Wed, Feb 26, 2014 at 8:09 AM, Bob Stewart b...@evoria.net wrote: I've been experimenting with digital thermal compensation on my GPSDO. The results have been favorable for a 14 bit dithered PWM-based DAC, but leaves a bit to be desired in the big picture. And it takes up a lot of program bytes on my PIC.. What's the general consensus on this? Should thermal compensation be completely analog? Bob - AE6RV ___ 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. -- Chris Albertson Redondo Beach, California ___ 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.