[time-nuts] Fury - Rubidium

[Brian Kirby]

I read the article on PID on Wikipedia last night.  I do not fully 
understand it, but I see/learning some of the relationship.

I did a test on the FRS-C rubidium.  The average frequency was 10 000 
000.0025 hertz at the rubidium 10 turn dial dial setting of 255, and the 
control voltage out of the pot was 1.7900 volts.  I recorded the 
frequency for a while and then changed frequency to see how long it took 
to get there.  I changed the dial setting to 516   (3.5800V) and it took 
8 seconds for the rubidium to change frequency and settle on a average 
frequency of 10, 000 000.0131 hertz.

I did another test and the rubidium dial setting was 000 for a control 
voltage of 0.068V and the average frequency was 9 999 999.9933 hertz. 
The dial setting was changed to 721 for a control voltage of 4.9999V and 
the average frequency was 10 000 000.0216 hertz

The measurements were taken with a HP5370B time interval counter 
referenced to a HP5065A rubidium oscillator.  The data was recorded 
using a ProLogix GPIB adapter.  The data was recorded in 10 minute 
intervals with the data coming in at one measurement a second.  When the 
frequency was changed, I allowed 20 minutes between the recordings.

Based on the above measurements, Said, can you recommend some starting 
point for the DAC Gain, EFC Scale, and the EFC Damping ?

Also from previous measurements, I know this particular rubidium was at 
  9x10E-11 at 0.1 sec, 1.8x10E-11 at 1 second, 5x10E-12 at 10 seconds, 
1.5x10E-12 at 100 seconds, 7x10E-13 for 1000 seconds, and 2.5x10E-13 for 
10000 seconds - running on a Shera GPS controller - which the PIC was 
modified for this rubidium (it was changed from a 30 second time 
interval measurement to 120 seconds, and Shera changed the sensitivity 
of the PIC to 1X10-9/volt).

Thanks to Don and Scott for the ops info.

Thanks

Brian KD4FM





On 7/27/2010 2:57 PM, SAIDJACK at aol.com wrote:
> Hi guys,
>
> it may help to increase DAC gain to get faster recovery times from "bumps"
> etc.
>
> On an OCXO, the frequency recovery from an upset should happen within a
> couple of minutes, definitely less than 15 minutes to achieve frequency  lock.
>
> The phase recovery (to 0ns offset) may take a couple of hours to do.
>
> If it takes a very long time to recover, then I think increasing the DAC
> gain, or alternatively the EFCS and PHASECO together may help.
>
> Wikipedia has some good instructions on how to optimize PID type controller
>   gains to get the fastest response with minimal noise...
>
> Also, please make sure to disable temperature compensation when using the
> external source, unless a thermistor is connected to the board, sensing the
> Rb  temperature. Otherwise the temperature compensation may add noise due to
> it  scaling the gain to huge values due to the missing thermistor.
>
> bye,
> Said
>
>
> In a message dated 7/27/2010 09:58:41 Pacific Daylight Time,
> true-cal at swbell.net writes:
>
> My  experience is very similar to Scott's. I ran many hours with both an
> LPRO-101
> and FE-5680A. The disciplining behavior and Fury settings  were the same
> for
> either Rb. My biggest disappointment was the  recovery time due to various
> common
> or intentional bumps or especially,  after power loss. I also had to let
> the
> "system" settle in for a week  before acceptable tracking smoothed out. Any
> long
> term slope to  the EFC trace (gpscon) caused excessive hunting and this
> didn't
> settle down until the Rb was VERY stable. My gpscon TI and stddev  was
> virtually
> the same as Scott's if I had EFCS set to 1.0 to 1.5 but  recovery was
> unacceptable (maybe 24-hours) so I usually ran at 2.0 or 3.0  with
> slight degrading of stddev to around 3.2. This EFCS setting  allowed a much
> better settling time around 3-hours.
>
> DACG=  1000
> EFCS = 2 to 3
> EFCD = 50 (25 allows little better settling  time)
> PHASECO = 15 (I favor 10 Mhz over  PPS)
> Regards...
> Don
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>