Hi, all I've done some experiments on my 53230A, and I've come across an issue I think could be worth knowing about. I have not tested this on other 53230A's - if anyone has a 53230A and some spare time, I'd be grateful if someone could try to replicate. In a nutshell, the first frequency measurement following an INIT is biased. The magnitude of the bias is dependant on measurement mode (CONT/RCON), and gate time. It is worth noting that every READ implicitly calls INIT - so any software collecting a time series using repeated calls to READ to fetch single readings will give biased results. Not for the first sample in the series, but all of them. As an example, the last time series I collected this way using RCON and 0.1s gatetime gave an average frequency error of -5.2e-10. A writeup is here: http://www.efos3.com/HPAK53230A-1.html Best regards, Ole

Ole, Is this part of the behavior that the Norwegian NMI was presenting at a poster session at EFTF 2016 in York? I remember they borrowed your 53230A for that work. Cheers, Magnus On 11/23/2017 09:05 PM, Ole Petter Ronningen wrote: > Hi, all > > I've done some experiments on my 53230A, and I've come across an issue I > think could be worth knowing about. > > I have not tested this on other 53230A's - if anyone has a 53230A and some > spare time, I'd be grateful if someone could try to replicate. > > In a nutshell, the first frequency measurement following an INIT is biased. > The magnitude of the bias is dependant on measurement mode (CONT/RCON), and > gate time. > > It is worth noting that every READ implicitly calls INIT - so any software > collecting a time series using repeated calls to READ to fetch single > readings will give biased results. Not for the first sample in the series, > but all of them. > > As an example, the last time series I collected this way using RCON and > 0.1s gatetime gave an average frequency error of -5.2e-10. > > A writeup is here: http://www.efos3.com/HPAK53230A-1.html > > Best regards, > Ole > _______________________________________________ > 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. >

Hi Magnus No, I believe this is a separate issue. The data for that paper was collected gap free, and was not affected by this particular issue. It may be related somehow - as presented in that paper, data collected in CONT mode shows a slope of tau^-1/2 on ADEV. Data collected gap free in RCON-mode (i.e. not by repeated calls to READ) shows the expected slope of Tau^-1. The issue I point out here is that if data is collected using repeated calls to "READ", even data collected in RCON-mode shows a slope of tau^-1/2, in addition to a frequency bias. The magnitude of the frequency bias is gate time dependent. Ole On Fri, Nov 24, 2017 at 2:45 PM, Magnus Danielson < magnus@rubidium.dyndns.org> wrote: > Ole, > > Is this part of the behavior that the Norwegian NMI was presenting at a > poster session at EFTF 2016 in York? > > I remember they borrowed your 53230A for that work. > > Cheers, > Magnus > > > On 11/23/2017 09:05 PM, Ole Petter Ronningen wrote: > >> Hi, all >> >> I've done some experiments on my 53230A, and I've come across an issue I >> think could be worth knowing about. >> >> I have not tested this on other 53230A's - if anyone has a 53230A and some >> spare time, I'd be grateful if someone could try to replicate. >> >> In a nutshell, the first frequency measurement following an INIT is >> biased. >> The magnitude of the bias is dependant on measurement mode (CONT/RCON), >> and >> gate time. >> >> It is worth noting that every READ implicitly calls INIT - so any software >> collecting a time series using repeated calls to READ to fetch single >> readings will give biased results. Not for the first sample in the series, >> but all of them. >> >> As an example, the last time series I collected this way using RCON and >> 0.1s gatetime gave an average frequency error of -5.2e-10. >> >> A writeup is here: http://www.efos3.com/HPAK53230A-1.html >> >> Best regards, >> Ole >> _______________________________________________ >> time-nuts mailing list -- time-nuts@febo.com >> To unsubscribe, go to https://www.febo.com/cgi-bin/m >> ailman/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/m > ailman/listinfo/time-nuts > and follow the instructions there. >

Although, as I read the paper again, it is possible that the frequency bias in the 10 second CONT-mode measurements could come from this effect. That particular series (and only that) was collected using TimeLab. I supplied them with a custom compiled driver that does not use repeated calls to READ - and therefore does not exhibit this particular bias. I always assumed that this driver was used to collect that series, but the paper does not state this explicitly so I might be wrong. The bias I am pointing to is gate time dependant, the bias getting smaller with longer gate times. It is possible it will be at the E-15 level at 10 second gatetime, like in the paper. Ole On Fri, Nov 24, 2017 at 2:45 PM, Magnus Danielson < magnus@rubidium.dyndns.org> wrote: > Ole, > > Is this part of the behavior that the Norwegian NMI was presenting at a > poster session at EFTF 2016 in York? > > I remember they borrowed your 53230A for that work. > > Cheers, > Magnus > > > On 11/23/2017 09:05 PM, Ole Petter Ronningen wrote: > >> Hi, all >> >> I've done some experiments on my 53230A, and I've come across an issue I >> think could be worth knowing about. >> >> I have not tested this on other 53230A's - if anyone has a 53230A and some >> spare time, I'd be grateful if someone could try to replicate. >> >> In a nutshell, the first frequency measurement following an INIT is >> biased. >> The magnitude of the bias is dependant on measurement mode (CONT/RCON), >> and >> gate time. >> >> It is worth noting that every READ implicitly calls INIT - so any software >> collecting a time series using repeated calls to READ to fetch single >> readings will give biased results. Not for the first sample in the series, >> but all of them. >> >> As an example, the last time series I collected this way using RCON and >> 0.1s gatetime gave an average frequency error of -5.2e-10. >> >> A writeup is here: http://www.efos3.com/HPAK53230A-1.html >> >> Best regards, >> Ole >> _______________________________________________ >> time-nuts mailing list -- time-nuts@febo.com >> To unsubscribe, go to https://www.febo.com/cgi-bin/m >> ailman/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/m > ailman/listinfo/time-nuts > and follow the instructions there. >