[time-nuts] Is Timelab with a Prologix-Eth and a PM6680 - working ?

[Azelio Boriani]

Yes, you are right: my example was, for this reason, limited to a couple
hundred nS. Usually testing PPSes or 10MHz sources leads to very small time
intervals so the resolution of the interpolator dominates. For time
intervals beyond the nS figures it is better to consider a stable
reference. Another problem of the start-stop mechanism is the lack of the
negative time interval capability of the PM6680/1: you can swap A and B
inputs on the fly but you loose one (or more) samples. I have seen my
PM8861 outputting wrong samples as the PPSes cross each other: usually a
4.999999999E-01 seconds sample is output before the correct reading
resumes. Very surprising was the correct negative output (I suppose when
the time interval is under the interpolator window) got when the two PPSes
were very very slowly crossing: a long list of negative values before the
usual 9.9999999E-01 output.

On Tue, Jul 3, 2012 at 9:42 PM, Bob Camp <lists at rtty.us> wrote:

> Hi
>
> That works out fine *if* the edges are close to each other.
>
> Consider one edge arriving a half second before the other one. Then the
> reading is off by a nanosecond for each 2 ppb the reference is off
> frequency. If you have a TCXO as the reference for your counter, you likely
> will be further off than that. There are a *lot* of surplus counters that
> originally went into setups with external references. They have (at best) a
> loose TCXO for a standard when the reference is not present.
>
> Bob
>
> On Jul 3, 2012, at 3:32 PM, Azelio Boriani wrote:
>
> > When used as a time interval counter, the PM6680 (as well as other
> > counters) counts the reference oscillator cycles + a residual got from an
> > interpolator. Suppose you have to measure 234nS: with a 100MHz oscillator
> > you count 23 cycles (the 100MHz is 10nS cycle) and then you are left with
> > an additional 4nS. To account for the 4nS, various digital and analog
> > techniques are available. Your 100MHz has to be stable at least only for
> > the time you make the measurement, 234nS in this example. For this short
> > time interval almost every oscillator will do and the final interpolator
> > doesn't take any advantage from a stable oscillator: must be well
> designed
> > and stable by itself. In short: if used as a frequency counter, yes, you
> > have to feed a good reference at the EXT REF, when used as a time
> interval
> > counter the good EXT REF is not so important, it is mandatory to use the
> > good reference at one of the two counting inputs. Don't forget that with
> > the PM6680 you can directly measure fractional frequency by using the
> MATH
> > menu and the K=, L= and M= keys and smooth the measure by the STAT key.
> > Later on you will discover that more resolution is necessary and you will
> > be interested in counters such as the HP5370B (20pS) or the SR620 (25pS).
> > Take a look at the PM6681 (50pS) too.
> >
> > On Tue, Jul 3, 2012 at 7:38 PM, cfo <xnews3 at luna.dyndns.dk> wrote:
> >
> >> On Tue, 03 Jul 2012 19:08:56 +0200, Azelio Boriani wrote:
> >>
> >>> Yes, you can ask here about these things. Anyway, the best is to use
> the
> >>> TBolt as a reference (not necessarily on the external ref of the
> >>> counter) and test another oscillator or GPSDO. For example: feed the
> >>> TBolt PPS to A, the GPSDO_under_test PPS to B and measure the stability
> >>> of the time interval using TIME A-B. Ideally it should be a constant
> >>> value (even if not exactly 0, but constant). The phase data (the time
> >>> interval errors) can be collected and an Allan deviation computed.
> >>
> >> Azelio , first ... Thanx for walking me through this.
> >>
> >> I'll try the above later on.
> >> But what you're saying is that when used as a TIC A->B , i don't need
> the
> >> TBolt on ext-ref ? , or do i need the TBolt on both Ext-Ref & on Chan-A
> ?
> >>
> >> If i don't use the TBolt on Ext-Ref , won't i get hit by having a
> Std-Osc
> >> installed ?
> >>
> >> I have a GPS Antenna-Splitter , another TBolt (just need to assemble the
> >> PSU) , and also a X72 Rubi & a few FEI-5680's.
> >>
> >> So i could try those when i'm a bit more familiar with the Counter &
> >> program.
> >>
> >>
> >>>
> >>> The default sampling time is 200mS
> >>>
> >> Ahhh ... Thanx now i know why my 1hr Trace Duration was so fast.
> >> I assumed that Timelab would set the sampling interval written in
> Timelab
> >> via GPIB. But i'll just select 1sec manually.
> >>
> >>
> >>
> >>> So, to try your setup (TBolt on EXT REF and A): first inform the
> TimeLab
> >>> that you have 1 second frequency samples, then set the sampling time to
> >>> 1 second (use FUNCTION button on the PM6680) and retry collecting
> >>> data.In this case the predicted uncertainty is 0.005 (5E-10) and sholud
> >>> average out to 5E-13 in 1000 seconds taking into consideratin that the
> >>> full 12 digit resolution is available at the PM6680 GPIB (that is, the
> >>> GPIB has more digits than the display can show, see page 11-18 of the
> >>> manual).
> >>>
> >>
> >> I will try this now , and i assume i should then select Frequency in the
> >> Data Type - In TimeLab , as i don't use A->B measurements .. Or ?
> >>
> >>
> >> I'll also dig into - How to calculate measurement errors , as you
> >> describe above.
> >>
> >>
> >> Regards
> >> CFO - Tnut-Beginner
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >> 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.
> >>
> > _______________________________________________
> > 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.
>
>
> _______________________________________________
> 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.
>