Bob The Phase noise of that GPDSO seems quite high and some form of filtering (bandpass or low pass) of its output may be necessary to reduce the resultant jitter when connected to the 5370A input.
There are commercial TDCs that claim that sort of performance (~10ps) but a comprehensive test of their actual performance appears as yet unavailable. HP used to sell a TDC with 5ps resolution but detailed specs are unavailable. Higher resolution and lower noise should be achievable but would require a lot of number crunching for each measurement. Bruce > On 05 March 2017 at 11:56 Bob Stewart <[email protected]> wrote: > > Hi Bruce, > > To me, it seems that the reasoning must be circular. If I measure three > units against each other with one piece of test equipment, then the one that > has the best noise performance is the one that is not being measured when the > ADEV plot is the worst. At least that's the way I see it. > > I've attached a screenshot of the Timelab plot for the GPSDO and the Cs. > What I consider the reference plot for my GPSDO is here on page 2: > http://ae6rv.com/GFS/Documents/Brochure.pdf . Note that that link downloads > a pdf file from my website. The unit under test there was a pre-production > unit. However, nothing of substance changed to the hardware, other than a > different OCXO. That unit did not output a proper 1PPS. The changes to the > production units takes a 1PPS pulse generated in the PIC and gates that with > the OCXO to generate a proper 1PPS signal. The tests on the 5370 use one > GPSDO to generate both the 1PPS reference and the 10MHz that clocks the 5370. > I don't see any particular difference in the results when a different GPSDO > drives the tests. I do see a pretty noisy plot up to about 80s when using > the 10811 in the 5370 to drive the clock. > > "An inexpensive replacement for the 5370A/B with an rms measurement noise > for short time intervals of around 10ps or better would probably be useful > for such measurements." > > What would such an "inexpensive" device be? What's inexpensive for you > would probably be way out of my reach. > > Bob > > > > > --------------------------------------------- > From: Bruce Griffiths <[email protected]> > To: Bob Stewart <[email protected]>; Discussion of precise time and > frequency measurement <[email protected]> > Sent: Saturday, March 4, 2017 4:28 PM > Subject: Re: [time-nuts] Cs standard stability > > Bob > > Your reasoning appears somewhat circular. > A typical 5370A may have around 35ps or so rms noise when measuring short > time intervals. This translates to an equivalent ADEV of around 4.3E-11 for > Tau = 1s due to the 5370A noise alone. The 5370A in particular may have > issues with DNL errors of the order of 100ps or so. > > The actual noise of a particular 5370A depends on how well the 200MHz > multiplier chain is aligned, whether a couple of changes to the original > circuit were implemented and the noise of its 10MHz frequency reference. > > Your setup is probably too noisy to be confident that the PRS45A doesn't > meet spec at Tau= 1s. > > For Tau > 100s or so the noise level of your setup should be small enough > that detection of deviations of the PRS -45A from its spec can be reliably > detected. > > Attaching plots of your GPDSO ADEV and the performance of your particular > 5370A would be helpful. > > An inexpensive replacement for the 5370A/B with an rms measurement noise > for short time intervals of around 10ps or better would probably be useful > for such measurements. > > Bruce > > On 05 March 2017 at 05:58 Bob Stewart <[email protected] > mailto:[email protected] > wrote: > > > > > > Hi Bruce, > > I'm aware that the GPSDO is better at low tau than the 5370. This is > born out by its low tau ADEV being right at the specs for the 5370, as well > as by measurements done some time ago on a Timepod. Unfortunately, I don't > have a Timepod, so I have to make some guesses based on what I can see; > limited as that is by the 5370. But, given that the GPSDO tests validate the > 5370, when I see an ADEV that is worse than the 5370's specs (i.e. worse than > the result of the GPSDO test), I take that as an indication that the > additional noise is from the DUT. Of course, I don't know anything more than > that. Correct? > > > > Bob > > > > > > From: Bruce Griffiths <[email protected] > mailto:[email protected] > > > To: Bob Stewart <[email protected] mailto:[email protected] >; Discussion > of precise time and frequency measurement <[email protected] > mailto:[email protected] > > > Sent: Saturday, March 4, 2017 3:44 AM > > Subject: Re: [time-nuts] Cs standard stability > > > > Bob > > > > That doesn't seem inconsistent with the specs given the contribution of > the 5370A to the measured ADEV is likely to be of the same order as the > PRS45A spec. The ADEV of the GPSDO also needs to be taken into account. The > contribution of the 5370A to ADEV will gradually become insignificant > compared to the PRS45A specs as Tau increases. To achieve an accurate > measurement of ADEV for Tau=1s you will need something better than the 5370A > together with a low ADEV source (for Tau = 1s). > > > > Bruce > > > On 04 March 2017 at 18:26 Bob Stewart <[email protected] > mailto:[email protected] > wrote: > > > > > > > > > I hooked up my PRS-45A yesterday, just to do some ADEV comparisons > with my GPSDO. Unfortunately, the GPSDO seems to be better by almost two > points at all tau as measured on my HP 5370A and plotted by Timelab. So, can > I infer from that that the OCXO in the PRS has a problem, or could it be > something as stupid as a bad cap? Or do I just not understand the specs and > the PRS isn't up to my expectations? The PRS docs show the unit as 3E-11 at > 1s tau, and I'm seeing about 4.25E-11 at 1s, so maybe it's the latter. > > > > > > NOTE: The test setup for this is 1PPS from a well-aged GPSDO to the > 5370's start input, and 10MHz from either a different GPSDO or the PRS to the > stop input. > > > > > > As best as I can tell looking through the holes in the case (still > running the tests), the PRS has an MTI 250-0827 5MHz OCXO. I've got one of > the Lucent KS units that I don't care much about, so maybe the MTI 260-0624 > 5MHz in that will work? Both are 5MHz, but I don't know whether both are 12V > at this point. And, of course, I don't know whether making the swap to an > MTI 260 OXCO will benefit the PRS. > > > Any comments or educated speculations are welcomed. > > > > > > Bob - AE6RV > > > _______________________________________________ > > > time-nuts mailing list -- [email protected] mailto:[email protected] > > > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > > > and follow the instructions there. > > > > > > > > > _______________________________________________ > > time-nuts mailing list -- [email protected] mailto:[email protected] > > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > > and follow the instructions there. > > > _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
