Time Nuts, I have a nice new-to-me 5372A that I'd like to use as part of a DMTD stability analysis system.  After getting a few odd glitches from a very stable source spaced 2 ns above or below the main time interval reading, I decided to run the performance verification checkout from the service manual.  The checkout calls for a BNC-T from the rear 10 MHz timebase output split to both A and B channels.  The spec is 100 ns +/- 200 ps but I am consistently getting either 101.6 or 101.4 ns in Time Interval mode on either channel.  So it seems to be high/off by exactly 1.4 ns.  The standard dev is very low (typically 0-60 ps over several 100 point runs).  Curiously, in Continuous Time Interval mode, it reads dead on (100.0 ns).  I've run the sensitivity cal because I got the 160 error (dead battery), but not the interpolator cal (I don't have the appropriate riser card and the manual calls for a specific cable).  Has anyone seen this?  I suspect maybe the interpolator board needs to be calibrated, but maybe something else?  Has anyone had experience trying to adjust the board without a riser?  Any other tests I might run to shed light on what might be causing this? Thanks,Adam MacDonaldShalimar, FL

Hi Adam, On 05/30/2018 12:26 AM, Adam MacDonald wrote: > Time Nuts, > I have a nice new-to-me 5372A that I'd like to use as part of a DMTD stability analysis system.  After getting a few odd glitches from a very stable source spaced 2 ns above or below the main time interval reading, OK, so the 5372A uses a 10811A as OCXO and then synthesize a 500 MHz coarse clock which then runs in custom high-speed chips. The interpolator is a delay-chain of DFFs. So, if you jump +/- 2 ns it is that you most likely have an issue with the start/stop and coarse-counting. Could also be that the interpolator misses one way or another. Regardless I will ponder on it. > I decided to run the performance verification checkout from the service manual.  The checkout calls for a BNC-T from the rear 10 MHz timebase output split to both A and B channels.  The spec is 100 ns +/- 200 ps but I am consistently getting either 101.6 or 101.4 ns in Time Interval mode on either channel.  So it seems to be high/off by exactly 1.4 ns. Consider that you need to swap the cables to verify that it is static with the setup and does not follow the cables. Also, there can be trigger voltage offsets which may be the cause, so that may be due to other reasons. >  The standard dev is very low (typically 0-60 ps over several 100 point runs).  Curiously, in Continuous Time Interval mode, it reads dead on (100.0 ns).  I've run the sensitivity cal because I got the 160 error (dead battery), but not the interpolator cal (I don't have the appropriate riser card and the manual calls for a specific cable).  Has anyone seen this?  I suspect maybe the interpolator board needs to be ca > librated, but maybe something else?  Has anyone had experience trying to adjust the board without a riser?  Any other tests I might run to shed light on what might be causing this? Let me ponder on this one. I don't think it is the interpolator boards, sure they might need calibration, but that should not be the issue here, you miss a cycle in the coarse counter on their the start or stop channel. That should be the focus. Cheers, Magnus

-------- In message <c10d6b2a-79c3-931d-1439-347fce06ff81@rubidium.dyndns.org>, Magnus D anielson writes: >Consider that you need to swap the cables to verify that it is static >with the setup and does not follow the cables. Run long series of measurements with the HP5372A OCXO free-running and with it slaved to the DUT clock. Any imperfection in the interpolator will show up as different noise characteristics in the two results. There is a test in the service manual which will give you a good indication how well tuned the interpolator is, run that before touching anything, and don't attempt to tune the interpolator unless you can follow the procedure in the manual *exactly*. PS: I do not recall it being mentioned, but the 1 MOhm input pods are horribly unstable compared to the 50 Ohm input pods, which are basically just a stripline and a BNC connector. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence.