PPS jitter and 6683H

[xl600]

I have a 6683H connected to a PPS source.  The PPS has a good long term (> about 1 minute) accuracy (About 20us) but a terrible short term stability (+/-250us).

When I use the 6683 to time stamp a test reference with a very high accuracy (1us) I get jitter in the resulting timestamps from the 6683 which looks about as bad as the incoming short term PPS jitter.

Is there a way to smooth out the 6683 so it adjusts more to the longer term stability of the PPS signal rather than the second by second jitter?

I don't have a better incoming PPS source so I'm looking for a configuration/software approach if one exists.

Thanks,

XL600

[FinchTrain]

Hi xl600,

Would it be possible to post a screenshot of your block diagram (or the VI itself) where you are timestamping the signal?

Based on your description, it sounds like after running for over a minute, the jitter does not settle down with the PPS signal and remains in the 250u range as opposed to the 20u? I just want to verify that I am understanding the issue fully.

One final question, where is the PPS signal coming from and what accounts for the jitter? Are there any steps we can take to possible improve the short term stability?

Regards,

Finch Train

[xl600]

The VI setup is a bit complex (Was using a RT loop from a DAQmx card outputting a 1s rising edge to a PXI trigger line as my test input to the 6683).  I would have expected a very steady 1s interval on the 6683H timestamps under any circumstances.  I'm only reading the timestamps using the Max test panel (Not a vi)

When I sync my 6683 to a very steady PPS, the test timestamps are very steady.

When I sync my 6683 to my real PPS (With the jitter) there's essentially the same jitter in the test timestamps.  If anything I would have thought the steadyness of the test timestamps would drift around slowly, but not at the level of the PPS jitter.  More to a gentle precession around the jitter long term stability.  It's almost like the 6683H is simply resetting its onboard clock to the second boundary on each PPS rising edge rather than drifting the clock like NI-TimeSync (or NTP) would.  I sincerely hope that isn't the case because that would guarentee discrete jumps in timestamp accuracy which could skew even short term measurements.

As for a better PPS.  The PPS is just a vxWorks CPU outputting a GPIO.  vxWorks is running at a 4kHz rate thus the +/-250us jitter.  On average though, the wakeup is spot-on to within about 20us (Interrupts etc).  I don't have control over that design.

[FinchTrain]

Hi,

The timing card adjusts its oscillator based on the PPS signal, so it makes sense that you are seeing the jitter. Essentially, because you are using the PPS as a hardware reference in your system, the system adjusts to it for the sake of synchronizing with external tasks associated with it.

Regards,

Finch Train

[xl600]

The oscillator itself is being adjusted?  In another experiment I did, I found that if I output my own PPS using the MAX generate signal (PPS), that PPS drifts away from the incoming PPS as if the oscillator isn't being adjusted.  It does this when creating a time based clock too using the create clock.vi. The create clock uses the oscillator which makes me feel like the oscillator itself isn't being adjusted (Something else is because timestamps are synchronized).

If I schedule output events at the seconds boundaries rather than using the PPS generate feature, it seems to stay right on top of the incoming PPS.  I posted this question here.

I wish NI would provide a better technical description of the 6683H and how it performs synchronization or what the elestrical interface limits are (The 6683H can't even see my PPS input signal so I have to route it via DAQmx... I think it's the slow rise time preventing the 6683 PFI from detecting edges).

Thanks,

XL600

[FinchTrain]

Hi,

The bottom of page 4-2 discusses how the card uses the incoming signal to set its timebase accordingly.

NI PXI-6683H Series User Manual

http://www.ni.com/pdf/manuals/373656a.pdf

Due to the jitter in your PPS signal, it is going to be difficult to have a consistent output or action time-based on the unstable input PPS.

Regards,

Finch Train

[xl600]

The manual doesn't actually say anything about how the synchronization occurs with respect to oscillator adjustments or how much jitter is too much jitter.  Regardless, I'm now working on an IRIG option which may provide either IRIG or a much better PPS.  I'm hoping for much better results.

Thanks for the help!

[FinchTrain]

Hi,

Jitter is not a specified measurement due to the number of factors that can contribute to it that are outside of the control of NI. Page A-1 of the manual does specify added jitter from the module itself on the CLKIN pin, however, factors such as quality of the signal and temperature provide too much variance to quantify the impact of jitter.

Any specific questions regarding the oscillator may be flirting with proprietary information of National Instruments, though I could attempt to answer them based on available resources.

Best of luck going forward,

Finch Train