cancel
Showing results for 
Search instead for 
Did you mean: 
Reply

Sampling rate/Conversion rate question

Solved!
Go to solution
Highlighted

Sampling rate/Conversion rate question

Hello everyone,

 

I am using Veristand 2016 on a PXIe-1078 chassis with a PXIe-8135 controller with a PXIe-6363 MIO card and 5 PXI-e 4330 cards.

All analog input channels in all cards are configured for single point input.

 

Due to the sampling rate limitation in the 4330 cards described in this KnowledgeBase article https://knowledge.ni.com/KnowledgeArticleDetails?id=kA00Z000000PA15SAG,  I have set the target rate for the controller to 711Hz.

 

When I set the Conversion Rate for the 6363 card to "Default", Veristand does not use the proper sampling rate of 711 Hz. When I sample a sine wave generated by an external function generator, it appears that Veristand is using a sampling rate closer to 500Hz.

 

However, when I set the conversion rate of the 6363 cards to "Maxiumum", it appears to be sampling at the specified rate of 711Hz.Setting the conversion rates of the 4330 cards to maximum or default has no effect.

 

I am wondering why setting the conversion rate in the 6363 cards affects the controller sampling rate, and how I can be sure that with the 6363 conversion rate set a Maximum, the controller is using the exact sampling rate specified.

 

Also, when I tested the attached VI in LabVIEW RT, I got slightly different actual sampling rates for the 6363 and 4330 cards. Is this happening is Veristand as well, are the sampling rates of the cards only approximately equal?

 

Thank you.

 

Siva

0 Kudos
Message 1 of 8
(440 Views)

Re: Sampling rate/Conversion rate question

Also, I forgot to mention, when I disable the 4330 cards, the 6363 card samples at the correct rate even when its conversion rate is set to default, which is puzzling.

 

Siva

0 Kudos
Message 2 of 8
(421 Views)

Re: Sampling rate/Conversion rate question

Hi Siva,

 

There's likely a bug in VeriStand somewhere that's leading to what you are seeing if the 6363 is actually sampling at 500 Hz when you've specified 711.  At a glance, here are my thoughts: 

 

  1. Changing background conversion rate on the 4330 not changing the measurement makes sense.  As you're probably aware, these modules have delta-sigma ADCs that oversample and then average the large set of samples to give you your single-point data.  The number of samples that the module averages is dependent on sample rate, so a slower sample rate will average more samples than a fast sample rate.  The number of samples averaged is ultimately a function of the frequency of the oversample clock, which is constant.  That is written into the DAQmx driver and will not be changed by VeriStand.  The fact that property is exposed in the VeriStand configuration window for the module may be a bug, depending on how you look at it. 
  2. The 6363 should definitely be able to sample at 711 Hz without enabling the maximum conversion rate.  If you can provide timestamped log files to back this up, we'll dig into it and get a more conclusive answer.  This behavior shouldn't be happening. 
  3. Again, the 6363 should be able to sample right around 711 Hz without issue.  Theoretically, if the 6363 were unable to get near 711 Hz by dividing down the 10 MHz backplane reference clock, it could slow down the control rate of the VeriStand system (assuming the 6363 is your master DAQ device and VeriStand is set up to use DAQ timing).  In the hardware-timed case, the controller in the system looks for a trigger sent out from the DAQ device to clock each iteration of VeriStand's PCL.  If you specify a control rate of 711 Hz and the DAQ board - for whatever reason - is actually sampling at 500 Hz and thinks that it's sampling at 711 Hz, your control loop will run at the erroneous 500 Hz.  I doubt that is happening here, but I suppose there's hardware where that may be possible.  
  4. The sampling rates should be a little different between the two when tested in LVRT.  The 6363 derives its sample clock by dividing down the 10 MHz backplane reference clock, while the 4330 is driven by its onboard oversample clock.  The 6363 can theoretically hit 711.035 Hz by dividing the clock down by a factor of 14,064.  Like the KnowledgeBase says, the actual sample rate of the 4330 is ~711.11 Hz because of the settling time associated with the ADC.  Sure, these are different, but assuming the 6363 is the master DAQ device, VeriStand will actually run at 711.035 Hz and ask for samples from the 4330 at that rate.  Because the sample rate of the 4330 is actually faster than the period at which samples are being requested from the 6363, you'll never get repeat samples into the control loop from the 4330.  That said, the two signals will not be phase-aligned when implemented in VeriStand and will drift to up to (1/711.035 sec). 

I'd get in contact with Support with log files an a simple representative System Definition corroborating the claim that the 6363 samples much slower than requested to validate that behavior and maybe get a bug report filed. 

 

If you want to independently validate the sample rate on the 6363, ensure you're using DAQ timing and that the 6363 is your master DAQ device.  Then, monitor an indicator the for the "Actual Loop Rate" system channel.  That number will give you the loop rate of the control loop as reported by the controller.  Since the controller is clocked by an interrupt from the master DAQ device - the 6363 - that indicator will show you the sample rate the 6363 is using.  

Matt | NI Systems Engineering
0 Kudos
Message 3 of 8
(387 Views)

Re: Sampling rate/Conversion rate question

Hi Matt,

 

Thank you for your detailed reply.

 

The 6363 is set as the master DAQ device for the chassis. Following your suggestion, I added an "Actual Loop Rate" indicator. When I run with the Covert Rate for the 6363 set to Maximum, the Actual Loop Rate reads ~711 Hz (it fluctuates from 710.7 to 711Hz). When I run the Convert Rate for the 6363 set to Default, the Actual Loop Rate reads 355.5Hz (half of the 711Hz, also not 500Hz like I thought).

 

Thank you also for clarifying the possible minor sampling rate discrepancy between the 6363 and 4330 hardware.

 

I also logged some data from the 6363 as you suggested, and with the Convert Rate set to Default, the sampling rate in the log file (attached as TDMS) is in fact 355.5 Hz.

 

So it appears that I have to set the Covert Rate for the 6363 to Maximum. While I am curious as to why this is the case, if the number in the Actual Loop Rate Indicator is reliable, then my purpose is served. Can I be sure that this number is right?

 

Thank you again.

 

Siva

0 Kudos
Message 4 of 8
(382 Views)
Solution
Accepted by topic author mvsiva
11-01-2018 12:52 PM

Re: Sampling rate/Conversion rate question

Hi Siva,

 

Thanks for the follow-up!  That behavior is admittedly very strange; I sent the AE you're working with the log file attached here.  He may want a simple reproducing System Definition to explore further, but I'll leave that up to him.  Personally, I'd be curious if this is happens with any number of 4330s in the chassis or if five is a magic number. 

 

You can be confident that the loop rate reported by the "Actual Loop Rate" system channel is accurate since it goes through two different layers - DAQ hardware and controller hardware.  The ~711 Hz timing signal is created by the master timing device (the 6363) and routed along one of the PXI_Trig lines that VeriStand reserves (either PXI_Trig0 or PXI_Trig1; not sure which your system is using without access to the source but it doesn't ultimately matter) to the controller.  The controller registers rising edges of that pulse and notes its internal system time at each rising edge.  Note that the timebase used by the DAQ hardware is completely separate from the controller's timing source.  The time delta between rising edges - measured by controller time - is what is displayed in the "Actual Loop Rate" system channel.  Hopefully that is clear, but let me know if it isn't and I can clarify. 

 

Though it sounds like you have a working system, this is absolutely something that we want to better understand internally.  Thanks for your diligence in investigating this!

Matt | NI Systems Engineering
0 Kudos
Message 5 of 8
(371 Views)

Re: Sampling rate/Conversion rate question

Hi Matt,

 

Your explanation for why the Actual Loop Rate is reliable makes sense. Thank you.

 

Following your input, I created a minimal System Definition that reveals the problem.

 

I have one 6363 card (set as the master DAQ device), and one 4330 with 1 channel (AI0) added.

 

The behavior is interesting.

 

1. For number of channels in the 6363 < 5, the loop rate is ~711Hz.

2. For number of channels between 5 and 10, the loop rate switches between 355.5Hz and 711Hz.

3. For number of channels > 10, the loop rate is 355.5Hz.

 

When I set the Convert Rate for the 6363 to Maximum, or when I disable the 4330, the loop rate becomes 711Hz even with > 10 channels.

 

I am attaching the minimal System Definition (zipped up, so I could upload it).

 

I will mark this issue as solved because I have a working system now, but I am curious and happy to contribute towards understanding the issue further.

 

Thank you.

 

Siva

0 Kudos
Message 6 of 8
(361 Views)

Re: Sampling rate/Conversion rate question

Hi Siva,

 

Thanks for attaching the SysDef!  I talked to the AE who you spoke to on the phone on Tuesday and he's picking up the hardware to try and reproduce this in-house.  Not sure exactly what the timeline there is, but I'll let him take point and communicate out what he finds.  I'll stay in the loop as an escalation resource for him; assuming we can reproduce this behavior, there's an almost 100% chance this will eventually make it into R&D's hands.  

 

I'm actively invested in finding a root cause since this behavior is so strange; it'll be interesting to see what we find out. 

Matt | NI Systems Engineering
Message 7 of 8
(358 Views)

Re: Sampling rate/Conversion rate question

Sounds good, thank you, Matt.

 

Siva

0 Kudos
Message 8 of 8
(348 Views)