Analog Input: Need help understanding "Timing Accuracy".

[Nabil E.]

Hello,

I have an application where I need to accurately determine the phase shift between two low-frequency repetitive signals. Do I need to invest in a simultaneous sampling DAQ, or could I get away with the an M-series DAQ (i.e. multiplexed).

I have looked at 4 products. The PXI-6221 (a 250 kS/s M-series DAQ), the PXI-6251 (a 1 MS/sec M-series DAQ), the PXI-6132 (a 2.5 MS/s S-series DAQ), and the PXI-6143 (a 250 kS/s S-series DAQ).

The timing accuracy of the two M-series DAQ cards is specified as "50 ppm of sample rate", with a resolution of "50 ns". I find this confusing. I calculated that for a 1 kS/s sampling rate, the timing accuracy is 50 ns. Does this mean that as long as I am sampling at 1 kS/s or higher, I will get the same analog input timing performance from either M-series DAQ card?

I then looked at the S-series cards, which were even more confusing. The timing accuracy of the PXI-6143 is specified as "±50 ppm for internal timebase". What is the internal timebase? Is it 20 MHz? If so, does that mean that the timing accuracy is independent of the sampling rate? Moreover, the interchannel "phase mismatch" is specified as "±0.5 deg at 100 kHz". What does that mean? Is the 100 kHz the sampling frequency, or is it the frequency of the signals? How much phase mismatch to expect when sampling a 1 kHz signal at say 10 kHz, or a 100 Hz signal at 10 kHz?

The PXI-6132, on the other hand, has an "Interchannel skew" specification of "1 ns typ". This is less confusing than the "phase mismatch" specification because it is clearly independent of the sampling rate or the timebase. However, I read the entire PXI-6132 datasheet and could not find any other specification that is related to the timing accuracy for analog input!

Could anyone please help me compare apples to apples?

Thanks.

[Sean_C.]

Hello Nabil,

"The timing accuracy of the two M-series DAQ cards is specified as "50 ppm of sample rate", with a resolution of "50 ns". I find this confusing. I calculated that for a 1 kS/s sampling rate, the timing accuracy is 50 ns. Does this mean that as long as I am sampling at 1 kS/s or higher, I will get the same analog input timing performance from either M-series DAQ card?"
The 50 ppm is just a percentage accuracy. The accuracy = 50/1000000 * sample rate. So, at 1000 Hz, Accuracy = 1000*50/1000000 = +/- 0.05 Hz. Because all M Series use the same oscillator for a timebase, all M series cards will have the same Timing Accuracy.

"I then looked at the S-series cards, which were even more confusing. The timing accuracy of the PXI-6143 is specified as "±50 ppm for internal timebase". What is the internal timebase? Is it 20 MHz? If so, does that mean that the timing accuracy is independent of the sampling rate?"
This 50 ppm is the exact same as for the M Series cards. It too is dependent on the sample rate. I agree that these specs are a little confusing.

"Moreover, the interchannel "phase mismatch" is specified as "±0.5 deg at 100 kHz". What does that mean? Is the 100 kHz the sampling frequency, or is it the frequency of the signals? How much phase mismatch to expect when sampling a 1 kHz signal at say 10 kHz, or a 100 Hz signal at 10 kHz?"
The 100 kHz is the input signal frequency. So to convert this to units of time: 100 kHz -> 10 us. 10 us * .5/360 = ~14 ns.

"The PXI-6132, on the other hand, has an "Interchannel skew" specification of "1 ns typ". This is less confusing than the "phase mismatch" specification because it is clearly independent of the sampling rate or the timebase. However, I read the entire PXI-6132 datasheet and could not find any other specification that is related to the timing accuracy for analog input!"
I could not see the "1 ns typical" spec. If you look at the 6132/6133 spec sheet at: http://www.ni.com/pdf/manuals/371231b.pdf, it lists a "Phase mismatch of ±2° at 1 MHz." You could then break that down into units of time the same way as with the previous question.

I hope this helps. Please let me know if you have any further questions.

Regards,
Sean C.

[Nabil E.]

Sean,

Thank you for taking the time to answer my questions.

Your calculated accuracy of +/- 0.05 when sampling at 1 kS/s does not make sense. How could the accuracy be +/- 20 seconds (which is 1 / 0.05 Hz) when a sample is being taken every 1 ms?

I calculated it as 50 ppm of the sampling period: (50 / 1,000,000) * (1 / 1000 s) = 50 ns.
This means that no matter which of the cards I use, and as long as I am sampling at 1 kS/S or more, the timing accuracy is the same as the timing resolution, namely 50 ns.

I conclude that for this particular application (slow repetitive signals), I can get away with a low-end (cheaper) M-series card, because the timing accuracy is will be the same (namely 50 ns) as long as I am sampling at 1 kS/s or faster. A more expensive card would not buy me any improvment in timing accuracy.

Is this making sense? Did I get this right?

As for the "1 ns typical", I got it from the PXI-6132 datasheet "http://www.ni.com/pdf/products/us/043918301101dlr.pdf"
It's on page 4 under "Dynamic Characteristics". It seems the interchannel skew is specified differently in the spec. sheet at +/- 2 deg at 1 MHz, which calculates to about +/- 5.6 ns, rather than 1 ns.

Thanks again for your time.

[Sean_C.]

"Your calculated accuracy of +/- 0.05 when sampling at 1 kS/s does not make sense. How could the accuracy be +/- 20 seconds (which is 1 / 0.05 Hz) when a sample is being taken every 1 ms?"
The accuracy is +/- 0.5 Hz, which would correspond to period difference of (1/1000 - 1/1000.05)= 50ns.

"I calculated it as 50 ppm of the sampling period: (50 / 1,000,000) * (1 / 1000 s) = 50 ns.
This means that no matter which of the cards I use, and as long as I am sampling at 1 kS/S or more, the timing accuracy is the same as the timing resolution, namely 50 ns."
The timing accuracy is 50 ppm for all M Series boards. Therefore, you will get the same accuracy with a low cost M Series as with a high end M Series.

As for interchannel skew, this is only spec'd out for simultaneously sampling boards. For E Series, it would be the same thing as interchannel delay, which is user configurable. What are you trying to do with this board? Is there a certain spec that you want to make sure to meet?

If the accuracy of the E Series boards is not high enough, you could always generate clock with a counter/timer board (6608) and use that as the sample clock for the E Series. The pci-6608 has a timing accuracy of 75 ppb (parts per billion).

Regards,

Sean C.

[acanivell]

Interesting discussion indeed.

I am actually thinking of using this NI 6608 for frequency (not phase) estimation.

I am also concerned about the clock stability, as I need:

- Accuracy: 0.1Hz

- Range: 1Khz-100Khz

If the stability of the clock of a 6608 is 75ppb, and its timebase is 10Mhz:

6608 clock accuracy:  75/(1,000,000,000)*10,000,000 = 0.75Hz

0.75Hz is already higher than my target accuracy 0.1Hz.

Can I get better accuracy on my clock in some way? (Maybe I can set timebase to be 1Mhz instead of 10Mhz? That would give me 0.08Hz accuracy)

Regards