PCI 5153 - Question regarding resolution/full scale range

[NablaQuabla]

Hello and I hope this is the right sub-forum.

I am currently looking at some data we have recorded using a NI PCI-5153 digitizer. One thing I noticed is that there is a discrepancy between the claimed resolution/full scale range and what was actually recorded.

We are running at 0.2 V pk-pk and in this configuration I see that each digitizer step is actually 0.8593 mV instead of the expected 0.2V/256 = 0.78125 mV.

Right now I don't have any data available that extends beyond -0.1V or +0.1V so I cannot probe if the full scale is actually closer to 0.22V pk-pk or if there are indeed only ~233 steps available for the digitizer. It's not a deal breaker but it would be nice to know what's going on. Can anybody help? Or did I just miss something in the manual?

And would the range for the other full scale settings also be bigger by 10%?

Thank you very much!

Bjorn

[James_Battleaxe]

Hi,

Can you tell me a little bit more about your test setup? How did you determine that 0.8593mV is one step? What kind of sampling mode was the digitizer in?  

[NablaQuabla]

Hi James,

Thank you for your reply!

Our setup is a Hamamatsu PMT connected to a Philips Scientific Gain Amplifier. The amplified and DC shifted output is then fed into CH0 of the PCI-5153 card. The host PC is running Win7 Enterprise (64bit). Has a Core i7-2600 and 8 Gb of RAM. No other PCI cards are used.

The digitizer was set to acquire 35000 samples at 500 MS/s. We are triggering on an external 60Hz signal. The settings for both channels are as follows:

CH0: 0.2 V, DC coupling

CH1: 0.5 V, AC coupling

We use niScope Fetch (poly) set to acquire a multiwaveform 1D cluster (floats) whenever there is a trigger. The y-values of the waveforms are then simply shown in a graph on the front panel. These values are supposedly already scaled to volts when we fetch them from the card. Now if I look at the difference between consecutive data points I only see multiples of 0.8593 mV, see below:

We noticed this as we also rescale the acquired data back to I8 for storage. We do so by scaling the float output (given in V) by 256/0.2V, as we are dealing with an 8 bit system and a range of 0.2V. But if we recover the data we see that some of the bits actually never occure. This can be seen in the following histogram which shows the values for thousands of waveforms (the peaks and everything is fully explainable on our side, it's really just the missing bits that give us a headache):

Those gaps can be mathematically explained if the digitizer steps were 0.8593 mV wide and we falsely assume them to be 0.2V/256 = 0.78125 mV.

It's not a huge problem, but still it would be nice to have a confirmation, that it is indeed the digitizer and not something we are doing wrong on our end. Also it would be nice to know the actual ranges for all full scale settings, such that we can properly scale the data.

Thank you very much!

Bjorn Scholz

[James_Battleaxe · Accepted solution]

Hi,

After some research, I found that our measureable range is typically 4-5% larger than the nominal range. I recalculated the step size with that in mind and I got results that are close to what you are seeing. So I believe you are seeing expected behavior on your device.

[NablaQuabla]

Hey James,

Using the numbers we've found you actually get pretty close to a 10% larger full scale range, i.e. 256*0.8593 mV = 219.9808 mV. Which again is fine, it's just something we'd like to know beforehand. So it would be nice to find an uncertainty on the full scale ranges in the manual. If there is already one and I just missed it I apologize, but the only thing I found is the DC accuracy which is not really the same thing. Given that there is no error noted on the ranges I would have expected it to be minimal.

One other thing: Is the actual measureable range for two different NI5153 cards comparable to maybe within~1%? Or could there be a huge (i.e. ~10%) variance between two cards of the same type? Because at some point we'll probably compare the voltage output of two cards. But if there is a large variance in the range of two cards we'd actually have trouble to compare 1 ADC count in one card to the other. Again we could actually measure the total range, but if we were to change ranges we'd have to do it again etc.

Thanks again for having a look into this!

Bjorn Scholz

[James_Battleaxe]

Hi,

There should not be a big variance between devices of the same model. Basically, if you have 2 PCI 5153 cards, then they should have the same measureable ranges. 

[NablaQuabla]

Thanks James!

Then I guess we just have to measure the ranges at some point. Or in principle we could just use the voltage per digitizer step to calculate it given it's an 8 bit card.

Thanks again,

Bjorn

[James_Battleaxe]

Yup! And glad I could help!