To measure event stamps using TTL signals into NI-4472

[sidoo]

Dear Sir/Ma'am

I'm using the NI-4472 data acquisiton card to sample sets of analog and digital data as part of an experiment.   The experiment consists of a person standing upright, pressing down on a switch, and releasing it upon a cue, raising their arm as quickly as possible. Inclinometers are used to measure the angular position of the arm.

A channel on the card is used to sample a TTL signal from the switch(either +5 or 0v) , where the voltage transitions due to the switch release are basically used as  time stamps to mark events of interest from the analog channels (e.g the inclinometer data).

I am currently looking for the transitions by using a threshold(50%).  However, I observe ringing at the step transitions, both before and after the step.  The ringing occurs on average for a period of 15-25ms. 

I believe that the ringing is caused by the digital anti-aliasing filters built into the card. Based on my understanding of the principles of such filters, i believe that the actual occurence of my event (beginning of the arm movement) should precede the ringing artifact. The inclinometer readings also seem to suggest that the time stamps acquired from the switch signals seem to be delayed.

Hence, i would like to know what the transfer function/step response of the digital filters on the DAQ are so that I account for the delay introduced.

Thank you

Siddharth Vedula

[Elizabeth W]

Hello Siddharth,

You are correct that this ringing is caused by the device's anti-aliasing filters.  The NI 4472 is intended for acquisition of sine waves in applications such as sound and vibration measurement.  It is not designed to acquire waveforms with sharp edges; thus the high-frequency components of these waveforms are attenuated by the anti-alisasing filters, causing the ringing that you observe.

To see the specifications on the frequency response of the input filter, refer to page 4 of the NI 447x Specifications.

Does this help at all?  Please let me know if I haven't answered your question!

[sidoo]

Hi Elizabeth

Yep, i understand what you mean. I guess im stuck with this acquisition system though, given that the rest of the channels are used to collect lower frequency analog input- so i have to find some kind of approximate workaround. 

But i'm still not clear on what i actually wanted to know. If it is anti-causal you would expect the ringing to occur AFTER the actual input step, followed by the output step- thereby leading to an input-output delay(if you are only looking for the 50% threshold, i.e. when the drop in voltage occurs)

Im not too sure how the bode plot in the spec sheet can be translated to the transfer function though.  Can you give me the actual mathematical transfer function of the filter?- that way i can simulate the step response(0-5v or 5-0v) and observe what the delay is.

Thanks again

Sid

[Robert_F]

Hi Siddharth,

There is a delay that will be introduced from the filter.  However the delay of the filter will also be the same amount of time as the other analog inputs.  This means that your digital signal will be synchronized with your other analog signals.  We can not give you the transfer function of that filter.  Look at page 5 of the document that Elizabeth mentioned earlier to find the delay of the filter. 

[sidoo]

Hi Rob

I spoke to the lab research assistant, and i think i got my wires mixed up in understanding the problem.  I agree with you that the delay will be the same across all the signals (analog or digital) given that they are all being passed through the same set of filters.

However, in terms of the dynamics of the signal, the fact that you see the ringing prior to the step indicates that the actual event in ABSOLUTE physical time occurs prior to what is recorded on the DAQ (otherwise the ringing on the DAQ system would be predicitive which is not possible).  Hence, i would need need to know how my actual event compares to my measured event (at the 50% threshold).  Since, i need to time stamp my analog data in the other channels i need to know what this mismatch is- which is why i asked for the transfer function.  Knowing the filter delay alone(from page 5) will not resolve the situation.

Please let me know if you have any suggestions.

Sincerely

Sid

[preston_johnson]

one item you can do to minimize the ringing is to set the sampling rate to 102.4kHz.  This will move the low-pass filter to 48kHz, allowing most of the TTL pulse content to pass. 

[sidoo]

Hi Preston

How would you go about that? 

However, oversampling at that rate doesnt seem like an acceptable solution for me as that would entail collecting huge amounts of data(i currently collect around 20-30 min of data sampled at 1 khz).

The ringing is not an issue though- all i really need to know is how the filter dynamics will displace the event in time(i.e. how far in absolute time does the ACTUAL event precede the ringing and the recorded step?)

Thanks again

Sid