modifying input filter

[cryo]

I need to modify the NI 6251 input filters as I would like to use a higher
frequency signal to do some interleaved undersampling measurements.
Basically inputing a periodic stream with high frequency (5 to 10 MHz) components
and interleaving the aquisition.
So would it be feasable (addmitedly I would lose the warranty on the card) to shift
the 1.7MHz acquisition input filter bandwith to 10 Mhz by replacing one or a few
capacitors on the board.

[SachaE]

Hi,
any modifications you chose to make on the board will invalidate the warranty, and we cannot cover that.
Your theory is sound enough, however, we cannot help with circuits or anything like that since it would be showing
information on our boards that we cannot share for legal reasons.

On a side note, the 6251 can only acquire data at 1MS/s, so to quote Nyquist, the highest frequency component you would be able to resolve is 500kHz when sampling over more than one channel.
You're referring to Random Interleaved Sampling, as you have correctly worked out, you therefore need to remove the filter which is limiting the bandwidth of the input signal so it will not be able to react to the higher frequency. This may not be the end of the story though. The on board pathways and the PGIA may not allow the signal bandwith either.
The real solution is to buy hardware appropriate for the signal. To be able to recover the 10MHz signal, you'll need 100MHz bandwidth on the board so you don't break the Nyquist theorem. (10 times oversampling to get some shape of the signal)
The balance to do this requires a filter which is very flat in the pass band, however the filters that do this tend to have a slow roll-off, so you end up with large aliasing problems. It's this kind of reason why there's a requirement for higher specification boards.

Have a look at the 5122, 5911, 5112 et or any of the higher speed / bandwidth cards. Also, consider what resolution you want to measure the signal to.

Sorry we can't be of more help in the modification of the board, however if you'd like more information on the higher speed cards, please call in to the nearest office, and discuss your requirements with our technical sales engineers.

Thanks

Sacha Emery
National Instruments (UK)

[rosaldorosa]

Looks to me, that bandwidth of ADC converter is limited by front RC filter.
This means, that if card has sampling rate of 1MS/s, I shouldn't bother to do undersampling of signals higher than, 500KHz
because they never arrive to ADC input at all.
Is it OK?
Is there any reason for using on board hardware LP filter in DAQ card except comercial reaons?

[SachaE]

Hi,

  the 1.7MHz isn't a specific input LP filter, it's simply the measurements we've taken on the board with a small signal to see what the half power point of reading can go up to. i.e. if you have a 1.7MHz signal going in, you'd read about half the voltage, however since the maximum sample rate acheivable is 1.25MS/s on one channel, then we wouldn't be expeting you to try to measure frequency based information content above 625KHz, and time based information probably around 200kHz (5 samples per cycle is a good rule of thumb for and absolute minimum number of samples per cycle).

Ordinarily you would place a 500KHz (or whatever half your sample rate is) real world low pass filter on the front end of the data acquisisiton setup to remove the possibility of aliased frequency content showing in your signal.

So it's just the layout of the ADC and the actual ADC itself that's giving the small signal bandwidth of 1.7MHz, not a specific low pass filter.

You can see the spectral response on page two of here : http://www.ni.com/pdf/manuals/371291d.pdf to give you a better idea of how it effects the signal.

Thanks

Sacha Emery
National Instruments (UK)