Forums

Anti Aliasing a NI 6259 BNC Daq System

Started by fsou...@yahoo.co September 15, 2011
Hello, 
I am trying to record 8 channels of analog data using an NI 6259 BNC Data
acquisition system. The Daq system has a built in lowpass anti alias
filter, but it has a much higher cut off frequency than what i need (i want
a 500Hz cutoff). 
Does anyone know of a good 8 channel programmable/non programmable anti
alias filter. 
I have also been suggested to use a software filter, but i thought anti
aliasing could only be done with front end analog hardware filters is this
right? 
Any information or direction to literature would be of great help. 
thank you for your help



Hello,

without knowing this particular card, the conventional approach would be to
use the highest sampling frequency that is possible (to minimize aliasing),
then lowpass filter in software and decimate to whatever rate you like.

=> What is your ADC sampling rate?
=> What sampling rate would you like? 
   preferrably division by integer n, or at least a rational factor.
=> how many bits do you use, can you tell anything about the dynamic range
of the signal. Or do you know an accuracy requirement such as "linear
distortion / aliasing < -60 dB"?

With that information, designing the filter is a standard problem.
On 9/15/2011 9:14 AM, fsouto2@yahoo.co wrote:
> Hello, > I am trying to record 8 channels of analog data using an NI 6259 BNC Data > acquisition system. The Daq system has a built in lowpass anti alias > filter, but it has a much higher cut off frequency than what i need (i want > a 500Hz cutoff). > Does anyone know of a good 8 channel programmable/non programmable anti > alias filter. > I have also been suggested to use a software filter, but i thought anti > aliasing could only be done with front end analog hardware filters is this > right? > Any information or direction to literature would be of great help. > thank you for your help
If the sample rate of the ADCs is programmable, it would make sense that the filter cutoffs are also. The anti-alias filter needs to substantially remove all frequencies not less than half the sample rate. An anti-alias filter suitable for use with a single channel at 1.25 MHz would be useless with a channel sampled at 125 KHz. You seem to want a low-pass filter with a substantially lower cut-off than would be needed for the sample rate you plan to use. Is that the case? If you need a 500 Hz low-pass, you don't need a much faster sample rate than 1 KHz, but you can, in fact use any higher rate. Provided the sample rate you choose avoids aliasing with an appropriate filter, you can apply a software low-pass filter after the signal is acquired. Higher sample rates impose higher computation loads, but that may not be an issue. Jerry -- Engineering is the art of making what you want from things you can get.
On Thu, 15 Sep 2011 08:14:00 -0500, fsouto2@yahoo.co wrote:

> Hello, > I am trying to record 8 channels of analog data using an NI 6259 BNC > Data acquisition system. The Daq system has a built in lowpass anti > alias filter, but it has a much higher cut off frequency than what i > need (i want a 500Hz cutoff). > Does anyone know of a good 8 channel programmable/non programmable anti > alias filter. > I have also been suggested to use a software filter, but i thought anti > aliasing could only be done with front end analog hardware filters is > this right? > Any information or direction to literature would be of great help. thank > you for your help
You can sample at a rate that matches your existing analog filter hardware, then filter and decimate in software. This can work quite well. If your noise level is low and the signal you want to measure is fast, anti-aliasing filters can introduce more error than they filter out: http://www.wescottdesign.com/articles/Sampling/sampling.pdf -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Jerry Avins <jya@ieee.org> wrote:
> On 9/15/2011 9:14 AM, fsouto2@yahoo.co wrote:
>> I am trying to record 8 channels of analog data using an NI 6259 BNC Data >> acquisition system. The Daq system has a built in lowpass anti alias >> filter, but it has a much higher cut off frequency than what i need (i want >> a 500Hz cutoff).
(snip)
> If the sample rate of the ADCs is programmable, it would make sense that > the filter cutoffs are also. The anti-alias filter needs to > substantially remove all frequencies not less than half the sample rate. > An anti-alias filter suitable for use with a single channel at 1.25 MHz > would be useless with a channel sampled at 125 KHz.
Sounds good, but maybe not. Sample rate is easily programmable in digital hardware. Programmable analog filters are much harder to make. For a low to mid range board, it wouldn't surprise me if it expected the user to supply the analog filter for lower frequencies. (Even more, lower frequencies tend to require bigger capacitors.) -- glen
On Thu, 15 Sep 2011 16:45:13 +0000, glen herrmannsfeldt wrote:

> Jerry Avins <jya@ieee.org> wrote: >> On 9/15/2011 9:14 AM, fsouto2@yahoo.co wrote: > > (snip) >> If the sample rate of the ADCs is programmable, it would make sense >> that the filter cutoffs are also. The anti-alias filter needs to >> substantially remove all frequencies not less than half the sample >> rate. An anti-alias filter suitable for use with a single channel at >> 1.25 MHz would be useless with a channel sampled at 125 KHz. > > Sounds good, but maybe not. Sample rate is easily programmable in > digital hardware. Programmable analog filters are much harder to make. > For a low to mid range board, it wouldn't surprise me if it expected the > user to supply the analog filter for lower frequencies. (Even more, > lower frequencies tend to require bigger capacitors.)
In this degenerate age it is often far less expensive to just sample fast. If necessary one can filter and decimate in software. In embedded systems, a processor that can sample in the tens of kilohertz then filter and decimate by a factor of 100 is often smaller and cheaper than the caps needed to make an analog filter with 16-bit accuracy. -- www.wescottdesign.com
On 9/15/2011 12:45 PM, glen herrmannsfeldt wrote:
> Jerry Avins<jya@ieee.org> wrote: >> On 9/15/2011 9:14 AM, fsouto2@yahoo.co wrote: > >>> I am trying to record 8 channels of analog data using an NI 6259 BNC Data >>> acquisition system. The Daq system has a built in lowpass anti alias >>> filter, but it has a much higher cut off frequency than what i need (i want >>> a 500Hz cutoff). > > (snip) >> If the sample rate of the ADCs is programmable, it would make sense that >> the filter cutoffs are also. The anti-alias filter needs to >> substantially remove all frequencies not less than half the sample rate. >> An anti-alias filter suitable for use with a single channel at 1.25 MHz >> would be useless with a channel sampled at 125 KHz. > > Sounds good, but maybe not. Sample rate is easily programmable > in digital hardware. Programmable analog filters are much harder > to make. For a low to mid range board, it wouldn't surprise me > if it expected the user to supply the analog filter for lower > frequencies. (Even more, lower frequencies tend to require bigger > capacitors.)
Then his other option is sampling fast, then filtering and decimating. With 8 16-bit channels, that might be hard to do in real time. Jerry -- Engineering is the art of making what you want from things you can get.