Anti aliasing filtering

[Amartansh13]

I want to buy a digital anti-aliasing filter (preferably from NI) with following specifications:
1) Digitally controlled cutoff frequency from 10kHz to 100kHz.
2) Real-time response.
3) 100dB reduction in gain in a span of 2.5 times cutoff frequency.

I want to use it for the following application:
I am using data acquisition board (USB 6356) at 500K Samples per seconds and I want to analyze frequencies in between 100Hz to 50KHz. I want to remove all higher frequency very efficiently using anti-aliasing filter before data acquisition step. This is important because I want to profile underwater noise. Therefore, I want to remove signals above 100Khz completely. I also have USB 6281 board with 40KHz filter but this filter is not sufficient as it is allowing signals upto 300KHz. 
If you can suggest better external anti-aliasing filter specifications (and product name) than it would be of great help for me.

I am sorry as I asked something not related to NI product. 

Thank you.

[johnsold]

"I want to buy a digital anti-aliasing filter..."

This is your first problem. The anti-aliasing filter cannot be a digital filter because digital filters are sampled data systems and thus subject to the same Nyquist criterion as any other.

100 dB attenuation at 2.5*fc will require a high order filter. An elliptic filter would probably be 6th or 7th order, a Chebyshev filter with 0.5 dB passband ripple would be at least 9th order. My handbook does not show how high the order would need to be for a Butterworth filter.

Achieving a true 100 dB dynamic range over a wide frequency range is not trivial. It requires attention to distortion, noise, gain-bandwidth product and so on.

I would probably get a set of high-performance active or passive filters for the bandwidths needed and use relays or high quality analog swtches to select the appropriate filter for a specific situation.

Lynn

[saratoga]

There is a much easier solution:

1)  Get an A/D with the dynamic range you need and significantly higher sampling range then you need.  I'd suggest the NI PCIe-6341, which can do 500 KHz at close to 100dB SNR (or perhaps slightly more with oversampling).

2)  Buy a normal antialias filter.  This one will work fine: http://krfilters.com/filter-docs/2928.pdf

You now have a 500 khz sampled waveform with a flat frequency response to 100 KHz, and the first alias to the highest frequency you want (50 KHz) is at 450 KHz (500khz -50khz) which is attenuated to nothing by the filter.

3)  If you want to reduce the bandwdith to 50 khz, use a digital low pass filter of your choice.  The ones generated by labview will easily attenuate everything above 100 kHz by 100 dB or more, or you can write your own.  

[LiorKi]

Hi,

First, unless there is some special reason, I suggest you focus on filtering only the spectrum suscpetible to alias onto you signal with no further DSP recovery option. In your case, since your signal resides in the 1st Nyquist zone- you should reject from 250kHz (=Fsample/2) and above with a low pass filter. The rest may be done with a DSP filter  (e.g; FIR) on the sampled signal.

I suggest you take a loook at Maxim's MAX274 or MAX275. They fit in terms of bandwidthand are programmable , but reject less than the 100dB required

Good luck,

Lior