how to determine antialising filter specification?

[liusupeng]

suppose i have an ADC working at a certain sampling rate fs, how to determine the specification of the anti-aliasing filter specification. like how much attenuation it should provide at fs/2?

 

[keith1200rs]

I would think you want maybe 1/(number of bits) attenuation.

Keith

 

[rampat]

it depends upon the bandwidth of the signal you feeding to the ADC.Basically antialias filter used to make input signal bandlimited..if you are sampling signal with fs,you need to figure out what kind of attenuation you want to see for fs components so that they fs components that fold back into your signal band is lesser than the noise level you are targetting...depending on the attenuation you can decide the order of the filter and depending on the resolution of adc you can decide the 3db badwidth of the filter ,passbandgain etc....

rampat

---------- Post added at 07:46 AM ---------- Previous post was at 07:43 AM ----------

Hi Keith,,

May i know the reason for 1/(no.of bits) for attenuation..

thanks
rampat

 

[liusupeng]

it depends upon the bandwidth of the signal you feeding to the ADC.Basically antialias filter used to make input signal bandlimited..if you are sampling signal with fs,you need to figure out what kind of attenuation you want to see for fs components so that they fs components that fold back into your signal band is lesser than the noise level you are targetting...depending on the attenuation you can decide the order of the filter and depending on the resolution of adc you can decide the 3db badwidth of the filter ,passbandgain etc....

rampat

---------- Post added at 07:46 AM ---------- Previous post was at 07:43 AM ----------

Hi Keith,,

May i know the reason for 1/(no.of bits) for attenuation..

thanks
rampat

thanks rampat. that make sense to me.

 

[FvM]

I think, 1/(2 raised NOB) is meant. A simple estimation of keeping aliasing components below the quantization threshold.

 

[keith1200rs]

I think, 1/(2 raised NOB) is meant. A simple estimation of keeping aliasing components below the quantization threshold.

Woops, yes, that is what I meant - sorry.

Keith