I'd like to have an optimal antialiasing filter to try and preserve the maximum fidelity and bandwidth on one of the input channels (within reason).
I see the delta-sigma ADC mentioned in a couple of places on the site and I understand that the oversampling of that converter would simplify the low-pass filter requirements... what is the ADC oversampling rate and how can that be used to calculate the necessary stopband attenuation required for an antialiasing filter?
Thanks,
Optimal antialiasing filter
Moderator: frank
-
- Posts: 18
- Joined: Tue Jul 22, 2008 3:54 pm
- Location: Ottawa, Canada
- Contact:
Optimal antialiasing filter
-mark
My blog: http://tubenexus.com
My blog: http://tubenexus.com
Had to talk to Keith about this one as he did the ADC/DAC, here is his answer:
The FV-1 includes an anti-alias filter already. The system cuts off at about 0.46*Fs, and has a very deep rejection level, on the order of 100dB. Both channels have the same response. If you clock the FV-1 at about 2.2 times the highest frequency you can hear (sine generator into a good tweeter), you will have excellent fidelity and no added limitation in terms of frequency or transient response. Also, the digital anti alias filters are constant phase (linear phase), so there will be no appreciable phase distortion.
The oversampling ratio, and the decimation filter for the ADC is 64:1, but the converter samples the input twice during each oversample cycle (averaging adjacent samples) I suppose you could say the converter is 64:1 with a double oversampling input, which would give 128:1.
The FV-1 includes an anti-alias filter already. The system cuts off at about 0.46*Fs, and has a very deep rejection level, on the order of 100dB. Both channels have the same response. If you clock the FV-1 at about 2.2 times the highest frequency you can hear (sine generator into a good tweeter), you will have excellent fidelity and no added limitation in terms of frequency or transient response. Also, the digital anti alias filters are constant phase (linear phase), so there will be no appreciable phase distortion.
The oversampling ratio, and the decimation filter for the ADC is 64:1, but the converter samples the input twice during each oversample cycle (averaging adjacent samples) I suppose you could say the converter is 64:1 with a double oversampling input, which would give 128:1.
Frank Thomson
Experimental Noize
Experimental Noize
-
- Posts: 18
- Joined: Tue Jul 22, 2008 3:54 pm
- Location: Ottawa, Canada
- Contact:
Thanks Frank!
That's just what I needed to hear!
Do the numbers quoted (e.g. 100dB) require the 'standard' low-pass RC on the inputs as per the datasheet drawing? It sounds like you would hardly need any filter at all.
That's just what I needed to hear!
Do the numbers quoted (e.g. 100dB) require the 'standard' low-pass RC on the inputs as per the datasheet drawing? It sounds like you would hardly need any filter at all.
-mark
My blog: http://tubenexus.com
My blog: http://tubenexus.com
I would still put it in as it has a few things it does, the 1u blocks any DC in the signal, etc. and rolling off the out-of-band high end a little prior to the ADC can only help.
Frank Thomson
Experimental Noize
Experimental Noize