Vladislavs Korehovs wrote on Wed, 30 May 2007 10:19 |
Tomas Danko wrote on Tue, 29 May 2007 14:40 |
Vladislavs Korehovs wrote on Tue, 29 May 2007 20:02 | Could you please point exact thread? Currently i don't see how FIR/IIR based filter can solve this design problem. As corresponding FFT transform will have band limited at Fs/2 as i remember, and convolution is just like multiplication in Freq domain. Anyway what will happen with responce in time domain? unpredictable phase rotations, distortions. I will try this in matlab filter design toolbox. Some nastalgy from school times:) And will give plots for phase spectrum and amplitude i i will succeed, for different filter coeficients.
|
It's not about FIR/IIR, and it's not convolution. It's about practical implementation.
http://recforums.prosoundweb.com/index.php/t/16683/2699/
|
Unfortunately on that theread are still not any "practical" implementation. And filters they are talking are "IIR" filters and they are using convolution by the way. Anyway thnks for pointing out interesting subject. I beleve if you don't have Ofxord filters you still don't want to lowpass after each stage of processing? Or maybe somebody have different oppinions? Anyway even idea of lowpassing after having already alisased impact is quite "amateur" to me.(sorry for going there.) Lowpass is done in ADC converters and you should be happy with that.
Best regards, Vladislavs
|
By implementation I didn't imply (heh) that it would provide you with a practical DIY walk-through. Still, it's about implementation in the sense that decramping as a method can be a more elegant solution than brute force oversampling which takes more resources.
Paul Frindle didn't talk about convolution, because afaik he doesn't do that.
I agree, of course it's a moot point to LPF anything above 20K to avoid aliasing if it's already in there.
Also, what are the penalty for using the same curve repeatedly across every channel, and then every next step that contains the same channels. I think the cure is worse than the problem, if one could even call it that. It's like HPF-ing everything @20 Hz instead of fixing grounding and getting a proper DC offset (ie none).
The way I see things, it's not a problem these days. There are bigger worries when trying to maintain some quality audio.