View Single Post
Old 06 December 2013, 01:47   #102
Registered User
Join Date: Jan 2012
Location: USA
Posts: 281
Originally Posted by pandy71 View Post
Nope - believe me, your example if antialiasing filter was used and reconstruction filter was used provide always sine - don't ignore Nyquist as it works perfectly.
Use opencalc (or excel) and verify your signal - use spline or cubic line smoothing to simulate lowpass filter.
I understand that. But real Amigas don't have perfect and infinitely adjustable reconstruction filters with variable cutoff frequencies.

A two point square wave output at 1000 points per second requires a low pass filter with a cutoff frequency of 500 Hz to construct a sine wave. There is no such filter on the Amiga. The cutoff frequency is much higher. To get perfect reconstruction of the original signal, the cutoff frequency must be half the the output frequency and since on the Amiga the output frequency is variable, we'd also have to have a programmable filter.

Originally Posted by pandy71 View Post
So this is why Amiga filter begin cut energy over 4.5kHz - if you sample signal with maximum DMA OCS speed - signal will be oversampled and antialiasing not present - remember - square wave have unlimited spectrum - lowpass filtering make square no longer square bu bandlimited signal thus Nyquist criteria are meet.
Oh sure. For the highest output rates, the low pass filter starts to be effective as a reconstruction filter. For lower rates, not much at all.

And maybe that's why this discussion has been so contentious. At higher output rates, the Amiga audio hardware performs more like a sample player. At lower output rates it looks more like a synthesizer. The Amiga gives us the choice.

Originally Posted by pandy71 View Post
Just take LTSpice and simulate analog circuitry behind Paula output - you will see how integrator and remain part of filter works.
Thanks for the link!
mc6809e is offline  
Page generated in 0.03997 seconds with 10 queries