I don’t know what the challenge is and why I’m mentioned here?
1 Like
Jussi asking what the challenge is may be the most rigorously proven result in this thread so far. It’s certified to P128, it just has no inputs. 
I have to get the attention of the king and his disciples somehow. 
You got the attention of his most zealous disciple 
ShannonPeriodic
The underlying approach is the same as Shannon64, but it uses periodic boundary conditions instead of zero extension. It’s reasonably fast: a 3-minute file takes less than 10 minutes to process on my PC. I’d be very interested in seeing independent measurement plots
There are many ways to reinvent the wheel.
The measurement results will look dramatically more perfect, but that doesn’t necessarily mean they are more correct.
Time is a line, not a circle.
Compared to? Zero-extension is right for individual tracks and wrong for gapless. Circular extension is wrong for both.
It’s actually a helix.
Fair enough — time may be a helix. But my SRC still needs to know what sample comes after the last One
The only offline way to do that correctly is to convert the full album with zero extension and chop back into tracks. The most practical way however, and audibly equivalent, is to do it at playback time. Even a Raspberry Pi can do that now.
1 Like
But how? It processes the entire file globally rather than block by block
Block-by-block corrects edge effects by overlapping regions of adjacent blocks. The only difference is that the impulse response of the filter has a fixed number of taps, smaller than the track’s sample count. The results are indistinguishable.
I’m only doing this because it represents the theoretical ideal, not because I believe anyone could actually hear a difference. No one can convince me that even 64 taps in the ASRC of a DAC or audio interface wouldn’t already be sufficient. A €10 quartz Casio can tell the time just as accurately as a hand-wound Rolex.
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.