Filter brewing for the Soekris R2R

Jump to Latest
#1,521
TNT said:
The author had a lot of knowledge about filters and signal processing and did not reveal all details on how they where made so I think it would be hard/impossible. Something in 2k can never be identical/ported to 4k.

//
Click to expand...

That's depressing... 🙁 I wish this were my field - it would make for such an interesting hobby. But as I'm mostly finished with my build I might be able to pick up some relevant knowledge from time to time.
 
#1,525
TNT said:
Something in 2k can never be identical/ported to 4k.
Click to expand...

If the filter is an FIR then this isn't correct. A shorter filter can be perfectly implemented in a longer tap implementation. The additional 2k coefficients are simply set to zero.

Audacity might be a useful tool to create the coefficients for FIR filters - in the version I'm using it goes up to 8k taps in the 'equalisation' pull-down.
 
Last edited:
#1,526
Of course one could use only 2k taps in a 4k system and this would be equivalent. But I felt here that this was a "higher number must be better" situation.... so something like "same but better" - now, we don't even have the input ideas for this filter or the biquad text file original. So a "port" is not possible for a number of reasons. I still stand by that it is not possible to fully use 4k taps (no "filling, but actual biquads) to equal a 2k filter i.e. to make them *identical*. Even if it was possible - why would you - you haven't gained anything 🙂

//
 
#1,527
TNT said:
Even if it was possible - why would you - you haven't gained anything 🙂

//
Click to expand...


Yes, you have: you can use pre 119 filters, as long as you have their txt files with the current firmware. To my knowledge none of the best sounding filters have been ported, so the only option to use them was to roll back to 106 firmware.

So, should we gather your claim "Something in 2k can never be identical/ported to 4k" was incorrect?
 
#1,533
Well, if you study the biquads resulting from a filter producing program like e.g, rephase, you will see that some filters do end up with a zero but others don't. I don't know why this is. But my reasoning is based on that if you want to do zero padding "un-noticed", your original filter better end with a zero or you will actually have a little "bump" transitioning into the padded part of the filter. This bump will have an impact on the analogue side of course.

//
 
#1,534
I was only talking about an FIR filter, not about biquads. I haven't any experience in implementing biquads. But '2k' and '4k' surely refer to the length of an FIR filter, biquads are only a couple of samples in length.

The 'zero ending' is no more of an issue with 4k than it is with 2k. Hopefully the coefficients at the extremes (furthest from the central impulse) taper away to zero anyway so the 'bump' effect is minimal. But all practical filters are finite in length so there will be inevitable 'bump' no matter how long the filter. Its not anything to get concerned about.
 
You must log in or register to reply here.
Top Bottom