In the mood 4 sharing. My 4 VST as OSM's 4 U :)

[tor]

Have a look at the first post. I have uploaded a tidy collection of my direct form 2 transposed filter with regular single precision, doble precision and over sampled implementations. I have also added allpass, notch and 2 bandpass filters in addition to the highpass, low shelf, peak, high shelf and lowpass filters. In my VST's there are alot of mess and wrong labeled modules. It might be somewhat confusing and I am sorry for that.

Enjoy the filters.

[tester]

Thanks. But I don't see the double precision version there or I'm missing something? I don't see dbprecision prims anywhere, and doing side by side comparisons on example (peak) filter - both have the same content (within the code of course) in "coeffs" and in "engine". Should there not be something different in the engine part?

[tor]

You are right

I managed to sync the two modules. Dont know why or how. But now fixed.

Guess I was not so tidy after all

[tester]

Yep, syncing in SM/FS something produces crazy results! Many thanks, and I guess I start checking it tonight or tomorow.

So let me understand this thing correctly. If I'd like to add "double precision" routine to these upsampled middle version - I would just need to take these equations from the engine code and convert them using double precision prims?

A project I work on - does not have band limitations for filtering routines, so I'm not sure how to approach it. My "monkey style" thinking tells me to do both: db-prec and oversampling, but I might be wrong in that concept.

[tor]

As I have mentioned earlier (think it was in the SM forum thread) it is not possible to have both oversampled and double precision biquads in SM/FS properly at the time beeing.

[tester]

Ah I see. I probably misunderstood. Okay, thanks for help. I see where I am after that.

[tor]

The compromize is to use what ever is best for your needs.
Double precision is good for low frequencies.
Over sampling is good for high frequencies.

[tor]

I have also explained why the up sampling is in front of the double percision biquads in Equilibre. But it does not add quality and if one tried to process high frequencies with that implementation you get alot of unwanted artifact in the sound.

[tester]

It's just a lot of new informations for me, so part of it - escapes my attention.

Generally the concept I'm working on is made of 5 slots of varions filters, plus stereo/midside/mix modification plus formant filtering plus wavetable based modulators for all parameters plus modulators for speeds and depths of primary modulators plus additional voices with delay (assigned to slots)... and modulators... From what I see - FS based app should handle it. As for modulators, these are generally slow ones, counted in seconds or few Hz. And various cross connections and rewirings. Right now I'm in touch with the gui part (...and after extensive optimization/reduction, I see there there will be around 2000 preset parameters to store anyway).

As for filters, LP/HP, LS/HS, Peak - are the basis here. As for notch, and bandpass versions I don't know if I will keep them or replace by peak + some switches. On the other hand, there will be multipole butterworths (rather sharper slopes) with some mods. Among other things - something like chebyshev in cool edit would be interesting thing, with dynamic ripple control, but I don't know if what's on SM fits it. Generally - I'm still looking for cool sounding concepts on that filtering part. Due to extensive modulation routines - I'm rather focusing on the differences within dynamic flow of filtering.

[tor]

I see, we have all been there and I am still
DSP is hard especially if you as I dont have the math knowledge it takes to grasp it all. I have not been working with DSP in a long time now and lots of things have to be refreshed to get going again.
I think you will probably find the most interesting sounding filters for modulation in Dozius' filter collection.

Best luck to you.