KVR Audio

kv331 · Post by **kv331** » Thu Oct 29, 2015 7:09 pm

chroma wrote:can't you just route the envelope going to the VCF to the VCA in the other direction (with smaller amplitube)? so that as the envelope drives the filter cutoff down, it raises the volume?

There's actually a filter gain parameter to adjust output volume of the filter so yes, you can route the envelope to the filter gain

If there's a will, there is a way in SynthMaster

kv331 · Post by **kv331** » Thu Oct 29, 2015 8:22 pm

and here comes the 12db/oct MS20 LP filter, implemented with zero delay feedback topology:

http://www.synthmaster.com/synthmaster/ ... filter.wav

the same pattern is played at 5 different resonance settings. input pregain and filter drive adjusted to give this nice resonant tone.

CPU usage is pretty good, close to the 2 pole digital filter

Now I'm gonna try I havent tried before: What if we calculate L/R channels together using SIMD instructions. that should hopefuly lower the CPU usage (in theory!)

Trancit · Post by **Trancit** » Thu Oct 29, 2015 11:51 pm

kv331 wrote: But compensating for volume drop due to filter sweep? Go use a compressor for it

Did you ever tried this with a dense supersaw sound with lots of delay and reverb... this sounds more than shitty...

can't you just route the envelope going to the VCF to the VCA in the other direction (with smaller amplitube)? so that as the envelope drives the filter cutoff down, it raises the volume?

The problem in first place is, that the volume loss of the filter is not linear and it is nearly impossible to compensate this with an envelope (imho)...

Why isn´t this compensation inside of the filter even worth to consider???
In therory, there is no better place to measure and to manage this...

Trancit · Post by **Trancit** » Fri Oct 30, 2015 12:20 am

Btw, no offense... just curious about something like that... would save me a lot of time

kv331 · Post by **kv331** » Fri Oct 30, 2015 4:24 pm

kv331 wrote:and here comes the 12db/oct MS20 LP filter, implemented with zero delay feedback topology:

http://www.synthmaster.com/synthmaster/ ... filter.wav

the same pattern is played at 5 different resonance settings. input pregain and filter drive adjusted to give this nice resonant tone.

CPU usage is pretty good, close to the 2 pole digital filter

Now I'm gonna try I havent tried before: What if we calculate L/R channels together using SIMD instructions. that should hopefuly lower the CPU usage (in theory!)

Just tried the SIMD implementation, holy sh*t! Why haven't I done this before

With nonlinearities and everything the MS20 LP filter is about 30% faster than our 2 pole digital LP filter!

and this is debug mode, in release mode it might be faster

iPlogger · Post by **iPlogger** » Fri Oct 30, 2015 6:28 pm

Wow! Great news. Any savings is greatly appreciated.

Not to derail this thread, but you had mentioned granular with links to examples. Is that on hold while you work on the filters or are they both being worked on in parallel?

kv331 · Post by **kv331** » Fri Oct 30, 2015 7:15 pm

iPlogger wrote: Not to derail this thread, but you had mentioned granular with links to examples. Is that on hold while you work on the filters or are they both being worked on in parallel?

Granular synthesis, direct disk streaming, sample editor, wavetable editor, etc will be new features in v3.0

iPlogger · Post by **iPlogger** » Fri Oct 30, 2015 8:20 pm

Thanks for the info. That is some great stuff to look forward to. Along with the new filters.

EvilDragon · Post by **EvilDragon** » Fri Oct 30, 2015 8:36 pm

kv331 wrote:Just tried the SIMD implementation, holy sh*t! Why haven't I done this before With nonlinearities and everything the MS20 LP filter is about 30% faster than our 2 pole digital LP filter! and this is debug mode, in release mode it might be faster

Hehe.

You should probably hang around more in the DSP section of KvR

kv331 · Post by **kv331** » Sat Oct 31, 2015 9:51 pm

EvilDragon wrote: You should probably hang around more in the DSP section of KvR

Maybe you're right but I don't have time

Anyhow, Ladder filter zero delay feedback implementation is also done, wow it sounds wonderful. now let's optimize this one

Dasheesh · Post by **Dasheesh** » Sun Nov 01, 2015 11:06 pm

I know you're busy , but this is just a very friendly little reminder that the phase knob for the low pass and high pass algo's still don't work on mac.

Very much looking forward to the updates. Have a nice day sir.

chroma · Post by **chroma** » Tue Nov 03, 2015 3:02 pm

Trancit wrote:The problem in first place is, that the volume loss of the filter is not linear and it is nearly impossible to compensate this with an envelope (imho)...

Why isn´t this compensation inside of the filter even worth to consider???
In therory, there is no better place to measure and to manage this...

well, if you think about it... the volume loss from the filter depends a lot on what you are feeding into it, and isn't necessarily predictable (without going ahead and completing the calculations) with FM or other modulations going on. and a lot of volume is in your head anyways e.g. apparent volume depends on frequency (it takes less sound pressure to sound loud for treble than for bass, etc), the 'smiley' fletcher-munson curve, etc. The filter may have some inside knowledge itself of nonlinearities etc, but it doesn't help much with a lot of this other stuff.

you could use overall RMS volume to chase an expected volume, but that's basically a compressor. although i mostly use dynamic EQs for things like this now.

kv331 · Post by **kv331** » Tue Nov 03, 2015 3:16 pm

chroma wrote: well, if you think about it... the volume loss from the filter depends a lot on what you are feeding into it, and isn't necessarily predictable

That's why you'd need a compressor. and it shouldnt be part of the filter, a filter should only be a filter

GHOST19 · Post by **GHOST19** » Tue Nov 03, 2015 5:38 pm

A compressor would do a lot of undesirable things on top of compensating for filter volume. If the filters are not gain compensated you can work around it by assigning cutoff and filter volume to a macro ("easy knob" in SM) and do your filter sweeps with the "easy knob". It's not exactly as good as gain compensated filters but should get you in the ballpark.

Trancit · Post by **Trancit** » Wed Nov 04, 2015 2:25 pm

I hear you and with some points I am with you all

...

But....

(there is always a "but", isn´t it???):

What I asked myself plenty of times, why are there some filters ( some are built in synths or samplers, some are fx...) which loosing only 3-4db of volume filtering a saw wave, while others loose 10-15 db filtering the same saw wave (while no sounding better or worse) ????

All with a 24db slope LP.... doesn´t this mean, that there is a way to let the filter loose less volume without having the need to compensate something manualy???

Ladder/acid/ms20/SVF 0df filters+Smooth Param Auto+Multichannel MIDI+Arp facelift in SynthMaster 2.8