Filter modulation (manually) quantizing is killing me!

acemonvw · Post by **acemonvw** » Mon Sep 16, 2013 2:54 am

Hey guys, posted below is a video I made using some of my gear (posted at the bottom). I worked super hard on this piece but find that I should almost remake the entire thing (video included) on the basis that when I modify the radias's cutoff and resonance, I can a hear very audible stepping on the cutoff from 65 to 64 (or any other single digit difference). Am I doing something wrong? Just listen from like 6:20 to 6:50, you'll hear it! Or at least, I hear it. My less audiophile friends don't notice it at all (or as they say, they might hear it, but it doesn't bother them...). Am I being too critical? Can you hear it? Any thoughts on how to reduce that problem?

I don't notice this stepping on the EMX like I do the radias. I could have used the Spectralis, you might say... and I did. I re-recorded it using the spectralis, but then felt it was too overpowering, so I removed it and re-re-recorded it using the radias again... This stepping doesn't happen when I use the LFOs to modify the sound. The Virus TI has a cool feature where it makes smooth transitions when changing settings, to avoid stepping... I don't THINK the radias has this.

Any thoughts? Thanks guys.

Oh, the part you're listening for is the 'acid-like' sound..

http://www.youtube.com/watch?v=CqsIymoTzOM

acemonvw · Post by **acemonvw** » Wed Sep 18, 2013 8:42 pm

Does no one have this problem of filter-quantizing or audible cutoff/res stepping? Or is it hard to hear in the song?

Timo · Post by **Timo** » Wed Sep 18, 2013 10:18 pm

I haven't seen your vid as I'm on an iPad with 3G, but yes the Radias - like many other digital virtual-analogues - have a tendency to alias audibly when stepping through discrete Midi values, creating the oft-called "zipper effect". This is always more pronounced when the resonance is set high, as the individual filter frequencies are intensified. LFOs and EGs use internal algorithms for smoother modulation, and are separate to knob twiddling.

The Virus is fairly unique in that it has algorithms that interpolate (smooth) between the discrete MIDI steps when knob twiddling too. However this does induce a certain amount of lag in doing knob modulations, but it's largely unnoticeable in most practices as it works intelligently depending on the speed you twist the knobs.

cibine · Post by **cibine** » Thu Sep 19, 2013 10:49 pm

I have the same issue.l. It is really hard to get rid of with the Radias. I am not sure if using a pedal makes a difference or using the Egs and LFos as it was mentioned before. It is even noticeable when you use automation in a DAW.

What I do is try to delete as many automation points as possible for longer filter transitions and keep the curves simple.

acemonvw · Post by **acemonvw** » Thu Sep 19, 2013 11:14 pm

Thanks for the reply guys

I did test to make sure that doing this using LFOs reduced (actually completely eliminated) this audible stepping.

I get some (IMO) beefy synth parts, but if I can't modulate the sound manually, it is sort of a waste. Maybe I just have to hook up my EMX and start using that for those types of sounds, but it's a shame. The radias SHOULD be better than that EMX. I made a song using only the radias back in april (the video is on youtube) and I have the same issue. Back then, I didn't actually notice it until I had already posted the song. For this one, I could hear it, but I thought it wasn't a big deal until I listened to it a few times and just couldn't stand it any longer.

Still would like some opinions on how much you guys can hear it in the song, if you get a chance. It's a long song, but you'd only have to go to the 6:00 minute point. Thanks

acemonvw · Post by **acemonvw** » Thu Sep 26, 2013 5:06 pm

Having gone back to the EMX and testing this... I find that it too steps pretty dramatically. I usually don't notice these things because I adjust the knobs fast enough to not notice it, it's only when you modulate the cutoff/res slowly does this stepping really stand out. I know that this is a common issue with VAs, but it's a shame, because they sound really nice. Access (Virus) really did a good job with its smooth algorithm for knob twisting.

The Spectralis, on the other hand, doesn't do this. Unfortunately its Asynth often hits you too hard over the head to fit in the mix, but it does depend on the song.

acemonvw · Post by **acemonvw** » Thu Sep 26, 2013 5:12 pm

If you want to hear the comparison, I had actually created the above song (on Youtube) using the Spectralis when I was completely frustrated with the Radias. Upon listening to the Spectralis's part, I didn't like it as much as the Radias, so I went back and re-re-recorded the song (and refilmed it) using the Radias (despite its stepping quality, it fit the mix better I felt)

This is the second rendition of this song, If you go into it about 2:42 min I believe is when you come across the Spectralis playing that same part...

https://soundcloud.com/midera/the-appro ... storm-with

X-Trade · Post by **X-Trade** » Thu Sep 26, 2013 9:26 pm

I can hear the stepping but I don't necessarily mind it. So much of our music is marked by characteristics of our instruments but it's not strictly a bad thing. In fact in this case it could almost sound intentional at points.

More generally speaking though, I find the Radias filter range and resolution a bit weird. 0-127 is normal but the highs seem to tend to cover a lot of nothing and it doesn't sometimes go low enough without external modulation or at the very least key tracking.

Another thought is, if you're only changing the filter by a fraction of it's full range, it may help to route via modulation the modwheel or another controller to cover just that range. It seems the problem at this point is not the filter parameter itself but the resolution of the input controller (the entire filter cutoff range spread over 127 steps).

X-Trade · Post by **X-Trade** » Thu Sep 26, 2013 9:40 pm

I can hear the stepping but I don't necessarily mind it. So much of our music is marked by characteristics of our instruments but it's not strictly a bad thing. In fact in this case it could almost sound intentional at points.

More generally speaking though, I find the Radias filter range and resolution a bit weird. 0-127 is normal but the highs seem to tend to cover a lot of nothing and it doesn't sometimes go low enough without external modulation or at the very least key tracking.

Another thought is, if you're only changing the filter by a fraction of it's full range, it may help to route via modulation the modwheel or another controller to cover just that range. It seems the problem at this point is not the filter parameter itself but the resolution of the input controller (the entire filter cutoff range spread over 127 steps).

CharlesFerraro · Post by **CharlesFerraro** » Fri Sep 27, 2013 3:57 am

Yeah the filter steps are purposefully calibrated to cover exact midi notes. Most MIDI controlled devices function this way, including many analog synths like Dave Smith Instruments Mopho for example. That is also why LFO or EG modulation won't cause stepping. If the manufacturer interpolates the stepped values of a knob then you can get a smooth transition like on the M-Audio Venom or DSI Prophet 12.

I've been meaning to reply to your original post for a while and tell you what the exact stepped note range is but my entire studio is in transit. Like X-Trade said some of the high notes smear and the lows stop at a certain point. This is due to some interesting band limiting checks Korg built into the RADIAS with its anti-aliasing technology.

acemonvw · Post by **acemonvw** » Fri Sep 27, 2013 4:27 am

X-Trade wrote:Another thought is, if you're only changing the filter by a fraction of it's full range, it may help to route via modulation the modwheel or another controller to cover just that range. It seems the problem at this point is not the filter parameter itself but the resolution of the input controller (the entire filter cutoff range spread over 127 steps).

That is a really good idea!

X-Trade wrote:I can hear the stepping but I don't necessarily mind it. So much of our music is marked by characteristics of our instruments but it's not strictly a bad thing. In fact in this case it could almost sound intentional at points.

Hmm... wow, I guess I never really thought of that. That is a good point. Glad to get some feedback on this, thanks!

X-Trade wrote:More generally speaking though, I find the Radias filter range and resolution a bit weird. 0-127 is normal but the highs seem to tend to cover a lot of nothing and it doesn't sometimes go low enough without external modulation or at the very least key tracking.

Yeah, I've noticed this too.

CharlesFerraro wrote:Yeah the filter steps are purposefully calibrated to cover exact midi notes. Most MIDI controlled devices function this way, including many analog synths like Dave Smith Instruments Mopho for example. That is also why LFO or EG modulation won't cause stepping. If the manufacturer interpolates the stepped values of a knob then you can get a smooth transition like on the M-Audio Venom or DSI Prophet 12.

I've been meaning to reply to your original post for a while and tell you what the exact stepped note range is but my entire studio is in transit. Like X-Trade said some of the high notes smear and the lows stop at a certain point. This is due to some interesting band limiting checks Korg built into the RADIAS with its anti-aliasing technology.

How the hell do you guys know so much about these synthesizers (obviously, this is why I come here, to be surrounded by geniuses). I didn't really expect an analog synth like the Mopho to step. That's really interesting.

Interesting stuff guys. Thanks so much for your feedback

CharlesFerraro · Post by **CharlesFerraro** » Fri Sep 27, 2013 4:44 am

acemonvw wrote: I didn't really expect an analog synth like the Mopho to step. That's really interesting.

Yup, analog synth, digital control. That's why korg is AWESOME for only implementing MIDI note input into the MS-20 rerelease. You get all that real analog resolution in the potentiometers.

X-Trade · Post by **X-Trade** » Mon Sep 30, 2013 8:51 pm

Yeah, the filter algorithm is perfectly capable of operating without stepping. As you would notice when modulating with a controller over a shorter range, or with an envelope or LFO.

However, the knob itself is either connected to a 7-bit ADC, or is otherwise limited at the digital stage to a 7-bit (0-127) data width.

It would be perfectly possible to create knobs with more resolution, which would get around the problem, but the finer resolution would not be addressable by MIDI (actually, it can, but both of these things are widely uninplemented).

CharlesFerraro · Post by **CharlesFerraro** » Tue Oct 01, 2013 1:33 am

X-Trade wrote: It would be perfectly possible to create knobs with more resolution, which would get around the problem, but the finer resolution would not be addressable by MIDI (actually, it can, but both of these things are widely uninplemented).

This is where we talk about HD-MIDI...

http://en.wikipedia.org/wiki/MIDI#The_future_of_MIDI

X-Trade · Post by **X-Trade** » Wed Oct 02, 2013 8:06 pm

I was thinking more along the lines of this:
http://en.wikipedia.org/wiki/NRPN

Which is entirely supported by existing MIDI standards. Unfortunately some DAWs (e.g. Ableton) don't implement it properly or at all, and other controllers and synths that do implement it often only use the 7bit implementation (ignoring the LSB/MSB part). I suppose if you do implement this extensively though you do at least lose half of the bandwidth in the case of controller changes.