John Bowen Synth Design

Ken, as I have mentioned here, the coder said that he needs to add an additional mode to the Comb filter model to work correctly for K/S. I quite appreciate your clarification in your post, but don't think he saw the current models as having a bug; rather that he just didn't program the mode needed to do proper K/S.

John Bowen wrote:So, we had our meeting yesterday, and Klaus states the best way to solve this will be to add another Mode to the Comb filter. I'm not sure when he can do this, but I am sure this will be easier to implement here.

Hello. I'm Ken Elhardt and saw my name brought up in regards to this thread. I don't know if people get email notifications or whether my post will go unnoticed, but I just randomly dropped by here and saw this thread. Let me clear up some of the confusion. When I had a Solaris here to program patches on, all 4 of the comb filters appeared to have negative feedback or were inverting the sound somewhere within the comb filter. That was one of the bugs I reported to John Bowen. Since that appears to have been forgotten, then that probably means it hasn't been fixed. Karplus Strong as far as I know wants a positive feedback loop. I can't remember whether I tried to setup a KS patch or not, or just happened to notice all comb filters seemed inappropriate for the task. That's not to say you won't get some kind of sound out of KS using negative feedback delay lines, but it may not give the results expected.

I did use some comb filters in some of my patches, but also used the built in digital delay (or was it the chorus, can't remember) for some comb filtering too IIRC. Using comb filters with negative feedback or negating the sound at some point within the comb filter results in a hollow sound (like a square wave) and all 4 Solaris comb filters had that hollow sound. To confirm visually that this is the case with the Solaris, you could set the comb filters to something like 1 ms (assuming they can be set based on ms or seconds) and turn the resonance up a bit, then run white noise through them and record to an audio file, then examine in an audio editor app that can display an FFT. You should see a series of peaks starting at 1000 Hz and spaced 1000 Hz appart like --> 1000 2000 3000 4000 and so on, in other words a natural harmonic series. If the comb filter is negating the signal at some point, you will instead see notches at 1000 2000 3000 4000 Hz, and so on. That means you'll see peaks in between those frequencies at 1500 2500 3500 Hz, and so on. That is not a natural harmonic series and produces a hollow sound.

Makes me wonder if the messed up voltage controlled envelope generators I pointed out to John were fixed or not, or any other bugs I reported.

I'd like to make another small point about these presets - and any presets where you want to use the output from a final VCA for some kind of control or modulating signal - that you don't need to have the Enable Part switch ON for it to work. This only mutes the Part in the final output (Synth) mix. So, for most of your Presets, you have Part 1 Enabled, but if you listen to only Part 1 by itself, you hear very faintly the little noise burst, and this is not a needed element to put into the final mix; it is used earlier in the signal path. (Of course, you CAN leave it in the audio path at the end, but disabling the Part with the switches does not remove it from the Mod Source and Signal Source paths.)

So, we had our meeting yesterday, and Klaus states the best way to solve this will be to add another Mode to the Comb filter. I'm not sure when he can do this, but I am sure this will be easier to implement here.

Well, I owe somewhat of an apology to Weston, it seems. My understanding of how the Solaris handles its Comb filter was incorrect - I asked the main coding engineer, Klaus, about how to create actual KS patches, and also, to read this thread...here was his response:

Interesting to read - getting a Karplus Strong patch seems to be more intriguing than I thought.

In short with the Solaris comb filter there are no means to pluck the chord (= fill the delay line with undamped noise burst).

In a KS patch this is usually done at every start of a note. So there is a special gate or initialization routine with a random generator.
Sometimes even a recording of a plucked chord instrument is used to fill the delay line.

It seems the people in the Forum fortunately found a solution.
Using envelopes for the excitation noise burst sounds like a good idea - I'm looking forward to listening to that on a Solaris.

There you have it. I've asked him if we could make a variation that would allow an easier solution than what Weston has put together in his example Presets, and we will be discussing this and other things when I'm there next week.

Nice work Weston, very good idea to reinject the second VCF in the first mixer, the sound is very punchy.
And the combination of two or three serial VCF does the job.
I was exploring another way, i shall give you my sounds when i'll have complete some more.

For the othes user : if you try to use those sounds : for three of them only sp/dif is assign, if as me, you are using output 1/2 you'll have to assign "synth" to them.

Weston : why do you reinject SP/dif in mixer 1 ? I don't use this connection and don't understand the reason.

There you go. Some of them may not be KS, all of the KS ones have KS in the name! - These presets could do with a better naming structure!

Should be a reasonable starting point.

Let me know what you think and post your improvements. Feel free to send cash and commissions.

BW

Weston Underwood wrote:I've been trying on and off during the day, in between,cutting the grass, strimming and training my hops (yes I know not very rock star).

Finally, at 20 minutes to one in the morning I've made it work, a nice plucked rich sound. I'll describe it tomorrow and try to get the hang of posting patches on the site. Basically it was a combo of comb filter, white noise and two mixers and filters, with the richness added by a rapid 10ms envelope strafing the cutoff filter in the comb.

Add some reverb and jobs a goodun.

BW

Nice, i've tried some hours and didn't manage to obtain it.
In fact, the only thing wich is missing to obtain KS sounds very easily is the possibility to modulate the effects parameters with all the cv family : here "note" will be enough.
I manage to obtain a nice KS sound using the delay, but it's a fix sound ; it doesn't follow the keyboard, obviously because i can't assign the rate of the delay to the note parameter.

I've been trying on and off during the day, in between,cutting the grass, strimming and training my hops (yes I know not very rock star).

Finally, at 20 minutes to one in the morning I've made it work, a nice plucked rich sound. I'll describe it tomorrow and try to get the hang of posting patches on the site. Basically it was a combo of comb filter, white noise and two mixers and filters, with the richness added by a rapid 10ms envelope strafing the cutoff filter in the comb.

Add some reverb and jobs a goodun.

BW

Sorry! I get your drift now. The plucked strings that he did aren't bad but the attack is a bit too soft. I love the crisp attack provided by true KS.

From the point of view of a KS engine, it should work, the components for KS are there, a burst of noise, a delay, a filter and a feedback loop. Somehow the effect isn't quite right and I wonder if its the inability to control the time dimension on the delay that is the issue.

I'll study Ken's patches in the meantime as a substitute!

I've no doubt that either someone will come along and demonstrate how to do it or in the fullness of time the OS will be tweaked to accommodate traditional KS even if it can't now. I love the KS output of the accelerator and it had been a close thing as to which machine I got. John's re-assurance to me regarding KS was one of the reasons I bought the Solaris in the end so I hope for solution. viewtopic.php?f=1&t=16035&p=18910&hilit ... ong#p18910

Jimmy V said that he'd managed it so maybe he'll chip in.

Best wishes

Weston Underwood wrote:Thanks Zahush

I thought of that, but then as you say Ken said he didn't use KS so I can't check his sounds for how to use it!

I think you didn't quite understand what i was trying to say. Of course Ken didn't use KS - since there isn't really a Karplus strong engine on the Solaris.
What he said was that he could imitate KS like sounds, using the features onboard the Solaris (i.e. clever use of envelopes, lfo's and other tricks). So go ahead and check out how he made his sounds.

I've been experimenting more with chains of comb filter and can get slightly unstable sounds which track at the upper 2 octaves of the keyboard but break down toward the lower ranges into harmonics. Still not the classic KP sound.

I wonder if the difference is that pitch / tune on the octatrack implementation is possibly cutoff and sample rate. Could the lack of control of sample rate be the limiting factor, allowing the ability to control pitch with either resonance toward the cutoff filter (which I would imagine will tend toward a sine wave - which is what it sounds like) or frequency controlled by changing the frequency of the delay, thereby keeping the wave itself in oscillation via feedback within the loop structure. (I am however making this up as I'm going along!)

I looked at the equations at the web and knew immediately they are way beyond my understanding!

And to add a little bit more, I believe the Solaris Comb filter is technically a filtered feedback or feedforward structure.

Weston Underwood wrote:Thanks JB. I agree that the comb filter doesn't sound quite right, not as fattened or hollowed out as others Ive used (like the one on the octatrack for example). On the OT the controls are:

PTCH (pitch) controls the resonance frequency of the comb filter.
TUNE changes the pitch by up to 2 semitones up or down.
LP controls the cutoff frequency of an lowpass filter in the feedback signal. FB controls the gain of the feedback signal.
MIX controls the dry/wet signal mix.

My problems are however more likely to be user error than anything else but if Ken can't make it work I'd suggest its broken!

Well, I would say it's not a thing of the comb filter being 'broken'; what we may have here is a difference in design philosophy and/or understanding of what the algorithm should be.

For your octatrack parameters, I see:

1) PTCH and TUNE as the Cutoff Frequency
2) LP as Damping
3) FB as Resonance control

MIX would be something we don't have for the Comb filter.

Does that seem correct to you?

Thanks Zahush

I thought of that, but then as you say Ken said he didn't use KS so I can't check his sounds for how to use it!

Thanks JB. I agree that the comb filter doesn't sound quite right, not as fattened or hollowed out as others Ive used (like the one on the octatrack for example). On the OT the controls are:

PTCH (pitch) controls the resonance frequency of the comb filter.
TUNE changes the pitch by up to 2 semitones up or down.
LP controls the cutoff frequency of an lowpass filter in the feedback signal. FB controls the gain of the feedback signal.
MIX controls the dry/wet signal mix.

My problems are however more likely to be user error than anything else but if Ken can't make it work I'd suggest its broken!

If its working maybe the developer can show you an example of a K-S sound, I find a developer demo a great way of confirming functionality!

Finally maybe an email to Elektron HQ might help, they seem a nice bunch and its likely they'd help out, the Solaris and octatrack aren't what you'd call natural competitors.

BW