quantized response of the P6's cutoff knob

[Christopher]

First of all, let me say that I absolutely love my P6. It has become a mainstay of my little studio.
That said, it is certainly not without its downsides. Most of those are minor subjective things that one can eventually get used to, but there are also a few issues which I can't help but consider objective design flaws.

One of those relates to what many players will probably consider to be the most important knob on an analog synth: the filter cutoff knob. On the P6 this just hasn't got the smooth response that I would expect from an instrument of the P6's caliber.
Particularly with high filter resonance, its response is so evidently stepped, that it is simply impossible to achieve a smooth filter sweep. Instead the result sounds very artificial and—dare I say it—digital.

To illustrate what I'm talking about, I have attached two short audio recordings of a self-resonant filter sweep. File A is from a Moog SUB37, file B is from the Prophet-6.
The difference is striking. While the sweep on the Moog sounds absolutely smooth, the Prophet's sweep is clearly quantized—to the point where I would say that the utility of the P6's cutoff knob as a live performance controller is severely limited.
Mind you I'm not using MIDI here; I am actually tweaking the knobs directly on the instruments.

I am aware that on an instrument like the P6 the potentiometers are not part of the analog circuitry. They are scanned by a processor which then generates the analog control voltages. That of course implies that all pots will inherently have a quantized response. But the same is true for the SUB37—yet Moog manage to make it sound completely smooth. Either they are scanning the knobs at a much higher resolution or they are smoothing the values out in software. Anyway, I wonder why the P6 couldn't do the same?

[dslsynth]

I am sure that DSI staff with reply here too. But just two things to notice while you wait for their response: (1) What DSI instruments usually do is to have fixed tuned filter frequency steps but also to interpolate the steps so that manual sweeps are possible. (2) For the Prophet 12 there have been some noise on the other forum about the filter steps not being in tuned steps so I wonder if that imperfection have traveled into Prophet-6 as well?

In any case this is a limitation of the software controlling the voices so it could in theory be resolved if DSI decides to allocate the resources to fix it.

[_ADSR_]

I thought this was a trade off to be able to tune and play the filter polyphonically?

[Christopher]

OK, so both of you are essentially saying that DSI are doing this on purpose. And indeed, the cutoff knob is quantized more or less exactly in semitones, which would be quite unlikely if it wasn't a deliberate thing.

I don't own any other DSI gear, so I don't know how this works on a P08 or P12 etc. but I can sort of understand that on those instruments it would indeed be preferable to have the cutoff knob quantized like that. That's because on those synths you have extensive modulation capabilities that allow you to simply use other controllers like the mod wheel or touch sliders if you want smooth filter sweeps for a live performance. You're not limited to the cutoff knob (I am assuming here that mod wheel or touch sliders would give you a smooth non-quantized sweep?)

On the P6 however you just can't do that. The mod wheel is hardwired to the LFO and the only other options are using aftertouch (difficult to control precisely) or hooking up a pedal to the "LP FILTER" input jack.
For all intents and purposes, grabbing the filter cutoff knob is the only practical option for sweeping the filter live on a P6.

That raises the question if the policy to always quantize the cutoff knob should really be a global standard for all of DSI's products, or if instruments like the Prophet-6 (and possibly the upcoming OB-6) would—because of their inherent limitations—benefit from a different approach in this particular case.

[blewis]

That's a pretty well worded description of why it matters for the P6. I had not thought of it specifically in those terms. But you've nailed it.

Another way of possibly addressing this is another feature request a couple of us had where the LFO, when set to a 0Hz triangle or 0Hz random would actually let the player sweep through values with the modwheel from 0-127. At least we'd have some sweeping capability with the mod wheel instead of it always doing its one trick - "wow wow wa wa wow wow" - which gets a bit old.

[_ADSR_]

...the only other options are .... or hooking up a pedal to the "LP FILTER" input jack.
For all intents and purposes, grabbing the filter cutoff knob is the only practical option for sweeping the filter live on a P6.

The "only other practical option for sweeping the filter live on a P6 is hooking up a pedal to the "LP FILTER" input jack."

As they say, "fixed"!   

edit:  maybe this has the same quantization issue.  I really don't know. 

[Christopher]

Another way of possibly addressing this is another feature request a couple of us had where the LFO, when set to a 0Hz triangle or 0Hz random would actually let the player sweep through values with the modwheel from 0-127.

That would be a really elegant solution. I like it a lot!
I have added a few thoughts to that thread.

[silver_surfer]

OK, so both of you are essentially saying that DSI are doing this on purpose. And indeed, the cutoff knob is quantized more or less exactly in semitones, which would be quite unlikely if it wasn't a deliberate thing.

I don't own any other DSI gear, so I don't know how this works on a P08 or P12 etc. but I can sort of understand that on those instruments it would indeed be preferable to have the cutoff knob quantized like that. That's because on those synths you have extensive modulation capabilities that allow you to simply use other controllers like the mod wheel or touch sliders if you want smooth filter sweeps for a live performance. You're not limited to the cutoff knob (I am assuming here that mod wheel or touch sliders would give you a smooth non-quantized sweep?)

On the P6 however you just can't do that. The mod wheel is hardwired to the LFO and the only other options are using aftertouch (difficult to control precisely) or hooking up a pedal to the "LP FILTER" input jack.
For all intents and purposes, grabbing the filter cutoff knob is the only practical option for sweeping the filter live on a P6.

That raises the question if the policy to always quantize the cutoff knob should really be a global standard for all of DSI's products, or if instruments like the Prophet-6 (and possibly the upcoming OB-6) would—because of their inherent limitations—benefit from a different approach in this particular case.

From this video of dave tweaking the filter knob on the ob6 it dosn't seem to have quant. cutoff, sounds smooth?
What do you guys think

https://www.youtube.com/watch?v=wsgeCajeb_Q

[Toby Dammit]

...the only other options are .... or hooking up a pedal to the "LP FILTER" input jack.
For all intents and purposes, grabbing the filter cutoff knob is the only practical option for sweeping the filter live on a P6.

The "only other practical option for sweeping the filter live on a P6 is hooking up a pedal to the "LP FILTER" input jack."

As they say, "fixed"!   

edit:  maybe this has the same quantization issue.  I really don't know. 

Yes I tried with an expression pedal linked to the cutoff and the quantization is present too.

[TomNemesis]

Another way of possibly addressing this is another feature request a couple of us had where the LFO, when set to a 0Hz triangle or 0Hz random would actually let the player sweep through values with the modwheel from 0-127. At least we'd have some sweeping capability with the mod wheel instead of it always doing its one trick - "wow wow wa wa wow wow" - which gets a bit old.

Just turn on LFO sync then use the square wave LFO and slowest speed. This gives you that result for any keys held for less than 7 seconds. It just sucks that you have to sacrifice the LFO section to do it.

Until a better solution can be found, I would very much agree with programming the square wave LFO to remain a static value when the frequency is set to zero. It would make the most sense of any of the LFO shapes to use for that.

[Hector Space]

 The oscillators have a similar issue because they're not continuously variable either. The osc frequency controls works in a similarly quantised way. Yet it is clear that you can do smooth pitch glides on the P6, but not directly from the osc control panel.

The p6 filters should track nearly perfectly and when the low pass filter is self oscillating (high resonance) it should be possible to use the low pass filter as an extra oscillator. But the fact is that the cutoff control doesn't provide a precise enough tuning control to enable exact tuning of the low pass filter in this mode. The quantisation of the cutoff control tends to place the filter pitch just off! This is even after a warm up and re calibration. I must say I wasn't expecting this limitation. Back in 1982 even my Juno 60's filters could hold a pitch, or be tuned just off.. Precisely!

The underlying question is exactly how the P6 translates voltage control into pitch. Typically a linear voltage control is translated into some sort of exponential law to match the way we hear pitch. Back in the good old days the conversion law required the use of a transistor held or pushed into its non linear control range. This always required the semiconductor to be heated into a temperature controlled environment. The stability of this technique is largely why most analog synths are out of tune a lot of the time!

The fact that the P6 requires temperature calibration suggests that it does use some sort of non linear semiconductor translation but I'll bet that it's 'bent straight' by software, (corrected) in order to produce the range of control the P6 has. This is probably part of the reason for the quantised step.

It would be good to provide some sort of scale magnification that would allow for smooth sweeps over say a couple of octaves, when using the cutoff.

A LFO DC setting at 0 Hz would allow the mod wheel to control oscillator sync sounds directly through the pitch of osc 1. Just like you could on the good old Moog Prodigy! Let's hear it for LFO DC... Please 😀😃😜

[bigbrass]

The fact that the P6 requires temperature calibration suggests that it does use some sort of non linear semiconductor translation but I'll bet that it's 'bent straight' by software, (corrected) in order to produce the range of control the P6 has. This is probably part of the reason for the quantised step.

P6 voices are implemented on 6 soundcards. Everything is surface mount.
The cards use a 240 pin DDR3 DIMM board edge connector format.
P6 VCO's use ordinary 2N3904 NPN matched transistor pairs for the exponential pitch cv generation, and temperature sensing resistors in conjunction with software to achieve this level of temperature stability. Unfortunately, it looks like an integral part of this "combo" scheme is always having the pitch control voltages quantized into a known and predictable state, so the software can easily apply corrections over the wide temperature range based both on calibration(s) and tracking "real time" drift. It's worth adding that although the transistor pairs used in the exponential converters are cheap, dual 2N3904's, the op-amp's are precision zero drift Analog Devices parts.

[Hector Space]

Thanks for the tech background, very interesting! Certainly stave off me looking inside for a while 

The conclusion on quanised nature of the P6's CVs would be fine except that I don't think all pitch and cutoff controllers produce quantised effects. Glide doesn't (It'd be pretty interesting if it did - like the old DX7 could!), nor does LFO or Polymod env 1 sweeps or indeed filter envelop sweeps. These produce smooth changes. So it isn't fixed, it just seems to affect control pots and pedal inputs (both volume-which overrides main volume, grrr! and filter). My guess is it could be enhanced to smooth out the zipper noise.

[BobTheDog]

The fact that the P6 requires temperature calibration suggests that it does use some sort of non linear semiconductor translation but I'll bet that it's 'bent straight' by software, (corrected) in order to produce the range of control the P6 has. This is probably part of the reason for the quantised step.

P6 voices are implemented on 6 soundcards. Everything is surface mount.
The cards use a 240 pin DDR3 DIMM board edge connector format.
P6 VCO's use ordinary 2N3904 NPN matched transistor pairs for the exponential pitch cv generation, and temperature sensing resistors in conjunction with software to achieve this level of temperature stability. Unfortunately, it looks like an integral part of this "combo" scheme is always having the pitch control voltages quantized into a known and predictable state, so the software can easily apply corrections over the wide temperature range based both on calibration(s) and tracking "real time" drift. It's worth adding that although the transistor pairs used in the exponential converters are cheap, dual 2N3904's, the op-amp's are precision zero drift Analog Devices parts.

DSI synths have the frequency knobs quantised, it has nothing to do with temperature stability but rather the ability to accurately choose a frequency from the front panel.

[Hector Space]

As already stated in this thread, pretty much none of the p6 panel controls act directly. Even the master volume is a digitised pot. It is very difficult to make 127 levels (typical midi cc e.g. volume) into a smoothly changing control. Which is why midi pitch bend is a separate two byte data, control message.

It makes good sense to make oscillator pitch selection quantised to semitones with a single knob. The Mini Moog needed two knobs to deliver a similar function. Having the additional osc 2 fine tune knob allow the p6 to cover most of what you might want.

The case for quantised filter cutoff isn't so clear. Perhaps a clue is the NRPn filter cutoff range is 0 to 164. Midi cc would only give you 127. The equivalent osc frequency NRPn is only 0 to 60. So more than twice the precision. The filter cutoff range covered is massive as well! I suspect the depth of control needed to make the discrete steps inaudible would be similar to the midi pitch bend message (14bits ~ 1,6383).

[blewis]

For those of you reading, if you want the mod wheel to be able to manually sweep the filter by setting a frequency of 0, please file a feature request with DSI through their support line.

I did and Carson was kind enough to explain that the feature had been discussed but had not made the cut. At least communicate with them what you want.

I am convinced via this thread that a 0Hz Square setting is the most logical - that way the mod wheel still controls the depth of modulation. If the square is stuck at 127, then you're controlling the depth of 0-127. I like it.

[Hector Space]

+1.. 😎