Polyphony... Too Little? Too Much? - How Do You Deal With It! Let's Talk About It

[Tom Williams]

On 9/26/2022 at 9:03 PM, Al Coda said:

 

In a Kurzweil PC3 series machine, you´re able cascading up to 32 layers to create a VAST program.

1 layer eats 4 "voices".

So, with a 32 layer program, your available "polyphony" is 4 "voices".

Huh.  I was / am pretty sure that you get 128 simultaneous sounding layers, which would provide a minimum of 8 note polyphony.  The VAST organ from Daniel at SW tricked a single layer into providing two or three sounds by combining a wave source with two SINE+ blocks.  I think.  Somehow he wrung more than 3 organ notes out of a K2000.

 

Also, since most multi-layer Kurzweil factory programs do velocity switching, three program layers may only play one at a time, allowing even more polyphony.  AFAIK, if only three layers are sounding at any time per note, the PC3 will give you ~40 note polyphony.  PC4 / K2700 would provide 80.

 

[Tom Williams]

On 9/26/2022 at 7:13 AM, Julius D Majestic Studios said:

1. What keyboard are you using?
   1. Kurzweil PC4-7, Roland Ax-Synth
      1. and what is the advertised polyphony for it?
   2. 256, 128
2. How important is ALOT of polyphony to you?
   1. 8 out of 10.
3. Do you think it's time (or well beyond past time) for keyboard companies to up the number of voices significantly?
   1. I think they already have, for digital 'boards at least.
4. Do you check for polyphony stats on a keyboard before you consider buying it? 
   1. Yes.
5. Do you often find yourself running out of polyphony? 
   1. Rarely, but non-zero.  Pipe organs with multiple stops can suck up oscillators pretty quickly.
6. How many sounds are you generally stacking in your layers?
   1. Typically two, with the pipe organ exception above.
7. How many sounds do you stack before you start running out of polyphony?
   1. 8 or more.
8. Would you stack more sounds if you had more polyphony or do you find the polyphony adequate for your needs?
   1. Quite adequate at this point.
9. What are some of the programming tips/cheats that you use to help with polyphony issues?
   1. One oscillator + a little chorus or flange = two oscillators.  Even better with multi-tap delays.
   2. Similarly, an LFO-modulated  pulse wave can have motion without consuming extra oscillators or even effects blocks.
   3. One multisample of eight oscillators uses one voice of polyphony.

 

[Al Coda]

7 hours ago, Tom Williams said:

Huh.  I was / am pretty sure that you get 128 simultaneous sounding layers, which would provide a minimum of 8 note polyphony.

 

A PC3 (not PC4 or K2700) provides a polyphony of 128 voices.

Now, when creating a 32 cascaded layers program (not setup/multi), you devide 128 by 32 and the result is 4.

Like the advertised "128 voice polyphony", this is a theoretical value,- but that´s how it works.

You´re free taking all kind of tricks into consideration like voice stealing algorithm perfection, more or less DSP FX in the ballpark and last but not least,- clever programming,- but that´s not what we´re talking about here.

 

Instead,- when we agree on polyphony is VERY important for a workstation machine,- that means, the more DSP power we get, the more "voices" we can play simultaneously and together w/ all kind of "bells & whistles" insert and aux FX, advanced DSP mangling and so on.

 

On my PC361 and w/ the programming I do for my own programs, I rarely can use more than 3 simultaneously sounding internal programs in a setup, just because I use a lot of VA, tweaked algorithms, individual DSP processing and complex insert FX chains for each program in a zone.

With a limit of 16FX I run out of DSP power soon,- which at the end of the day IS a limit of polyphony as well.

So, I appreciate PC4 and K2700 offer 256 "voices", which in real world means they at least doubled the DSP power in these machines.

It´s also the reason why we can use a total of 32FX now.

 

In fact, in a Kurzweil and when it comes to "voices", programs using FM synthesis eat most DSP power while a program playing back a multisample, not so much.

But this is for the pure tone, doesn´t consider complex FX and other DSPmod treatment and can turn into the opposite as well, depending on programming.

 

And an advertised polyphony of 300 or 400 voices is a bluff package too,-

With less sonic quality, only 4 zones and limited insert FX per patch/zone, it´s possible to get such numbers of theoretical polyphony from the same CPU or DSP chip other manufacturers use for less polyphony.

 

IMO, your example above needs the amount of DSP power theoretically necessary for 1024 voices (128 x8) because you say "128 simultaneous sounding (polyphonic !) layers / 8 note poly minimum.

 

☺️

 

A.C.

[MushMusic]

I remember in Grigsby's studio, one of the first projects she had everyone do on her Fairlight was programming the march from L'Histoire du Soldat. The score is on 7 staffs plus drums, but violin and contrabasse has double stops. Getting around the 8 voice limit was easy. While programming all your sounds, the various double stop intervals were done as additional separate patches. As it ran, the violin and the contrebasse parts had to switch patches to play the double stops.

[J.F.N.]

On 9/28/2022 at 11:19 AM, Al Coda said:

 

☺️

 

In a Kurzweil PC3 series machine, you´re able cascading up to 32 layers to create a VAST program.

1 layer eats 4 "voices".

So, with a 32 layer program, your available "polyphony" is 4 "voices".

 

This doesn't make sense, IIRC, 1 layer consumes 1 voice, which then add upp with what you say about a 32 layer cascade, this would then give you 4 voices in total, 128/32 = 4, right?

 

If one layer consumed 4 voices and you create a 32 layer cascade, you would end up with exactly 1 voice, 4x32 = 128, right?

 

 

[Stokely]

In my experience, I don't really agree about polyphony being that important, within reason.  I used to gig a lot with a Roland JX10 as my only keyboard, with the max polyphony of 12 with one tone, doubled to 6 or split to 6 in each zone.  I made it work.

Within reason...I mean a piano with 20 notes of polyphony is likely to have issues.   People have been making great recordings with Prophet 5s for decades, though I doubt anyone is going to throw a Prophet 10 out of bed  

In the studio, it's moot especially with software synths.  If I was recording with hardware synths I could always record each part as audio and move on.  I tend to do multiple, simpler parts at home vs larger layered things.

I have no idea of whether any of my current keyboards other than my Summit even have polyphony limits.  I've never hit them knowingly if they do.  Granted, I don't sequence live and I don't really do much layering beyond adding a pad to piano, or organ to horns etc sometimes--I feel that often "less is more" sonically and layering can actually detract from the sound.

[b3plyr]

23 minutes ago, Stokely said:

In my experience, I don't really agree about polyphony being that important, within reason.  I used to gig a lot with a Roland JX10 as my only keyboard, with the max polyphony of 12 with one tone, doubled to 6 or split to 6 in each zone.  I made it work.

Within reason...I mean a piano with 20 notes of polyphony is likely to have issues.   People have been making great recordings with Prophet 5s for decades, though I doubt anyone is going to throw a Prophet 10 out of bed  

In the studio, it's moot especially with software synths.  If I was recording with hardware synths I could always record each part as audio and move on.  I tend to do multiple, simpler parts at home vs larger layered things.

I have no idea of whether any of my current keyboards other than my Summit even have polyphony limits.  I've never hit them knowingly if they do.  Granted, I don't sequence live and I don't really do much layering beyond adding a pad to piano, or organ to horns etc sometimes--I feel that often "less is more" sonically and layering can actually detract from the sound.

Well, I gigged for years with a K2600 that had 48 voice polyphony and had no issues. Chalk that up to either me being a poor musician or the Kurzweil voice stealing algorithm being great. Of course I now have much more polyphony, but I also do more splits, layering, etc.

[JoJoB3]

I'm not committing just yet but perhaps the most exciting thread title about polyphony I've seen here yet. 💯

Listen, this is all the more reason you are more than justified adding more hardware synths in your dwelling.

[RABid]

I remember buying a Roland SH-32. It was rated as 32 note polyphony. You could layer 4 sounds, each being stereo and using 2 polyphony, and then hit a 4 note chord. The sounds would kick in one at a time, each new sound playing a couple seconds after the previous. It was horribly underpowered, even for a 32 note polyphony engine. I have owned Roland ROMplers that were rated as 64 note polyphony. Each stereo part used two polyphony, and a voice could use up to 4 parts. In this case, one note = 8 polyphony. For most sounds you were limited to 8 notes total. This was actually fine with me. I don't often need more than 8 notes on a single keyboard. The exception is playing piano parts with the sustain pedal in use.

 

The modern Fantom line is a bit more complicated. It is listed as 256 notes when using the sound core engine. From here it gets a bit sketchy but as I understand it, the two chips work together and share the load, avoiding the Yamaha problem of having two chips with separate 128 poly. It gets a bit foggy when using the V Piano or ACB instrument like the Jupiter 8 ACB. The piano is listed as unlimited polyphony meaning you can hit every note on an 88. The ACB synths are limited to 8 notes. The V Piano and the ACB synths must be a singular instance and reside in channel one of the system. What Roland does not say is what happens to the sound core polyphony when you play a V Piano or ACB synth. I did see one video in which the Roland rep said that an ACB synth took 4 times the processing per note as the sound core engine. (Don't ask me which video. I cannot remember and you can search through the various YouTube demonstrations as easy as I can.) Anyway, I've seen complaints that the amount of power on the Fantom is not enough. One YouTubber had 8 patches layered and was playing large, sustained chords and complaining when notes dropped out. My opinion, learn some restraint. If you really need 8 complicated sounds layered and played with two handed chords with sustain on, add more hardware, or listen to your music and ask yourself why you need so much mud.

[Thethirdapple]

Generally, less is more…

 

Polyphony being directly tied to cpu load, it makes perfect sense that a sweet spot has to be found.  I often wonder if “we” have reached unlimited polyphony, given the cpu advances. I would love to see a company go back to installable “cards” which could off load cpu calculations. Much like a modern GPU, or similar to installing more memory in the Triton Rack for increased sampling size…

 

We’ve been using computers disguised as musical instruments for decades now.
 

Not so much needed while playing, but for our goals, polyphony is crucial in generating midi effects.
 

The idea (fantasy) of creating a true midi reverb unit, might be closer than imagined. Not audio signal processing but actually midi signal processing. From a theoretical (dreamers) perspective a midi event created reverb effect would enable each “event” to be modulated as it “reflects”.
 

A wave-sequencing reverb reflection would be super fun!

 

 

PEACE

_
_
_

 

[RABid]

3 hours ago, Thethirdapple said:

A wave-sequencing reverb reflection would be super fun!

About as close as I have found in hardware.

 

[Thethirdapple]

@RABid at around the 13 minute mark, the operator hits really nice spot for me…

 

although I'm not generally looking to eurorack modules, some really groovy gear is coming out!

 

 

 

 

PEACE

_
_
_