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

[Julius D Majestic Studios]

As stated early in this video, I believe that polyphony is one of the single most important stats of a keyboard... especially if it's being considered for live use.  It can make or break a deal as far as functionality goes.  We've gotten used to 128 voices and have been at this limit for quite a while.  Maybe 256 should become the standard now...  I think that number would be massive and take us forward for quite a while as far as usability and needs for playing live.

 

While the features shown in this video are specific to Korg flagships (Kronos, Nautilus, etc...) some of the idea and concepts can be applied to other keyboard models too!

 

Questions to consider... (for conversation)

1. What keyboard are you using? and what is the advertised polyphony for it?
2. How important is ALOT of polyphony to you?
3. Do you think it's time (or well beyond past time) for keyboard companies to up the number of voices significantly?
4. Do you check for polyphony stats on a keyboard before you consider buying it? 
5. Do you often find yourself running out of polyphony? 
6. How many sounds are you generally stacking in your layers?
7. How many sounds do you stack before you start running out of polyphony?
8. Would you stack more sounds if you had more polyphony or do you find the polyphony adequate for your needs?
9. What are some of the programming tips/cheats that you use to help with polyphony issues?

 

 

[Julius D Majestic Studios]

Here are my answers to the questions:

 

1. My main keyboard is the Kronos.  But I also use the Nautilus, Korg M3, Motif Rack ES, I recently sold a MODX, Roland Integra 7 sound module....

The Korg boards all have dynamically allocated polyphony except the m3.  The main engine has 120 voices, but I have an expansion card (Radius) that has it's own 24 voices of polyphony to add... so 144 voices combined.  The Integra 7 has 128, and considering the scope of the instrument, I believe it is well under where it should be.  They definitely should have made it 256 voices (Like the Jupiter 80).

 

2. Polyphony is EXTREMELY important to me.

3. Yes well beyond time for an increase, we should be at 256 as standard now.

4. Yes, I always check for that stat.  It's always interesting to me to see if the number has been boosted from 128 yet.  LOL.

5. I use keyboards more in studio than live... because I play organ at church more than anything at the moment.  But I do use keyboards too.  I try to set my sounds up so I don't have polyphony issues.

6.  I try not to cloud my layers with too many textures but still try to make sure there is a thick sound.

7.  When I try to push the limits (just cause), I think 5-6 is around the point that by default, you start seeing issues.  In some extreme cases 4-5 sounds.

8. Probably not more sounds, but the sounds I'd choose may change to utilize the increased polyphony.

9.  Check the video. 

[Delaware Dave]

Many top quality keyboards are at 256 voices, two examples that come to mind are the Kurzweil PC4 and K2700.  The Forte is at 128 voices but KB3 uses 0.  The General Music Pro Mega keyboard with its DRAKE technology introduced in the early 2000's allows for 320 voices.  It is also how it is implemented which is just as important; some keyboards remove the oldest notes first but they may be the most prominent and therefore when they drop they are noticed; Kurzweil used audio level approach where the least prominent notes sustaining are dropped first. Recording into DAW's are probably more noticeable since you may use the internal k/b sounds for bass, drums and other sounds (chewing more polyphony) than playing live with a band.  I use my k/b to play live and  I've not had polyphony issues with my PC3 and its 128 voices. 

[Stokely]

Hmmm, I think it may depend on what you are playing and in what style.

I did tons of gigs in the 80s with my main keyboard being a Jx-10, most of my patches were layered so it was six notes.   You have to pick and choose your parts, and honestly even with more modern keyboards with xx notes of polyphony (I have no clue what my Modx has, I've never run out that I've noticed), I try to be pretty particular about what I'm playing.   In a five piece band that does a lot of background singing there's little reason I can think of to fill up all that much space.  I don't sequence live I should mention.

[Al Coda]

1 hour ago, Stokely said:

You have to pick and choose your parts, and honestly even with more modern keyboards with xx notes of polyphony (I have no clue what my Modx has, I've never run out that I've noticed), I try to be pretty particular about what I'm playing.

 

This exactly !

I learned in the 80s when poly-synths offered 4, 5, 6 or 8 voices at max and while I used a 8-voice poly in split and dual mode.

I still believe the DX7´s success was it´s 16 voice polyphony for it´s,- at that time,- low price.

Unfortunately, most users gave up being creative and programming patches.

 

1 hour ago, Stokely said:

I don't sequence live I should mention.

 

  you´re welcome !

 

My rule:

No sequences and no backing tracks if possible.

 

But when I got hired, I often was not the one who decided.

 

☺️

 

A.C.

[Julius D Majestic Studios]

3 hours ago, Delaware Dave said:

Many top quality keyboards are at 256 voices, two examples that come to mind are the Kurzweil PC4 and K2700.  The Forte is at 128 voices but KB3 uses 0.  The General Music Pro Mega keyboard with its DRAKE technology introduced in the early 2000's allows for 320 voices.  It is also how it is implemented which is just as important; some keyboards remove the oldest notes first but they may be the most prominent and therefore when they drop they are noticed; Kurzweil used audio level approach where the least prominent notes sustaining are dropped first. Recording into DAW's are probably more noticeable since you may use the internal k/b sounds for bass, drums and other sounds (chewing more polyphony) than playing live with a band.  I use my k/b to play live and  I've not had polyphony issues with my PC3 and its 128 voices. 

Please list the top quality boards that are 256... 👀

[stoken6]

I've never noticed a problem with polyphony since I graduated from my 6-voice Roland Juno 2. I've had keyboards with 32, 40, 64 - I genuinely can't remember what my Nord Stage 2 can do (although I remember it's 18 synth voices), but I know it's never held me back.

 

Cheers, Mike.

[David Emm]

 

1. What keyboard are you using? and what is the advertised polyphony for it?

 

CME XKey 37. Just a USB controller. My last piece of  hardware was a Korg TR61, which I finally gave to a music student.

 

2. How important is A LOT of polyphony to you?

 

Semi-irrelevant in a DAW, short of hitting a choke point if you're streaming  a big piano.

 

3. Do you think it's time (or well beyond past time) for keyboard companies to up the number of voices significantly?

 

If you mean in one instrument, yes. Nord has a synth section that note-steals for want of more voices. Its one super-rare bad from them. Kurzweils do well there.

 

In general, no. One DP with a modest poly on top is potent. A synth with 4 split points driving 2 outboard synths, moreso. Voice management is not the same as voice limitations. Korg's Pi-based Wavestate has 64 voices and zero reports of choking.                

 

4. Do you check for polyphony stats on a keyboard before you consider buying it?
  
Yes, but once they hit 64, I had an organasm and never really hit that wall again.   

 

5. Do you often find yourself running out of polyphony?

 

Nope, see above.

 

6. How many sounds are you generally stacking in your layers?

 

It rarely went beyond 3 for me, often just 2. I did more splits than layers and my sub-layers often had a broader zone, so pads went further, for example.  

 

7. How many sounds do you stack before you start running out of polyphony?

 

I no longer have that issue, but when I was learning the Triton, I made it choke a few times.  

 

8. Would you stack more sounds if you had more polyphony or do you find the polyphony adequate for your needs?

 

I learned NOT to stack unless really called for, because some of my lovely Combis were so spectrum-filling, they became better for bong sessions than orchestration! Sometimes the enemy of polyphony isn't the voice count, its the clutter.  

 

9. What are some of the programming tips/cheats that you use to help with polyphony issues?

 

Huh, good question. Above all, learn how Less really is More. A few voices come to life better with an EQ boost or complimentary layer, but effects can be just as beneficial. In 2022, the fix for low polyphony is just a little more gear and a quick snort of testosterone, especially if you get to whiff the air at EMEAPP.   

 

***

 

The query misses me by a few measures, as I've been all ITB with Logic Pro for several years now. You can overload anything, but I don't seem to hit that kind of limit, even less so with an M1 Mac. I can relate to Korgism as a long-time 01W and Triton hound. I learned to make a $@#! track sheet and do the math beforehand. I've come a downright hallucinatory distance since first pounding on my toy piano as a toddler.

 

[Julius D Majestic Studios]

10 hours ago, Stokely said:

Hmmm, I think it may depend on what you are playing and in what style.

I did tons of gigs in the 80s with my main keyboard being a Jx-10, most of my patches were layered so it was six notes.   You have to pick and choose your parts, and honestly even with more modern keyboards with xx notes of polyphony (I have no clue what my Modx has, I've never run out that I've noticed), I try to be pretty particular about what I'm playing.   In a five piece band that does a lot of background singing there's little reason I can think of to fill up all that much space.  I don't sequence live I should mention.

For some reason, I think the MODX has 128 for each engine... (Maybe making it 256 total).  Don't quote me on that though!

[Julius D Majestic Studios]

5 hours ago, stoken6 said:

I've never noticed a problem with polyphony since I graduated from my 6-voice Roland Juno 2. I've had keyboards with 32, 40, 64 - I genuinely can't remember what my Nord Stage 2 can do (although I remember it's 18 synth voices), but I know it's never held me back.

 

Cheers, Mike.

Yeah stepping up from those lower numbers... You are probably in heaven with current systems that are available! 

[Julius D Majestic Studios]

4 hours ago, David Emm said:

 

1. What keyboard are you using? and what is the advertised polyphony for it?

 

CME XKey 37. Just a USB controller. My last piece of  hardware was a Korg TR61, which I finally gave to a music student.

 

2. How important is A LOT of polyphony to you?

 

Semi-irrelevant in a DAW, short of hitting a choke point if you're streaming  a big piano.

 

3. Do you think it's time (or well beyond past time) for keyboard companies to up the number of voices significantly?

 

If you mean in one instrument, yes. Nord has a synth section that note-steals for want of more voices. Its one super-rare bad from them. Kurzweils do well there.

 

In general, no. One DP with a modest poly on top is potent. A synth with 4 split points driving 2 outboard synths, moreso. Voice management is not the same as voice limitations. Korg's Pi-based Wavestate has 64 voices and zero reports of choking.                

 

4. Do you check for polyphony stats on a keyboard before you consider buying it?
  
Yes, but once they hit 64, I had an organasm and never really hit that wall again.   

 

5. Do you often find yourself running out of polyphony?

 

Nope, see above.

 

6. How many sounds are you generally stacking in your layers?

 

It rarely went beyond 3 for me, often just 2. I did more splits than layers and my sub-layers often had a broader zone, so pads went further, for example.  

 

7. How many sounds do you stack before you start running out of polyphony?

 

I no longer have that issue, but when I was learning the Triton, I made it choke a few times.  

 

8. Would you stack more sounds if you had more polyphony or do you find the polyphony adequate for your needs?

 

I learned NOT to stack unless really called for, because some of my lovely Combis were so spectrum-filling, they became better for bong sessions than orchestration! Sometimes the enemy of polyphony isn't the voice count, its the clutter.  

 

9. What are some of the programming tips/cheats that you use to help with polyphony issues?

 

Huh, good question. Above all, learn how Less really is More. A few voices come to life better with an EQ boost or complimentary layer, but effects can be just as beneficial. In 2022, the fix for low polyphony is just a little more gear and a quick snort of testosterone, especially if you get to whiff the air at EMEAPP.   

 

***

 

The query misses me by a few measures, as I've been all ITB with Logic Pro for several years now. You can overload anything, but I don't seem to hit that kind of limit, even less so with an M1 Mac. I can relate to Korgism as a long-time 01W and Triton hound. I learned to make a $@#! track sheet and do the math beforehand. I've come a downright hallucinatory distance since first pounding on my toy piano as a toddler.

 

Nice.... A computer person perspective.  

 

So your polyphony is based on CPU and processing power vs a concrete number.  That makes sure more complicated then right?  Because it would vary based on what else is going on.  Kinda like the Kronos and Nautilus system....

 

Very interesting that people are at home with anything over 64 and happy.  I would be dead in the water with that amount (at least for main keys) BUT...I could make it work if I had to! 

 

Lol @bong session vs orchestrations.  

 

but I will ask this, what classifies as called for with stacks?  And yes less can be more.  You can really cloud a mix vs just making stuff sound big and thick. 

[Al Coda]

10 hours ago, Delaware Dave said:

I've not had polyphony issues with my PC3 and its 128 voices. 

 

That´s why Kurzweil offers the (possibly) fastest voice allocation algorithm since the PC3 series exists.

I still love my PC361 for that.

 

☺️

 

A.C.

 

 

[Al Coda]

7 hours ago, Julius D Majestic Studios said:

Please list the top quality boards that are 256...

 

I´ll try to answer for Dave,-

 

LMGTFY

 

[Al Coda]

9 minutes ago, Julius D Majestic Studios said:

So your polyphony is based on CPU and processing power vs a concrete number.

 

Today, there aren´t such fixed numbers anymore.

For a digital DSP or CPU driven instrument, the number you read in advertisments is the MAX of polyphony you can get,- depending on complexity of patch, number of insert FX in use, number of parts/programs used in a MIDI-multitimbral setup and so on.

In today´s hardware instruments work more or less powerful processors,- some being ancient stock Intel (Korg Oasys, Kronos, Nautilus), Motorola (Access Virus series), ARM, FPGAs (HX3) and proprietary DSP/FPGA (MARA / LENA) in Kurzweil´s machines,- all just only examples because there are more and I don´t know everything.

Ressources are limited always and depending on what you do, you´ll run out of CPU or DSP power sooner or later always.

Compensation is stealing voices and/or FX.

Some do better, other´s don´t.

 

In hi-end digital pianos, especially sample based ones, everything needs "voices",- the so called physical modelled "pedal noise" and "sympathetic resonance", release samples, separately sampled "bells & whistles" sonic artefacs related to the authenticity of the tone and every velocity layer being sampled for a multilayer sample.

In software, 20 layers or more for a single note isn´t a rarity.

 

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".

 

256 "voices" in theory doubles the max. available polyphony while number of FX in use cost  CPU/ DSP power too.

 

Again,- how this is successfully masked is matter of quality of the voice stealing algorithim in the sense of what it does when and why,- and how fast that processing is.

The target is inaudibility always.

 

B.t.w., we´re beyond the theoretical 256 voices limit since some time,- at least w/ a Korg Kronos (up to 400 w/ most humble demands) but also w/ the younger Studiologic Numa X (GT) which is advertise w/ (up to) 300 voices of polyphony.

 

☺️

 

A.C.

[Tusker]

Thanks for your thoughtful questions. This is my response to your question #5 ...

 

I usually use mixed rigs with laptop support, so polyphony is practically unlimited. Or I do simpler DP gigs where there is sufficient polyphony for my needs.

 

In my musical world, having more than three things going on at once is taxing for an audience. I don't remember worrying about polyphony since keyboards hit 128. Sure if I layer strings, pads and piano, then hold down the pedal and play an ascending arpeggio I can hear note stealing before I get to the end of a keyboard. I've done that, but I tend not to play in that style anymore.

 

What is far more interesting to me today is the ergonomics of a keyboard and most of all the sound. Give me a good sound and I'll make it speak the way it needs to, without much need for layers.

[Mighty Motif Max]

My minimum for a main keyboard is going to be 128 mono voices (i.e. Motif XF line). Maybe 120. Anything lower than that and I run out of polyphony for stereo piano + pad layers (even worse if I want to add a bass patch, or anything on an expression pedal). I've had all sorts of keyboards spanning from 256 notes of polyphony. Even 128 mono notes can start dropping if you layer two pads and, say, a string sound (since you have anywhere from 4-15 stereo layers possible in a single sound depending on the keyboard). A lot of current keyboards have switched to having 128 stereo voices, which is actually 256 mono voices. So oftentimes the specs might just say that it's 128 voices, but it's stereo voices rather than the older mono voices. Or you could be like Korg with the Kronos and make it really confusing with dual-mono voices and all that jazz.

 

If we're talking about a secondary keyboard, i.e. one used for synth leads, some organs, various other sounds where you aren't layering anything, 64 is the minimum unless you're only using monophonic/duophonic sounds. This is because it takes a good number of waveform layers in the patches to get the realism we expect nowadays for acoustic instruments and/or the complex pads and synths that are heavily used today.

 

Even the current Roland Fantom has this issue big time - the SuperNatural pianos have pretty low polyphony, and as soon as you layer a pad or something you're down to Yamaha YPG-235 territory (which is a 32-note polyphonic keyboard). Heck, all I wanted to do was use one patch, the preset "Shimmer Pad", but you couldn't layer it with anything at all without frequent note drops (this is with very judicious sustain pedal use).

 

My #1 technique to try to minimize polyphony issues is to be really efficient with sustain pedal usage, kind of like with some classical piano pieces where you smooth the transitions but let up immediately. After that I have to try to edit the patches down to less elements/waveforms and things, or simply just not use them. But generally the sound quality suffers in some way once that is on the table.

 

To answer a few of your other questions:

 

I always check polyphony stats on a keyboard I'm considering, and brand architecture/note-stealing tech differences can play into that as well.

If I'm layering I'm probably layering a minimum of three things. The most I've split/layered for a single regular song is 11 I believe, but that was a pop tune and most of those sounds weren't being played concurrently (i.e. heavy on the splits).

 

My definition of a lot of polyphony on a new keyboard is minimum 256 stereo. Dexibell has exceeded this and some have found ways around it (like Roland's unlimited polyphony in the V-Piano patches). But 128 stereo/256 mono is perfectly good, and 120-128 mono is the baseline for me.

 

 

[AnotherScott]

I have found polyphony to be a pretty useless spec when comparing boards, because it can mean something different on each board. A few examples:

 

I had a Yamaha NP-30 with polyphony of 32, and a MOX6 with a polyphony of 64. Playing their default piano sound, and playing something not particularly musical but designed specifically to see how much polyphony I could get out of them before hearing an audible dropout, I got more out of the NP30. I suspect it's because the MOX6, being a multi-purpose sample playback board, handled polyphony on a simple first-in-first-out basis, whereas the NP30, being a dedicated piano board, had a different algorithm that accounted for--on a piano sound in particular--which previously played notes were least likely to be noticed when dropped, e.g. depending not simply on how long ago they were played, but also possibly things like how forcefully they were played and where in the keyboard range they were.

 

Kronos and Montage both list polyphony for each engine... but Kronos polyphony is subtractive (a second layered sound played by another engine reduces polyphony available for the first) whereas Montage polyphony is additive (a second layered sound played by its other engine has its own polyphony and so does not reduce polyphony available for the first).

 

MODX's sample engine claims 128 polyphony, regardless of whether you're playing a mono or stereo sample. Fantom-0 has 256 polyphony, but each sampled sound take two units of polyphony to allow for stereo... and it reserves those two units of polyphony even if you're only playing a mono sound. This means that, in terms of their ordinary sample playback engines, Yamaha's 128 and Roland's 256 are exactly the same -  128 mono or stereo samples. (Though Yamaha's could conceivably seem to go faster because a single part can have up to 8 elements, whereas on the Roland a single tone can only have up to 4 partials.)

 

Korg Kross (1st gen) had polyphony of 80, or 40 in dual mode. But its piano used 4 units of polyphony per note. So its piano was only capable of 20 notes before a drop-out, which you would never have guessed from its spec of 80 or 40. So at the time, you could choose a MOX with a max polyphony of 64 and get 32 stereo piano notes, or you could choose a Kross with a max polyphony of 80 and get 20 stereo piano notes.

 

In real world use, I have very rarely hit a polyphony limit. I don't run sequences or make use of arps. 99% of the time, the only things you hear are what I'm playing with my fingers. (On rare gigs, sometimes a drum pattern.) I don't use the sustain pedal excessively, and rarely on a sound that isn't some kind of piano. I'm not sure I've ever layered more than two sounds.

 

So, no, I have never considered polyphony when thinking about buying a board. It's almost a non-issue for me. Especially now that you'd be hard-pressed to find any board that doesn't have at least 128, which, regardless of how it's allocated, is likely to be enough for whatever I'd do.

 

 

[Julius D Majestic Studios]

10 hours ago, Al Coda said:

 

I´ll try to answer for Dave,-

 

LMGTFY

 

Lmbo.... Thanks.  

 

My question was somewhat  because I didn't think "many" of the top brands did .... So it was a genuine question as to maybe something I missed.  

 

Just thought maybe since he made the statement and left it unsupported that maybe he knew them or at least some.  

 

Sorry for posting a question on a post (to generate conversation) in a forum where everyone asks questions 😂😂😂. 

 

Maybe there should be a pinned thread with that in the subject line (LMGTFY)? Hahaha.🤷🏾‍♂️

[Julius D Majestic Studios]

9 hours ago, AnotherScott said:

I have found polyphony to be a pretty useless spec when comparing boards, because it can mean something different on each board. A few examples:

 

I had a Yamaha NP-30 with polyphony of 32, and a MOX6 with a polyphony of 64. Playing their default piano sound, and playing something not particularly musical but designed specifically to see how much polyphony I could get out of them before hearing an audible dropout, I got more out of the NP30. I suspect it's because the MOX6, being a multi-purpose sample playback board, handled polyphony on a simple first-in-first-out basis, whereas the NP30, being a dedicated piano board, had a different algorithm that accounted for--on a piano sound in particular--which previously played notes were least likely to be noticed when dropped, e.g. depending not simply on how long ago they were played, but also possibly things like how forcefully they were played and where in the keyboard range they were.

 

Kronos and Montage both list polyphony for each engine... but Kronos polyphony is subtractive (a second layered sound played by another engine reduces polyphony available for the first) whereas Montage polyphony is additive (a second layered sound played by its other engine has its own polyphony and so does not reduce polyphony available for the first).

 

MODX's sample engine claims 128 polyphony, regardless of whether you're playing a mono or stereo sample. Fantom-0 has 256 polyphony, but each sampled sound take two units of polyphony to allow for stereo... and it reserves those two units of polyphony even if you're only playing a mono sound. This means that, in terms of their ordinary sample playback engines, Yamaha's 128 and Roland's 256 are exactly the same -  128 mono or stereo samples. (Though Yamaha's could conceivably seem to go faster because a single part can have up to 8 elements, whereas on the Roland a single tone can only have up to 4 partials.)

 

Korg Kross (1st gen) had polyphony of 80, or 40 in dual mode. But its piano used 4 units of polyphony per note. So its piano was only capable of 20 notes before a drop-out, which you would never have guessed from its spec of 80 or 40. So at the time, you could choose a MOX with a max polyphony of 64 and get 32 stereo piano notes, or you could choose a Kross with a max polyphony of 80 and get 20 stereo piano notes.

 

In real world use, I have very rarely hit a polyphony limit. I don't run sequences. 99% of the time, the only things you hear are what I'm playing with my fingers. I don't use the sustain pedal excessively, and rarely on a sound that isn't some kind of piano. I'm not sure I've ever layered more than two sounds.

 

So, no, I have never considered polyphony when thinking about buying a board. It's almost a non-issue for me. Especially now that you'd be hard-pressed to find any board that doesn't have at least 128, which, regardless of how it's allocated, is likely to be enough for whatever I'd do.

 

 

I'd never heard the Kronos engine categorized like that (subtractive vs additive)... 

 

I've heard that the Fantom 0 isn't truly 256 voices too... I haven't tried it myself though.  But I've heard people saying that voices run out much quicker than what you'd think with that many advertised voices "available". 

 

Again, very interesting to see another person where polyphony limits are a non issue.  I've actually had a bit more of these types of responses than I really figured I'd get.  128 is enough.  Another guy said 64 was good for him. 

 

I've had guys say they are running out after 4 or so voices and really need more.  The answers have been very varied enough to make me think that MAYBE the keyboard companies (in their research) have found that considering the target market of their boards that 128 is plenty for enough people to not warrant the extra cost associated with increasing the polyphony numbers too much.

 

But in church it's quite a bit different... Often times layering as much as is allowed... Minimum 3-4 and often playing big 4-6 key chords in one hand and adding 2-3 from the other...  Polyphony gets eaten up quickly.  

[Julius D Majestic Studios]

11 hours ago, Al Coda said:

 

Today, there aren´t such fixed numbers anymore.

For a digital DSP or CPU driven instrument, the number you read in advertisments is the MAX of polyphony you can get,- depending on complexity of patch, number of insert FX in use, number of parts/programs used in a MIDI-multitimbral setup and so on.

In today´s hardware instruments work more or less powerful processors,- some being ancient stock Intel (Korg Oasys, Kronos, Nautilus), Motorola (Access Virus series), ARM, FPGAs (HX3) and proprietary DSP/FPGA (MARA / LENA) in Kurzweil´s machines,- all just only examples because there are more and I don´t know everything.

Ressources are limited always and depending on what you do, you´ll run out of CPU or DSP power sooner or later always.

Compensation is stealing voices and/or FX.

Some do better, other´s don´t.

 

In hi-end digital pianos, especially sample based ones, everything needs "voices",- the so called physical modelled "pedal noise" and "sympathetic resonance", release samples, separately sampled "bells & whistles" sonic artefacs related to the authenticity of the tone and every velocity layer being sampled for a multilayer sample.

In software, 20 layers or more for a single note isn´t a rarity.

 

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".

 

256 "voices" in theory doubles the max. available polyphony while number of FX in use cost  CPU/ DSP power too.

 

Again,- how this is successfully masked is matter of quality of the voice stealing algorithim in the sense of what it does when and why,- and how fast that processing is.

The target is inaudibility always.

 

B.t.w., we´re beyond the theoretical 256 voices limit since some time,- at least w/ a Korg Kronos (up to 400 w/ most humble demands) but also w/ the younger Studiologic Numa X (GT) which is advertise w/ (up to) 300 voices of polyphony.

 

☺️

 

A.C.

Yeah the voice stealing algorithm is KEY!!! you are right about that! 

 

It's weird to me though that although advertised it seems like people aren't getting the numbers that companies state... That's why I'm glad Kronos/Nautilus actually show it to us, so we can see what's happening while we play and adjust accordingly.  

 

And if they can make a 400 voice piano... Seems other worldy, but also overkill at the same time.  Lol

[Julius D Majestic Studios]

11 hours ago, Tusker said:

Thanks for your thoughtful questions. This is my response to your question #5 ...

 

I usually use mixed rigs with laptop support, so polyphony is practically unlimited. Or I do simpler DP gigs where there is sufficient polyphony for my needs.

 

In my musical world, having more than three things going on at once is taxing for an audience. I don't remember worrying about polyphony since keyboards hit 128. Sure if I layer strings, pads and piano, then hold down the pedal and play an ascending arpeggio I can hear note stealing before I get to the end of a keyboard. I've done that, but I tend not to play in that style anymore.

 

What is far more interesting to me today is the ergonomics of a keyboard and most of all the sound. Give me a good sound and I'll make it speak the way it needs to, without much need for layers.

Interesting.... "Without the need for layers"

[Julius D Majestic Studios]

11 hours ago, Mighty Motif Max said:

My minimum for a main keyboard is going to be 128 mono voices (i.e. Motif XF line). Maybe 120. Anything lower than that and I run out of polyphony for stereo piano + pad layers (even worse if I want to add a bass patch, or anything on an expression pedal). I've had all sorts of keyboards spanning from 256 notes of polyphony. Even 128 mono notes can start dropping if you layer two pads and, say, a string sound (since you have anywhere from 4-15 stereo layers possible in a single sound depending on the keyboard). A lot of current keyboards have switched to having 128 stereo voices, which is actually 256 mono voices. So oftentimes the specs might just say that it's 128 voices, but it's stereo voices rather than the older mono voices. Or you could be like Korg with the Kronos and make it really confusing with dual-mono voices and all that jazz.

 

If we're talking about a secondary keyboard, i.e. one used for synth leads, some organs, various other sounds where you aren't layering anything, 64 is the minimum unless you're only using monophonic/duophonic sounds. This is because it takes a good number of waveform layers in the patches to get the realism we expect nowadays for acoustic instruments and/or the complex pads and synths that are heavily used today.

 

Even the current Roland Fantom has this issue big time - the SuperNatural pianos have pretty low polyphony, and as soon as you layer a pad or something you're down to Yamaha YPG-235 territory (which is a 32-note polyphonic keyboard). Heck, all I wanted to do was use one patch, the preset "Shimmer Pad", but you couldn't layer it with anything at all without frequent note drops (this is with very judicious sustain pedal use).

 

My #1 technique to try to minimize polyphony issues is to be really efficient with sustain pedal usage, kind of like with some classical piano pieces where you smooth the transitions but let up immediately. After that I have to try to edit the patches down to less elements/waveforms and things, or simply just not use them. But generally the sound quality suffers in some way once that is on the table.

 

To answer a few of your other questions:

 

I always check polyphony stats on a keyboard I'm considering, and brand architecture/note-stealing tech differences can play into that as well.

If I'm layering I'm probably layering a minimum of three things. The most I've split/layered for a single regular song is 11 I believe, but that was a pop tune and most of those sounds weren't being played concurrently (i.e. heavy on the splits).

 

My definition of a lot of polyphony on a new keyboard is minimum 256 stereo. Dexibell has exceeded this and some have found ways around it (like Roland's unlimited polyphony in the V-Piano patches). But 128 stereo/256 mono is perfectly good, and 120-128 mono is the baseline for me.

 

 

Yeah I may just skim over it, but I don't ever even notice them advertising mono vs stereo voices as it relates to polyphony. 

[AnotherScott]

10 minutes ago, Julius D Majestic Studios said:

I'd never heard the Kronos engine categorized like that (subtractive vs additive)...

Me neither. ;-) But I thought it was a good short-hand way to describe it.

 

10 minutes ago, Julius D Majestic Studios said:

I've heard that the Fantom 0 isn't truly 256 voices too... I haven't tried it myself though.  But I've heard people saying that voices run out much quicker than what you'd think with that many advertised voices "available".

Roland's figure is deceiving when comparing boards. I gave one example, that when playing a PCM (sampled) sound, it uses 2 instances of polyphony per sound played, even if the sound is mono, meaning its 256 figure is the same as MODX' 128 figure for this kind of sound. Another way it is deceiving is that while it is a multi-engine board where different engines have different polyphony, Roland only provides that one "256 max" figure, and doesn't tell you how much less it is if you're using a particular SuperNATURAL Acoustic or modeled vintage synth engine. This is unlike Kronos, for example, where it tells you that its max polyphony for the HD1 PCM (sampled) sound engine is 140, but if you use the MS-20EX VA engine, the polyphony drops to 40. Roland similarly has sound engines with reduced polyphony, they just don't tell you which ones and what those figures are. So that's another reason its apparent 256 polyphony can go much more quickly than you expect, depending on which sound engine is being used for a particular sound. Korg tells you which sound engine sounds will task the max polyphony, Roland doesn't.

 

10 minutes ago, Julius D Majestic Studios said:

The answers have been very varied enough to make me think that MAYBE the keyboard companies (in their research) have found that considering the target market of their boards that 128 is plenty for enough people to not warrant the extra cost associated with increasing the polyphony numbers too much.

If I was looking at a board with 128, and the manufacturer offered an option to increase it to 256 or even 512 for $100, I'd pass. No need. Zero value to me.

 

Yamaha just increased the FM engine polyphony from 64 to 128 in the transition from MODX to MODX+. I know, some people do all kinds of interesting things with FM, and this will be valuable to some people. But the way I see it, my DX7 had polyphony of 16. So even 64 would allow me to layer four DX7 sounds and play them at their original "full" polyphony, which in fact was always enough for me when I was using one, I never dropped a note on it. (Unless FM polyphony is calculated differently on the MODX than it was on the DX7?)

 

10 minutes ago, Julius D Majestic Studios said:

But in church it's quite a bit different... Often times layering as much as is allowed... Minimum 3-4 and often playing big 4-6 key chords in one hand and adding 2-3 from the other...  Polyphony gets eaten up quickly.  

Using the MODX/Montage as an example, another variable in those figures is that playing a single note could use as little as one unit of polyphony, or as many as 8 for a single-part sound, or even more than 8 for multi-part single instrument sounds. So your choice of sounds, again, can make a huge difference in how quickly you hit a board's max polyphony figure. Which again points to how it is nearly useless to compare the polyphony figures on these boards. A single note of a single-element sampled synth sound on the Yamaha will use one unit of its 128 polyphony. But if you choose that one "9 drawbar" sound that was specifically designed to allow you to control all its drawbar levels, the 9 drawbars plus key click and leakage tones would mean that pressing a single key would consume 11 instances of polyphony out of your 128 available. So again, the rate at which you consume polyphony playing 4-6 key chords using 3-4 layers can vary tremendously depending on the particular sounds you choose. And that's even before you get to the additional variables regarding your use of sustain pedal, or how long the release stage of a sound's envelope is.

 

In the end, outside of very extreme examples, I think it may be nearly impossible to figure out whether a board has more or less polyphony than another for your particular usage, except by playing the boards and trying it out. I mean, think it's quite possible that two people could walk up to a Kronos and a Montage, create their personal favorite layers on both, and one person might run out of polyphony more quickly on the Kronos while the other person might run out of polyphony more quickly on the Montage. Or Fantom, whatever.

 

12 minutes ago, Julius D Majestic Studios said:

Interesting.... "Without the need for layers"

 

Yeah, I don't layer much. Like I said, two, maybe three, rarely if ever more than that. Some examples where I layer: A piano and pad or strings (not both). A horn or string patch where I load two instances an octave apart, typically so I can play a bigger sounding unison line. As coincidentally just mentioned in another thread, layering two string patches in a Fantom-0 for a fuller sounding ensemble. A bass line with a synth line for some songs where I play LH bass, and the bass and some other sound are doing basically the same thing. Maybe layering a piano and EP, or two pianos slightly detuned (can give you a better honky-tonk effect than a chorus effect does). But 95% of the time, I'm playing a single sound. My multi-sound patches are much more often used to create splits for quick access to different sounds I'll need at different parts of a song, than layers. And splits don't inherently use any more polyphony than single sounds.

 

But maybe the church situation you're talking about calls for different things. For me, if I'm doing a live performance, it's usually with a full band, where there's no reason for me to try to fill up the entire sonic landscape. If I'm playing solo, it's usually solo piano. (I've heard wonderful gospel solo organ, though it's not a style I play.) If I'm recording, I'll put down parts individually. So there's never a situation where I'm playing tons of sounds simultaneously.

 

30 minutes ago, Julius D Majestic Studios said:

Yeah I may just skim over it, but I don't ever even notice them advertising mono vs stereo voices as it relates to polyphony. 

 

Sometimes you do need to read the fine print! As I mentioned, in sampled sounds, the mono/stereo distinction shows that Fantom's 256 is no more than Yamaha's 128. (And actually less, since on the Yamaha, adding a sound from its second engine increases total polyphony wheres employing sounds from other Roland engines can reduce effective polyphony. But that's back to other issues, not stereo/mono.) This particular comparison used to be a bit more straight-forward, it used to be common that a mono sound took one instance of polyphony and a stereo sound took two. But unlike earlier Yamahas and Rolands, Montage/MODX and Fantom/Fantom-0 don't work that way anymore. Nord was an earlier outlier here, with pianos that would be 60 stereo and 40 mono (not the 2:1 ratio one would have expected). And as I mentioned, Korgs have been using 4 instances of polyphony per stereo note. So even to the extent that comparing polyphony figures may be useful (a somewhat dubious proposition to begin with, as I've described), it can be that much less accurate if you look at figures without taking stereo/mono into account.

[Stokely]

I will say, despite my comments about not worrying about it--the reports of noticeable voice stealing on the Fantom has me a bit concerned. I'm considering getting one and envision it being the only board at some gigs (maybe) the way my Modx is now...and while I don't do a ton of layering, I do some.  Mostly with a pad to give piano parts a bit of help.   I say a "bit concerned" because again I don't normally do any huge sustain-pedal playing especially if there's a pad on there in my rock band gig....

If you do use your workstation as an actual workstation (sequencing), then yeah polyphony could be a big deal.

[ProfD]

Considering I could get through most any gig situation with 8 notes of polyphony, 128 is a buffet. 😁

 

Of course, the type of music one plays will determine how much polyphony and how many KBs they need.😎

[Delaware Dave]

18 hours ago, Al Coda said:

 

I´ll try to answer for Dave,-

 

LMGTFY

 

 

LOL ....

[RABid]

In the 80's I was happy with my Chroma (8 voice when using 2 osc's, which was always for me), MemoryMoog (6 voice) and Moog Source (1). When companies with ROMplers advertised 32 voice max, but 2 stereo samples took 4 voices you learned to be weary about advertised polyphony. I just considered them 8 voice units and went with it. 

 

Moving from the Virus TI to some of the DSI polly's felt very restrictive to me in a home studio setting. The same polyphony that I was happy with on stage in the 80's was no longer enough for me when doing electronic music with lots of layers. I ended up selling all of my DSI units. Now I have a Jupiter Xm that will play 4 parts plus drums and has up to 256 note polyphony. Even if the sounds I use cuts that by 1/4, I'm in good shape. At this point I am satisfied. 

[stoken6]

6 hours ago, AnotherScott said:

Nord [...] pianos that would be 60 stereo and 40 mono

Yeah, how does that work? Are they encoded M/S with the Sides channel at half the sample rate?

 

Cheers, Mike.

[AnotherScott]

14 minutes ago, stoken6 said:

Yeah, how does that work?

Beats me!

[David Emm]

20 hours ago, Julius D Majestic Studios said:

Nice.... A computer person perspective.  

 

So your polyphony is based on CPU and processing power vs a concrete number.  That makes sure more complicated then right?  Because it would vary based on what else is going on.  Kinda like the Kronos and Nautilus system....

 

Very interesting that people are at home with anything over 64 and happy.  I would be dead in the water with that amount (at least for main keys) BUT...I could make it work if I had to! 

 

Lol @bong session vs orchestrations.  

 

but I will ask this, what classifies as called for with stacks?  And yes less can be more.  You can really cloud a mix vs just making stuff sound big and thick. 

 

My polyphony is based on whatever strikes me, now, thanks to Apple's zooming M1 chip design and being all-SSD. Its LESS complicated, being what I call grab-&-go. Its a cousin to a problem one can solve by throwing enough money at it. Its vastly easier than squinting into an 80s-type EL display. 

 

Yeah, 64 voices, managed carefully, will do a lot. As an at-homer, I have time to proceed at my leisure. Live playing is another world, so I hear you. That's one reason I like and encourage two-board solutions. You gain far more in power than you lose in management-sweat. 

 

As to bong sessions, I submit this excellent NSFW moment from the great Bill Hicks. 

 

As for stacks, mine have generally been larger versions of piano/EP/strings/choir. My 01W was only a 32-voice tool, but by splitting & layering, I eventually came up with a very playable live-rock AC piano. It had balls on the low end because I included 2 octaves of bass strings. Its also useful to remember that tweaking the filter envelope can give you a touch-sensitive element, even in hardware not equipped for it. Early players got a lot of quality mileage from that one. If you landed a vintage synth, that would make it much easier to slot into a modern rig.