Too much polyphony??

[Dave Bryce]

So, right now the industry standard for polyphony is pretty much between 64 and 128 voices. If you put the piano and GM expansion cards into the Yamaha S80, I believe that you can even get it to do 192 voices of polyphony.

 

How much higher does this need to go? Do you think that we'll see 256 voice units? 512 voices? How would *that* sound coming out of 2 outputs? http://www.musicplayer.com/ubb/wink.gif

 

What do you think?

 

dB

[DBENNVAhotmail.com]

Dave Bryce:

 

 

It could sound fantastic although I believe

that expanding the polypony in a system has got to have an increased number of effects (insert/sub-group/master) designed to be a part of those types of systems. I like playing whole combinations(Korg)&

performances(Roland) in the same manner that

I would play a standard program in synths.

It helps you to get the production to sound

thick quickly.

 

Quantum! C/O

DBENNVA@hotmail.com

[Erik Norlander]

Originally posted by Dave Bryce:

How much higher does this need to go? Do you think that we'll see 256 voice units? 512 voices? How would *that* sound coming out of 2 outputs? http://www.musicplayer.com/ubb/wink.gif

 

Good topic, Dave.

 

To quote a friend, "If less is more, think how much more MORE would be!"

 

For an accurate piano, I suggest that you might need 382 voices. Why 382?

 

First of all, each note(modeled, sampled, etc.) should be stereo, with a dedicated representation (model, sample or combination) for every note. So I first ask for 176 voices (88 keys x 2 channels ... duh).

 

But why do you need to be able to sound every note at once? Am I a big Jerry Lee Lewis fan? Actually, no ...

 

When you play one note, every other note across the keyboard should react in some way, even if it is what we engineers call "very, very soft". Hence, 176.

 

Now consider that when you play a second note, the rest of the notes should react in a different way. You shouldn't just choke the reaction of the first wave, you should do some sort of crossfade so it sounds natural. Hence, 382.

 

That's just for the piano. So in an ideal keyboard, 382 should be the minimum. Of course I then want to add stereo strings to the piano, and maybe even a Rhodes to the layer. So how many voices is that? As we engineers say "a lot".

 

Not to mention multitimbral applications. Yikes.

 

When you start to think like this, 1,024+ voices isn't all that outrageous.

 

Cheers,

 

Erik

[Jeff Klopmeyer]

Originally posted by eriknorlander@thetank.com:

 

When you start to think like this, 1,024+ voices isn't all that outrageous.

 

Cheers,

 

Erik

 

Sometimes you frighten me, Erik. http://www.musicplayer.com/ubb/eek.gif

 

Here's another twist on this topic. Erik brings up a great point in regard to the best possible performances using sample-based emulation of sounds like pianos, strings and such. But for certain types of music, does a limited number of voices actually cause you to stretch your creativity in composition and performance?

 

Hmmm. I don't know the answer. Talk amongst yourselves.

 

- Jeff, TASCAM Guy

[marino]

More like that, Jeff: a limited number of available voices *forces* you to stretch!

My personal experience is, now that I have quite a bit of synths and computer power, everything is easier and faster. The only drawback is having to learn many different operating systems. But the advantages of having many sounds/voices at once are obvious.

A lateral question: All those voices should be able to sound with good timing, and that's another stretch on the processor. See the EX5 for example, with 128 voices and choking when you try to play more than 20/30 at once.

 

[This message has been edited by marino (edited 10-27-2000).]

[Jim Aikin]

Marino mentioned the main points I was going to bring up, but I think they're worth emphasizing. The "128 voices!!!" spec is partially hype unless the instrument can trigger, let's say, a minimum of 40 or 50 of those voices **all at once** within a 10-to-15 millisecond window. I haven't performed rigorous tests on the current crop of 128-voice modules, but I'd bet money that the majority of them would fail this test.

 

The other issue Marino mentions is equally apropos: Why would you want to use a single 128-voice module when you could get much more sonic variety (and in all likelihood, more simultaneous effects) by using four 32-voice modules?

 

Erik's idea that you'd need all those extra voices for a realistic piano emulation is a bit over the top, in my view. It's a brute force approach to modeling. A nice fast processor and a decent physical model of the piano (which, in spite of a few advertising claims, I have yet to see actually implemented) should be able to provide both harp resonance and pedal up/down effects (including half-pedaling!) without eating up any polyphony. I'd settle for 64 stereo piano notes; I don't think there's much music that calls for all 88 keys to sound at once, and if you insist on being able to do that, you can always buy two modules.

 

--Jim Aikin

[Pim]

First asked question from my clients in the shop is: How many sounds? http://www.musicplayer.com/ubb/eek.gif

Second asked question: How many voices?

The specs are dictating what is good and what is bad.

The public wants more: the public will get more.

Our so called synthesizers from today, are degenerated to disposable objects, where more is better...

I am not forty years old yet, but am already a "in my time it sounded better" musician.

Unfortunatly! http://www.musicplayer.com/ubb/rolleyes.gif

[Erik Norlander]

Originally posted by Jim Aikin:

Erik's idea that you'd need all those extra voices for a realistic piano emulation is a bit over the top, in my view. It's a brute force approach to modeling. A nice fast processor and a decent physical model of the piano (which, in spite of a few advertising claims, I have yet to see actually implemented) should be able to provide both harp resonance and pedal up/down effects (including half-pedaling!) without eating up any polyphony. I'd settle for 64 stereo piano notes; I don't think there's much music that calls for all 88 keys to sound at once, and if you insist on being able to do that, you can always buy two modules.

--Jim Aikin

 

I was not suggesting that all 88 notes sound simultaneously at a similar volume, I was suggesting that each note should react in some way, even if it was only very softly. This means that one note could be at 100% volume while the other 87 are at only a fraction of 1% volume. When you sound the highest string on a piano, my guess is that the lowest string still vibrates a bit. That's where I was coming from with this idea.

 

As far as using two modules, the phase relationship would likely be unreliable. I even found that when using two mono L/R instruments in SampleCell, they do not remain phase aligned. (However, an instrument made up of stereo samples does stay aligned.) Add the slowness and inconsistency of MIDI to this mix, and you are guaranteed an out of phase experience.

 

But I do acknowledge that my idea is indeed a brute force approach.

 

Cheers,

 

Erik

[Dave Bryce]

Here's a thread that died back in October. I'd really like to hear what some other folks have to say on the subject - anyone else want to chime in?

 

dB

[joegerardi]

I don't know about all this...

 

I mean, a *real* piano is not a stereo instrument, so why must a sampled version of it be? I understand the sympathetic string resonance, but that only occurs during pedalling for the larger part of the piano, so it wouldn't require extra voices except for that, and it *should* be included on the upper keyboard range in the initial sample. Additionally, why couldn't it be a crossfade sample triggered only by sustain for the lower notes? Again, we're not using up any extra voices there.

 

I can understand the use of 128 voices: Long arpeggiated runs on a piano with sustain, like the Chopin C# minor prelude, will eat up voices quickly, however, some of the greatest synth music ever created was done on mono synths. I think we're getting selfish with the need for polyphony. As for effects, is the day of outboard processing gone the way of the Dodo?

 

Processor speed is another matter. I believe it only a matter of time before we see multi-processor instruments offered, with the extra processors handling voice duty to speed up lag. So too, I think we all agree that there's no 1 synth that does everything well, hence the reason we are all multi-synth players. I hear people bemoaning the QSR's strings, this one's piano, that one's organs, etc., so on sequencer playback, I would think you'd assign the synths or modules to the parts that they do best.

 

Finally - and this should be taken in the context of an old gigging musician - can audiences really capture all this nuance? Really, with 60-cycle hums coming from crappy mains, noise being generated by the guitarist's effects boxes, with the drummer mumbling under his breath all the time trying to get a clue - well, studio work aside, is all this modelling of sound really necessary? And if someone is that critical of a synth's piano sound, wouldn't it behoove that musician to get a real piano for the session?

 

Dave:

That oughtta get a flame war going to revive this thread! http://www.musicplayer.com/ubb/smile.gif

[Stephen Fortner]

Jim Aikin,

Maxing out a synth's ability to address its claimed number of voices is one of my favorite pre-purchase torture tests. Your conjecture would be correct: there's nothing on the market claiming to be 128 voices that won't choke fairly readily. Boxes that claim this, such as the Roland VX's, Proteus 2K, or even a Yamaha S-80 with the piano boards in, are doing the equivalent of putting a 350-hp engine upstream of a terribly inefficient transmission. (i.e. MIDI) FWIW, until the industry settles on a higher-bandwidth standard for note and control information, well-thought-out voice allocation is what it's all about. In terms of sitting in a mix, fooling the ear, etc, the Kurzweil K2500, with it's relatively skimpy 48 voices, still does this better than anything.

[musicman1ovation.net]

What about the Nemesys GigaStudio? With all this information about how 128 voices of polyphony might not *really* be as manufacturers claim (because of choking MIDI streams and latency issues), what does anyone make of the GigaStudio's ENORMOUS polyphony claims in the highest level product?

 

And I don't mean how it will ultimately come out on a recording, but how does it stack up when using it in a strictly hardware way. For instance, if you use a MAC for sequencing/DAW, but get GigaStudio to be a massive hardware sampler, can it really put out such a stream of polyphony with MIDI controllers, etc. as it is played live in the studio?

 

[PS - On FEB 17, 2001, I opened a topic with the unfortunately very vague name, "Need Help! ROM/RAM - Hard/Soft? Read On... ", which addresses some of this. I'm afraid the name sounded so broadly ridiculous that nobody cared to check it out and answer. I do apologize. Should have known better than to try to get answers with a name like that! http://www.musicplayer.com/ubb/biggrin.gif Also, I know that post's background information rambles, but I was just trying to be clear about my specific needs. Probably overdid it! I'd still love some feedback, though. It's not as confusing as the name implies, though. It's just a query about whether to get a ROMpler w/sample RAM, a sampler, or GigaSampler (w/PC) as my next sound source. I could really use some thoughtful replies, because my setup goes back to the mid-seventies, and I am attempting to be as efficient a helpful as possible on this particular purchase. I would have already made the purchase, but I'm hoping for some opinions before I leap. I'll be going somewhere between $2,000.00 and $3,000.00, which is enough to cover Roland XV-5080, Kurzweil K-2600R at current prices, and more. I've still got digital audio stuff to update, so I want to be really careful here! http://www.musicplayer.com/ubb/rolleyes.gif]

[Dave Bryce]

Originally posted by musicman1@ovation.net:

It's just a query about whether to get a ROMpler w/sample RAM, a sampler, or GigaSampler (w/PC) as my next sound source. I could really use some thoughtful replies, because my setup goes back to the mid-seventies, and I am attempting to be as efficient a helpful as possible on this particular purchase. I would have already made the purchase, but I'm hoping for some opinions before I leap. I'll be going somewhere between $2,000.00 and $3,000.00, which is enough to cover Roland XV-5080, Kurzweil K-2600R at current prices, and more.

 

Why just get one?

 

If you want to spend up to three grand, I'll bet you can get the Gigasampler, a JV1080, and either a K2500 or a K2000 module if you look around. The Gigasampler is under $300, the JV1080 still sells new for around $900, and there's a whole bunch of new and used Kurzweil options under $2k. If you're willing to get a K2000RS, you will even have a bunch of money left over for CD-ROMs and other accessories. Then you basically get two different samplers, two different ROMplers, a bunch of polyphony and outputs and a pretty varied palette to boot.

 

I think it's good to have ROMplers around - many of the sounds onboard are quite useful and inspiring, and it's nice not to have to take the time to load samples in. Plus, the Kurzweil V.A.S.T.engine is so unbelievably powerful that it is simply amazing what it can do with the raw samples once they're in there. So you get the JV1080 for polyphony, variety and the Roland sound, the Nemesys for processing power, RAM size and polyphony, and the Kurz for its sound, its power and access to an immense library (Kurz samplers read just about everyone's samples).

 

Hope that helps,

 

dB

[Steven Denike]

Another one from 'King of the Dumb Questions', and thank you again, David, for the informative and sagacious answers you have given me in the past, but -

 

again, if 'more is more', then where does something like the Moog performance synth fit in? I kow it's eagerly expected, and once David et al get one, I'd love to know if it will be worth buying, but

 

THAT is MONOphonic...

 

Maybe Les is Moore?

 

pps. I love the sympathetic vibration excursus

[Dave Bryce]

Originally posted by Steven Denike:

if 'more is more', then where does something like the Moog performance synth fit in? I kow it's eagerly expected, and once David et al get one, I'd love to know if it will be worth buying, but

THAT is MONOphonic...

 

So are quite a few of the instruments in the orchestra, and the human voice is as well. Those all rely more on quality than quantity.

 

My favorite of all of my synthesizers is my Minimoog. Not only mono, but no velocity, no MIDI, no presets. Why is it my favorite? Because it sings to me like nothing else that I've ever played. I can lose myself in the thing for hours. Bass lines...lead lines...tuned chords....long filter-sweeping drones... http://www.musicplayer.com/ubb/cool.gif

 

More may be more; but, for me, killer tone is what it's all about.

 

dB

[musicman1ovation.net]

Originally posted by Dave Bryce:

Why just get one?

 

If you want to spend up to three grand, I'll bet you can get the Gigasampler, a JV1080, and either a K2500 or a K2000 module if you look around. The Gigasampler is under $300, the JV1080 still sells new for around $900, and there's a whole bunch of new and used Kurzweil options under $2k. If you're willing to get a K2000RS, you will even have a bunch of money left over for CD-ROMs and other accessories. Then you basically get two different samplers, two different ROMplers, a bunch of polyphony and outputs and a pretty varied palette to boot.

 

I think it's good to have ROMplers around - many of the sounds onboard are quite useful and inspiring, and it's nice not to have to take the time to load samples in. Plus, the Kurzweil V.A.S.T.engine is so unbelievably powerful that it is simply amazing what it can do with the raw samples once they're in there. So you get the JV1080 for polyphony, variety and the Roland sound, the Nemesys for processing power, RAM size and polyphony, and the Kurz for its sound, its power and access to an immense library (Kurz samplers read just about everyone's samples).

 

Hope that helps,

 

dB

 

Good points http://www.musicplayer.com/ubb/biggrin.gif! It's my bad, but all the pertinent background info exists in the non-responded-to topic called "Need Help! ROM/RAM - Hard/Soft? Read On...". I think the title was so vague as to cause it to be readily skipped. Anyway, it has information about my current setup and what I specifically am looking for in the next purchase. For instance, such things as this: if I bought a Kurzweil, I don't need the sampling option (K2x00S), because I only want to read the CD-ROMs and play them back. Also, I don't already have a PC that would be suitable for Gigasampler/Gigastudio, so that I'd have to get a decent PC with an audio card (with multi outs) PLUS the three hundred bucks. I'd want a PC that could handle it effortlessly. Also, that post went into my needs of polyphony, which aren't that great, because I prefer multi-track overdubs to MIDI stacks and multi-channel/port sequences. MIDI is a great sketch pad to try things out, but I generally like to play things live onto multi-track one at a time, or at least keep the MIDI load light enough so it will perform with its best timing.

 

I've got a JD990 with the Vintage Synths expansion, and although I've though about a JV1080 to bring in some of the other expansions, I'm thinking the sound I want are most likely to come through sample playback. I am leaning towards a Kurzweil, but I'm not sure about how much difference there is between the K-2000, K-2500 & K-2600. The Kurzweil PC2R, if it had RAM for sample playback, would be my choice. Considering the current prices, the K-2600R has me very interested. I'd like to have the Orchestral ROM and maybe the Contemporary ROM (necessitating the daughter-board), but then, sample CDs might cover those bases better. Over time, I'd get an internal hard drive to hold by favorite stuff. The odd thing about all of this, to me, is that I really don't tend to use such unusual sounds. I like electric pianos, horns, strings (and an extended orchestral/symphonic soundset). "World" instruments appeal to me, too; but good acoustic simulations. Guitar sounds, no. For actual synthesizer sounds, I can use a modeled analog, but in that case I am using it as a little spice. I don't get into the real of total synth soundscapes.

 

I know I'm not unique in that many of these devices are filled with sounds that do not interest me. I remember when I first got my DX-7 way the heck back. The dealer also included literally thousands of patches. I thought JACKPOT!! Then I spent hours (days, weeks, months) going through them, and when I was finished, there were no more than three or four banks (96 to 128 sounds) that I could envision using. That took a lot of time that might have been used actually making music!!:rollingeyes: Ultimately I came up with about a bank of my own sounds (more time), and over time I've ended up using those and one bank's worth of the thousands that came with it. To varying degrees, that's always been the case with modules. I suspect that with certain CD-ROM sample discs, there's a finer point on the focus and quality, and therefore more usable sounds. Do I ever learn?!"

[Steve44]

I wanted to revive this old topic because I thought of something new: I don't want more voices anymore, I want more sources, more envelopes, more filters per voice. I've recently been reading about Absynth, a VA software synth that runs on the Mac and is soon to be distributed by Native Instruments. While it does fairly good polyphony, what makes it stand apart is that every voice has up to six oscilators, three filters, and four or more envelopes.

 

I think it's wonderful that most of the current crop of VA synths have dual filters with parallel/serial and other types of routing. And I love the cool Andromeda envelopes, and I'm glad that the Virus made a four oscillator algorithm and that both the Supernova II and the Nord 3 have special unison modes. However, none of the ROMplers feature any of this - they just keep playing canned samples and push the sampling capacity higher and the number of cvoices higher. I want those dual filters and extra oscilators on the Sample playback synths!! I want those dozens of effects processors on the Multitimbral modes of the sample blayback synths too! Even 6 effects processors is barely enough for 32 channels and 128 voices. I'd much rather have a synth that's 64 voices and 16 parts but has 12 effects and dual filters.

 

Furthermore, envelopes have gone back to ADSR's even though an instrument as old as a Roland S550 had six or seven break points. Envelopes are so important! They need to be complex and flexible, at least as cool as those on the Andromeda, if not better! I think it's much more important for manufacturers to put back the power and flexibility into the components of synthesis itself than to put the emphasis on Polyphony #'s.

 

 

Yamaha is one company that could really stop pushing the Numbers. And they push the Multitimbral numbers as much as the polyphony. I just saw a review of their new USB interface that does 96 channels, but only has one midi port? It claims to have 80 channels on the serial "To Host" connector. I think they only make one module that can receive that many channels anyways, and who gives a damn! Why would I want one stupid synth to play 80 different instrument sounds at once? Especially with only 5 effects processors. I would much rather settle for a single sound that actually has some character to it.

 

 

Of course, I should've issued a small caveat: I do think 64 voices at least is necessary (assuming we're talking about oscillators and not Kurzweil style full patches) so that I can play at least a dozen notes with up to six oscillators each.

[Tedster]

Too much polyphony? Wait a minute...I've only got ONE wife, here, Dave.

 

http://www.musicplayer.com/ubb/biggrin.gifhttp://www.musicplayer.com/ubb/biggrin.gifhttp://www.musicplayer.com/ubb/biggrin.gifhttp://www.musicplayer.com/ubb/biggrin.gif

 

I can hear you all muttering: "GUITAR players...what a buncha nimrods"...