Here is the answer Yamaha gave me concerning the polyphony issue...

[keynote]

Okay, I had a talk with a Yamaha technical support representative concerning the polyphony issue on the Genos and I wanted to let everyone know what he told me. The polyphony is in fact separate i.e. 128 note polyphony for the preset sound engine and 128 note polyphony for the expansion voice sound engine. FWIW, the Tyros 5 polyphony is "shared" polyphony for a total of 128 note polyphony. Back to the Genos. He told me that the AWM engine on the Genos (both preset and expansion voices) are 'stereo' samples and therefore it is apparently the same as the Montage which also has 'stereo' polyphony for the AWM2 engine.

Now as we know Yamaha stated specifically on its website the Montage had stereo polyphony for its AWM2 sound engine whereas the FM sound engine has 128 note mono polyphony. For whatever reason Yamaha did not specifically say that the Genos has stereo polyphony like they did for the Montage so perhaps in the future Yamaha won't specifically say the word 'stereo' polyphony on future products going forward. But according to the Yamaha tech rep the Genos does have stereo polyphony for both the preset and expansion voices for a total (Max) of 256 note polyphony. That's what he strongly inferred anyway. The Genos is so new that tech support is apparently playing catch up regarding the intricacies and workings of its newest Flagship arranger keyboard. He had to look at his service manual as a reference point regarding samples and the AWM/AEM sound engine. Now unless a Yamaha design engineer comes on this forum and states otherwise we can conclude the Genos does indeed have 'stereo' polyphony for a total of 256 polyphony. Hip hip hooray!!

Mike

[zionip]

I hope the Yamaha technical support engineer really found the answer from the Genos Service Manual - I could not find whether the stereo sample counts as 1 polyphony, or 2.  The Yamaha hardware design engineers of Genos will definitely know.

The only portion mentioned about polyphony and stereo sample is on the Genos Service Manual is on P.3, Specifications, which is the same as on P.123 of Genos Owner's Manual:


An alternative of finding the real answer is conditioned tests on polyphony on Genos.

Heratch (Mazeka) from Yamaha probably knows the answer.

Thanks,
Paul

[pjd]

Hi Mike and Paul --

Guys, guys, I think you're wringing your hands about this a little too much. 

The implementation of stereo tone generator channels is a fundamental feature of the SWP70 tone generator IC. This is clear from looking at the Montage service manual and the Montage specs.

The hardware tone generation micro-architecture (based on SWP70) is the same in Genos, modulo differences in wave memory. Thus, I've never really doubted that Genos has stereo tone generation channels and 128 voice stereo polyphony on the preset side and 128 voice stereo polyphony on the user wave memory side.

Yamaha is totally mum about the tone generator IC and hasn't even revealed the chip name "SWP70" in public. I seriously doubt if they will ever say anything much about it except that "It's new and wonderful."  That's not a slam on Yamaha; they have eager competitors.

Counting polyphony may or may not be as straightforward as you would like. Taking Montage as the example, the newest Montage voices are probably "native stereo" and require one stereo channel, i.e., one stereo voice of polyphony. Some legacy Motif voices may still require two channels (each channel panned hard right or left) as this was the way in olden times. The authoritative source in Motif-land is the Data List PDF which has an Element count for each voice. This info was dropped and is no longer published in the Montage Data List. I suspect that Genos has the new vs. legacy voice issue, too.

It's a fun subject, but the picture is soft-focus and not likely to get any clearer...

-- pj

[markstyles]

I think the blanket statement 256 voice polyphony is misleading.. Fact is if you use complicated voices, (with more layers in them) the polyphony is significantly reduced even more.

Fact is Yamaha has done an extremely decent job of 'voice stealing' to assign new voices, without noticing older notes cutting out..  When using a DAW to supply live tracks to Tyros, along with playing a style, the note stealing was noticed.  But in live performance it is non-issue..

But saying it's 256 voice seems a bit dis-honest to me

[Joe H]

It really depends on the number of "Elements" each Voice uses.  I assume many use 4 Elements. When you layer R1, R2 and R3... you are using many polyphony for just one sound.

Joe H

[pjd]

I think the blanket statement 256 voice polyphony is misleading.
But saying it's 256 voice seems a bit dishonest to me.

Hi Mark --

I think knowledgeable people -- and I certainly count you as one of those! -- realize that each active element consumes a single voice of polyphony. The problem, as I tried to address it in my earlier message, is counting active voices of polyphony. The counting has always been somewhat difficult on AWM2 because of element triggering rules (Expanded Articulation and such) and the use of layers. The problem of counting active voices of polyphony is not unique to Yamaha; it's a problem in any digital synth with layering.

A few posters are concerned about the split between preset (128 voice polyphony) and expansion voices (128 voice polyphony). Some writer at Yamaha naively said "128+128=256," ignoring the practical implications of the actual split. I regard that as a mistake, not dishonesty. This should be corrected in their specs. As the aphorism goes "Never ascribe to malice that which can be explained by incompetence." 

All the best -- pj

[Joe H]

... "Never ascribe to malice that which can be explained by incompetence." 

All the best -- pj

Well, at a distance stupidity and ignorance look pretty much the same.  I think most people who write the documents for Yamaha keyboards never see them... let alone use them or play them.

Joe H

[pjd]

Well, at a distance stupidity and ignorance look pretty much the same.

Hi Joe --

Yeah, Hanlon's razor (the aphorism) is sometimes stated with "stupidity" or "ignorance" inserted. Such is the world. 

Hey, hey, take care -- pj

[voodoo]

Counting polyphony may or may not be as straightforward as you would like. Taking Montage as the example, the newest Montage voices are probably "native stereo" and require one stereo channel, i.e., one stereo voice of polyphony. Some legacy Motif voices may still require two channels (each channel panned hard right or left) as this was the way in olden times.

Hey PJ,

I want to emphasize this point. Looking at the UVF-XML definitions of YEM voices, one sees, that there a two possibilities of creating stereo:

* One stereo sample played in one voice element. Here still two mono samples are played internally, but the control of the tone generator (filter, envelope, lfo etc.) is done equally for both samples.
* Two mono voices are assigned to two different voice elements and are panned left and right.

In the first case one voice is needed and we have full stereo polyphony. In the second case the keyboard does not know about stereo and plays to mono samples on separate tone generators. Here the number of polyphony is being the half.

Uli

[pjd]

Hi Uli --

I agree completely. The difference is apparent, too, when scanning the Montage waveform list. (Sure wish we had a waveform list for the Genos. )

The "Two mono voices are assigned to two different voice elements and are panned left and right" is the legacy case.

The mix of these two different voice implementations complicates polyphony voice counting in practice, or even experiments that we would want to try. We would have to be careful to identify voices using the legacy approach and the new stereo approach.

As you say, at least the UVF XML is definitive.

Hey, hey, thanks -- pj

[voodoo]

Hi pj,

I have still two remarks:

* When we want create stereo voices by ourselves, we should take care to use stereo samples and put them in one voice element. So we can create real stereo voices if we want. That's the good message. Unfortunately YEM has no indication whether a used sample is mono or stereo. But YEM definitely supports stereo samples. This can be seen in the corresponding XML.

* There are voices, that use only mono samples, put these to two elements, pan them left and right, and give them a slight detune. This is how to make a chorus type stereo voice from a mono sample. Of course, this uses two voices of polyphony count. (This is used in the expansion pack EasySounds FM Xpanded for example.)

Uli