Topic: Seamless Sound Transitioning

[tyrosrick]

It's my assumption that there are no workarounds for the G2 to perform seamless smooth transitions from one style or voice to another. The Korg PA5X had this, and that is the ONLY thing I miss about that board. Are there any suggestions on even half-you know what ways to do this on the Genos2?
thx

[strandkam]

That's why I don't switch from G1 to G2 

Strandkam

[pjd]

Let's say you want to choose and play three lead voices during a song. Put each lead voice into RIGHT1, RIGHT2, RIGHT3, respectively. This (essentially) reserves the tone generation and effect resource for all three voices.

A half-way measure, use the same chorus and reverb DSP effect throughout. It helps avoid DSP-related glitches.

I'm sure other folks will chime in -- pj

[DaPaleRider]

Genos 1 and Genos 2 does not support Seamless Sound Tradition.

[BogdanH]

We need to be specific when we claim Yamaha keyboards don't have seamless transition.
Because:

...that there are no workarounds ... to perform seamless smooth transitions from one style or voice to another.
...

-in above context Yamaha keyboards do have seamless voice transition and the workaround is: properly organized RH voices.

That is, on all Yamaha keyboards which have three RH voices, we can seamless transition:
-one voice to another one voice,
-one voice to another two voices, and
-two voices to another one voice.
If sum of used voices (from/to) is bigger than three, then seamless transition is currently not possible.

Bogdan

[tyrosrick]

Thx all for comments/suggestions. Most of what I meant was to transition smoothly from one style to another, rather than voices within one style.

[pjd]

As Bogdan mentioned, that's a tougher problem -- style to style transitions.

In this case, I would try to use the same DSP effect types and settings for variation, chorus and reverb effects. I don't think there is much that can be done with the insertion effects because the style parts are usually distinctive from one style to another.

Take care -- pj

[Amwilburn]

It actually *is* possible, but takes more work on our part (and no, it really shouldn't be up to us). But you can do SSS even using all 3 voices, but it's *very* complicated. You have to match the internal volume of each part, additionally, to avoid sample buffer flushes or sudden DSP switches (which can make quite a loud pop) you probably want to use a legacy or gm/xg as one or 2 of the 3 voices (fun fact, if you use all 3 voices as GM/XG, you don't even need to do any special programming in that case! They *all* used to be SSS until around the 2000's)

Long story short; just assign R1, R2, R3 as your different voices that you turn on and off, would be the simplest.

Mark

[ton37]

About Seamless Style Transition:
When changing the style variations you just have to keep in mind that occasionally something can go wrong in terms of SST. Fortunately it is rare and if it arises you will need to adjust your playing skills. You can create a flawless transition without getting into trouble and based on your knowledge you sometimes know 'where' the problem (SST) is. This may be because your timing for a style change is not on time. Then it is sometimes a solution to press the break key before you go from, for example, variation 4 (usually a very busy variation) to variation 1 (usually a very quiet variation). Registrations are great, but be alert to the behavior of the saved style variation. Sometimes it is better to switch to another variation manually (with the break key) than to have this done (without break) by the registration. Just look at the various videos where the keyboard player first plays a break before varying. The Yamaha demonstrators are certainly aware of the pitfall and respond to it to prevent problems with SST. It's just part of operating the keyboard and you also have to be skilled in this... Jm2c 

[tyrosrick]

It appears that the song player, which has two players can SST, but it can only do it with audio styles or MP3's, not with regular styles one uses from the main categories.

[ton37]

Hi @tyrosrick, Can you show what exactly you mean, for example by means of a link to a video fragment (please with time)? Your question doesn't seem entirely clear (at least to me).

[DerekA]

It appears that the song player, which has two players can SST, but it can only do it with audio styles or MP3's, not with regular styles one uses from the main categories.

Well, that's because it's audio and it's easy to cross-fade between them. The issue the arrangers have is SST on midi voices, because all the dsp is being generated in real time by the processor, and it can't switch horses mid flight without glitching (if you'll forgive the mixed metaphor)

[tyrosrick]

to TON37    https://www.youtube.com/watch?v=KfCQ2a_oOUA&list=PL3rjqgoqj7LKXjqNI6vvXuhZc0O0_9aWm&index=3  4;50 minutes into video

[ton37]

@Tyrosrick, thnx. for clarifying. I misinterpreted your question. 

[AndyB]

There was a video done by Casper Tutorsynth that will explain the issues with ALL Yamaha keyboards.

The video link is below, hope it helps and explains the issue correctly. The problems are with the DSP effects etc, if you edit the DSP effects, so ALL sounds have the exact same DSP and effects the there is no pop when changing from one sound to another.

There was another very in-depth video on YouTube I found a few months ago, that showed the issue and examples of how to edit the DSP and effects, to have a smooth and seamless transition sound from various instruments, but cant remember the channel, sorry.
 We should NOT have to jump through hoops to get the transition working, but we have to because Yamaha cut corners with the DSP engine.
https://www.youtube.com/watch?v=O5iUjXMDuhc

[BogdanH]

...
We should NOT have to jump through hoops to get the transition working, but we have to because Yamaha cut corners with the DSP engine.

-I absolutely agree on that.
To use the computer terminology: voices should be multi-threaded. Means, a voice should not be interrupted (changed) as long it is running -regardless of how voice is used (as keybed-OTS voice or in style). The only situation where I would expect that rule to break, is if we exceed polyphony limit.
But as it seems, Yamaha don't care about solving that... it's easier to sell recycled stuff.

Bogdan

[soundphase]

But as it seems, Yamaha don't care about solving that... it's easier to sell recycled stuff.

I disagree.

Our keyboard is before all, a real-time tool.
Yamaha programmers perfectly knows how to manage multi-threading, and pool of common resources.
For me, Genos is so demanding in real-time sound computation, that even such optimizations would not lead to satisfying results.
We certainly need more DSP power, ... a lot of expensive DSP power just to manage these clean transitions (so with an additional price for the client)

[BogdanH]

hi soundphase,
I'm not sure on what you disagree... I said that "a voice should not be interrupted as long it's running", so yes, Yamaha needs better/different DSP architecture obviously. Here I prefer the word "different" instead of "more powerful". For analogy: if we struggle at drilling a hole into concrete, we don't necessary need more powerful drill machine.. we need different drill for concrete.

Genos is no special demands keyboard... PSR-SX600, which is considered as a simple low demands arranger, also doesn't have seamless sound transition.
Korg arrangers however, seems to have no problem with that and so I think that solution is not necessary price related... I would say it's more a "let's do it right" thing.

Just my opinion,
Bogdan

[DerekA]

I don't have any real technical understanding of the details of how Yamaha's DSP technology works.

But, I think it's safe to say that since neither the latest Genos2, or Montage M, can do SST "properly" - it must be a significantly more difficult problem to deal with than it may seem from the outside.

[DerekA]

Incidentally, I have not tried this myself, but I wonder if it would work.

Many people know about the "V Console" product, and some know that it works by directing the MIDI messages from the keyboard parts to the Song parts (using a MIDI cable from OUT to IN). Since the Genos has 28 DSP, it is possible to allocate voices including DSP to each of the 16 parts (Tyros and below don't have enough DSP to do this)

So, it should be possible to set each of the 16 Song parts, each will full DSP.. Then switch off local control, set R1 to transmit on Channel 1, and use registrations to switch between MIDI setups that select which parts listen to MIDI channel 1. In theory, this should give full SST.

Anyone fancy giving that a shot??

[soundphase]

a voice should not be interrupted as long it's running

It's not analogic, it's numeric. For a sound to continue properly while other sounds are added and removed, requires dynamic computation.

At the end of the computation, it is a numeric wave for left and a numeric wave for right.

The more there are polyphony + simultaneous effects, the more it requires computations... with real-time constraint (result in less than X ms ....) and some internal power must be kept to capture all what can happen on the full interface (keyboard, buttons, screen ....) without any latency.

[KurtAgain]

I have great confidence that Yamaha will solve the problem soon - simply because Korg advertises that they have solved the problem with the Pa5X.

[pjd]

To use the computer terminology: voices should be multi-threaded. Means, a voice should not be interrupted (changed) as long it is running -regardless of how voice is used (as keybed-OTS voice or in style). The only situation where I would expect that rule to break, is if we exceed polyphony limit.

I don't mean to pick on Bogdan, but folks should not assume that the DSP is performed on a conventional programmable processor, i.e., that there are threads, let alone multi-threading. Yamaha use a lot of dedicated hardware pipelines...

The issue is keeping voice synthesis and effects alive (operational) until a held/sustained note and its effects complete. Montage/MODX restrict SSS voices to 8 (MODX: 4) parts or less. It's like pre-reserving the hardware resources needed for the new voice to be switched in, while the active resources complete the old voice. Genos/PSR does not have a pre-reservation scheme like SSS.

As to using the same effects in the original and target (new) voices, Yamaha's control programming may check to see if the new effects are the same as the old in order to avoid unnecessary reconfiguration of the hardware effect pipelines. I don't think Yamaha have ever guaranteed seamless switching even in this case.

This problem in hardware is more complicated than folks might think or believe because Yamaha distributes synthesis across the TG pipelines and the effect processors.

That doesn't mean I wouldn't like to see a solution. Another workaround is to switch voices via RIGHT1, RIGHT2, RIGHT3. That is effectively a pre-reservation scheme.

All the best -- pj

[BogdanH]

hi pj,

I don't mean to pick on Bogdan...

-I would never take it that way from you 

...but folks should not assume that the DSP is performed on a conventional programmable processor, i.e., that there are threads, let alone multi-threading...

-I was only using PC terminology for easier understanding -and I failed, I guess 

This problem in hardware is more complicated than folks might think or believe because Yamaha...

Yes, I can imagine it's complicated. But the thing is, as it seems, the problem exist "because Yamaha..." -I mean, Korg solved the problem.

Another workaround is to switch voices via RIGHT1, RIGHT2, RIGHT3. That is effectively a pre-reservation scheme.

YES!
Greatly simplified: each voice should have additional channel (pipeline, buffer, whatever) which is invisible to user and is only used in background for "pre-reservation" purpose.
Or even more simplified: I have no idea how to do it, but is for sure possible (Korg again) -and we're not in year 2000 anymore.

But to be honest, I'm not really complaining about (lack of) seamless sound transition: at style creation I just pay more attention by trying to avoid that to happen (and similar for R1,R2 and R3 voices).

Best wishes,
Bogdan

[pjd]

Hi Bogdan --

Thanks for all of the help you offer on the Forum!

I think Yamaha should have treated all of the insert effect DSPs as a common pool from which hardware resources are assigned. For example, there is an insert effect unit assigned to each style part (somewhat sensible) and to each song part (possibly unnecessary).

If someone isn't using a MIDI song and its 16 parts/insert effects, those effects could be assigned elsewhere dynamically. Oh, well. Maybe they will think of something clever since they know the internal design (and I sure don't). 

Take care -- pj

[soundphase]

I have great confidence that Yamaha will solve the problem soon - simply because Korg advertises that they have solved the problem with the Pa5X.

I didn’t know. SST is now common.

Difficult to understand Yamaha doesn’t implement it on its new flagship arranger.

I hope something will be done on Genos2 without having to wait Genos3, 4 to 6 years more.

[sh]

I didn’t know. SST is now common.

Difficult to understand Yamaha doesn’t implement it on its new flagship arranger.

I hope something will be done on Genos2 without having to wait Genos3, 4 to 6 years more.

To my understanding this is a hardware limitation of the Yamaha custom chips which can't be fixed with a software update.

[pjd]

To my understanding this is a hardware limitation of the Yamaha custom chips which can't be fixed with a software update.

Hi --

The Montage/MODX and Genos/SX use the same underlying hardware chips (the SWP70 tone generator integrated circuit). The software controls the allocation of the hardware resources. Yamaha could adopt a solution similar to Montage/MODX SSS.

This could be fixed through a software update, but it would require a different allocation of DSP units. Right now, DSP units are pre-allocated to RIGHT1, etc., style parts and song parts. An SSS-like solution might (would?) not be backward compatible to the original Genos allocation/assignment scheme.

Hope this info helps -- pj

[sh]

Hi --

The Montage/MODX and Genos/SX use the same underlying hardware chips (the SWP70 tone generator integrated circuit). The software controls the allocation of the hardware resources. Yamaha could adopt a solution similar to Montage/MODX SSS.

This could be fixed through a software update, but it would require a different allocation of DSP units. Right now, DSP units are pre-allocated to RIGHT1, etc., style parts and song parts. An SSS-like solution might (would?) not be backward compatible to the original Genos allocation/assignment scheme.

Hope this info helps -- pj

You are right that it could be somehow solved this way in theory, but since this would break backward compatibility it is not a real solution IMHO.

[soundphase]

As long as OS manages both modes, it’s  only a « registration » additional parameter that can be set to « on » for new registrations saved with new OS.

[janrhansen]

It's the price you pay for using a VLSI instead of virtual software sound generation. It's much easier to implement in software. They would need to reprogram the whole base OS System of how the SWP70's is set up  to somehow give us real SST. Like other posted, you can prealocate different voices in each part to have the effect of SST betwen different keyboard parts. Think Yamaha is the last of the major Synth companies that still insists on using VLSI based tone generation, most other synth manufactorers are using software only based sound generation on either some kind of off the shelf computer parts or their own proprietary hardware. Many smaller synths focused on a particulary kind of sound generation, like all the smaller Korg Synths run on just a small Pi compute module, maybe with an added dsp chip for sound processing.

Not that there arent plusses with using them, but with their new software plugin of the Montage M I hope Yamaha finally kills this archaic way of designing their synths for the next generation. They have to use these VLSI's chips for several generations to pay back the cost of designing and producing them, and it's still a software programable chip, so why keep using it when todays cpu's is hundreds of times faster than what they were back when VLSI's was first introduced. Back then it was a necessity as the CPU's werent fast enough without adding expensive DSP chips.