Jason wrote:The SWP70 is the very definition of dedicated logic gates. It's not field programmable.

Do you have any sources for that that you can quote? I would be very surprised if that was the case, to be honest. That was most probably the case for good old DX-7 but I think those times are long since gone.

A DSP is usually just a processor which has been designed to be good at crunching numbers at the highest possible sustained speed. And a tone generator DSP would typically add dedicated (digital) audio channels which can interface with other such chips and/or AD and DA converters. Maybe some dedicated features for audio processing in general, but that's it. For that a lot of features of a general programmable CPU are dropped which are essential for general purpose operating systems. That's why in most cases there is an additional general CPU. But still all DSPs I came across are programmable - in most cases using C or assembler.

And since the SWP70 is used in several very different keyboards (PSR-S970, PSR-S770, Montage, ...) they can obviously be used for very different tone generation types. In the Montage there are (according to the link I mentioned) two of them, apparently one for AWM2 and one for FM-X. So they are obviously programmable. The fact that Yamaha was able to add more effects in OS 1.50 (which are most probably done in the SWP70) seems to indicate that in fact they are field programmable. Note that the predecessor (SWP51L) has been used for the Motif, the Tyros 5, the MOX series and others. And the SSP2 another Yamaha DSP has been used (apparently) for all the Reface instruments, so they are clearly widely programmable.

To be clear: those are not proofs but I guess they are strong indications...

Wish I could private message some of this stuff - because not sure how much it helps the thread at all to get into the weeds clarifying statements which are taken out of context. But, the system is what is - so hope there's something useful that comes out of this.

First, let me start with a positive: you're not saying anything incorrect about DSPs. They are processors that excel at certain functions which general purpose processors do not in order to do complex "math" in an efficient manner. And they do execute machine code. I never stated code could not be programmed.

However, in a discussion about transistors - and looking at the "sand" part - the "bare metal" stuff - this is the context where "field programmable" is invoked - that the transistor (array, if you will) is not field programmable. CPLDs and FPGAs (field programmable gate array) are examples of non-dedicated transistors. You can "fuse" an FPGA or CPLD to be a DSP if you wanted to (lots of sequential/state-based logic) - or you could fuse the transistors of an FPGA/CPLD just to take half of the signals and invert them out the other side (purely combinatoral). These transistors would not therefore be dedicated - like in a DSP ("ASIC") where transistors are dedicated to registers, memory, ALU, pipelines, etc (different components of a processor). Therefore being able to interpret/execute/"load" software code or firmware ("programming" as an act of writing the code) is a false equivalency to "programmable" in the context of fusing/arranging transistors.

I think you're shooting at the wrong targets here. That said, it is important there is a base-line understanding, and to help clarify if there's anything that needs clarification. My motivation wasn't to write a thesis on electrical component fabrication and related terminology.

Obviously, no need to quote sources because the question (calling for this) was predicated on an incorrect "read".

And again, the architecture (what goes on inside the black box) doesn't change the original statement that the cost of the black box doesn't change significantly if you put different stuff inside the box because the box fits more stuff (more transistors, "gates") you have to pay for anyhow whether you use them or not. This is called not being "gate limited". As a generalization, it's open for scrutiny but I've already addressed some of the other (cost) dimensions of this statement previous.

I do appreciate your dedication to making sure the information is accurate - just be careful not to "latch on" to one term and "run with it". I know, I've made this kind of mistake before - so it's no big deal. Just causes this kind of back/forth until someone wins the Star Wars trivia battle (nerdy speak that may not help everyone).

Hi Jason, I am not sure I understand all of the above. Programmable for me means, it is possible to load changing software on it. That's it. But I agree, that the discussion would be better in a private conversation.

Anyhow, the point I was trying to make is that I am quite confident, that the current Montage hardware could be used for implementing the FSR-1 style FM synthesis with only a new OS version. Or that they could extend FM-X a bit to add the formants. Maybe with reduced maximum number of voices. But I guess, that's a choice Yamaha has to make. I would actually hope that they do! That's why I am arguing here.

If you want, you can do more research on the terms I've given to understand how "programmable" takes on different meanings depending on the context. And that "programmable" does not always imply "code" which is executed/interpreted - but sometimes describes a fuse map which is used to arrange hardware blocks. The disconnect may be from a lack of background in hardware - which can be remedied with a few clicks and light reading - at least for this.

The capacity for doing formants is there because we see effects that use formants. Whether or not the entire engine is there to completely support the FS1R feature set is harder to determine. Maybe or maybe not. As I've mentioned before, I'd even like just support for glissando (DX7 stuff) to make the DX7 "patches" more true to original programming intent. As well as the way the pitch bend wheel could be programmed to work in steps vs. continuous (without employing motion control) again to help match the DX7 - this one is less of a wish because if I wanted to make that happen - I could use motion control. But gliss is not something I can program.

Back to the FS1R - (sounds like a Beatles tune) - I agree it'd be great if a firmware upgrade could extend FM-X. Not saying it cannot be done - or can. But if the limitation is because hardware doesn't support this - my comment continues to be that adding such hardware support is relatively inexpensive given the process curve (available gate count) as time goes on.

IF the limit is hardware, then we wait for a "respin" - probably in the form of a new product. Yamaha's past for introducing follow-on flagship products is relatively slow vs. competition. I like Yamaha's approach better because it allows for a platform to get a long useful life and also "forces" targeting the right mix of features instead of relying on a frequent product refresh schedule.

I guess we'll see what pans out. About 3-4 years left if the average product life-cycle trend continues.