...because of the analogue sections?

If my understanding is correct, the math involved is identical; a sine wave at x frequency, being modulated by another sine at x frequency would produce the same results, regardless of which calculator is being used.

Unless I'm missing something about the actual synth engine, it seems the only sonic difference would come from the DAC and amps in the DX7.

Can anyone confirm or deny?

Maybe someone can back up my answer as being correct or not but I would say yes.

The math is the same, yes. Now, it depends on the calculator which does that math. DX7 had limited CPU power and I would think that samples in it weren't calculated as 32 or 64-bit floats, which is what FM8 is doing. That's why FM8 sounds "cleaner" and not as "grungy" - it has more precise calculations.

The digital processing can be done in different ways.

A digitally generated sine wave is always an approximation at some resolution (certainly amplitude-wise, and when done with lookup tables like in the DX, also in the other dimension). This will affect the sound. There's also a difference between doing maths as fixed and floating point. Additionally, at least some Yamaha chips[1] do multiplication via lookup tables (because a*b = exp(log(a)+log(b)), so you can do multiplication with addition, which is a lot simpler to do in a chip, and tables for the exponential and logarithmic functions) adding further non-linearities.

This can all be modelled exactly, of course, but a naïve software implementation on a modern processor where you can calculate any number of sines per second and do all your processing as 32 bit floating point is going to be different to the significantly coarser calculations in the Yamaha ASICs.

[1] https://docs.google.com/document/d/1...nwo/edit?pli=1

Interesting!

I wonder if there's a way to demonstrate the difference... like if someone could build a simple FM plug that would let you switch calculation modes.

Absolutely. Grungy vs cleany. Old school vs modern. Noisy vs minimal.

Firmware ROM:
Source: Ted's Yamaha DX7 Page

By far I prefer the sound of FM8 in comparison to a DX7. For me FM really shines with FM8.

Isn't there already threads covering this ?

FM8 is more similar to the DX7II / TX802 than the original DX7 / TX7.

Still, I like the sound of my TX802 more than FM8 (I have both.) The attack on the TX802 is definitely different. It doesn't have the same digitalish high frequency harmonics as FM8.

The TX802 is also punchier at the bottom end. And I can get cooler sounding Eno-esque moving pad sounds with the TX802.

I don't know why they sound different, but they do. Filters? Output amps? Who knows.

The placebo effect is very powerful.

It is indeed.
However, there are audible differences between the DX7 and FM7/8 that have nothing to do with the placebo effect.

While FM8 can load DX sysex and patches it is not really intended to be an emulation of the DX7, it can do soooooo much more than a DX7, for example you get 32 different wave forms rather than just one

A plugin that makes it sound '3D'...??

?????????

Likely.



I thought there were no filters.

can anyone recomend some digital converts from early 80s?

I'm sure it has reconstruction filters after the DAC.

Who cares. Long as you can see what you're doing on the plugin, its a nobrainer for me.

Not all filters come with a cutoff knob.

It could've been easy: build an emulation of the ASIC Yamaha uses, compare outputs. Bit for bit identical? Sweet, you've got an emulation. But that's lots of work to get "reasonably close".

FM7 is already ancient. FM8 was the same engine, more FX, improved UI (but not improved by much). But it's hard to create an effective interface for an FM synth.

I'm assuming he meant the standard audio filters.

It's a discussion forum. I wanted to discuss!

Also, because I have analogue outboard, if the main difference between the two comes down to something like the amps in the DX7, I'll just stick with an FM8, and just run tracks through some external gear. The FM8 is far more powerful and easier to program.

But there are no user-controllable filters, or filters used in the actual synthesis.

Yeah, there are filters (as someone else pointed out) at the ADC, for example, but that would fall under the blanket of my original question (components that are part of the hardware chain, like amps).

I got the impression that the person who mentioned filters meant the kind that are part of subtractive synthesis.

It's different. I'm not 100% sure why. It might be because the audio comes out of the DX7 through a transformer and then through another transformer at my ADC but the DX7 is "warm" where FM8 is "Bright."

It's not a fair comparison because FM8 is a lot more powerful but the things the DX7 does, it does well and just sounds better to my ears.

I'm pretty sure OP accepts the audible difference from the outset.

So, I think it's the fact that the audio is taking a trip through the analog realm that helps beef it up, but there's also a lower harmonic/sub-harmonic on the classic EP patch that's missing in the FM8 one so clearly there IS some difference in the way they're calculating the sounds.

Yeah it seems pretty widely accepted that they are, so I'm just trying to understand if that has to do with the effect of the physical components, or if the actual calculations are producing different results.

I've used the FM8 a bunch, and have a TX7 coming my way, so I'll be able to compare directly, I just want to understand what is causing most of the differences.

I've only just got FM8 for the first time, as part of Komplete 9 Ultimate.

For me it's one of the highlights of the whole bundle. It's sounds gorgeous.

And I'm speaking as someone who has never been without a Yamaha DX synth of some kind since the 80s.

And if you think it can't sound grungy, I used it to make the sound that comes in right at the beginning of this track:



D.

I haven't run the tests, but I'd like to compare the sound of FM8 going through a trip in the analog realm and back in. I think that would level the playing field of the the transformers I was talking about. That's been a trick I've used in the past to "warm up" plugins before mixing.

But, as other posters have mentioned. This level of minutiae is academic. Make good music with the tools you have.

This. I'm pretty sure the DX7 is like the OPLx series that have been extensively reverse engnieered for very precise emulation. You're right that the FM chip is waaay too slow to do even 16-bit floating point calculations on the fly. In the YMF262 OPL3 case (kinda of a single chip child of the 81z 4-op) it uses fixed integer lookup tables for not only frequency divisions but also for sinusoid calcs. Hence it's quite rough in the maths where sin accuracy (phase) functions are quite audible.

Quite ingenious, its all just lookup tables, very fast response w/ no real cpu interaction. Remember this is 1980s tech!

I own/owned many FM synths over the years, aliasing is really audible in the older gear. My PreenFM for example has loads of aliasing which gives it this sharp edgy grit moreso than your average Yammy FM. Also I used to have one of the original brownie DX7, might be the 12-bit DAC but it sounds quite different than say a 81z on top of having a high noise floor.

Check it out, someone completely de-encapsulated a OPL3 and reversed it:

https://docs.google.com/document/d/1...1rIm42nwo/edit

Great post and link. It's a bit of an eye opener to see some of the techniques used to make fm digital synthesis affordable - encoding just the first 256 samples of a sine wave for example!

Similar thing was done by PPG and Waldorf for single-cycle waves in their wavetables - they store only half of a wave then mirror the other half in realtime.

What irks me is that you could apply U-He levels of "let's get to the bottom of this" and end up with something that's bit-by-bit identical but nobody's done this yet for some reason.

KVR: VOPM by Sam - Details

Perhaps?