Yamaha TX81z

To those using/or who have used the Yamaha TX81z, do the FM sounds
have harmonic harmonics (or do most of them?)... Having read a lot
more on timbre, I don't think most wavetable synthesis will do what I
want, especially sampled piano timbres which will simulate the chorus
effect of multiple strings and the inharmonicity of the sampled
instrument... With acoustic instruments I prefer bassoons, oboes,
trombones, voices, etc.: strong harmonics. Will the FM sounds work
well with these for JI? If so, I'm probably finally convinced to get
a TX box. -Justin

on 8/7/03 9:13 AM, Justin Weaver <improvist@usa.net> wrote:

> I don't think most wavetable synthesis will do what I
> want, especially sampled piano timbres which will simulate the chorus
> effect of multiple strings and the inharmonicity of the sampled
> instrument...

Curiously its much easier and cheaper (less RAM) to do wavetable synthesis
*without* chorus and inharmonicity. This requires only a sample of a single
waveform cycle. The very assumption of a waveform cycle which repeats
guarantees no inharmonicity.

This doesn't mean that people interested in sampled synthesis usually have
much interest in such sonic simplicity, so I don't know whether you will
readily be able to find such samples. But if you have a sampler I would
assume there would be a recording mode that would let you create single
cycle samples.

Would you consider a software solution? It is extremely easy to do
purely-harmonic sinusoidal additive synthesis in software, and I assume
someone has done either this, or a software sampler that would let you do
the equivalent thing with a single-cycle sample. I have not looked into
what is available because I tend to write my own software for very specific
purposes.

There is also Max/MSP which will let you do exactly what you want (including
control over tuning), at the cost of a bit of learning curve. But there is
a lot of community support there, e.g. an email list, plus in some areas
there are users groups.

But Max/MSP is IIRC still Macintosh-only, though they have been working on a
Windows version for some time. What platform would you want, if indeed a
software solution would work for you at all?

> With acoustic instruments I prefer bassoons, oboes,
> trombones, voices, etc.: strong harmonics. Will the FM sounds work
> well with these for JI?

Probably will work fine, if you can stand FM sounds. You can even do a
limited amount of sinusoidal additive synthesis, though I forget the
details. IIRC on the DX7 you can just add 6 (sinusoidal) oscillators with
algorithm 32, but that may be rather limiting.

> If so, I'm probably finally convinced to get
> a TX box. -Justin

-Kurt

I guess my real question is: do FM sounds have harmonic harmonics? -
Justin

>
> > With acoustic instruments I prefer bassoons, oboes,
> > trombones, voices, etc.: strong harmonics. Will the FM sounds work
> > well with these for JI?
>
> Probably will work fine, if you can stand FM sounds. You can even
do a
> limited amount of sinusoidal additive synthesis, though I forget the
> details. IIRC on the DX7 you can just add 6 (sinusoidal)
oscillators with
> algorithm 32, but that may be rather limiting.
>
> > If so, I'm probably finally convinced to get
> > a TX box. -Justin
>
> -Kurt

on 8/7/03 11:34 AM, Justin Weaver <improvist@usa.net> wrote:

> I guess my real question is: do FM sounds have harmonic harmonics? -
> Justin

Yes, I was indirect in answering because I am a little vague.

So all I can really offer is something to focus the question a little
further, so that others who may answer might give more complete
information...

I think you can set the center frequencies of individual oscillators to be
non-harmonic if you like. But if you set the center frequencies of
individual oscillators to harmonics, then is the resulting spectrum also
fully harmonic?

And I was also trying to steer you away from FM becase the timbre
possibilities are restrictive, in my experience. Anything purely harmonic
that you can do with FM you can do more directly and with better control in
other ways. I personally think it was a mistake of history and marketing
that FM became so popular. It was relatively cheap to implement (like other
possibilities), but it also had a patent (unlike other possibilities), which
Yamaha bought or licensed exclusively, to its advantage.

-Kurt

>
>>
>>> With acoustic instruments I prefer bassoons, oboes,
>>> trombones, voices, etc.: strong harmonics. Will the FM sounds work
>>> well with these for JI?
>>
>> Probably will work fine, if you can stand FM sounds. You can even
> do a
>> limited amount of sinusoidal additive synthesis, though I forget the
>> details. IIRC on the DX7 you can just add 6 (sinusoidal)
> oscillators with
>> algorithm 32, but that may be rather limiting.
>>
>>> If so, I'm probably finally convinced to get
>>> a TX box. -Justin
>>
>> -Kurt

Justin Weaver wrote:
> I guess my real question is: do FM sounds have harmonic harmonics? -

Yes, if you set them harmonic they're harmonic. You can also get inharmonic timbres as an option. In some cases, inharmonicity gets added in from aliasing, but I've only ever noticed that for timbres that started out inharmonic.

Graham

--- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...>
wrote:
> To those using/or who have used the Yamaha TX81z, do the FM sounds
> have harmonic harmonics (or do most of them?)...

maybe most of them, but it's quite possible to use FM to get
inharmonic timbres.

> Having read a lot
> more on timbre, I don't think most wavetable synthesis will do what
>I
> want, especially sampled piano timbres which will simulate the
>chorus
> effect of multiple strings and the inharmonicity of the sampled
> instrument...

i don't think FM can be used to generate the type of inharmonicity
found on the piano, unfortunately.

> With acoustic instruments I prefer bassoons, oboes,
> trombones, voices, etc.: strong harmonics. Will the FM sounds work
> well with these for JI?

sure -- any periodic waveform will have the same series of possible
partials as the acoustic instruments you're using.

--- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...>
wrote:
> I guess my real question is: do FM sounds have harmonic harmonics?

some do, some don't. in theory, if you use random parameters for the
FM, you have a zero probability of getting a periodic
(harmonic) waveform.

Do FM synthesizers like the Yamaha TX series have preset timbres with
harmonic harmonics or do you have to create all the sounds yourself
from scratch? -Justin

--- In tuning@yahoogroups.com, "Paul Erlich" <perlich@a...> wrote:
> --- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...>
> wrote:
> > I guess my real question is: do FM sounds have harmonic
harmonics?
>
> some do, some don't. in theory, if you use random parameters for
the
> FM, you have a zero probability of getting a periodic
> (harmonic) waveform.

--- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...> wrote:
> Do FM synthesizers like the Yamaha TX series have preset timbres
with
> harmonic harmonics or do you have to create all the sounds yourself
> from scratch? -Justin

TX81z has 128 factory presets and the majority of them are harmonic
although some of them are a bit detuned. You can adjust the factory
programmed sounds so that they are exactly harmonic if that is what
you want. I have FM7 soft synth which can load TX81z and DX7 patches
and I have to say that those TX81z-patches sound really crappy when
compared to the full capabilities of FM7 itself. I haven't heard the
original though but everyone says they sound the same. TX81z must
also be a nightmare to program because of its spartan controls while
FM7 is a dream.

Kalle

--- In tuning@yahoogroups.com, Kurt Bigler <kkb@b...> wrote:

> And I was also trying to steer you away from FM becase the timbre
> possibilities are restrictive, in my experience. Anything purely
harmonic
> that you can do with FM you can do more directly and with better
control in
> other ways.

I don't think this is true. You just can't beat the expressiveness of
FM when you have key velocity or some other control assigned to
operators. Samples/oscillators with filters just don't offer such
subtlety of timbral colors.

Kalle

What sort of FM synth would you recommend, if not the TX81z? Two
things are important: 1) tunability resolution and 2) affordability. -
Justin

--- In tuning@yahoogroups.com, "Kalle Aho" <kalleaho@m...> wrote:
> --- In tuning@yahoogroups.com, Kurt Bigler <kkb@b...> wrote:
>
> > And I was also trying to steer you away from FM becase the timbre
> > possibilities are restrictive, in my experience. Anything purely
> harmonic
> > that you can do with FM you can do more directly and with better
> control in
> > other ways.
>
> I don't think this is true. You just can't beat the expressiveness
of
> FM when you have key velocity or some other control assigned to
> operators. Samples/oscillators with filters just don't offer such
> subtlety of timbral colors.
>
> Kalle

on 8/7/03 3:57 PM, Kalle Aho <kalleaho@mappi.helsinki.fi> wrote:

> --- In tuning@yahoogroups.com, Kurt Bigler <kkb@b...> wrote:
>
>> And I was also trying to steer you away from FM becase the timbre
>> possibilities are restrictive, in my experience. Anything purely
> harmonic
>> that you can do with FM you can do more directly and with better
> control in
>> other ways.
>
> I don't think this is true. You just can't beat the expressiveness of
> FM when you have key velocity or some other control assigned to
> operators. Samples/oscillators with filters just don't offer such
> subtlety of timbral colors.

Yes, I wasn't really thinking of samples in relation to dynamics. For that
some form of algorithmic synthesis is needed, I believe.

I was speaking too generally perhaps, about what "can" be done, rather than
the more practical question about what "has" been made available to the
musician who is not a programmer. Nowadays software sinusoidal synthesis is
very cheap (but probably not available off-the-shelf) and you could program
it to replicate any given FM pattern you wanted, but you could also program
it to replicate an entire universe of other possibilities, without any of
the restrictions inherent in FM. I felt FM provided expressiveness, but so
often with the wrong degrees of freedom, i.e. to get one thing you want you
always had to get something else you don't want. That was my experience.

Meanwhile the fully parameterized sinusoidal additive thing or numerous
other approaches could be done in Max/MSP but probably requires more than a
small learning curve, whereas the perhaps-limited FM approach is already up
and running.

Partly I speak up because I figure non-developers are not necessarily aware
of what is readily possible right now, but simply not on the market. But
this may be misplaced. If enough people were ticked off at FM synthesis
then I suppose we would be seeing more algorithmic alternatives. So I'll be
quiet now.

-Kurt

Kurt Bigler wrote:

> I was speaking too generally perhaps, about what "can" be done, rather than
> the more practical question about what "has" been made available to the
> musician who is not a programmer. Nowadays software sinusoidal synthesis is
> very cheap (but probably not available off-the-shelf) and you could program
> it to replicate any given FM pattern you wanted, but you could also program
> it to replicate an entire universe of other possibilities, without any of
> the restrictions inherent in FM. I felt FM provided expressiveness, but so
> often with the wrong degrees of freedom, i.e. to get one thing you want you
> always had to get something else you don't want. That was my experience.

I don't agree with the middle bit. Noisy FM patches would require a very large number of sinusoids that no additive approach could provide in real time. In addition, the result probably depends on the sample rate and depth of the original, so mathematical predictions won't be enough, and you may as well use samples.

Also, as FM in practice includes AM there are no inherent restrictions. The TX81Z itself has all kinds of restrictions, because of what was practical at the time it was made. If you've got enough power to implement 100 additive sine waves, you can also implement 100 FM operators. Both will be hugely complex to work with unless you have sone kind of meta-algorithm.

> Meanwhile the fully parameterized sinusoidal additive thing or numerous
> other approaches could be done in Max/MSP but probably requires more than a
> small learning curve, whereas the perhaps-limited FM approach is already up
> and running.

I can do it in Kyma, but each sample is expensive. Perhaps it would be better if there were an efficient sinusoid generator, instead of using samples of sine waves.

> Partly I speak up because I figure non-developers are not necessarily aware
> of what is readily possible right now, but simply not on the market. But
> this may be misplaced. If enough people were ticked off at FM synthesis
> then I suppose we would be seeing more algorithmic alternatives. So I'll be
> quiet now.

I don't think FM's prominent enough now to hold anything back. Perhaps analog emulation is. But what are the alternatives? Additive is there, but more complex to program. Physical modelling is on the market, but is also complex to implement and program. Granular needs samples to work with. There are soft synths to do all of these things.

A TX81Z is still a good, cheap, standalone, microtonal synth module. It is limited in sound quality, but what are the alternatives?

Graham

>I can do it in Kyma, but each sample is expensive. Perhaps it would be
>better if there were an efficient sinusoid generator, instead of using
>samples of sine waves.

Wha? There must be sine oscillators in kyma.

>I don't think FM's prominent enough now to hold anything back. Perhaps
>analog emulation is. But what are the alternatives? Additive is there,
>but more complex to program.

Resynthesis is a good way to quickly set up additive patches, but I'm
not aware of any consumer synth that supports it. SoundDiver claimed
to do it, but the free version that shipped with my Kawai K5000 has
that feature torn out.

>Physical modelling is on the market, but is also complex to implement
>and program. // There are soft synths to do all of these things.

I don't know of any consumer synth that does polyphonic phy. mod.
piano. Actually, the only polyphonic keys I've found are BigTick's
TickyClav (soft) and the Korg Oasys board had a electric piano and
organ, I think.

-Carl

on 8/7/03 8:26 PM, Graham Breed <graham@microtonal.co.uk> wrote:

> Kurt Bigler wrote:
>
>> I was speaking too generally perhaps, about what "can" be done, rather than
>> the more practical question about what "has" been made available to the
>> musician who is not a programmer. Nowadays software sinusoidal synthesis is
>> very cheap (but probably not available off-the-shelf) and you could program
>> it to replicate any given FM pattern you wanted, but you could also program
>> it to replicate an entire universe of other possibilities, without any of
>> the restrictions inherent in FM. I felt FM provided expressiveness, but so
>> often with the wrong degrees of freedom, i.e. to get one thing you want you
>> always had to get something else you don't want. That was my experience.
>
> I don't agree with the middle bit. Noisy FM patches would require a
> very large number of sinusoids that no additive approach could provide
> in real time.

I can't say I haven't heard of sinusoidal modelling of noise being done, but
someone indicated to me that most music can be approximated extremely well
using FFT of surprisingly limited frequency resolution, and that this would
include effectively modelling of pitch/consonance/dissonance as well as
noise. So I don't know directly, but I have heard that argument. Since
pitch is so important here I'll try to learn more about this, and report
back if I find anything interesting.

> In addition, the result probably depends on the sample
> rate and depth of the original, so mathematical predictions won't be
> enough, and you may as well use samples.

I don't follow this.

> Also, as FM in practice includes AM there are no inherent restrictions.
> The TX81Z itself has all kinds of restrictions, because of what was
> practical at the time it was made. If you've got enough power to
> implement 100 additive sine waves, you can also implement 100 FM
> operators.

This might not be true if you use FFT to generate the sine waves, though I
haven't thought deeply about it.

> Both will be hugely complex to work with unless you have
> sone kind of meta-algorithm.

Yes, the meta-algorithm would be what I had in mind.

>> Meanwhile the fully parameterized sinusoidal additive thing or numerous
>> other approaches could be done in Max/MSP but probably requires more than a
>> small learning curve, whereas the perhaps-limited FM approach is already up
>> and running.
>
> I can do it in Kyma, but each sample is expensive. Perhaps it would be
> better if there were an efficient sinusoid generator, instead of using
> samples of sine waves.

I would expect that even without using FFT, a state-of-the-art computer
would now be capable of 10,000 or more simultaneous sine waves. I'm doing
500 or so simulataneous (very simple) physical model instances on a dual
1GHz G4, and that is severely limited by memory access, whereas sinusoidal
would not have that problem, and is also generally much simpler (per
oscillator), so I give it at least another factor of 10 for 5000
simultaneous sines, and therefore at least 10,000 on the soon-to-release
dual 2GHz G5. Of course parameters controlling the sine waves in real time
would add a bit to this, but still, it would be a lot. Not that this isn't
a very expensvie route (at the moment), but you may not need the latest
hardware to get what is needed.

>> Partly I speak up because I figure non-developers are not necessarily aware
>> of what is readily possible right now, but simply not on the market. But
>> this may be misplaced. If enough people were ticked off at FM synthesis
>> then I suppose we would be seeing more algorithmic alternatives. So I'll be
>> quiet now.
>
> I don't think FM's prominent enough now to hold anything back. Perhaps
> analog emulation is. But what are the alternatives? Additive is there,
> but more complex to program.

I would think the alternative might be Additive with better abstractions (as
you said "meta-algorithm"). Try to match the terse functionality of FM,
with perhaps just a couple more voicing options that would affect how the
frequency spectrum "blooms". I think FM often blooms wrong.

> Physical modelling is on the market, but
> is also complex to implement and program. Granular needs samples to
> work with. There are soft synths to do all of these things.
>
> A TX81Z is still a good, cheap, standalone, microtonal synth module. It
> is limited in sound quality, but what are the alternatives?

Yes, as I said, I should rest my case on the truly practical point. I
always wish for better things for musicians, but alas development is
limited!

-Kurt

> Graham

--- In tuning@yahoogroups.com, Kurt Bigler <kkb@b...> wrote:

> I would think the alternative might be Additive with better
abstractions (as
> you said "meta-algorithm").

I think this would be a very nice approach if people came up with
economical, effective and musical meta-algorithms. Also additive
synthesis has the benefit of completely bandlimited signals with no
aliasing.

But at least in the past additive has required enormous amounts of
input data if done from scratch. I've read about the meticulous
programming of Wendy Carlos making additive sounds but many of those
sounds she made with so much pains sound thin, lame and lifeless in
my ears. Just my opinion.

Try to match the terse functionality of FM,
> with perhaps just a couple more voicing options that would affect
how the
> frequency spectrum "blooms". I think FM often blooms wrong.

I think there is a certain quality to FM sounds. If you have just a
single modulator-carrier-pair and broad and fast changes in the
modulation index I would most often describe this quality as
irritating. Something to do with those bouncing Bessel functions I
guess. But with more complex algorithms the results get much better.
One simple but effective algorithm is three pairs of modulator-
carrier with different ratios. Good bell and electric piano sounds
come out of this configuration. And FM is a bliss with bass sounds.

FM7 has completely programmable matrix for programming your own
algorithms. Yamaha FM synths have preprogrammed algorithms. It has
six operators (TX81z has only four) with 32 different waveforms.
Every operator can be both a modulator and a carrier and anyone of
them can modulate itself in feedback (only one with Yamaha products).
Ratios can be set with 4 decimals and the operators can be detuned
with Hz offset. In addition FM7 has dual filters and distortion unit
with a noise source. The operators, pitch, filter and distortion have
all loopable envelopes with 31 breakpoints. It has also quite good
programmable effects section which can do delay, pseudo reverb,
chorusing and flanging type effects. With all this and more its sound
capabilities go way beyond any previous commercial implementation of
FM.

And most importantly: registered users can download an update with
the ability to load Midi Tuning Standard files (you can create them
with Scala) as SysEx into the machine. I'm not sure about the tuning
resolution though. It has a 12-tone microtuning section which seems
to have a one cent resolution but my listening experiments indicate
that it goes much beyond that in the SySex mode.

Kalle

>I can't say I haven't heard of sinusoidal modelling of noise being
>done, but someone indicated to me that most music can be approximated
>extremely well using FFT of surprisingly limited frequency resolution,

If you have Wendy Carlos' *Digital Moonscapes*, you can hear what
was possible in the early 80's with the GDS Synergy -- a box capable
of, IIRC, 16 simultaneous partials. They didn't have to be harmonic.
With all of Carlos' painstaking programming, the results are still
my favorite synth sounds to date.

-Carl

Kurt Bigler wrote:

> I can't say I haven't heard of sinusoidal modelling of noise being done, but
> someone indicated to me that most music can be approximated extremely well
> using FFT of surprisingly limited frequency resolution, and that this would
> include effectively modelling of pitch/consonance/dissonance as well as
> noise. So I don't know directly, but I have heard that argument. Since
> pitch is so important here I'll try to learn more about this, and report
> back if I find anything interesting.

Oh, yes, if you use FFTs you can reproduce any sound. I was thinking of literally adding together a set of sine wave generators. And I remember producing fractal noise using FFTs on a 486, faster than the duration of the resulting files, so it must be possible in real time on modern equipment.

Noise patches will require more resolution than most music.

>>In addition, the result probably depends on the sample
>>rate and depth of the original, so mathematical predictions won't be
>>enough, and you may as well use samples.
> > I don't follow this.

You can theoretically predict the harmonic structure of FM timbres using Bessel functions. So that would be your meta-algorithm for additive synthesis. But for noise patches, the partials are still significant up to a high number. Anything approaching or exceeding the sample rate will alias, which you also have to model. Perhaps that can be done by bouncing below the Nyquist frequency, but I suspect the only way to reproduce a particular FM implementation would be to model it exactly and do an inverse FFT. If you can do that, you're really doing FM or sampling, and the additive bit doesn't save you anything.

>>Also, as FM in practice includes AM there are no inherent restrictions.
>>The TX81Z itself has all kinds of restrictions, because of what was
>>practical at the time it was made. If you've got enough power to
>>implement 100 additive sine waves, you can also implement 100 FM
>>operators.
> > > This might not be true if you use FFT to generate the sine waves, though I
> haven't thought deeply about it.

Yes, I think you're right. When the complexity gets into thousands, FFTs will be more efficient than FM synthesis. FM is more efficient for small numbers of operators.

> I would expect that even without using FFT, a state-of-the-art computer
> would now be capable of 10,000 or more simultaneous sine waves. I'm doing
> 500 or so simulataneous (very simple) physical model instances on a dual
> 1GHz G4, and that is severely limited by memory access, whereas sinusoidal
> would not have that problem, and is also generally much simpler (per
> oscillator), so I give it at least another factor of 10 for 5000
> simultaneous sines, and therefore at least 10,000 on the soon-to-release
> dual 2GHz G5. Of course parameters controlling the sine waves in real time
> would add a bit to this, but still, it would be a lot. Not that this isn't
> a very expensvie route (at the moment), but you may not need the latest
> hardware to get what is needed.

10,000 sines over 128 voices is only 78 sines per voice. So there are still practical limits, but a TX81Z will run out first. You may want to add reverb, chorus, etc, but that would only be a linear increase.

Kyma advertises 192 sine waves (with independent envelopes) for a basic system and 1743 fully loaded. That's old hardware, so the orders of magnitude are correct.

I could try resynthesising a TX patch with 78 sines to see how well it does come out. I've a feeling it won't work because of the analysis overhead.

> I would think the alternative might be Additive with better abstractions (as
> you said "meta-algorithm"). Try to match the terse functionality of FM,
> with perhaps just a couple more voicing options that would affect how the
> frequency spectrum "blooms". I think FM often blooms wrong.

If you can come up with such abstractions. I don't think anybody has with results much better than FM. But then I don't know of any soft synths with 10,000 sinusoids, so perhaps you're onto something there.

Graham

>Kyma advertises 192 sine waves (with independent envelopes) for a basic
>system and 1743 fully loaded. That's old hardware, so the orders of
>magnitude are correct.

I see they're no longer advertising any numbers for additive...

Also they apparently *still* don't have NT kernel compatibility.
That's a joke. If I were them, I'd start on porting it entirely
over to Windows now.

-Carl

How many tuning user-saved settings (vs. presets) are available on
the TX81z -- I read that the DX-11 has only 2 user-saved tuning
settings, and 10 presets. Is it the same for the TX81z? Would this be
crippling or is it quick to save a new setting? -Justin

--- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...> wrote:

/tuning/topicId_46261.html#46278

> Do FM synthesizers like the Yamaha TX series have preset timbres
with
> harmonic harmonics or do you have to create all the sounds yourself
> from scratch? -Justin
>

***There are, of course, presets, but you can also design your own
sounds with a patch editor. Midi Quest (now version 9) is quite
good. And you can create sounds with *both* harmonic and inharmonic
content.

J. Pehrson

--- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...> wrote:

/tuning/topicId_46261.html#46288

> What sort of FM synth would you recommend, if not the TX81z? Two
> things are important: 1) tunability resolution and 2)
affordability. -
> Justin

***The good ol' TX81z. Reasons? Points 1 and 2 above...

J. Pehrson

P.S. Tunability: *flexibility* in the full range. The resolution is
the 1.56 cent "standard" discussed frequently on this list.

--- In tuning@yahoogroups.com, "Justin Weaver" <improvist@u...> wrote:

/tuning/topicId_46261.html#46352

> How many tuning user-saved settings (vs. presets) are available on
> the TX81z -- I read that the DX-11 has only 2 user-saved tuning
> settings, and 10 presets. Is it the same for the TX81z? Would this
be
> crippling or is it quick to save a new setting? -Justin

***There are only two. One is called "octave" and you set only one
octave and the Z duplicates it throughout the range. The other is
called "full" and is the full keyboard.

Using SCALA, new settings can be loaded in in a moment of seconds.
It might be difficult, though, to do this in a live performance
situation, although it could be possible if something else were going
on. (The drummer has a big riff... :)

J. Pehrson