ION oscillators NOT analog modelling?

[Jeebus]

While playing the Ion yesterday, I noticed strange behaviours on the INIT patch, can somebody explain this to me?

 

It sounds like the Ion oscillators are simple ROM playback waveforms, and not analog modeling.

On high notes on the sawtooth wave, there is very harsh aliasing, just like a sample beeing played back too fast. Actually this is the worst aliasing that I have ever noticed on a

synthesizer.

 

On low notes the sawtooth looses its high end the lower you play, like a very slow sample playback. The filters don't bite anymore, as they are not fed the high frequencies.

 

I have an Access Virus, its oscillators create a thick and full spectrum even on the lowest notes, just like on analog synths.

 

There is other evidence for the sample theory:

When you move the oscillator shape very slowly from -100% (Sawtooth) to Triangle, you can hear little steps. It sounds like the shape is actually selector into a digital wavetable.

 

When 500 MIPS is so much processing power, why is the oscillator a simple digital wavetable instead of analog modeling?

[Dave Bryce]

Originally posted by The Jeeebus:

While playing the Ion yesterday, I noticed strange behaviours on the INIT patch, can somebody explain this to me?

 

It sounds like the Ion oscillators are simple ROM playback waveforms, and not analog modeling.

On high notes on the sawtooth wave, there is very harsh aliasing, just like a sample beeing played back too fast. Actually this is the worst aliasing that I have ever noticed on a

synthesizer.

 

On low notes the sawtooth looses its high end the lower you play, like a very slow sample playback. The filters don't bite anymore, as they are not fed the high frequencies.

 

I have an Access Virus, its oscillators create a thick and full spectrum even on the lowest notes, just like on analog synths.

Under which conditions did you evaluate the Ion?

 

There is other evidence for the sample theory

 

So, you think Alesis is intentionally being deceptive when they say it's a modeling synth, and that it's really just a ROMpler? Really?

 

dB

[Pim]

If you don't like Ion, avoid it. Put your energy in things you like.

[Dave Bryce]

Originally posted by Pim:

If you don't like Ion, avoid it. Put your energy in things you like.

Agreed. Well said.

 

dB

[aeon]

It is interesting, because I have heard some serious aliasing issues/complaints from ION owners on a couple of mailing lists.

 

I am going to check one out very soon to hear for myself.

 

That said, Jeeebus, the Virus is hardly any paragon to hold up when it comes to this issue given the fact that it aliases in the high end, and aside from its pulse and sawtooth waves, it uses PCM samples for playback.

 

Hell, it may use PCM samples for the pulse and sawtooth as well, given I have not disassembled the ROM, LOL, but my guess is that they are generated on the fly, if even with coefficients precalculated.

[RABid]

It seems that most software VA's use tables so this does not bother me. The stepping that can be heard may just be the resolution of the controller that is changing the parameter. As for aliasing most any synth that does not use filtering at the high end will do this. Just because some like Virus hide it a bit better, it does not mean they produce high frequencies any better. Open up the filter and play some very high notes on those synths. You will either hear aliasing or filtering unless it is running at an extremely high processing rate.

 

Robert

[aeon]

so here is the easy way to show it to the world, or this forum...

 

someone with an ION, record each note from C5 on up using a single sawtooth osc with the filter wide open, or with no filter if possible, and give us 1 second per key...

 

if it aliases, I am sure we will be able to hear it!

[Dave Bryce]

Originally posted by aeon:

so here is the easy way to show it to the world, or this forum...

 

someone with an ION, record each note from C5 on up using a single sawtooth osc with the filter wide open, or with no filter if possible, and give us 1 second per key...

 

if it aliases, I am sure we will be able to hear it!

Okay, I just checked it. Jeeebus is right.

 

There is definitely aliasing, and it's actually fairly noticeable on the top octave of the highest oscillator setting (transpose set to +3 octaves) with the filter wide open. By the time you get to the last few keys, it's really pretty bad.

 

I reviewed the programs that I wrote, and none of them had the oscillators set that high - guess that's why I didn't notice it before.

 

I stand corrected. My apologies for blindly defending without checking it out.

 

dB

[aeon]

OK, it is interesting, especially in light of the fact that this is a fairly basic (and I do not mean easy, at least computationally) thing to avoid in a modeled oscillator.

 

I also find it interesting in light of the reviews that did not speak about it.

 

After all, given that sawtooths and pulse waves have (theoretically) infinite harmonics, if a digital synth clearly aliases in the top octaves, it means it is aliasing throughout the range of the keyboard, even if the fundamental and harmonics are masking it in those ranges. In any event, it is trivial to check and graph such things with a quality spectrum analyzer...it would be great to see that done with a wide range of VAs, ION included.

 

The best antialiasing oscs I know of that run on DSP in hardware are the analog emulations on the Korg OASYS PCI...there is 480 MIPS doing it near-perfect.

 

That said, it doesnt mean the Alesis ION cant do the job for many musical uses...after all, like I said, the Virus is no shining star in the aliasing department, and it has sold boatloads, appeared on many recordings, and perhaps most importantly, has become known for having a signature sound...something all synth designers strive for, and something that can be said about most synths that have become classics.

[Jeebus]

Thank you Dave. I'm glad you hear it too. It just bothered me that Alesis has been touting it as an incredibly smooth synth with no aliasing, and everyone around me seeming to love it without hearing what I was hearing.

 

Sorry if I came across too harsh. Just trying to make people aware.

[VanZea]

On the Ion mailinglist somebody explained this 'aliasing' by telling us that in the INIT patch OSC Sync is set to ON. So it's not aliasing you guys are hearing.

 

Grtz,

Jeroen

[VanZea]

Here's the quote:

 

===============

 

>Still, on the high notes there is tons of aliasing: Set the octave to

> +2 and play the highest key while moving the pitch bender. The

weird

> noise you hear is aliasing.

> It is clearly perceptual on lower ranges, too. My Virus C is fine

in

> the same key ranges!

 

You need to turn off oscillator sync. That is the "noise" you are

hearing. I've made the same mistake myself starting patches from user

127. My ears may be getting on in years, but I cannot hear ANY

aliasing from the Ion oscillators during slow to fast shape or pitch

sweeping in any octave. The dedicated fine pitch knobs on osc 2 & 3

are real sensitive, too. Unless you want beating make sure all 3

oscillators are set to 0. In addition to sounding "real" to me, the

oscillators are dead on tune, track perfectly, and don't need

calibration! (Of course, that's all probably true of most digital

oscillators;-)

 

===============

Grtz,

Jeroen

[Dave Bryce]

Originally posted by VanZea:

On the Ion mailinglist somebody explained this 'aliasing' by telling us that in the INIT patch OSC Sync is set to ON. So it's not aliasing you guys are hearing.

The program I started with for the test was not an initialized program, it was one of mine.

 

I turned off all but one oscillator (set to saw) and one filter, and opened the filter all the way up. Oscillator sync is not on, and it definitely aliases on the last octave and a half. I just double-checked, and turning oscillator sync on and off didn't seem to affect it at all. I just tried the same thing on three other programs - each one yielded exactly the same results.

 

One thing I did find was that it's much more noticeable on certain filter types than it is on others.

 

dB

[coyote]

Is the aliasing something that can be corrected in a future OS update?

 

Also, in the OS update you might want to turn off OSC SYNC on the init patch. Seems kinda strange to have an init patch that's not giving you a pure, unadulterated waveform.

[Pim]

Originally posted by Dave Bryce:

I reviewed the programs that I wrote, and none of them had the oscillators set that high - dB

Your programs aren't set that high - mine either. There is something there that doesn't occur on analogues. Hats off for Jeebus! No one digged that deep.

 

On the other hand, what are we talking about? It's a fine musical instrument. There are so many fine musical instruments where you can place several remarks about.

 

I recorded with several Sequential Prophets which all had filter stepping when I turned the filter frequency knob. A DX7 has a lot of hiss on the bass side - my Mini has to be re-tuned when I switch to higher octaves; it has to be re-tuned when the central heating switches on.

 

You know, even every great guitar has a "dead" spot on the neck. Well, every great synth too. I am not looking for design and marketing errors from any factory. I want to play with great synths. If I don't like a synth I simply ignore it. I ignore a lot of synths!

[marino]

You guys definitely got my attention.

 

I remember a rumor about the OB-12 when it first came out, that its raw waveforms were sampled rather than modeled. I never got to know the truth, but when I played it, they did sound like samples to me...

 

Anyway I did a few tests on my synths, and here's what I've found:

 

The Matrix-12 does not alias. Period.

 

The SE-1 shows some artifacts in the very high register, but that's not aliasing, just the analog oscillators getting out of their tracking range. It's another kind of noise.

 

I did expect the the AN1x to show some kind of aliasing, but it holds very well, and I can hear only a tiny bit of aliasing just before it enters the inaudible range. Chapeau.

 

All synths show some strange cyclic behavior in the very high register when LFOs and some kind of envelopes are used, but I guess that doesn't count as aliasing - I did the tests with a naked sawtooth wave, then a square wave, no filter, no Q, no effects, no envs, no LFOs.

 

Of course all my ROMplers alias like mad... But as the former owner of a PPG, I can stand that.

 

The DX7 was horrid too. Which reminds me, I didn't test the TG77! Back in a few minutes...

 

(edited for a mistake: I wrote OB-8, I meant the OB-12 of course! Sorry)

[marino]

OK I tested the TG77, initialized patch, FM section only, one operator, no filter, no nothing.

 

It does alias rather badly. The amount of aliasing is strongly dependent on the kind of waveform used. Not surprisingly, the sine and sine-like waves are almost devoid of aliasing. The pulse-like waves show quite a bit of noise, and the saw-like waves are pretty bad - you can hear it even on the last *octave and a half* of an 88-note keyboard.

 

Let me just add that I like this instrument a lot. It (and the SY77/99) is the only FM synth which is really programmable without a computer. (I used to use an editor, but the panel is not bad either)

With the added waweforms and filters, it sounds great to me. The ROM samples suck bigtime, but when you FM them with the synthetized waves, they can help too. It's very good at analog-type sounds, too!

 

Bottom line: It aliases, but I love it.

 

Back to the music now.

[coyote]

I see what you are saying. Yet having someone look for such potential issues is not a terrible thing - it might even prod Alesis to mak the product better in the future. Had everyone said "the Minomoog is perfect as is" there'd be no Ion, no SE-1, no Andromeda.... perhaps their QA team could put "check sound at high registers" on their checklist?

 

It's a natural oversight to miss extreme high-end stuff like that because we mostly play & hear in the middle registers. But I like the idea of people checking for such problems, so that the times we do use high registers we have good sound.

 

 

Originally posted by Pim:

Originally posted by Dave Bryce:

I reviewed the programs that I wrote, and none of them had the oscillators set that high - dB

Your programs aren't set that high - mine either. There is something there that doesn't occur on analogues. Hats off for Jeebus! No one digged that deep.

 

On the other hand, what are we talking about? It's a fine musical instrument. There are so many fine musical instruments where you can place several remarks about.

 

Well, every great synth too. I am not looking for design and marketing errors from any factory. I want to play with great synths. If I don't like a synth I simply ignore it. I ignore a lot of synths!

[Geoff Grace]

Aha! So it's not always wise to apply the truism that a chain is only as strong as its weakest link!

 

Generally, I'd say that a keyboard is only as strong as its strongest link. In other words, it's the good features of a keyboard that I count on when finding or making sounds. The weaknesses are almost never "fatal flaws" as long as the strengths are compelling.

 

Nonetheless, I always appreciate a heads up. Good work Jeeebus!

 

Best,

 

Geoff

[Tusker]

Originally posted by marino:

I did expect the the AN1x to show some kind of aliasing, but it holds very well, and I can hear only a tiny bit of aliasing just before it enters the inaudible range. Chapeau.

Marino, forgive my ignorance. What does chapeau mean? I looked it up in an on-line dictionary ... and it says it's a "hat". I remember you using this expression before and it went over my head ( ) that time too. It's really bugging me not to know. Would you please end my torment as I search for the right meaning? Keep it under your hat? This synth is completely toppers? My synth it has three corners?

 

Originally posted by marino:

The DX7 was horrid too. Which reminds me, I didn't test the TG77! Back in a few minutes...

Interesting note about the TG. I notice a huge escalation in aliasing in Roland (XP) synths when using ring modulation. With the ring mod, I've wondered if the sum frquencies (as in sum plus difference) just blow past the capabilities of the D/A convertors. And it depends on the sample used too. You can drive the aliasing down by using the simple (triangle, sine) waveforms. This enables you to play higher. But those rich plucked string waveforms are great for digital ring mod (or I would guess FM) type effects. Just so long as one plays the low end of the keyboard.

 

Best,

 

Jerry

[Dan South]

Just because a trumpet doesn't sound like a tuba when played in the tuba register doesn't mean that a trumpet is not a useful instrument. Highest octave of an oscillator? I don't give a sh*t. I'm not writing music for dogs.

[Jeff Klopmeyer]

Originally posted by Dan South:

I don't give a sh*t. I'm not writing music for dogs.

Dan,

 

Rowf bow wow roof roof growl. Rowr row row, bow rowf woof woof hooooooooowwwwwwwwllllll!

 

(translation: we dogs don't need your steenkin' music anyway.)

 

Anyway, I guess it's one of those things...if that high register aliasing bugs you, you shouldn't get this keyboard. If not, it still seems pretty cool.

 

- Jeff

[Postman]

Um, guys, all digital waveforms alias, modelled or not.

 

Nyquist's theorem: A theorem, developed by H. Nyquist, which states that an analog signal waveform may be uniquely reconstructed, without error, from samples taken at equal time intervals. The sampling rate must be equal to, or greater than, twice the highest frequency component in the analog signal.

The "error" in the definition is aliasing.

 

If the sampling rate of Ion's oscillators are less than two times the highest frequency they're asked to produce, you'll hear aliasing. Alesis doesn't claim that the Ion uses digital control of analog oscillators (that's the Andromeda).

 

Aliasing has nothing to do with modelling. If Alesis is generating the waveforms using some sort of algorithm, they can claim modelling. If they're just playing back digitized analog signals, then they're lying.

 

Either way, the presence of aliasing isn't indicative of either. The sample rate of the ION's oscillators are just a little too low.

[Bobro]

Originally posted by Geoff Grace:

Generally, I'd say that a keyboard is only as strong as its strongest link. In other words, it's the good features of a keyboard that I count on when finding or making sounds.

 

Substitute "musical instrument" or "voice" or even "musician" for the word "keyboard" and it's even better.

 

-Bobro

[F7sound -]

Originally posted by Postman:

Um, guys, all digital waveforms alias, modelled or not.

 

If the sampling rate of Ion's oscillators are less than two times the highest frequency they're asked to produce, you'll hear aliasing. Alesis doesn't claim that the Ion uses digital control of analog oscillators (that's the Andromeda).

 

Aliasing has nothing to do with modelling. If Alesis is generating the waveforms using some sort of algorithm, they can claim modelling. If they're just playing back digitized analog signals, then they're lying.

 

Either way, the presence of aliasing isn't indicative of either. The sample rate of the ION's oscillators are just a little too low.

We never see what the "sample rate" of a VA synth is from it's spec sheets. Does it make any difference? I think so. And apparently so do a lot of you.

 

All I know is, my JP8000 aliases quite a bit. It also "stair steps" and does a lot of other kinda strange things. That's what gives it it's "sound".

 

Maybe we'll begin to hear more about synth's sample rates in the future.

 

Michael Oster

F7 Sound and Vision

[Postman]

Originally posted by F7sound:

We never see what the "sample rate" of a VA synth is from it's spec sheets. Does it make any difference? I think so. And apparently so do a lot of you.

Yes, I've wondered about that, too. The synth industry pushes resolution statistics (8 bit, 16 bit, etc.) as if its the only measure of digital accuracy, but there are really two measures, resolution and sample rate. I'm guessing here, but it seems to me that the best measure of processing power in a synth is the sample rate. Do any of the modelling synths achieve anything close to 44.1KHz for instance?

[aeon]

coyote said:

It's a natural oversight to miss extreme high-end stuff like that because we mostly play & hear in the middle registers.

Dan South said:

I don't give a sh*t. I'm not writing music for dogs.

What you must keep in mind, however, is that the aliasing will occur in the middle registers as well if the instrument is a modeled analog with waves like sawtooth and pulse waves...because those waves have harmonics extending to infinity, any note on the keyboard will alias to some degree.

 

The ear is very sensitive to inharmonic frequencies, and I think the presence of them, through aliasing or otherwise, is a major factor in why VAs have remained as far from emulating analog sound as they have, among many other reasons.

 

Postman said:

Um, guys, all digital waveforms alias, modelled or not.

Not true! It is entirely possible to create digital waveforms that will not alias across the keyboard...and it has been done near-perfectly already in a couple of DSP synths and has been done to pefection in a few softsynths.

[Dave Bryce]

Originally posted by aeon:

It is entirely possible to create digital waveforms that will not alias across the keyboard...and it has been done near-perfectly already in a couple of DSP synths and has been done to pefection in a few softsynths.

Also, if this aliasing is in fact as obvious as it appears to me, I'm actually not sure why Alesis' web site description of the Ion would say (as Jeeebus pointed out), "extraordinary precision void of any undesirable digital artifacts and a very analog sound throughout the entire frequency range."

 

I must be missing something...

 

dB

[Postman]

Originally posted by aeon:

Postman said:

Um, guys, all digital waveforms alias, modelled or not.

Not true! It is entirely possible to create digital waveforms that will not alias across the keyboard...and it has been done near-perfectly already in a couple of DSP synths and has been done to pefection in a few softsynths.[/QB]

Sorry, I should have said "all digital waveform generators will exhibit aliasing at some frequency though that point may be above the threshold of human hearing (~ 20Hz-20KHz) if the sampling rate is high enough." That's why the sampling rate of CDs is 44.1KHz.

[marino]

Originally posted by Tusker:

Marino, forgive my ignorance. What does chapeau mean? I looked it up in an on-line dictionary ... and it says it's a "hat". I remember you using this expression before and it went over my head ( ) that time too. It's really bugging me not to know. Would you please end my torment as I search for the right meaning? Keep it under your hat? This synth is completely toppers? My synth it has three corners?

Jerry,

'Chapeau' is French for 'Hats off'. It was a compliment to the AN1x!

 

Also, your comment about the ring mod on the Rolands rang a bell. I went back to the SE-1, and sure enough, there was a bit of ring mod engaged! Strangely enough, the SE-1 ring modulator works even when the oscillator are off... strange, if you ask me.

In short: The oscs on the SE-1 have a tracking range limit, but they don't produce that nasty noise I had noticed. It was operator error. Sorry!