KVR Audio

aciddose · Post by **aciddose** » Thu Apr 23, 2015 10:01 pm

Okay, assuming they didn't actually dive in and modify the synthesizer and that they were indeed using the OB-Xa and not something completely different (entirely possible), the filter tracking adjustment uses a 10k trimpot in series 44k or 50k... so lets just say 16k trim in series 44k into 1k.

So:

min: 1/44 = 0.0227
mid: 1/52 = 0.0192
max: 1/60 = 0.0166

So these are our scaling coefficients for the frequency exponent, which is normally = 2.

So we'll simply normalize to 2: coefficient = 104.166.

Now, normalized exponential coefficients (assuming mid = 2)

min: 1.729
mid: 2.0
max: 2.364

And we want to try to estimate the minimum exponent, so our answer is 1.729.

Now with 100% key track, each octave will actually be only 1.729 times rather than 2 times the frequency.

Each semitone is: 1.729 ^ (1/12), or 1.0466856 vs. 1.059463.

So now I can set Xhip's asymptote to 130% and the filter scale coefficient to 1.0466856 and we can see what happens... not even close.

I'd be willing to say that there is absolutely no way van helen used an OB-Xa for that sound unless it was heavily modified or damaged.

But the alpha version you uploaded today is good enough. Would be nice to have a slider controlling the curve, but i can wait for that.

Well, the problem is that there is no real way to get this sort of result with any practical synthesizer.

You mentioned:

Elektrostudio Model Mini in exponential mode

I have no idea what this would be doing... To get a similar curve the asymptote needs to be something like 100.0001%. No shaping of the resulting envelope could produce this unless the shape were very extreme. It may be using a linear envelope with a function like x/(x+y) where y is very low such as 0.0001.

We can be certain though that to get this sort of curve, it is impossible the original synthesizer was the OB-Xa. It is also impossible the envelope were any envelope based upon a functional CEM-3310 as the asymptote is determined by the internals of the chip.

You may get closer with this http://xhip.net/temp/xhip_alpha_100_tiny_asymptote.7z

Still though, we're talking about a broken synthesizer here, not a special feature. Even if Xhip had an adjustable asymptote, this would be a value as far to the side as you could possibly put it and most likely way further.

Mutant · Post by **Mutant** » Thu Apr 23, 2015 10:06 pm

fluffy_little_something wrote:I also had a problem understanding the diagrams. It is unusual to have time on the y axis. The x axis is traditionally the time axis in many applications.

The analyzer gives time on Y because if i remember correctly it started as just a simple frequency analyzer and oscilloscope and kept the window size and format as the developer was adding new features which the sonogram was one of the last.

Mutant · Post by **Mutant** » Thu Apr 23, 2015 10:11 pm

aciddose wrote:Okay, assuming they didn't actually dive in and modify the synthesizer and that they were indeed using the OB-Xa and not something completely different (entirely possible), the filter tracking adjustment uses a 10k trimpot in series 44k or 50k... so lets just say 16k trim in series 44k into 1k.

So:

min: 1/44 = 0.0227

mid: 1/52 = 0.0192

max: 1/60 = 0.0166
So these are our scaling coefficients for the frequency exponent, which is normally = 2.

So we'll simply normalize to 2: coefficient = 104.166.

Now, normalized exponential coefficients (assuming mid = 2)

min: 1.729

mid: 2.0

max: 2.364
And we want to try to estimate the minimum exponent, so our answer is 1.729.

Now with 100% key track, each octave will actually be only 1.729 times rather than 2 times the frequency.

Each semitone is: 1.729 ^ (1/12), or 1.0466856 vs. 1.059463.

So now I can set Xhip's asymptote to 130% and the filter scale coefficient to 1.0466856 and we can see what happens... not even close.

I'd be willing to say that there is absolutely no way van helen used an OB-Xa for that sound unless it was heavily modified or damaged.

But the alpha version you uploaded today is good enough. Would be nice to have a slider controlling the curve, but i can wait for that.
Well, the problem is that there is no real way to get this sort of result with any practical synthesizer.

You mentioned:
Elektrostudio Model Mini in exponential mode
I have no idea what this would be doing... To get a similar curve the asymptote needs to be something like 100.0001%. No shaping of the resulting envelope could produce this unless the shape were very extreme. It may be using a linear envelope with a function like x/(x+y) where y is very low such as 0.0001.

We can be certain though that to get this sort of curve, it is impossible the original synthesizer was the OB-Xa. It is also impossible the envelope were any envelope based upon a functional CEM-3310 as the asymptote is determined by the internals of the chip.

You may get closer with this http://xhip.net/temp/xhip_alpha_100_tiny_asymptote.7z

Still though, we're talking about a broken synthesizer here, not a special feature. Even if Xhip had an adjustable asymptote, this would be a value as far to the side as you could possibly put it and most likely way further.

Ill check this version now.
Also please reread my last post on 2nd page, i edited it 1 minute after you wrote this post.

fluffy_little_something · Thu Apr 23, 2015 10:15 pm

How does Sonic Projects' OP-X behave? If it produces similar diagrams, Van Halen's synth was not broken

Mutant · Post by **Mutant** » Thu Apr 23, 2015 10:22 pm

fluffy_little_something wrote:How does Sonic Projects' OP-X behave? If it produces similar diagrams, Van Halen's synth was not broken

IDK i don't have it.

But as i said i highly doubt that tutorial guys synth was broken the exact same way, so maybe the spec sheet was wrong instead.

---

And the tiny asymptote version is perfect ! I can't see anymore where A connects to S.
Good night everyone.

[edit]
More OBX-a from Retrosound:
https://www.youtube.com/watch?v=byxQ6aQRGis

aciddose · Post by **aciddose** » Thu Apr 23, 2015 10:26 pm

fluffy_little_something wrote:How does Sonic Projects' OP-X behave? If it produces similar diagrams, Van Halen's synth was not broken

What!?

You mean how does a real OB-Xa behave, right? Van Halen didn't use Sonic Projects' OP-X

It is entirely possible the OP-X doesn't model the OB-Xa correctly, but it is impossible for the OB-Xa to behave in the way the synthesizer in 1984 behaves if it is tuned correctly. Not only that, I won't say I'm absolutely certain about this but I see no way it is possible for the OB-Xa to behave that way even when tuned to its most extreme settings.

I also can't see one being modified to work that way, although there are a lot of clever ways to modify/break a thing and I'm sure I can't imagine all of them, so I won't say that's impossible.

nineofkings · Post by **nineofkings** » Thu Apr 23, 2015 10:27 pm

Synthmaster can set any amount of exponentiality for the envelopes. Not sure what spectrum analyzer that is though so I can't provide pretty graphs

aciddose · Post by **aciddose** » Thu Apr 23, 2015 10:31 pm

Wait, mutant did you use the original CD master of 1984 to produce the graph, or did you rip that off youtube?

You're probably just looking at the frequency get warped into the range of the encoding used by youtube for that clip if so. I see a normal envelope, perfectly tuned with 130% asymptote on my vinyl copy.

Sharp edges.

fluffy_little_something · Thu Apr 23, 2015 10:48 pm

I know they didn't use the software (was the developer even born, yet?

), but it is said to be a very good emulation, so the developer probably also took those things into account.

Anyway, I doubt VH used a broken or modified synth, maybe there are other aspects to consider...

aciddose · Post by **aciddose** » Thu Apr 23, 2015 10:54 pm

I think that we need to start with the most pure recording we have available. If anyone has the CD containing 1984 and can verify it matches the vinyl release (not heavily processed or remastered) we can just use the digital rip of that master (CD) to measure.

What I'm seeing from most of the sounds is a sharp edge and often the cutoff is going above 20khz just as you'd expect. In these cases if we apply another filter like the steep filter at 18khz applied for mp3, we can actually see a line that looks rounded off quite a lot like mutant's graph... although not exactly the same.

Also I'm not sure exactly which "wow" that graph was plotted from, most of them are polyphonic with a lot more than two voices (usually three or four) and it sounds layered also.

So:

Are we certain the synthesizer is a OB-Xa?
Which synthesizers were used for the layering? The same OB-Xa or multiple different instruments?
Do we have a better (clean) source to analyze?
Are we sure the source used so far hasn't introduced anything weird into the graph?

I don't really feel comfortable with it, but if nobody has a CD I can try to record the vinyl. A bit of a pain in the arse though and vinyl almost definitely has a more limited frequency range than the digital rip from CD will.

aciddose · Post by **aciddose** » Thu Apr 23, 2015 11:06 pm

Also: I'll try to find a free tool we can all use to plot a good spectrograph of a .wav file. I may even write one later on if the only tools I can find (audacity?) produce results I don't like.

I'll make sure the graph is in decades and that we can all get the best looking "standard" graph. The one you posted mutant is not really standard at all

The extra curve I thought at first must just be from the spectrograph, but after reading your edit it seems the version of Xhip I posted is so close that it is hard to argue the asymptote is 130% as it should be in the OB-Xa.

So... the only thing I'm left with is to be almost certain that 1984 is not a recording of an OB-Xa, but a completely different synthesizer.

Mutant · Post by **Mutant** » Fri Apr 24, 2015 12:58 am

aciddose wrote:Wait, mutant did you use the original CD master of 1984 to produce the graph, or did you rip that off youtube?

Ripped from CD.
Which should be enough for peaks of the sweeps that go to about 16000-19000hz ?

I see higher notes go over the range (i believe ketrack was set to 100%), but i can see the full attack shape on lower octave notes.

aciddose · Post by **aciddose** » Fri Apr 24, 2015 1:09 am

Yep, that's as I thought. My ears aren't accurate enough to judge that by listening I guess, although I could clearly hear the peaks go over my range of hearing (about 18.5k max) and they seemed to go over 20k. I've got the scope, but I didn't bother to connect a PC to that electronics gear to record it.

I can't hear the frequency "hanging out" around the peak of the envelope very much but it is a bit hard as there are a lot of chords and I'm just trying to judge by ear. I suppose it may be.

The skew of the frequency graph (log I guess?) may have an effect on the "rounded" shape, but as I said if the 102% asymptote version I posted at first comes that close... then, well, the asymptote is about 102%...

So I suppose we're just forced to admit the recording is not a OB-Xa at all. It couldn't be.

To my ears it sounds like the key tracking is tuned fine, so unless they somehow modified the asymptote of the chip (impossible?) or as you say if the datasheet is incorrect (I'd also say improbable) we can't include a more linear frequency response to account for it.

Where did the idea that it was an OP-Xa come from? Is this just legend? Have I assumed this on my own? I thought that it seemed agreed upon.

Mutant · Post by **Mutant** » Fri Apr 24, 2015 1:17 am

aciddose wrote:What I'm seeing from most of the sounds is a sharp edge and often the cutoff is going above 20khz just as you'd expect. In these cases if we apply another filter like the steep filter at 18khz applied for mp3, we can actually see a line that looks rounded off quite a lot like mutant's graph... although not exactly the same.

Also I'm not sure exactly which "wow" that graph was plotted from, most of them are polyphonic with a lot more than two voices (usually three or four) and it sounds layered also.

Most go to release stage before hitting sustain.
The ones that go over 20Khz are the top octave notes.
You have to find a longer note from the lower octave to see how its attack goes into sustain.
100% sure it is just 1 base note + 1 note 1 octave up.
There is a delay effect active.
My graph was plotted from the first "wow", which doesn't show full attack stage.

aciddose wrote: So:

Are we certain the synthesizer is a OB-Xa?

Which synthesizers were used for the layering? The same OB-Xa or multiple different instruments?

Do we have a better (clean) source to analyze?

Are we sure the source used so far hasn't introduced anything weird into the graph?

1. If we can believe that RetroSound guy. My ears tell me that his synth comes very close.
2. I believe there was no layering of the high resonance sound, just 2 note chords.
3. Do we need it ? The attack shape is not a very complicated curve and we dont need to nail it 100%. IMHO

4. How can lossless audio ripped from a CD change frequencies of a high resonance filter sweep ?
Remember that the sound is not important, only the resonance curve that shows us how the envelope shape looks like.

aciddose wrote: I don't really feel comfortable with it, but if nobody has a CD I can try to record the vinyl. A bit of a pain in the arse though and vinyl almost definitely has a more limited frequency range than the digital rip from CD will.

Do we really need/want to do that ?

Is your goal to emulate that synth ?
If you include a slider that an user could change the envelope shape with, then the user will tune it to his needs and you wont have to go through the pain to emulate the envelope just to make this sound.
Linear attack shape is important for many good sounds too.

Mutant · Post by **Mutant** » Fri Apr 24, 2015 1:21 am

aciddose wrote:So I suppose we're just forced to admit the recording is not a OB-Xa at all. It couldn't be.

I don't know.

I wasn't there.
But:
http://www.google.com/search?q="OB-Xa"+1984+"van+halen"

Give me NONlinear ADSR !