SQ80 emulating VSTi available

[djsubject]

thanks for sharing this gem!!! & to r_buchty (i don't know who did what?)

Kudos for programming all go to Sigi. My part (since 1997 ) was the reverse-engineering of the machine.

well Thanks to both of you

http://www.skirize.adbe.org/SQ8.mp3

[r_buchty]

The reason for that is fairly simple. Even though you can emulate an old hardware unit fairly well (at least on the digital basis - not analogue cause you'll never get there - circruity can't be turned into bits and bytes) there's always another factor: the input/output converters.

No, that's not true. After all, *any* analog equipment has a transfer function which you can derive and implement. However, getting this transfer function right, is incredibly hard work and will also eat up precious CPU time.

So usually you will not do the real model but an approximation which will be most suitable/tolerable.

And since this plugin started as a semester project on requirements of digital audio algorithms with respect to processor architecture, we decided to go for some average digital filter instead of an exact emulation. Developing a proper CEM3379 model would easily be an own project and was also out of the scope of this very task.

For those interested, there are some nice papers out by fellow SQ80 list member Antti Huovilainen describing how to model typical filter circuitries and how e.g. a Moog Cascade differs from a OTA-style filter.

[tomg]

Now I do
Sigi told me about the SQ8L thread here, and I got curious.
And yes, it is that very project.
Rainer

Hi Rainer, Welcome to KVR.

[3*s]

crashes sonar and ext after pressing a couple keys

edit: nm, it just doesn't like 20 ms latency.


Sounds great.

[tomg]

Hi, I'm Siegfried Kullmann, the developer of the SQ8L.
Sorry for the confusion about the installing procedure; just wasn't sure the installer would work, so I described the manual steps.

Toshu, what's your CPU? When does the crackling occur?
Perhaps, when moving the mouse over the GUI, or really all the time, even when you do nothing?

Hope you all fun with it!

Hello, Siegfried. Thanks so much. It's terrific!

Tom

[Compyfox]

You mean Antii from Smartelectronix who coded the Taurus 1?

Well okay, analog circruity can be portet, but it's way more work than with digital synths cause you need the nouances of changing resistors/capacitors, etc. A lot of people complained about my list that it's not accourate and I had to defend myself with several examples (for example Nubi which is a Hammond Emulation, but seing as every hammond sounds different right from the start - then you can't nail it with an emulation, same with Moogs).

I really appreciate your work - thanks to both of you. It's really one of the clostest emulations around. And offering it to us for free in the first place is a very nice move and really appreciated.


*Und? Könnt ihr zwei vielleicht schon mehr über den SQ8X sagen? *

[tomg]

crashes sonar and ext after pressing a couple keys

edit: nm, it just doesn't like 20 ms latency.

Sounds great.

I'm running at 5ms here. eXT and Sonar are ok.

[tomg]

No, that's not true. After all, *any* analog equipment has a transfer function which you can derive and implement. However, getting this transfer function right, is incredibly hard work and will also eat up precious CPU time.

So usually you will not do the real model but an approximation which will be most suitable/tolerable..

It may be theoretically possible but the off the shelf PC stuff is a little too slow to accuratly emulate transistor switching through a DAC in real time.

Are you saying that you know of someone that has actually done 1:1 on stock PCs, using VST technology, in real time or just that in theory it's possible? I have yet to see this result out of anything on my scope, so I'm interested.

[tomg]

No, that's not true. After all, *any* analog equipment has a transfer function which you can derive and implement. However, getting this transfer function right, is incredibly hard work and will also eat up precious CPU time.

So usually you will not do the real model but an approximation which will be most suitable/tolerable..

It may be theoretically possible but the off the shelf PC stuff is a little too slow to accuratly emulate transistor switching through a DAC in real time.

Are you saying that you know of someone that has actually done 1:1 on stock PCs, using VST technology, in real time or just that in theory it's possible? I have yet to see this result out of anything on my scope, so I'm interested.

Because I want to be fair, I'll give you 100ms to get it spot on. Thrill me!

EDIT: This is a trap.

[mistertoast]

tomg, PCs will get faster. It's a matter of time. More than the CPU speed, though, I still think it's a trade-off of closeness vs. programmer time. Taking a day or two to write a couple K of code for a filter that's 90% there is more likely than taking a year to write 100K of filter code that's 99.9% there.

I'm happy with most emulations in a mix. Less happy (but still happy) with some of them on their own.

[1-2-Many]

Welcome to KVR Rainer & Siegfried!

Thank you just doesn't cover it. Like I said, I've been aware of this project for over a year now, and was starting to wonder if it would ever see the light of day. . . This really makes me happy! An ESQ-1 was my first synth (OK, my first synth with MIDI, my true first was a Juno-60). I sold the ESQ in the mindset up updating my equipment, and always regretted it. I bought an SQ-80 off ebay about 6 years ago. I've always loved this synth.

I guess I can kick my SQ80 VST project to the curb now Siegfried, because you really nailed it! Great job, and thanks for all the time you've put into this.

And Thank You Rainer for all your help, and also for keeping up the great ESQ website resource all these years - I've gotta say it, to anyone who may not know already - this man is a genius.

[bengeorge]

Love this synth. Comments:

1. CC map in eXT doesn't list any parameters.. Any one else having this problem?

2. I agree that the GUI could be much better, you don't have to be limited by a two line readout.

Thanks again for a great synth.

[tomg]

tomg, PCs will get faster. It's a matter of time. More than the CPU speed, though, I still think it's a trade-off of closeness vs. programmer time. Taking a day or two to write a couple K of code for a filter that's 90% there is more likely than taking a year to write 100K of filter code that's 99.9% there.

I'm happy with most emulations. Less happy (but still happy) with some of them on their own.

Transistors switch at near light speed. This switching forms the leading edge of oscillator saw and square waveforms. If the waveform isn't correct it won't matter what follows.

PCs are going to have to get a lot faster and we're not even talking about aliasing yet.

Don't get me wrong. I'm a VST fanboy but there's theory and then there's reality.

[r_buchty]

It may be theoretically possible but the off the shelf PC stuff is a little too slow to accuratly emulate transistor switching through a DAC in real time.

Well, what you're describing here is the brute-force approach like building a synthesizer model using SPICE and then trying to run SPICE as your VST plugin. Little too slow would be just plain understatement

But: If you run a thorough analysis you will be able to either directly get the overall transfer function or at least develop an approximation function which comes as close to the desired result as necessary. Therefore you don't need the brute-force approach.

So instead of modeling e.g. a filter circuit based on single transistor models, you will rather determine the filter's true characteristics and just use the filter model instead.

Are you saying that you know of someone that has actually done 1:1 on stock PCs, using VST technology, in real time or just that in theory it's possible?

The latter. I was just objecting your point that authentic digital emulation of analog circuitry is impossible. It's not. You *can* capture the analog spirit, but you also will have to pay a price -- either by doing brute-force calculations on lowest level, or determining the effective transfer function.

Besides: would have anyone thought about convolution reverbs being practical for home use a couple of years ago? So I wouldn't object the idea that the brute-force attempt, i.e. doing a full analog simulation rather than just an emulation, might become feasible in a couple of years.

[1-2-Many]

Hope you don't mind Rainer, just wanted to share some more info on you.

http://itec.uka.de/~buchty/index.php

I know you're to modest to toot your own horn - so I'll do it for you