PG8X (inspired by the JX8P): new beta version uploaded

[martin_l]

But the new version sounds much better, much less aliasing than default Synthedit oscillators!

That has nothing to do with SynthEdit, though. The old SynthEdit version did not use the standard SE oscillators, and had a custom antialias filter.

In fact, the new version uses the same code for the oscillators, and I only changed the code for the anti alias filters in the current version.

So it's not the oscillators that sound better but the filters? Anyway, I hear slightly less phasing than in other Synthedit synths!

To be precise, even the (VCF) filter code is the same. The only big change in the audio stream is the anti-alias filter before the downsampling stage. In the old code (including the SE version) it was a 8 pole SVF and now it is a 16 pole Butterworth filter.

I changed the mapping of parameters to filter frequencies and the resonance slightly, not much more. As said before, in the Synthedit version, the oscillators and filter (including anti alias filter) were hand coded custom modules. In the new C++ version I am reusing most of the original code, only wrapped into a different framework.


Cheers,
Martin.

[martin_l]

For me (reaper on Win7, Intel i7 2600) it takes a few percent on one core.

Yes it takes few percent when playing.
The problem starts when the synth is not producing any sound.
REAPER, Win 7 too.

What is your processor?

[Mutant]

For me (reaper on Win7, Intel i7 2600) it takes a few percent on one core.

Yes it takes few percent when playing.
The problem starts when the synth is not producing any sound.
REAPER, Win 7 too.

What is your processor?

Intel Core2 4300 on Win 7 64 Home premium and Athlon 64 X2 5200 on Win XP.
Long time passed since i last encountered that denormal fault in any synth.

[DrewDale]

Getting better all the time, cheers Martin.
I'm thinking of creating a bank of 32 patches to upload here in the next few days, will there be any internal changes to the sound architecture before the first non beta or any more parameter changes ?
Or to put it another way:
Is it worth holding on for the non beta before making any patches ?

[martin_l]

For me (reaper on Win7, Intel i7 2600) it takes a few percent on one core.

Yes it takes few percent when playing.
The problem starts when the synth is not producing any sound.
REAPER, Win 7 too.

What is your processor?

Intel Core2 4300 on Win 7 64 Home premium and Athlon 64 X2 5200 on Win XP.
Long time passed since i last encountered that denormal fault in any synth.

I will look into that. I don't see it on my machine. Most parts are written with SSE intrinsics, of which I thought they would not have the problem.

Can anybody else confirm the problem?

[martin_l]

Getting better all the time, cheers Martin.
I'm thinking of creating a bank of 32 patches to upload here in the next few days, will there be any internal changes to the sound architecture before the first non beta or any more parameter changes ?
Or to put it another way:
Is it worth holding on for the non beta before making any patches ?

I do not intend to change the format, but I am not yet entirely sure whether I overlooked something.

On the other hand, if you export the bank as Sysex, they should definitely be compatible with any new version, as they should always be compatible with the hardware.

Cheers,
Martin

[albertodream]

Can anybody else confirm the problem?

Yes, I have the same problem.
I have a very old machine: 1 core Intel Pentium 4, 2,8 GHz, Win XP, Reaper 32 bits, and the same with SaviHost. (I think my processor supports SSE extensions: x86 Family 15 Model 2)
When playing it's ok, but when iddle, consumes all the cpu.
The previous version don't have this problem.

[layzer]

on my musicroom xp x32 system that has a spidif connection that forces the daw to 48khz, loading up PG8X causes the output on other
vst's to crackle. .....WEIRD! on my xp x32 HP Elitebook it seems fine at 41khz. so there may be a issue with running it a 48khz sample rate.

edit: confirmed... just switched EXT2 on the notebook to 48khz and it did the same thing as the musicroom PC.

[Tale]

I will look into that. I don't see it on my machine. Most parts are written with SSE intrinsics, of which I thought they would not have the problem.

Generally whenever you use a IIR filter you might run into denormals. You could however tell the CPU to automatically zero denormals. Another way to prevent them (assuming WDL/IPlug) is to #include WDL/denormal.h, and use one of its filter or fix functions. Or you could add a tiny bit of DC, so your filter outputs move towards the DC value rather than zero.

Note that some hosts will automatically reduce denormals (REAPER e.g. has an option for this), for testing purposes you won't want that.

[squidlr]

Probably it's already mentioned in this huge thread, didn't have time to read thru - PG-8X forgets loaded presets once you save a project including PG-8X in Logic Pro X, so you have to load the presets back each time you open project in Logic.

[martin_l]

Probably it's already mentioned in this huge thread, didn't have time to read thru - PG-8X forgets loaded presets once you save a project including PG-8X in Logic Pro X, so you have to load the presets back each time you open project in Logic.

I assume you are talking about the AU..?
I don't have Logic, I will try to test later with GarageBand, which hopefully will behave similarly.

Cheers,
Martin

[DrewDale]

32 patches here http://www.kvraudio.com/product/pg-8x-b ... /downloads in syx & PG-8X file format

The seems to be a problem when naming and saving patches in syx format, the last few letters of the patch name is missing and the LFO Bend Range slider sets itself to maximum, this doesn't happen when saving and naming patches in PG-8X format and also the value I set in the LFO Bend Range remains as how I set it. Just thought I would let you know.

[martin_l]

32 patches here http://www.kvraudio.com/product/pg-8x-b ... /downloads in syx & PG-8X file format

The seems to be a problem when naming and saving patches in syx format, the last few letters of the patch name is missing and the LFO Bend Range slider sets itself to maximum, this doesn't happen when saving and naming patches in PG-8X format and also the value I set in the LFO Bend Range remains as how I set it. Just thought I would let you know.

Yes, this is a limitation of the JX8P Sysex format. In JX8P speak, the LFO bend range is not part of the TONE but part of the PATCH. A Patch consists of a TONE number (without the actual data) and some other parameters, such as Polyphony Mode, Aftertouch settings, and the LFO bend range. Also the length of preset names are shorter in the hardware.

In the native pg8x formats, all information are stored.

I should add that I am not intending to change the native format, but I also cannot guarantee at the moment, that any change could be necessary for whatever reason, until I release the official first release.

Chunks from the previous betas will not be read correctly.

Cheers,
Martin

[DrewDale]

Yes, this is a limitation of the JX8P Sysex format. In JX8P speak, the LFO bend range is not part of the TONE but part of the PATCH. A Patch consists of a TONE number (without the actual data) and some other parameters, such as Polyphony Mode, Aftertouch settings, and the LFO bend range. Also the length of preset names are shorter in the hardware.

In the native pg8x formats, all information are stored.

Cheers,
Martin

Thanks for clearing that up Martin, many thanks

[squidlr]

Probably it's already mentioned in this huge thread, didn't have time to read thru - PG-8X forgets loaded presets once you save a project including PG-8X in Logic Pro X, so you have to load the presets back each time you open project in Logic.

I assume you are talking about the AU..?
I don't have Logic, I will try to test later with GarageBand, which hopefully will behave similarly.

Cheers,
Martin

Yes, the AU version - seems like the above problem is apparent in Garageband too.