[Q] export does not generate valid vst plugin

[tulamide]

I thought about it. You said that in Synthmaker the s&h mosule won't work. That's absolutely possible, since it uses highly optimized Assembler code. There might be instructions added way after your Synthmaker version. What about the dsp code version? Does it work?

[zucker1]

Sadly also the dsp code version does not work as supposed.
It won't be pitch preserving at all and sounds the same as using a simple interpoltaed delay with a dezipper

I talked whith lawo, but the don't care about sdk 2.4 atm.

i really need that... (i only have 1 more week to get the thing up and working)

it's a shame that the old SM forum has gone.
I hoped that an SM user could have an working pitch preserving delay....

now im begging for help

[zucker1]

What about the dsp code version? Does it work?

The dsp Version delivers on the xfade Output extreme values (1.233e+0007 and upcounting), delay1 always stays 0 and i have absolutly no idea why.

That's the dsp code

Code: Select all

streamin delay;
streamin freq;
streamout delay1;
streamout delay2;

float saw;

saw = saw + freq;
delay1 = delay1 + (delay - delay1)&(saw>0)&(saw<=freq);
delay2 = delay2 + (delay - delay2)&(saw>1);

saw = saw - 2&(saw>1);
xfade = abs(saw);


[tulamide]

The dsp Version delivers on the xfade Output extreme values (1.233e+0007 and upcounting), delay1 always stays 0 and i have absolutly no idea why.

That's the dsp code

Code: Select all

streamin delay;
streamin freq;
streamout delay1;
streamout delay2;

float saw;

saw = saw + freq;
delay1 = delay1 + (delay - delay1)&(saw>0)&(saw<=freq);
delay2 = delay2 + (delay - delay2)&(saw>1);

saw = saw - 2&(saw>1);
xfade = abs(saw);


Í don't know much about dsp code, but it doesn't look right to me. xfade isn't defined anywhere, and this looks wrong:

Code: Select all

(delay - delay1)&(saw>0)&(saw<=freq)

If you mean something like "if saw greater than zero and lower than or equal to freq", shouldn't it then be written like so?

Code: Select all

(delay - delay1)&((saw>0)&(saw<=freq))

But maybe someone with better knowledge can say something about this.

[martinvicanek]

@zucker1 pm me your email address and I'll send you the fsm file with the pitch preserving delay.

[zucker1]

Many thanks for all the help.
now it works pitch preserving (converting the Zipper Output to integer for the delay)

BUT one last Problem:

The delay value changes from 0samples to e.g. 3500samples and i'd need to make an smooth transition between these values (and back again).
so i take 0 or 3500 into an dezipper with transition time 1000ms.
but the output of the Dezipper jumps from 0 to about 200 and then smoothly increases until it reaches the endvalue of 3500. same when going back from 3500 to 0, the first step in an jump of about 100 to 200.

But i'd need a way to prevent These "large" step at the beginning of the transition.

is there a way to Change the behaviour of the dezipper or a "code-Version" of an dezipper that i could modify???

i found the slider module... i'm currently checking if this works for my need..but it seems so

MANY MANY thanks for all the help!!!!
YOU ARE AN GREAT COMMUNITY!!!

[zucker1]

using the slider instead of the Zipper has not worked for me well.

this is what i've put together right now.
it's far awaay from perfect and maybe you'd have a look (or an ear) at it and give me some hints how to do things better.
when chaning the delay with transition times smaller than 2.5seconds the Audio Signal jitters.
It will be needed to Change while someone is peaking.

i've added an gate so when noone is speaking, the transition gets 10 time faster.

thanks for any further Input.

[martinvicanek]

I need a way to prevent These "large" step at the beginning of the transition.
Is there a way to Change the behaviour of the dezipper or a "code-Version" of an dezipper that i could modify???

The dezipper will respond to a sudden change by linear ramping from the last value to the new value. If you have only positive values (=delay times) to consider and you want slower ramping for low values and faster ramping for high values, you could sandwitch the deziooer by a nonlinear function and its inverse, respectively.The screen shot below illustrates the idea fora log/exponential pair (both base 2):
yellow - input step function
green - plain dezipper
blue - sandwiched dezipper

nonlinearDezipper.png (30.26 KiB) Viewed 15455 times

If you are concerned about CPU load you can do two things:
1. use my optimized stream functions at Flowstone Guru
2. depending on your ASM skills - skip code if value does not change. KG's blog post at FS Guru tells you how to do that.

[adamszabo]

would you be able to post that schematic Martin?

[zucker1]

HI Martin,
due i'm stuck on the old synthmaker, there's no 2^x and no log2(x) module.
i've tried to create an code block with

Code: Select all

streamin xin;
streamin xout;
xout = log10(xin) / log10(2)

and

Code: Select all

streamin xin;
streamin xout;
xout = pow(2,xin)

but the Output of the 2^x block delivers creepy negative values...