Some new plugins to test...

[fritzman]

Hi Alex,

thanx a lot for your Chorus Sounds really good.


All the best, FRitz

[Jafo]

Awesome! CH-1 has an exquisite sound, particularly when set to more subtle settings. It works fine in Tracktion 1 and Aodix 3. It technically works in Audacity, but it silences all input to which it is applied... but such oddities are common in Audacity. Thanks!

[alex.g]

I've tested your new version. No problems at all in WL4.
I think you could put it online.
I sent you an e-mail with detailed results.

Thank you so much Peter,
I owe you a beer
The new version will be uploaded in minutes.

[PeterL]

I've tested your new version. No problems at all in WL4.
I think you could put it online.
I sent you an e-mail with detailed results.

Thank you so much Peter,
I owe you a beer
The new version will be uploaded in minutes.

[alex.g]

OK THE DOWNLOAD PAGE IS UPDATED
New versions work fine in wavelab.

Thank you all for your patience, Wavelab users!!!

Cheers,
Alex

[alex.g]

Higher FFT sizes and zoom levels can be useful for analyzing bass parts. 8192 hardly gives enough detail to see what's played in the low end of a 5-string bass, while at 32768 you can easily see what's going on.

Ok I forgot bass players!
Perhaps I should learn and see if I could use Wavelets. That should be the optimal solution.

Software that I use at work to analyze automotive sounds uses a variable FFT size to analyze different parts of the spectrum so that 'visually' the spectral lines look similar in width on a logarithmic scale.
So, bass end of the spectrum gets a 'zoom' fft with high number of points, while higher end of the spectrum uses less number of fft points in the windowing.
Is that possible with your code?

Another solution could be splitting the input signals into frequency bands with filters, and then analyzing at different resolutions...
Perhaps I could do a plain DFT for the frequency bins of interest only.

You could downsample the signal and use this lower resolution signal for the detailed bass FFT. This way you'd get a pretty high bass resolution with a small FFT size.

Sounds interesting, can you give me much info about this solution? a reference or something to start with

Anyway, can't wait for SG-2! Will there be a CH-2 as well? (with a GUI)

Yes, it will be! It should feature the "future development" list at the end of ch-1's page.
Anyway, I can't promise nothing about the release date, since I'm very busy this period... I'll do my best!

[alex.g]

Awesome! CH-1 has an exquisite sound, particularly when set to more subtle settings. It works fine in Tracktion 1 and Aodix 3. It technically works in Audacity, but it silences all input to which it is applied... but such oddities are common in Audacity. Thanks!

I'm happy you experimented with it! Have you tried the "Sci-Fi movie" preset with male voices? it sounds demoniac

Can you confirm that the Audacity thing happens with other plugins as well? It really sounds strange referred to ch-1, since it always passes dry signals through, simply copying inputs to ouputs (scaled with gain)...

[NeoVXR]

hi alex,
long time ago I built an analog chorus with a particular idea about the intonation issues. perhaps you like it
I constructed 2 independent delay lines.
there was 1 main LFO.
it produced a triangle, but I managed some exponential function of it, as input to the transport VCO control voltage. the purpose was, that the single delay line should have a fairly constant frequency shift, more like a harmonizer, surely with some modulation, but no howling.
the 2 delay VCOs got opposite signals. so one delay time was increasing, the other decreasing at the same moment, shifting tones down and up in parallel. (this cures the typical ugly detune e.g. of a strong flanger while preserving maximal effect depth.)
during the reverse of triangular wave direction, the delay lines change place in their f shift roles. this was a critical point in modulation results, but came out fairly pretty, with much luck and little math. it was audible but in a nice way, like a string orchestra when many bows change direction distributed during a short time interval.

then of course there was the direct signal that had an on/off switch.

the next nice feature was an additional sine wave LFO (also with a switch) that was running faster than the other, it produced kind of vibrato. I coupled this LFO into the 2 VCO controls, with ~90 degree phase shift. so the 2 voices had a vibrato that was kind of time shifted between each other. also the vibrato was modulated in intensity by the slower LFO, in opposite ways. (this is nice but trivial because the longer the delay is, the more influence it has)

there was abit EQ going on. the anti-aliasing cutoff was between 12-13 kHz. the direct signal had full range. the delayed signals also had a small boost in the "presence" tone range (maybe 5-8 k, can't remember) and a bit of phase shift because of an unexact preemphasis/deemphasis that gave the boost.

this effect unit was particularly wonderful for keyboards (esp.DX7) and guitars (fender style).
and it was a hell of construction work to get rid of the chirp and hiss noise hehe

in case you want to make use of these ideas one day...
I would be happy

keep on and thanks for the free plugs!!

[autloc]

Perhaps I should learn and see if I could use Wavelets. That should be the optimal solution.

If you're really motivated, it might be worth looking into Wigner-Ville distributions (DFT of the covariance function, seem to be pretty popular in time frequency analysis). I haven't look into using them seriously, and I think the CPU load might not really justify their use for practical real-time application, but it would be pretty awesome to have that in an analysis plugin. For a web site showing the high resolution you can achieve with them, check out http://www.ecf.caltech.edu/~case/wv/ .

[alex.g]

hi alex,
long time ago I built an analog chorus with a particular idea about the intonation issues. perhaps you like it
[...]

Hi NeoVXR,
how much work did you do to your chorus effect! it seems quite complex
You had some original ideas, in effect in ch-1 I tried with a saw wave LFO too, which should produce the effect you say (inverting triangles LFOs), but it sounded clippy to me...
Anyway, if I use some of your ideas, I'll tell you first
I don't mind equalizing the wet signal, I'd rather keep it as similar as possible to the dry one, in its frequency components...
The vibrato idea is nice. You are talking about amplitude modulation, aren't you? this could be nice.
Thanks for your post!

[alex.g]

If you're really motivated, it might be worth looking into Wigner-Ville distributions (DFT of the covariance function, seem to be pretty popular in time frequency analysis). I haven't look into using them seriously, and I think the CPU load might not really justify their use for practical real-time application, but it would be pretty awesome to have that in an analysis plugin. For a web site showing the high resolution you can achieve with them, check out http://www.ecf.caltech.edu/~case/wv/ .

I didn't know about this!
It's kind experimental but it seems interesting.
Thank you for your suggestion!
I wish I had 48hours a day, to be able to do everything...

[autloc]

I didn't know about this!
It's kind experimental but it seems interesting.
Thank you for your suggestion!
I wish I had 48hours a day, to be able to do everything...

Yeah, well I'd be keen to try it out if you get it working. Same goes for the 48 hour day!

[NeoVXR]

hi alex,
well, I was an electronic nerd, took a few weeks, full PCB and everything...
the triangle was inverted, then each VCO had got an exponential transfer characteristic (steeper side for shorter delay time), and also the edges of the triangle were smoothed.
the fast LFO was just added to the control voltage. no deliberate tremolo, but the wide sweep made for big phase cancellings over most part of the spectrum. it was a bit psychedelic, esp. when I added overall feedback. imagine the cross-feed effects of the (anti) parallel lines.
--
btw I noticed serious amplitude upbuilding in your implemetation (which is natural). I needed a limiter after the effect.
--
did you implement fractional sample granularity for the delay time sweep? should sound much rounder, particularly for small delay times.

[rockstar_not]

I would stay away from Wigner Ville transforms.

In automotive sound analysis, they were the rage about 10 years ago. Problem is that they would always leave visible (we used them for time/frequency sonograms) artifacts that looked like either broadband impulses or resonances (straight lines along either a time slice or frequency slice). Nobody was doing them real-time, either - certainly not fast like FFT.

Anyhow, none of the big technical sound analysis programs used to analyze noise seem to use them these days. (programs from Bruel & Kjaer, LMS, Head Acoustics, etc.)

-Scott

[NeoVXR]

addendum...

a ramp (sawtooth) is not useful.
it needs 2 delay lines anyway for a harmonizer-like effect, and some switching trick.

back to the control feed. because it was available from the semiconductor transfer curves (something similar to i = a * e ^ (u-0.6) ), I used the exponential function to approximate the required curve. it should have been a 1/x function, with always x>0 because a delay=0 is not physically possible. the deliberate functional identifiation error was perhaps significant to give it an "organic" sound. you might introduce an additional function instead of this, to prevent pitch shift from being exact and constant.

hardware uses fixed sample count with varying clock speed, software in plug-in context uses fixed clock (e.g. 44.1k) with varying length, that should be interpolated for continuous and soft sweep function.
so, a linear-triangular modulation shape for the sample length count per time should be ok, but I would put a saturation effect on the edges of the triangle to make them perfectly round. you want to avoid an exact rectangular output pitch of a shrinking/growing delay line. pitch shift should have a transitional ramp like:
<neg. constant> <ramp up> <constant> <ramp down>. ___/---\___/---
it is important that you derive the saturation start and stop points from the _duration_ of the desired ramp. (should be constant, to set somewhere between 30-300 msec, and make everything a bit round)
this duration should be independent of LFO speed. (I used a RC lowpass for the triangle, which naturally turned into an attenuated sinus for the highest LFO-freq that the control knob could set.)
without a good transitional ramp it will sound clicky.

If you really dig into this, I would recommend to provide for a very wide sweep range, where the control parameters are minimum delay (e.g. 5 ms) and maximum delay (e.g. 500 ms), and as usual the LFO freqencies (sweep and vibrato). sweep LFO should have an automatic mode, where effect intensity is maintained when the user changes delay ranges, this means LFO time period becomes a function of (max-min) delay time.
you might also provide for a negative delay range in the "min" settings (which was not possible in the hardware), where the transition through zero might yield a very "techno" effect. this works when direct signal is on. with the 2 delay lines having the same delay value at certain moments (one running forward, the other backward so they meet twice per interval) this will have 2 different kinds of "zero" points, one with the direct signal, one among the delays, which is interesting because you can use the HF roll-off for the delay lines that you already have in the LF path - it will have kind of a softer analog sound then.

LFO should have a sync option to the music beat, plus a start offset. there is definitely a "sweep" sound in this effect, that should work _with_ the music.
I can promise you it is exciting!

(you might provide two general modes, the original that you have, and the new one called sweep/string chorus, or just make another plugin from the same basic code, the main work has been done already and I think it is great)