Stomps/Reverb1 bug/issue

#1

Hey smile

Finally got Axoloti working. Nice tool smile

But I think I encountered a bug in the stomp/reverb1 object. When you turn the reverb object on, it seems like the feed of the reverb tilts to the right side. As soon as you turn it off the stereo field slowly goes back to normal. At higher decay the sound first tilts to the right side and the sound gradually fades out.. I feed the same signal to both sides so there shouldnt be any difference.

You can recreate it the simply by adding the reverb to a signal chain and turn it on. Or check the attached patches.

I build a synth and added the reverb and noticed the bug. Then I tried loading up the guitar multi fx patch and I added an osc, an ADSR, a vca, connect them and feed them into the reverb, to check if it was my first patch that was the issue. And as soon as i turned the reverb on, same thing happened again. The feed tilts to the right side.

I am on Mac, OSX 10.9.5, No HUB, Usb cable directly to computer
Axoloti version 1.0.3-0-gc248736

Verb SawOsc Buggy.axp (4.6 KB) Verb SineOsc Less Buggy.axp (4.5 KB)

#2

I tried the patch, reverb panning seems fine to me. Which firmware version are you using right now? Have you tried with others sources other than phasor? (which provides a unipolar waveform instead of a bipolar one)

#3

Ok. I will open computer and check again. Maybe make a small video to show the issue, since you cannot recreate it.

I will look for firmware. I did post the axoloti version number in first post.

#4

Oh, right, i did not see that.
Also, are you using headphone out or the main jack?

#5

I think you're seeing a weakness of the stomp/reverb1 algorithm. No reason to doubt about firmware, OS, cables...
The only solution is a better reverb object, that will arrive over time. There are many different approaches to reverb implementation, each with their own color and character. The stomp/reverb1 was a first quick shot aimed to add reverb to the object library, and certainly not perfect.

#6

yeah, I had a quick look at it last night, and can be see with sine waves
the offset didn't seem to be coming from the fqn4 object, but I guess that is still the most 'likely' source, as the delays lines are mono, and the mixing is pretty simple.

@jaffasplaffa, the stomps can be edited, by using 'edit object definition' as they are sub patches, you are free to go in there, and see if you can find the cause, or at least narrow it down.
(Id always encourage community involvement, in some of this leg work... many hands make light work etc)

#7

I will do that. I think I have an idea what it might be smile

#8

And I think it is very good that you involve us. To me the goal is to be able to patch/program everything on my own and to learn that, only way is to get my hands dirty smile And listen closely to people in this community smile

I did open up the subpatch and I think I can see some issues.
First one: The left and right stereo inputs of the sub patch are feed into the same mono mixer to create the feedback network. There is no stereo signal chain in the feedback path and that might has something to do why they sound different on the outputs. I will check signal paths later. and also try and compile a new reverb, with full stereo signal path. But Now I promised the kids to go outside. In Denmark a non rainy day is a good day smile

#9

I think thats part of the 'design'... you will see that although the delay lines are mono, the fdn output is not, i.e. it is creating a stereo field which is mixed in with the original stereo....

and of course thats part of the fun on reverbs, how you mess with the stereo field and delays to introduce space, you can do also sorts of funky things messing with this, and create your own effects.
I nearly started on topic on this, because this is one area Id love to see the community get stuck into, Id quite happily dedicate one board, to just having a stack of reverb algos on an SDcard. (johannes granular verb is my favourite at the moment)

one small question I had for @Johannes is, how much headroom is there over +/- 1 in an audio signal? as this patch adds 2 signals together (so could go to +2/-2) and the does a div 2... is there 1 extra bit to cope with this, or will it overflow/saturate? if so the div2 should come first, but that would reduce precision by 1bit...
so whats the range of a audio signal is 1 already 32bit (signed) or only 31 bit?

#10

Ahh ok smile Learning everyday.

But still there is something, atleast on my system that, that makes the reverb do something weird, especially when the phasor osc is send into the reverb . Check the video I posted. It shows a phasor oscillator and the cheap saw feed into the reverb with a low pass filter. There is huge difference between the output of the reverb, depending on which oscillator is feed into the reverb. Sputnki it might thave something to do with the phasor osc being feed into the delay anyway wink. Maybe some phase canceling going on in the feedback network? It is almost the reverb first feeds alot and then phase cancels and totally dissapears..... until i turn down the phasor osc. Watch the RME mixer inputs on the right how the input reacts, when I change from cheap saw to phasor osc.

I have also been trying to build my own reverb in Nord G2 cause the one that is there could be better. But for this I am not really using math and algorithms, just my ears.

Yeah I also got a couple of them, so I could use one for effects and one for synth smile

Video(maybe not the best, but can you see what is going on):
https://vid.me/BT5J

For now, I will just use another osc than the phase oscillator.