delay loop artifacts

Tom

I perform with delay loops on my voice and ebow electric guitar. I now have a G2X (os 1.40). I would like to control my entire sound with the G2X, using one slot to loop the guitar, one slot to loop my voice, one slot to generate a drone, and one slot for lead synth. When I created a simple patch to capture and loop my guitar continuously, I found that some background noise built up to an audible level in just a few minutes. I wondered if the G2X was capable of performing looping functions, so I googled the net and found Rob Hordijk's tutorial. Rob states the following;

"But by using digital memory the audio can loop forever without any degradation. In the preparation for this article a two second loop on the G2 was made to recirculate for two days, and still sounded as fresh and crisp as when it was captured."

The loop I made degenerated rapidly. To invistigate this I created a patch with 3 parallel loops fed by an oscillator instead of audio in. Loop 1 is nearly identical to the one in Rob's tutorial. All three loops produce audible artifacts. When the feedback level is set below 100%, the original sound decays to silence but the artifacts remain audible indefinitely.

The source signal is designed to produce a sine wave pulse lasting approximately half the duration of the loop. One interesting thing is that the artifacts overlap the sine wave only. The silent half of the loop remains silent.

Am I doing something wrong?
Is my synth malfunctioning?
Can anyone create clean loops that don't degenerate?
Does my patch produce noise on anyone's synth that is also capable of producing clean loops?

My patch is attached.

Thanks,
Tom

Delay artifacts.pch2

Description:

Three parallel delay loops that produce audible artifacts, even after the original sound has decayed to silence.

Download (listen)

Filename:

Delay artifacts.pch2

Filesize:

3.63 KB

Downloaded:

994 Time(s)

blue hell · Posted: Sun Oct 08, 2006 4:33 pm Post subject:

The patch seems to be defective, I can't load it (and the header looks funny when I open it in a text editor), don't know what went wrong, could you maybe try to repost it Tom ?

Tom · Posted: Sun Oct 08, 2006 10:03 pm Post subject:

I am reposting the patch. I downloaded the patch from my first post and it works fine on my machine. This patch was made on a Mac with the NMG2Editor VERSION 1.40 BUILD 266. I don't know if there is a Mac/PC issue. Let me know if it still doesn't work.

Also, I forgot to mention that this patch works best if you play just one note. Playing more than one note sometimes produces a pop. I didn't bother to fix this since the patch is intended to demonstrate an effect, not to be played musically.

Tom

Delay artifacts.pch2

Description:

Three parallel delay loops that produce audible artifacts, even after the original sound has decayed to silence.

Download (listen)

Filename:

Delay artifacts.pch2

Filesize:

3.63 KB

Downloaded:

962 Time(s)

ian-s · Posted: Mon Oct 09, 2006 12:29 am Post subject:

"Can't load file, file format is not supported!"

Never seen that message before. Question

sebber · Posted: Mon Oct 09, 2006 2:05 am Post subject:

Same thing here.

G2DREAM · Posted: Mon Oct 09, 2006 3:48 am Post subject:

Me too!
_____________
cyber-evolution

gurk · Posted: Mon Oct 09, 2006 1:45 pm Post subject:

here is a fixed version of the file, so it now opens...

had some kind of header added to the file, which i deleted..

delay_artifacts_122.pch2

Description:

Download (listen)

Filename:

delay_artifacts_122.pch2

Filesize:

3.5 KB

Downloaded:

1042 Time(s)

gurk · Posted: Mon Oct 09, 2006 2:15 pm Post subject:

I hear artifacts in the sound also..

They appear to increase in amplitude when the feedback decays right out..

Quantisation error maybe? Something to do with the 16-bit precision of the delays possibly?

Tom · Posted: Tue Oct 10, 2006 12:37 am Post subject:

Thanks for making an readable patch. Mac osx has file permissions. Maybe it is only allowing my computer to open the file. Eventually I will need to figure this out so I can post readable patches.

Fozzie · Posted: Tue Oct 10, 2006 2:30 am Post subject:

Can't check or test myself right now, but could it be (though it shouldn't) that the delaytime settings have anything to do with this? I think the looper patches that have been posted so far mostly use the Clocksync delaytime settings. I use Tim's magnificent looperpatch a lot, and have not noticed artifacts (the source I use is a noisy guitarist, so yes, that could play a role Laughing

; lot's of distortions Rolling Eyes

)

Tim Kleinert · Posted: Tue Oct 10, 2006 3:51 am Post subject:

Let me join in briefly...

When developing the DeluxeLooper patch, I let it run for hours to check the longterm stability against time drift. When I came back to check, the loop was still in sync, but sounded a bit degraded. So I guess the same holds here too.

I'm pretty convinced that it has to do with the 16 bit delay memory. That old devil. Converting the 24 bit audio signal down to 16 bits and back once or twice doesn't do that much harm. But within a feedback loop, the subtle harmonic distortions resulting from the conversion will build up. That's the way it sounds to me at least, when listening to the demonstrational patch.

The 16 bit delay memory is the most severe design flaw of this synthesizer.

There are also other artefacts connected to the delays, which come from inacurracies of the memory readout pointers -delay lines occasionaly grabbing samples from other delay lines allocated to the same memory chip. I documented those in the "bugs" section. Those could be easily fixed by a software update -if there ever was one... Rolling Eyes

cebec · Posted: Tue Oct 10, 2006 7:31 am Post subject:

Did we ever check to see if the degradation occurs on the G2 Demo? It shouldn't, should it, since the Demo uses system RAM?

Tim Kleinert · Posted: Tue Oct 10, 2006 10:59 am Post subject:

blue hell · Posted: Tue Oct 10, 2006 2:23 pm Post subject:

The demo version is said to basically be the same code as the hardware version and to be running in a DSP emulation environment (which makes it slow). Nevetheless there are some minor differences in behaviour.

Tom · Posted: Wed Oct 11, 2006 1:48 am Post subject:

Now that I know this is not a problem with my machine, but a problem in the code, I decided to take some noise floor measurements. If we call the maximum output before clipping 0db, then the output of a sine wave oscillator is down 2 bits at -12db. A delay module with internal feedback, such as Delay A, decays to silence without producing any lingering artifacts whenever the feedback parameter is set at or below 50 (64). At a setting of 50.8 (65) after the source signal has decayed to silence the artifacts are at -78 db. When compared to the original signal at -12 db this yields a s/n ratio of 66 db. Keeping in mind that these artifacts persist indefinitely, any patch using a delay like this can have a s/n ratio as low as 66 db, roughly the same sound quality as the 12 bit samplers produced by companies like ensoniq back in the 1980's. At higher settings it gets worse, up to 16 db worse, yielding a s/n ratio of only 50 db. When the feedback loop is set to unity, I can't measure the noise floor because the source signal is still present, but by listening I guess that the artifacts are only about 20 db quieter than the source signal. So much for 24 bit audio. Even if the delays are 16 bit, properly coded feedback will produce artifacts at -93 db. Delay modules that do not include internal feedback, and have feedback externally applied perform about 4 db better than delay modules with internal feedback.

In general I find the sound quality of the g2 to be ok, but I have not seen any looping patches made by anyone else. Since looping is my favorite performance technique, this is the natural place for me to start programming my own patches. What a surprise. Confused

Tom

Fozzie · Posted: Wed Oct 11, 2006 2:52 am Post subject:

Tom · Posted: Wed Oct 11, 2006 6:18 pm Post subject:

Tim Kleinert · Posted: Thu Oct 12, 2006 12:45 am Post subject:

Tom · Posted: Thu Oct 12, 2006 1:41 pm Post subject:

ian-s · Posted: Thu Oct 12, 2006 4:14 pm Post subject:

A delay module does not seem to be very difficult, a block of memory addressed by a pair of pointers. Data goes in at the head and is taken out at the tail. For a fixed delay with unity gain feedback, it would be hard to imagine how distortion could occur out side the initial bit reduction.
The G2 delay times can all be altered at run time, even the ones with no modulation inputs. This requires some form of interpolation and most likely this is where the problem is. I suppose a future software upgrade might use an improved interpolation technique, or offer a mode which does not use interpolation at all for a small selection of fixed delay times.

The poor delay implementation also ruins some non loop related patches. I had a feed forward comb filter where the ‘sweet spot’ got hammered by some really nasty grit.

Are the delay effects in other Clavia keyboards any better?

dorremifasol · Posted: Fri Oct 13, 2006 1:52 am Post subject:

Though I haven't had my G2 for much time, I have definitely heard those delay clicks and noises. I haven't had enough time to check all the patches but I have heard them in some organ leslie effects, and a great nl3 unison patch.

This is the biggest bug I have found yet, and I sincerely don't understand how can be that it hasn't be fixed by now...

Cheers,
Albert

Antimon · Posted: Fri Oct 13, 2006 5:25 am Post subject:

3phase · Posted: Fri Oct 13, 2006 7:56 pm Post subject:

they should at least remove the zippery noises on integrated delay modules...
you allways can have that on the basic delay module..
now its the other way around..i found the integrated delay modules pretty redundant because i like to generate my zippery noises myself and have clean delaytime sweeps available.

i agree that the delay modules are underdeveloped yet.
I dont know what clavia is up to...but i really hope that they will do a bugfix at one point...