Nord Micro, is this 'aliasing' noise supposed to happen?

masstronaut · Joined: Nov 20, 2009 Posts: 6 Location: lahndahn

Hey Nord kids, I wonder if someone could check this for me...

Here's a brief sound example. It's just a low sine osc rising in volume. What you'll notice is a higher tone the accompanies it as it reaches a certain level. It sounds like aliasing or PWM or something.

Audio - http://www.mediafire.com/?z0nihtknkyo

The patch - http://www.mediafire.com/?nnwyo52mdw1

Thanks a lot if you can help with this.

ark · Posted: Fri Nov 20, 2009 7:19 pm Post subject:

I don't have an NM1, but if you post a screenshot of the patch, and it's not too complicated, I'll try to put together an equivalent G2 patch and post audio from it.

ark · Posted: Fri Nov 20, 2009 7:32 pm Post subject:

By the way, here's an image that results from applying Ableton Live's spectrum analyzer to your audio. You'll see that there appear to be various high partials, presumably distortion components, but they're all at least 60 dB down from the main signal.

Spectrum.jpg

Description:

Spectrum analysis of nordo.wav

Filesize:

143.98 KB

Viewed:

209 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

blue hell · Posted: Fri Nov 20, 2009 7:43 pm Post subject:

Hmm .. only when I make it clip somewhere it does make the high tone, I hear nothing odd on the mp3 either ... again unless I make it clip.

oh .. and welcome

of course!

masstronaut · Joined: Nov 20, 2009 Posts: 6 Location: lahndahn

Hello and thanks.

I don't think there's any clipping. The levels in the patch should be OK and I've normalised the audio to -1db. It's actually more pronounced at low volumes, as if it's digitising noise from a low bit depth.

I can hear it on the headphone out and the main out from the MM.

Here's the revised patch, I've made it as simple as possible just to demonstrate the effect. You can see the level on the mixer is right down. The x-fade is just to automate the level of the oscillator.

Patch - http://www.mediafire.com/?zcwguttylio

The audio is a .wav so it's not an mp3 artefact.

New audio - http://www.mediafire.com/?goymkmqzy1d

Recorded at a lower level, hence the hiss, but you can plainly hear what sounds like a square wave with PWM.

Thanks again.

nordio2.JPG

Description:

Filesize:

30.64 KB

Viewed:

9500 Time(s)

blue hell · Posted: Sat Nov 21, 2009 8:20 am Post subject:

Not sure what happened here ... I was replying Ark's post in which he said there was clipping ... but the post is gone ... I hope I did not delete it by accident?

ark · Posted: Sat Nov 21, 2009 9:00 am Post subject:

blue hell · Posted: Sat Nov 21, 2009 10:36 am Post subject:

Ok, was about to point that out and then the post went away Wink

anyway, I see nothing odd on either the recording or the patch ... so either my ears fail me or there is clipping in an external device.

masstronaut · Joined: Nov 20, 2009 Posts: 6 Location: lahndahn

There's definitely no clipping happening externally, this happens at quite a low level. The patch is very simple intentionally so there's no confusion as to where the distortion is coming from. The reason I have the level rising is that the effect 'kicks in' at a certain point. It's very apparent on more complex patches and is actually ruining the sound of it for me, it's not a minor thing - I'm quite sure there's something wrong with the outputs now.

Did you listen to the second clip Blue Hell? It's unmistakable there, bear in mind this is supposed to be a low sine tone, the 'Nintendo' effect is not intended. And although that sounds kind of interesting, in a more complex patch it just sounds like scratchy noise. Don't worry about the hiss - that's just because it was quiet.

Thanks chaps. I'd just like to know what the cause could be - a faulty component I guess, but which one?

Kassen · Posted: Sat Nov 21, 2009 11:33 am Post subject:

Blue signals act at a 4th of the sample-rate of red ones. When you use them for am (which this essentially is) you'll get stepping at a 4th of the sample rate in your audio signal which is indeed unpleasant to sensitive ears.

In some cases a LPF over the modulation source is a good solution but structurally speaking I think you are better off using red osc's pitched way down for am. If sound quality is dear to you, and it sounds like it is, it pays off to make sure blue signals *never* touch the actual audio without smoothing in place. Often a first order lpf will service to push the errors below the audible, especially in the context of a mix.

This is a especially delicate issue with envelopes, I find, as it interferes with snappy attacks. A variable filter can be used there; I find the exact spectral content during extremely fast attacks doesn't matter so much. However, when you take the filter down again beware that the modulation source for that should likely be smoothed as well. Overall it's quite annoying, and compensating for this can easily take up a quarter or more of your patch. It's worth it though.

masstronaut · Joined: Nov 20, 2009 Posts: 6 Location: lahndahn

It's not that, I wouldn't mind if it was. The duty cycle of that LFO is around 5 seconds or something - it's just there to provide a slow rising and falling level for demonstration purposes, not audio rate AM. I should clarify, this is not an issue with a particular patch, that patch is just a test.

The same thing happens with an osc going straight to an output, or into a mixer module first. It goes away if I turn the physical main volume knob down A LOT, but kicks in again when it's turned up - so it's related to the overall output level.

Thanks for looking at it.

blue hell · Posted: Sat Nov 21, 2009 12:00 pm Post subject:

Kassen · Posted: Sat Nov 21, 2009 12:08 pm Post subject:

Kassen · Posted: Sat Nov 21, 2009 12:11 pm Post subject:

masstronaut · Joined: Nov 20, 2009 Posts: 6 Location: lahndahn

masstronaut · Joined: Nov 20, 2009 Posts: 6 Location: lahndahn

blue hell · Posted: Sat Nov 21, 2009 12:34 pm Post subject:

Kassen · Posted: Sat Nov 21, 2009 12:38 pm Post subject:

I suspect a broken DAC. The effect reminds me of sounds I got emulating bad DACs in a otherwise anti-aliased system. This gives harmonic (for practical purposes) overtones as the effect will repeat with the cycle of the signal.

Looking at the actual wave confirms this. For the positive half of the cycle there are discontinuities on both the way up and down at the same amplitude. cycles that stay below that level look like decent enough sines, give or take some recording hiss.

Kassen · Posted: Sat Nov 21, 2009 12:45 pm Post subject:

blue hell · Posted: Sat Nov 21, 2009 12:56 pm Post subject:

Kassen · Posted: Sat Nov 21, 2009 1:10 pm Post subject:

Oh, yes, the harmonics were no doubt quite similar as it's basically a periodic higher-order discontinuity. When Masstronaut said it reminded him of PWM that was quite understandable as well, if you look at the place of the artefact and how it only affects one half of the cycle.

It's the DAC though, I bet.

That's bad news, I imagine paying somebody to solder in a new chip would cost more than a new MM.

blue hell · Posted: Sat Nov 21, 2009 1:21 pm Post subject:

ark · Posted: Sat Nov 21, 2009 2:54 pm Post subject:

Aha! Here's the problem.

If you look at the nordio2 clip, you'll see that the overall envelope has a weird plateau in it, and once the level passes that threshold, there's clear defect in the waveform. Look at the attachments for more detail.

I don't seem to be able to add more than five attachments; so I'll put the last one in a separate message.

nordio2 envelope.jpg

Description:

The overall envelope (i.e. amplitude vs. time) for the second sample has a strange shelf in the positive half.

Filesize:

101.74 KB

Viewed:

172 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

nordio high-level detail.jpg

Description:

The waveform for the first signal looks normal in the high-level portion too.

Filesize:

123.28 KB

Viewed:

176 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

nordio low-level detail.jpg

Description:

The first signal's waveform looks normal in the low-level portion.

Filesize:

114.15 KB

Viewed:

164 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

nordio2 high-level detail.jpg

Description:

Looking at a small portion of the waveform when it's loud, we see that there are artifacts of some kind in the positive part of each cycle.

Filesize:

105.62 KB

Viewed:

165 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

nordio2 low-level detail.jpg

Description:

Looking at a low-level part of the second signal, the waveform appears normal.

Filesize:

136.03 KB

Viewed:

184 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

ark · Posted: Sat Nov 21, 2009 2:56 pm Post subject:

As promised, here's the missing attachment. It shows the overall envelope for the first sample.

nordio envelope.jpg

Description:

The first signal has the kind of envelope you'd expect: It rises linearly until the fader has reached its limit, stays there for a while, then falls linearly back to zero.

Filesize:

103.1 KB

Viewed:

171 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

Oli · Posted: Mon Nov 30, 2009 7:27 am Post subject:

Sorry about the issue, but it is an interesting read.