Old 03-25-2020, 05:09 PM   #1
phoo
Human being with feelings
 
phoo's Avatar
 
Join Date: Dec 2010
Location: Hampstead, NC
Posts: 76
Default Inharmonic or Intermodulation Distortion Issues

I have a very complex mix in the works. It has a lot of high frequencies that are much like sibilances, but not all are from voices. Some are "airiness". Yes, these are up there, and yes they have been EQed as much as would normally be done. The problem is not tone, nor them occupying the same space. This has turned onto a technical issue. Compromise is expected, but I'm trying to understand why this is happening.

The problem is when some tracks are mixed there is what sounds like intermodulation distortion being produced to the point it sounds like inharmonic distortion. Two clean smooth tracks mix and sound like they are being run through a clipper. The sound is grainy at best, and has crackles at worst.

Tracks are 24 bit float, playing as 32 bits - 44.1K sample rate. Mixing to the same.

This is not a level issue as all tracks are well below any kind of clipping. The mixed tracks are around -12db. For that matter, since mixing is done in 64 bits I don't see how it would be possible to clip during the mix.

I'm guessing intermodulation, which will create frequencies not present in any of the source. This could and would cause what I'm hearing. Lower "new" frequencies would sound grainy. Higher "new" frequencies could and would cause terrible distortion is those frequencies are above the nyquist frequency. I can justify (via my ignorance) that this is what's happening.

It's also similar to dropping to 8 bits, or lower.

Solutions? Well, yes, EQ out the HF of the sources until it goes away. That's impractical, as I'm having to EQ out everything above 5k or so for it to be completely avoided. That's OK for some tracks, but completely ruins the sound of others, but to cut off the cymbals at 5K, or vocal sibilance? Not really.

Solutions that seem to have an effect, after EQ had failed is quite interesting. These have worked with a good bit of success. Each causes a good number of questions to arise. These don't work at all on some tracks and with more than two tracks isn't much more complicated when any pair of tracks is OK.

Solution:

1) Flip the phase of one track.

2) Shift the timing of a track a few msecs, or samples.

3) Use the RePitch plug-in to change the pitch just a little, or just the formant.

Think of this problem as being "how to mix multiple tracks of white noise" in such a way as to not create the artifacts that intermodulation can cause.

In the real world this is happening when the snare hit, there is a cymbal crash, or there sibilances, all while a smooth airy mellotron is sustaining chords.

What is actually causing this, and are there better solutions to avoid it?

I know this is a lot to read, but I hope I explained the problem well enough for discussion.

THANKS.

(please forgive any typos)
__________________
Experienced enough to know - Too old to care

http://www.phootoons.com
phoo is offline   Reply With Quote
Old 03-25-2020, 05:25 PM   #2
vdubreeze
Human being with feelings
 
vdubreeze's Avatar
 
Join Date: Jul 2011
Location: Brooklyn
Posts: 2,055
Default

In one of these smooth mixes, what happens when you go down the mixer and mute tracks, both one at a time and cumulatively? And in various strategic combinations?
__________________
The reason rain dances work is because they don't stop dancing until it rains.
vdubreeze is offline   Reply With Quote
Old 03-25-2020, 06:05 PM   #3
ErBird
Human being with feelings
 
Join Date: Jan 2017
Location: Los Angeles
Posts: 552
Default

Quote:
Originally Posted by phoo View Post
I'm guessing intermodulation, which will create frequencies not present in any of the source.
That's true, but inter-modulation distortion requires a non-linearity to be applied to the sum of the signals.
Do you have a non-linear plugin on a bus or the master? If you don't it can't be inter-modulation distortion.
ErBird is offline   Reply With Quote
Old 03-25-2020, 06:15 PM   #4
phoo
Human being with feelings
 
phoo's Avatar
 
Join Date: Dec 2010
Location: Hampstead, NC
Posts: 76
Default

-What application you're using to play those files "as 32-bit float" (VLC or other media player?)

-Did you record the files as 24-bit float in Reaper (is that even a possible driver setting in Reaper?),

-What audio hardware you have,

-What your driver settings are (in Reaper especially).

THANKS! Good points, all of them.

I misspoke about the format being float. It is 44.1khz, 24 Bit PCM. The default rendering format of Reaper, and the recording format.

I find the combinations of tracks by muting tracks one at a time until the problem goes away, then it's bring back tracks one at a time until it shows back up. It's not always the obvious combinations.

I have no idea is the plug-ins are non-linear. All are Reaper installed by default. ReaEQ mostly, to roll off the lows and highs to limit bandwidth. That said, the issue it there with no plug-ins in the path at all. Since this should rule out intermodulation, what's the remaining possibilities?

All individual tracks were recorded in Reaper. Audio interface is a ZOOM UAC-8 (USB3) via a main recording desktop, Windows 10 (sigh), and onboard audio in an HP Envy Laptop. The artifacts happen on both machines. Record on the desktop, then the majority of post recording work is done on the laptop. Everything is then passed back to the desktop for final "real" mixing through good speakers and final rendering. (I haven't noticed any differences between rendering on either machine if that matters - all that should stay in Reaper software)

Playback of the rendered files is in Reaper, VLC, Adobe Audition, and ultimately in Sony CD Architect, and in the car via blutooth directly from the Reaper on the laptop. (love mixing in the car - grin number one)

Track mixing in Reaper is 64 bit float.

The artifacts are audible during realtime playback in Reaper on both machines. What's being heard is what's rendered. This definitely makes it much easier to isolate. Because of this it seems like it would be irrelevant what the rendered format is, unless Reaper changes it's realtime playback amd mixing based on the rendering setting.

It also seems to not be computer or hardware specific since it happens on both the laptop (with really crappy audio) and the desktop (much better and drastically different hardware). Also, the fact that it's audible in every place I've tried to play these on may also help rule out it being something app specific. It's happing in Reaper, and it's not in any of the individual tracks.

Yes, I do understand how downsampling for playback could be relevant. Some apps, such as CD Architect has a dithering plug-in enabled by default, specifically for burning to 16 bit CD. That does make playback sound better, but it can't remove pops and clicks in the source (which is ultimately how this ends up sounding in the song unfortunately). At least there's no samplerate conversions going on.

These artifacts are subtle. My wife can't hear them at all. Unfortunately, I've "heard them" and now they are all I can hear. (grin two)

(I'll edit this response with the driver settings shortly. Not where the desktop is at the moment.)

Thanks for the responses!
__________________
Experienced enough to know - Too old to care

http://www.phootoons.com
phoo is offline   Reply With Quote
Old 03-25-2020, 08:56 PM   #5
Philbo King
Human being with feelings
 
Philbo King's Avatar
 
Join Date: May 2017
Posts: 1,923
Default

Since it seems to affect the highs specifically, there is a fair chance you are running into aliasing in plugins. One solution is to change the project sample rate to 48k or even 88.2k.

Aside from this, intermod distortion can be caused by any nonlinear processing that is fed more than 1 frequency at a time. compressors, limiters, saturators, clippers, even clipped samples.

Start by putting a spectrum analyzer on the master track. Solo each track, one at a time, and look for audio frequencies that aren't a multiple of the fundamental note. Those are the culprits. If these only show up when 2 or more tracks are soloed, it is caused by intermod distortion.
__________________
Tangent Studio - Philbo King
www.soundclick.com/philboking - stream music
www.facebook.com/philboking - gigs and news
Philbo King is online now   Reply With Quote
Old 03-26-2020, 05:07 PM   #6
phoo
Human being with feelings
 
phoo's Avatar
 
Join Date: Dec 2010
Location: Hampstead, NC
Posts: 76
Default

Quote:
Originally Posted by Philbo King View Post
Since it seems to affect the highs specifically, there is a fair chance you are running into aliasing in plugins. One solution is to change the project sample rate to 48k or even 88.2k.

Aside from this, intermod distortion can be caused by any nonlinear processing that is fed more than 1 frequency at a time. compressors, limiters, saturators, clippers, even clipped samples.

Start by putting a spectrum analyzer on the master track. Solo each track, one at a time, and look for audio frequencies that aren't a multiple of the fundamental note. Those are the culprits. If these only show up when 2 or more tracks are soloed, it is caused by intermod distortion.
Thanks!!! I believe you may have hit on the most likely culprit with the aliasing. The tracks in question all have a good amount of high harmonics, as mentioned earlier. It's the combinations of those in very specific places.

I've done some experimenting with raising the samplerates of the projects and done some rendering, but haven't done enough to see if the problem is affected for the better. One step at a time - changing the project first, and making sure there are no obvious differences because of that that. So far none.

I opened the files rendered at higher samplerate in Audition and looked at the spectrum in there. I can see hints of energy above 22K in them. That to me does seem to point to there potentially being some energy above the nyquist frequency. Audition won't display anything above that for any given samplerate (grin - well duh).

Thanks for your thoughts. I'll post here if I find something definitive. I suppose others have run into this, so I this helps them. That said, my guess is most folks won't be mixing projects with this much HF. Airy mellotrons are special.
__________________
Experienced enough to know - Too old to care

http://www.phootoons.com
phoo is offline   Reply With Quote
Old 03-30-2020, 07:05 PM   #7
phoo
Human being with feelings
 
phoo's Avatar
 
Join Date: Dec 2010
Location: Hampstead, NC
Posts: 76
Default

Big BINGO on the aliasing. After converting the project in question to 88.2khz and rendering out, and opening it in Audition, the HF showed up in all its glory.

The screenshot of Adobe Audio in spectrum view shows a lot of energy above the nyquist frequency of 44.1khz. There also appears to be some (maybe) dithering up around 35khz, even though dithering in the rendered mix was not enabled. This is straight out of Reaper.



Converting the project after the fact didn't get rid of all the artifacts, unfortunately. Some audible aliasing artifacts are in a few the original wav files, as those are 44.1k. Can't fix these with filtering or resampling, but I can go back and rerender a few as 88.2khz. Most of these are from a couple of mellotron and synth VSTs. I still have the originals in Reaper, having rendered them to wav for the main project.

While experimenting with this I discovered that the main desktop recording computer can't play the project at 88.2hz because it hits the CPU way too hard. The desktop does support it, as I have recorded a few tracks at that sample rate this week. There is just too much going on, resampling of the underlying wav files probably. That said, the laptop is happy playing at 88.2khz. The onboard audio, as crappy as it is, supports that playback sample rate and the CPU is indeed more powerful than the desktop and has more RAM. Might be time to start thinking about a replacement for the desktop. The motherboard is from 2004. It's still doing great at 44.1k, and even higher sample rates if there aren't many tracks. Because of all that, I have to render on the laptop, but have to mix at 44.1khz on the desktop for the good speakers. It's a compromise I must learn to live with.

But, now I know what the problem is and can easily avoid it in future projects.

Thanks everyone for your help, thoughts, questions, and answers.
__________________
Experienced enough to know - Too old to care

http://www.phootoons.com
phoo is offline   Reply With Quote
Old 03-31-2020, 08:42 AM   #8
not-relevant
Human being with feelings
 
Join Date: Jun 2019
Posts: 92
Default

In your example above, is the sampler you're using not simply filling the frequency range it has available to it? I.e. because you're rendering at 88kHz, it's generating output with frequencies up to 44.1kHz, i.e. the Nyquist rate, for 88kHz.

I mean that just because there is energy above 22kHz here, doesn't mean your sampler is broken, just that it is correctly pre-filtering up-pitched samples according to the Nyquist rate for your 88kHz render.

If you were rendering at 44.1kHz, it should be pre-filtering the samples before pitching them up, such that by the time they've been pitch shifted, they contain no frequencies above the Nyquist rate for the target 44.1kHz sample rate.

Basically, if your source sample is at 44.1kHz and you're going to pitch it up an octave, the sampler should pre-filter it to remove content below about 11kHz, then decimate it/interpolate it to basically remove every other sample. The resulting sample is an octave higher, but with no harmonics above 22kHz, and no aliasing should result. Any sane sampler will do this pre-filtering based on the Nyquist rate for the sample rate it's currently running at. I'm not sure your spectrograph proves that the sampler you're using is not doing that

Last edited by not-relevant; 03-31-2020 at 08:49 AM. Reason: Clarify
not-relevant is offline   Reply With Quote
Old 05-10-2020, 09:36 AM   #9
phoo
Human being with feelings
 
phoo's Avatar
 
Join Date: Dec 2010
Location: Hampstead, NC
Posts: 76
Default

Responding to not-relevant, I believe some of what you are mentioning is going on, in that the filtering is there but it could be better. That said, there is more going on that just that. There seems to be something "fundamental interesting" going on as well.

While doing some reanalyzing of the studio monitors, playback and recording of white and pink noise played through the monitors with no EQ, some suspicious artifacts are being seen in the captured wav files.

They can be seen in the linked images as what looks like a line of dithering noise, HF about 3/4 the way up the frequency spectrum. That is there 100% of the time in all captured audio when recording, regardless of sample rate, and is always about 3/4 of the way up. When recording at 44K that noise is visible around 34K for example.

What's disturbing is that it's in ALL wav files, even when there is no enabled microphone. It's in files that should be dead silent. Mute the mics, which are plugged directly into the ZOOM UAC-8, add a track to reaper and hit record. It's there. Not only that, it's a strong signal compared to the expected noise floor. It shows up a a sharp peak in apps like Har-Bal, spectrum analyzers, etc. It's well above the nyquist of whatever the sample rate happens to be and it's there for all sample rates. Record at 88K and there's noise around 75K, etc.

This is irrelevant to samplers, plug-in synths, etc, but it could be adding to the original issue if this noise isn't filtered out at the source. I'm needing to manually roll off the highs in a wave editor to get rid of them so they never get into the reaper digital pipeline. Digital EQ filtering, plug-ins, aren't getting rid of it, and it can be seen in the spectrum monitoring. Put in an EQ plug-in followed by a spectrum display plug-in - the EQ can't remove it. Very odd and totally unexpected. But, once it's gone from the wav it's gone. The trick with the synths is to freeze them and filter the freeze file in a wave editor.

I have not been able to track down the source. It SEEMS to be the ZOOM when recording with mics, but that's just anecdotal based on the recorded tracks from that device. Since it's in all wav files captured in reaper using mics, what is the source of the noise in the first place. Why is it also seen in many freeze files from some plug-in synths?

I'll need to get another sound device to be able to remove the ZOOM before I can take that step, but that doesn't explain some of the plug-in synths. Could very well be two totally different reasons there noise if getting generated.

This isn't the only noise issue, or issues, I've run across in the last months trying to figure out these issues. Another post coming up later. It seems unrelated, or at least different enough that it needs its own thread.
__________________
Experienced enough to know - Too old to care

http://www.phootoons.com
phoo is offline   Reply With Quote
Old 05-10-2020, 02:03 PM   #10
Philbo King
Human being with feelings
 
Philbo King's Avatar
 
Join Date: May 2017
Posts: 1,923
Default

It's quite possible the source is SMPS (switch mode power supply) noise, either in the Zoom or in the USB port. This can require some electronics expertise to sort out.

Simple stuff to try:
If the Zoom has a wall wart, try swapping it with another wall wart that has the same voltage, connector polarity and GTE (greater than or equal) current rating.

If it's USB powered you could try using a powered USB hub.
Another simple diagnostic test would be to carefully measure the noise level at high sample rate, then unplug any nonessential USB devices to see if the level or frequency of the noise changes. If it does, at least one noise source is the USB power supply of the PC.


Beyond that, you'd need a good oscilloscope to trace the noise to the source.

If you're handy with small scale soldering, you could tack-solder a 1000 pFd capacitor across the power rails of the Zoom supply or the USB 5V/Gnd. Don't try this unless you know what you're doing.
__________________
Tangent Studio - Philbo King
www.soundclick.com/philboking - stream music
www.facebook.com/philboking - gigs and news
Philbo King is online now   Reply With Quote
Old 05-11-2020, 02:13 AM   #11
Allybye
Human being with feelings
 
Join Date: Jan 2020
Posts: 183
Default

Must admit reading through the thread I too was wondering induced noise along the lines of Philbo's post.

Several threads show up people having problems with induced digital noise from usb comms, PSUs, exacerbated by grounding problems on the digital side. This is usually at higher frequencies and would seem it should be a first port of call when doing odd 'noise' diagnostic investigations.

As posted above aliasing ought not to be an issue when inputs are properly filtered.

Intermodulation distortion by it's very nature requires non linearity in tge signal path (maybe a good reason to do distortion effect on individual tracks before any mixing?)



I did wonder in passing why there is so much very high frequency content in the recorded material. A very small fraction of the population can even hear 20kHz; 22kHz? Must be music for dogs...... :-)

Many a transducer for listening (speaker, headphone) will not handle high levels of those frequencies without themselves distorting owing to their non linear responses! Is it worth including them?

Last edited by Allybye; 05-11-2020 at 02:20 AM.
Allybye is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 11:07 PM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.