Noise floor level changes depending on which app is in foreground

[Tony359]

Hi all,

I am experiencing a weird issue with my Focusrite 18i20 in Windows 11. I use this interface to check the Noise Floor of some devices I repair and over time I discovered that the Noise Floor would develop a "bump" around 400Hz once Audition had been run. The workaround for that was to close my RTA (Smaart) and unplug/replug the interface.

Since a few weeks, this workaround stopped working. I fiddled with drivers and settings for hours but could not find a way to remove that bump. It's NOT there after I re-install the drivers but once I open Audition, it appears and doesn't go away.

Only recently I've discovered that if I have Audition running in the Foreground while the RTA still runs, the "bump" goes away.

I made a short video to explain. I am switching between Ardour and Smaart and you can see that the Noise Floor Smaart reads, do change depending on what app is in the foreground.

In fact, I discovered that as long as I run Ardour in the background, then the "bump" doesn't manifest itself.

All apps are using ASIO drivers. Audio is set to 24bit 48KHz everywhere including Windows (though it shouldn't matter) but I also tried with different combinations as I believe it was how I managed to make the bump go away a couple of years ago when I first noticed that.

I'm a bit lost now. I tried all possible combinations, different drivers, reinstalling, changing. This has - in my opinion - something to do with some re-sampling somewhere but I cannot figure out what it is.

Can someone maybe recognise this behaviour and give me a pointer?

 

[Philbo King]

Sounds like Adobe snuck a worm or trojan into your system. Try finding a copy of Cool Edit (What Audition was called before Adobe bought it). Or download Reaper. See if either of those exhibit the same issue.

 

[wwenze]

Do you have OS-wide EQ?

 

[KSTR]

I'd do a double-check with REW, which I can generally recommend as a measuring suite. And using ASIO driver (which must support channel pair selection).

 

[KSTR]

Try finding a copy of Cool Edit (What Audition was called before Adobe bought it).

CoolEdit and Audtion < V3 don't support ASIO.

 

[Sokel]

For measuring gear with different ASIO drivers, better use Multitone Analyzer, it supports every combination.

 

[RandomEar]

What an odd problem! Could you tell us a little more about the setup?

• What components are in the source PC? Desktop or notebook?
• What monitor are you using? Does it have an external power brick?
• How is everything connected?
• What are we seeing in the video, a loopback recording or a mic connected?

 

[Tony359]

Thanks all for the replies!

I'd like to use Smaart for my tests, it's not a cheap software and I use it for work too.

This is a Lenovo laptop, the 18i20 is wired via USB to it. I don't have a monitor, I am using headphones. I see where you are going but this doesn't seem to be a ground loop. It also happens with NOTHING connected to the 18i20 and with the laptop on battery. What you see on the video is a multichannel device, if I unplug it, the noise floor lowers but the weird effect remains.

Connections: Laptop on battery, Focusrite direclty on USB. Focusrite doesn't have to be plugged into anything to show the issue.

@wwenze
Uhm... good point. I don't think so but I might have played over time with EqualiserAPO, it might still be there causing havoc. Let me check, thanks!

I will try another RTA to rule out Smaart. (Edit: I tried and it shows the same behaviour)

Thanks for now!

 

[RandomEar]

Thanks all for the replies!

I'd like to use Smaart for my tests, it's not a cheap software and I use it for work too.

This is a Lenovo laptop, the 18i20 is wired via USB to it. I don't have a monitor, I am using headphones. I see where you are going but this doesn't seem to be a ground loop. It also happens with NOTHING connected to the 18i20 and with the laptop on battery. What you see on the video is a multichannel device, if I unplug it, the noise floor lowers but the weird effect remains.

Nah, this doesn't look like a ground loop at all. Just trying to get a better general understanding of the problem. I think some weird audio driver interaction would be the prime suspect. Could also be some weird timing/interrupt or ASIO-buffer related thing - you could definitely try increasing the buffer sizes.

Connections: Laptop on battery, Focusrite direclty on USB. Focusrite doesn't have to be plugged into anything to show the issue.

@wwenze
Uhm... good point. I don't think so but I might have played over time with EqualiserAPO, it might still be there causing havoc. Let me check, thanks!

I will try another RTA to rule out Smaart. (Edit: I tried and it shows the same behaviour)

Thanks for now!

I am not an expert in audio interfaces and ASIO drivers, so I would focus on troubleshooting the less likely solutions just to rule them out:

• I would tinker around with thread priority in Task Manager (go to Taks Manager-> Details -> right click your RTA software or Audition or Ardour and choose "real time" priority)
• If VRR is enabled in the graphics settings, I would turn that off temporarily
• If your notebook does have a dedicated GPU, I would turn that off and test with the iGPU only (and vice versa if the dGPU is currently off -> turn it on and use that)
• If your notebook uses and Intel CPU, I would turn off all e-cores and all C-states for testing

I don't think these are likely culprits, but at least you can rule them out quickly.

 

[phofman]

IMO precise description of your setup is needed. IIUC: you have two audio apps, both using ASIO. To what audio device? The same at the same time? That would require multi-client ASIO with mixing. Or the apps output to different devices? Playback/Capture?

What does that spectrum show - Focusrite analog audio loopback? You say pink noise - what device plays the noise, Focusrite analog loopback?

On your video Audition does not seem to be capturing (no time change there) while in the foreground.

Also exact wiring would be required to know (esp. the loopback wiring).

An app at foreground can change noise profile produced by the computer, which a high-Z input could pickup. Your db values show it's a tiny signal.

 

[phofman]

Could also be some weird timing/interrupt or ASIO-buffer related thing - you could definitely try increasing the buffer sizes.

IMO (and IME) a buffer issue would produce glitches which look very different. Here it looks like no issue with buffers.

 

[phofman]

I do not know the RTA software but is the pink noise generator really on? What spectrum is the RTA measuring? Noise at -17dB being recorded at -150dB? That does not look right...

 

[Tony359]

Nah, this doesn't look like a ground loop at all. Just trying to get a better general understanding of the problem. I think some weird audio driver interaction would be the prime suspect. Could also be some weird timing/interrupt or ASIO-buffer related thing - you could definitely try increasing the buffer sizes.

ASIO buffer already at maximum and I have tried with different settings

I am not an expert in audio interfaces and ASIO drivers, so I would focus on troubleshooting the less likely solutions just to rule them out:

• I would tinker around with thread priority in Task Manager (go to Taks Manager-> Details -> right click your RTA software or Audition or Ardour and choose "real time" priority)
• If VRR is enabled in the graphics settings, I would turn that off temporarily
• If your notebook does have a dedicated GPU, I would turn that off and test with the iGPU only (and vice versa if the dGPU is currently off -> turn it on and use that)
• If your notebook uses and Intel CPU, I would turn off all e-cores and all C-states for testing

I don't think these are likely culprits, but at least you can rule them out quickly.

Good tests! Will try.

Discrete GPU is disabled via BIOS - the software implementation was hideous and causing all sort of issues with Windows.
CPU is AMD Ryzen 7 6800 so no efficiency cores.

IMO precise description of your setup is needed. IIUC: you have two audio apps, both using ASIO. To what audio device? The same at the same time? That would require multi-client ASIO with mixing. Or the apps output to different devices? Playback/Capture?

What does that spectrum show - Focusrite analog audio loopback? You say pink noise - what device plays the noise, Focusrite analog loopback?

On your video Audition does not seem to be capturing (no time change there) while in the foreground.

Also exact wiring would be required to know (esp. the loopback wiring).

An app at foreground can change noise profile produced by the computer, which a high-Z input could pickup. Your db values show it's a tiny signal.

I am not playing pink noise - did I say pink noise? What you see on the RTA is the NOISE FLOOR of a device which is not playing anything. The ON on the generator is currently disabled (I think it becomes RED when it's engaged!)

But please ignore the HW connections: the issue happens with NOTHING connected to the focusrite and nothing connected to the laptop. So the wiring and the device under tests are not relevant.

What you see on the RTA are the 8 inputs of the FocusRite, each one independently traced.

Audition only needs to be open to cause that effect. It doesn't need to play or record. As soon as I close it, the bump in the RTA appears and doesn't go away.

No loopback wiring with the FocusRite. I have 8 cables going in the inputs and 8 cables going out. Smaart is configured to analyse each one of the 8 inputs. FocurRite Mixer is set to pass each input to the relevant DAW channel - plus it makes a mono mix for the headphone output.
What you see in the video happens with or without something connected to the interface.

An app at foreground can change noise profile produced by the computer, which a high-Z input could pickup. Your db values show it's a tiny signal.

Interesting theory. Something running which is changing the electromagnetic noise the laptop is producing. Though I can run whatever I want on the laptop and nothing changes on the RTA (besides another DAW!). If what you suggest was the case, I think I'd be able to induce lots of visible noises on the RTA by just doing different things on the laptop, like run a stress test.

In the past I was able to get rid of the "bump" by simply closing Smaart, power cycling the interface and re-opening Smaart. That would work well until Audition was used again.

I know, quite weird

 

[phofman]

Well then you are indeed measuring FFT of an unconnected (i.e. also not shorted to ground) hi-Z input, i.e. the internal noise and whatever the input picks up. So the Audition CPU/GPU processing changes (power and ground) currents in your computer, changing the computer noise profile, and your hi-z input picks that up (at -140dB which is extremely small). It adds some 300Hz component which is perfectly possible with OS timing, etc. My 2 cents if you short the analyzed analog input to ground (e.g. with a shorted cinch etc) the effect will be gone or different.

Typically noise level of interface input is measured with that input shorted to ground, to avoid any induced noise which is not internal to the audio device. Unconnected inputs are never part of the measurement chain. When some source is connected to the input, the output impedance of that source is hundreds of ohm which is much lower than the 100s of kiloohm of the input impedance - the noise from the PC would not be induced into such low impedance.

 

[Tony359]

Well then you are indeed measuring FFT of an unconnected (i.e. also not shorted to ground) hi-Z input, i.e. the internal noise and whatever the input picks up. So the Audition CPU/GPU processing changes (power and ground) currents in your computer, changing the computer noise profile, and your hi-z input picks that up (at -140dB which is extremely small). It adds some 300Hz component which is perfectly possible with OS timing, etc. My 2 cents if you short the analyzed analog input to ground (e.g. with a shorted cinch etc) the effect will be gone or different.

But the video shows the issue WHILE all 8 channels are connected to a device. If what you say is right, then I should only be able to see the issue with nothing connected to the interface.

My point in saying "it ALSO happens when nothing is connected to it" was to rule the equipment/wiring out of the equation.

Typically noise level of interface input is measured with that input shorted to ground, to avoid any induced noise which is not internal to the audio device. Unconnected inputs are never part of the measurement chain. When some source is connected to the input, the output impedance of that source is hundreds of ohm which is much lower than the 100s of kiloohm of the input impedance - the noise from the PC would not be induced into such low impedance.

I'm not measuring the noise of the Focusrite inputs. I am measuring the noise of the outputs of an 8 channel device using the Focusrite. I appreciate it's not going to be a super accurate measurement but I have a "good reference" and as long as that matches, I am happy.

So no, the "no input" was just an example but I don't think it's the issue here.

 

[phofman]

But the video shows the issue WHILE all 8 channels are connected to a device. If what you say is right, then I should only be able to see the issue with nothing connected to the interface.

Well, you never explained your connection here. I would assume "nothing was connected to focusrite" would mean the inputs were not connected. IMO without explaining your setup fully first (both software and hardware) this discussion is going nowhere.

 

[Tony359]

@phofman

I'm new here so I don't want to sound arrogant but I explained my setup (SW and HW including connections) in post #8

Nine lines below "Nothing was connected to focusrite" I also say:

"What you see in the video happens with or without something connected to the interface."

 

[phofman]

Connections: Laptop on battery, Focusrite direclty on USB. Focusrite doesn't have to be plugged into anything to show the issue.

I do not see anything about "WHILE all 8 channels are connected to a device".

So please describe your complete setup here (SW+HW) first, for me and others.

 

[Tony359]

If a summary helps, happy to help.

HW: Lenovo laptop, AMD Ryzen 7 6800HS CPU, 32GB RAM, iGPU ony. Focusrite 18i20 directly wired via USB 2.0. Laptop powered by battery.
SW: Smaart 8, Audition (can't remember the version number but similar behaviour observed with other DAW)

Audio Connections: irrelevant. The issue is observed WITH or WITHOUT the inputs wired to something. When wired to a multichannel sound processors, all input/outputs are balanced.

Issue: When reading the Noise Floor of the INPUTS of the 18i20 on Smaart, I see a bump around 400Hz on all channels which seem to disappear when another DAW (in this case Audition) is running AND brought to the foreground. If using Ardour instead of Audition, Smaart DOESN'T shows the bump AS LONG as Ardour is running, even in the background.

Closing Smaart and Audition and restarting Smaart only, WILL show the bump.
Power cycling the 18i20 USED TO remove the bump but not anymore
Rebooting the laptop doesn't change anything
Reinstalling the drivers DOES remove the bump, until Audition/Ardour are run, then the bump comes back and stay.

The video shows the following setup
INPUTS: an 8 channel sound processor, not playing any audio
Smaart running in RTA mode, Generator not running
Ardour just open, not playing.

Anything else you might need?

 

[phofman]

OK, what is the digital path of Smaart to the Focusrite? The other DAW - you say running, also capturing? I do not see it capturing on your video. If capturing, what device and what digital path?

What are the audio connections you call irrelevant? When disconnected - that means Hi-Z inputs, susceptible to picking up the PC noise. When connected - to what? How is that device further connected? Example - the device could be a pre-amp with disconnected inputs, again picking up the PC noise with its hi-z inputs.

The first test - short the Focusrite inputs and test its real internal noise profile, without any external influence.