Persistent AUDCLNT_E_UNSUPPORTED_FORMAT error on WASAPI Shared input

[Makstuff]

Had FlexASIO happily working for a year now, didn't use it for a few months, and since I started recording again some weeks ago, Reaper always throws the error "There was an error opening the audio hardware: Error initializing ASIO buffers" (see screenshot).

I didn't change anything on my hardware at all. I had version 1.4 before, updated it to 1.6 after I saw the error for the first time, but this didn't help. Googling the error also isn't very helpful. Any idea what could be wrong here?

FlexASIO.log

I should probably mention that ASIO4ALL v2 still works normally.

[dechamps]

Interesting:

Selecting input device
Using default device with index 15
Selected input device: PortAudio device index 15 (name: 'Mikrofon (USB Audio CODEC )', host API: 2, default sample rate: 44100, max input channels: 2, max output channels: 0, input latency: 0.003 (low) 0.01 (high), output latency: 0 (low) 0 (high))
Selecting output device
…
Getting WASAPI device default format for device index 15
WASAPI device default format for device index 15: WAVEFORMAT with format tag 65534 [EXTENSIBLE], 1 channels, 44100 samples/second, 88200 average bytes/second, block alignment 2 bytes, 16 bits per sample, 16 valid bits per sample, channel mask 4 [Front Center], subformat {00000001-0000-0010-8000-00AA00389B71} [PCM]
…
Opening PortAudio stream with...
...input parameters: PortAudio stream parameters for device index 15, 2 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 68 [UseChannelMask], channel mask 4 [Front Center], host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...output parameters: PortAudio stream parameters for device index 12, 2 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 68 [UseChannelMask], channel mask 3 [Front Left, Front Right], host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...sample rate: 48000 Hz
...frames per buffer: 480
...stream flags: 8 [PrimeOutputBuffersUsingStreamCallback]
...stream callback: 00007FF9760F1C85 (user data 0000000000619BB0)
[PortAudio] WASAPI: IAudioClient2 set properties: IsOffload = 0, Category = 0, Options = 0
[PortAudio] WASAPI ERROR HRESULT: 0x88890008 : AUDCLNT_E_UNSUPPORTED_FORMAT

Something's definitely off here - PortAudio's WASAPI backend is reporting an input channel count of 2, but the default wave format only mentions 1 channel, so FlexASIO ends up trying to open the input device with 2 channels but a channel mask of [Front Center], which makes no sense. That could quite possibly be the root cause of the AUDCLNT_E_UNSUPPORTED_FORMAT error, and looks like a bug in FlexASIO and/or PortAudio.

This is the first time I see this problem so I'll have to investigate further. In the mean time, if my diagnosis is correct, you should be able to work around the issue by explicitly specifying the input channel count in your configuration file, which will have the effect of disabling channel masks altogether. Just add channels = 2 in your [input] section. Does that solve your issue?

[Makstuff]

You mean like this?

No, seems it doesn't change anything. New log file attached.

FlexASIO.log

[dechamps]

No channel mask this time as expected, but still the same error:

Opening PortAudio stream with...
...input parameters: PortAudio stream parameters for device index 15, 2 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 64, channel mask 0, host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...output parameters: PortAudio stream parameters for device index 12, 2 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 68 [UseChannelMask], channel mask 3 [Front Left, Front Right], host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...sample rate: 48000 Hz
...frames per buffer: 480
...stream flags: 8 [PrimeOutputBuffersUsingStreamCallback]
...stream callback: 00007FF977851C85 (user data 00000000006661F0)
[PortAudio] WASAPI: IAudioClient2 set properties: IsOffload = 0, Category = 0, Options = 0
[PortAudio] WASAPI ERROR HRESULT: 0x88890008 : AUDCLNT_E_UNSUPPORTED_FORMAT

Can you try with channels = 1? It might be that it's the PortAudio max input channel count that's wrong and WASAPI might only accept a 1-channel configuration in this particular situation.

[Makstuff]

Still the same, unfortunately. New log attached. Thanks for all the effort by the way, really appreciate it.

FlexASIO.log

[dechamps]

Okay. Well it may be that the error has nothing to do with channel counts at all. The log states that the input device uses a 44100 Hz sample rate, but you're trying to use it at 48000 Hz. Can you try using 44100 Hz instead (with both channels = 1 and channels = 2)? In theory that's not supposed to matter because WASAPI is supposed to resample automatically, but I'm not sure how well that works on the input side.

Even better would be to run FlexASIOTest (with both channels = 1 and channels = 2) and attach the resulting log. FlexASIOTest probes for all common sample rates and probes both devices separately so it will cover more stuff in a single run.

[Makstuff]

FlexASIO.log (channels=1)
FlexASIO.log (channels=2)

[dechamps]

It looks like all sample rates and input channel count combinations are rejected with the same error 😕 This is starting to smell like a PortAudio bug or some obscure problem deeper in the stack. I have no idea what could cause this, and I'm not sure how I could possibly reproduce. Out of curiosity, can you describe the hardware input device you're using (brand, model, configuration)?

I'll see what I can do, but in the mean time, here are a few things that might help work around the issue:

• Using a different input device;
• In the Windows input sound device properties, fiddling with the "default format" (channel count, sample rate, sample type)
• Using WASAPI exclusive mode on the input side;
• Using another backend such as DirectSound or WDM-KS;
• Fiddling with the sampleType option (I'd be extremely surprised if it made a difference in WASAPI shared though)

[Makstuff]

Is this something that could be caused by a Windows update? Cause that's the only change to my system I can possibly think of. Same Reaper version, same hardware, same everything. But if it's Windows, it would seem pretty strange that I'm the only one with this problem.

I'm using the Behringer U-Phoria UMC22. Not sure what you mean by configuration. Probably other devices have some options built in the hardware? Mine doesn't, as far as I'm aware.

[Makstuff]

No channel mask this time as expected, but still the same error:

Opening PortAudio stream with...
...input parameters: PortAudio stream parameters for device index 15, 2 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 64, channel mask 0, host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...output parameters: PortAudio stream parameters for device index 12, 2 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 68 [UseChannelMask], channel mask 3 [Front Left, Front Right], host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...sample rate: 48000 Hz
...frames per buffer: 480
...stream flags: 8 [PrimeOutputBuffersUsingStreamCallback]
...stream callback: 00007FF977851C85 (user data 00000000006661F0)
[PortAudio] WASAPI: IAudioClient2 set properties: IsOffload = 0, Category = 0, Options = 0
[PortAudio] WASAPI ERROR HRESULT: 0x88890008 : AUDCLNT_E_UNSUPPORTED_FORMAT

Can you try with channels = 1? It might be that it's the PortAudio max input channel count that's wrong and WASAPI might only accept a 1-channel configuration in this particular situation.

I did some random testing today and figured out that channels = 1 actually works, but seemingly only if the "Enable inputs -> last" setting in Reaper is 0 (see screenshot).

I was playing around with different audio systems in Reaper and different backends in FlexASIO, and at one point the "Enable inputs -> last" setting probably changed itself from 1 to 0 (at least I didn't do it intentionally). Since then, channels = 1 works, and the option "1" at "Enable inputs -> last" in Reaper that was previously there is now gone.

Obviously the second input of my interface isn't working with this setting, but at least the first input works fine. First input btw works with every different backend (MME, Windows DirectSound, Windows WASAPI and Windows WDM-KS), although with differing latency, while the second one is never accessible.

New log file is attached. Some part seems to be repeating itself again and again. If it matters, all I did to produce it was open Reaper and close it again after about 2 seconds.

FlexASIO.log

[dechamps]

The only difference between that last log and back when you tried with Reaper and channels = 1 is that in that last log, the sample rate is 44.1 kHz instead of 48 kHz. That's it. There is no other difference in parameters, including the channels used, so I don't think what you did with the channels made a difference. The difference in sample rate doesn't explain everything either, because you provided a log where FlexASIOTest did try 1 channel at 44.1 kHz and that didn't work:

CFlexASIO::OpenStream(inputEnabled = 1, outputEnabled = 0, sampleRate = 44100, framesPerBuffer = 0, callback = 00007FF959E4E110, callbackUserData = 0000000000000000
Using shared mode for input WASAPI stream
Enabling auto-conversion for input WASAPI stream
Opening PortAudio stream with...
...input parameters: PortAudio stream parameters for device index 15, 1 channels, sample format 2147483649 [Float32, NonInterleaved], suggested latency 0s, host API specific: 56 bytes structure, type 13 [WASAPI], version 1, WASAPI specific: flags 64, channel mask 0, host processor output 0000000000000000, host processor input 0000000000000000, thread priority 0 [None], stream category 0 [Other], stream option 0 [None]
...output parameters: none
...sample rate: 44100 Hz
...frames per buffer: 0
...stream flags: 8 [PrimeOutputBuffersUsingStreamCallback]
...stream callback: 00007FF959E4E110 (user data 0000000000000000)
[PortAudio] WASAPI: IAudioClient2 set properties: IsOffload = 0, Category = 0, Options = 0
[PortAudio] WASAPI ERROR HRESULT: 0x88890008 : AUDCLNT_E_UNSUPPORTED_FORMAT

So it's still unclear what the root cause is or why you were suddenly able to make it work.

You say that you can only make it work with channels = 1. Are you able to use both input channels successfully in other non-ASIO software (e.g. basic recording software)? ASIO4ALL is a special case because it uses WDM-KS so it's not as interesting, I'm more interested in what works in "boring" Windows software. If the second channel doesn't work in basic Windows apps either, then it's very likely to be a hardware driver issue.

[Makstuff]

What kind of "boring" Windows software would that be? I wanted to try Audacity, but for some strange licensing reason it doesn't seem to support ASIO. I also looked at settings in the Windows built in Voice Recorder app, but didn't find an option to change input channels. Same with a third party voice recorder app from the Microsoft Store.

I don't know what exactly you mean by "hardware driver", but this interface, as some other smaller models by Behringer, doesn't have it's own dedicated driver. Instead, Behringer advises to install ASIO4ALL directly from http://www.asio4all.org/ (already tried reinstalling that, however without any success).

[dechamps]

What kind of "boring" Windows software would that be? I wanted to try Audacity, but for some strange licensing reason it doesn't seem to support ASIO.

That's precisely the point. We need to know what's happening in software that is using the Windows audio driver normally, without using ASIO. Audacity is not a great example because it uses PortAudio behind the scenes so it has shared code paths with FlexASIO.

I guess another option would be to try FlexASIO with, say, the default DirectSound backend. What happens if you just use the default FlexASIO config (i.e. empty or missing FlexASIO.toml file)? Can you address both input channels independently in that case and get separate audio signals from each?

I don't know what exactly you mean by "hardware driver"

I mean the Windows WDM hardware device driver. In your case that might be the Microsoft USB Audio Class driver.

[Makstuff]

Using an empty config file, I don't get an error when opening Reaper and can select both input channels in Reaper, but they are the same (both have the mic signal, and the instrument input isn't showing up at all).

FlexASIO.log