ALSA multi plugin error when combining two soundcards

[anthonio9]

My goal is to mix the output of two sound cards (RAVENNA from bondagit and hifiberry), then pass it to a DSP pcm plugin. As it's not possible with dmix (dmix only supports output to hw devices) the reasonable next step was multi plugin, but my tries always end up in a failure. Below is a quite standard multi configuration inspired by alsa docs that I'm using:

pcm.quad {
    type multi
    slaves.a.pcm "hw:RAVENNA"
    slaves.a.channels 2
    slaves.b.pcm "hw:sndrpihifiberry"
    slaves.b.channels 2

    bindings.0.slave a
    bindings.0.channel 0
    bindings.1.slave a
    bindings.1.channel 1
    bindings.2.slave b
    bindings.2.channel 0
    bindings.3.slave b
    bindings.3.channel 1
    master 1
}

pcm.quad2 {
    type plug
    slave.pcm "quad"
}

This is run with alsaloop to be input into an extplug plugin that easily accepts 4 channels of audio - alsadsp:

$ alsaloop -C quad -P alsadsp -c 4
Rate 48000Hz not available for capture quad: Invalid argument
Unable to set parameters for capture quad stream: Invalid argument
Loopback start failure.

This results in a failure, moreover running quad2 also results in the same type of error.

Next step was to try plug pcm:

pcm.quad_plug {                                                                                                                              
    type multi                                                                                                                               
    slaves.a.pcm "plughw:RAVENNA"                                                                                                                
    slaves.a.channels 2                                                                                                                      
    slaves.b.pcm "plughw:sndrpihifiberry"                                                                                                        
    slaves.b.channels 2                                                                                                                      

    bindings.0.slave a                                                                                                                       
    bindings.0.channel 0                                                                                                                     
    bindings.1.slave a                                                                                                                       
    bindings.1.channel 1                                                                                                                     
    bindings.2.slave a                                                                                                                       
    bindings.2.channel 0                                                                                                                     
    bindings.3.slave a                                                                                                                       
    bindings.3.channel 1                                                                                                                     
}

pcm.quad_plug2 {                                                                                                                             
    type plug                                                                                                                                
    slave.pcm "quad_plug"                                                                                                                         
}  

With quad_plug the same error occurs, with quad_plug2 the error is as below:

$ alsaloop -C quad_plug2 -P alsadsp -c 4
Broken configuration for capture quad_plug2 PCM: no configurations available: Invalid argument
Unable to set parameters for capture quad_plug2 stream: Invalid argument
Loopback start failure.

Then the last thing I tried was to force the 48kHz sampling rate directly on the hardware devices with the help of the following config:

pcm.quad_48k {                    
    type multi                    
    slaves.a {                    
        pcm {                     
            type plug     
            slave.pcm "hw:RAVENNA"
            slave.rate 48000      
        }                         
        channels 2                
    }                             

    slaves.b {                    
        pcm {                     
            type plug             
            slave.pcm "hw:sndrpihifiberry"
            slave.rate 48000              
        }                                 
        channels 2                        
    }                                     

    bindings.0.slave a            
    bindings.0.channel 0          
    bindings.1.slave a            
    bindings.1.channel 1          
    bindings.2.slave b                    
    bindings.2.channel 0          
    bindings.3.slave b                    
    bindings.3.channel 1                  
}

and again it ends up with a failure:

$ alsaloop -C quad_48k -P alsadsp -c 4
Sample format not available for capture quad_48k: Invalid argument
Unable to set parameters for capture quad_48k stream: Invalid argument
Loopback start failure.

I know that both these cards are capable of running 48kHz, I'm doing that all the time. So what could be the cause of this error? alsadsp wants them to be 48kHz and if run separately it all works well, both cards set themselves according to alsadsp requirements. It works even if there's only one card in the multi plugin that occupies all 4 output ports - my conclusion is that it's not a problem of alsadsp. Are there any other steps that could help solving the issue?

[perexg]

Sample format issue? Try -f S32_LE for alsaloop or check the supported formats by arecord -Dhw:RAVENNA --dump-hw-params for both hw: devices.

[perexg]

... and basically it's a wrong idea to combine two clocks (two different cards) with the multi plugin ... The clocks difference will increase in time and the stream won't be usable.

[anthonio9]

Thanks, it seems to work quite well with the -f S32_LE option, but I've only tested it for a minute or so. Would a sound server (like pipewire) make this combination work better? I know that pipewire does have some tricks, but have no idea what would that turn out to be in the long run. Is there a way to inspect that clock differece? That would actually help me a lot.