Greetings,
I'm incorporating an algorithm that upmixes a stereo input signal to
multichannel output into an APO, using the MFX SwapAPO from Sysvad as
reference. For the time being, I'd like to save the multichannel output as a
.wav file for testing (Sysvad's normal behaviour). I'm having trouble
understanding how to correctly change the number of channels and hence the
interleaved buffer size I receive during the APOProcess call. So far I have
done the following:
1. Removed the assertion lock from ValidateCacheAndConnectionInfo in the MFX
cpp file
"_ASSERTE(UncompOutputFormat. dwSamplesPerFrame ==
UncompInputFormat.dwSamplesPerFrame);"
2. Added only (APO_FLAG_BITSPERSAMPLE_MUST_MATCH |
APO_FLAG_FRAMESPERSECOND_MUST_MATCH) in the CRegAPOProperties
3. In isOutputFormatSupported() I have tried a 'hacky' approach with explicitly
setting the channel count as follows:
hResult =
recommendedFormat->GetUncompressedAudioFormat(&uncompOutputFormat);
IF_FAILED_JUMP(hResult, Exit);
uncompOutputFormat.dwSamplesPerFrame = this->numberOfPlaybackChannels;
// explicit fixed channel count
hResult =
CreateAudioMediaTypeFromUncompressedAudioFormat(&uncompOutputFormat,
ppSupportedOutputFormat);
IF_FAILED_JUMP(hResult, Exit);
hResult = S_OK; // S_FALSE;
Exit:
if (recommendedFormat)
{
recommendedFormat->Release();
}
return hResult;
However when checking the channel count for outputs in
ValidateCacheAndConnectionInfo it is still set to 2. (I am also getting the
feeling that this is not a good idea/approach to begin with).
Should I be making edits in the kernel code? Please correct me if I'm wrong,
but from inspecting the Sysvad solution, I see that CMiniportWaveRTSTream::Init
is where the saving occurs, and I presume therefore nothing exits the output
pin of the Wave KS filter to the corresponding topology filter at this stage -
in other words, I would assume explicitly setting the channel count on the pin
here would be redundant for the current purpose of saving multichannel output
(and therefore I must be missing something else?)
How can I make this work and ensure that the APOProcess output buffer I get has
the same number of frames but different number of channels? I presume the
correct approach would also automatically set the desired channel count in the
RIFF wave header for output (I see that the PKSDATAFORMAT DataFormat_ supplied
to CMiniportWaveRTStream::Init is responsible for this).
Apologies for my confusions, I'm still wrapping my head around a lot of these
concepts
Thanks!
