A
aerosplat
I do a lot of work with audio signal path artifact correction and, specifically, correction of Windows Sound Mapper audio for specific listening hardware and environments. Audio signal path latency is not an issue for me because I can correct for it in the video players (VLC Media Player, Cyberlink PowerDVD) I'm using.
I've recently transitioned from Windows 7 to Windows 10 and have noticed some problems with audio glitches. I've followed all of the advice out there to fix this (driver update, windows update, BIOS update, stopping unnecessary processes/services, etc., etc.) but have not succeeded in getting glitch-free audio.
The evidence suggests that the culprit may be how Windows 10 handles audio device drivers that weren't specifically written for Windows 10. In Windows 7, the OS used a fixed 10ms buffer from which the audio device driver would fill its own buffer. As long as the device buffer wasn't larger than the OS buffer (resulting in buffer underruns or a total failure of the sound device to function at all), audio was glitch free, but, as expected, it included a minimum of 10ms of latency because of the fixed buffer size within the OS. For real-time audio processing where you're trying to eliminate latency wherever possible, 10ms of fixed latency was a problem that Microsoft addressed within Windows 10. Under Windows 10, the OS allows the audio device driver to specify the desired buffer size up to a maximum of 10ms. Because Windows 10 eliminates the fixed 10ms audio buffer on the OS side, audio latency can be significantly reduced... if the audio device driver knows how to inform the OS of the desired buffer size.
Now, you would think that Windows 10 could still work with legacy audio drivers if, unless told otherwise, it still assumed a 10ms buffer from which the audio device driver could draw its data. Unfortunately, this does not appear to be the case. I have evidence from two specific cases of attempting to use legacy audio drivers with Windows 10 that suggest the OS is not defaulting to a 10ms buffer.
Example 1: Using Synchronous Audio Router (SAR, an app that allows creation of synthetic ASIO audio input/output devices) to capture Windows Sound Mapper output to reroute it to Ableton, through a VST Finite Impulse Response correction filter, and finally through a Roland Octa-capture to speakers. Under Windows 7, this configuration worked flawlessly with ASIO buffer sizes up to 10ms. Under Windows 10, however, this configuration is not stable (results in continuous buffer underruns) beyond an ASIO buffer size of 5.3ms. The Roland Octa-capture needs an ASIO buffer size of at least 8ms to guarantee glitch-free audio, so the 5.3ms that Windows 10 is forcing upon this configuration makes glitch-free audio impossible.
Example 2: Once again, using SAR to capture Windows Sound Mapper output to reroute it via ASIO4ALL to Ableton, through a VST Finite Impulse Response correction filter, and finally through ASIO4ALL to a laptop internal sound device to internal speakers. Just like with example 1, under Windows 7, this configuration provided glitch-free audio with ASIO buffer sizes up to 10ms. Under Windows 10, however, this configuration is not stable (results in continuous buffer underruns) beyond an ASIO buffer size of 1.3ms. Just as with example 1, a 1.3ms buffer within ASIO4ALL is not long enough to guarantee glitch-free audio.
In both of these cases, the actual ASIO drivers (Roland Octa-capture and ASIO4ALL) are legacy drivers that weren't specifically written for Windows 10 and its new "latency-reducing" audio buffering architecture. Instead of defaulting to the legacy 10ms buffer to accommodate the legacy drivers, however, the evidence suggests Windows 10 is still attempting to employ "latency-reducing" buffer sizes that simply aren't appropriate for the legacy drivers.
Finally - the question: Is there any way to force Windows 10 into "legacy audio mode" where it always uses a 10ms buffer, just like Windows 7 did?
I apologize for being so verbose; it just seems that there is a metric ton of advice online for correcting Windows 10 audio problems and none of the advice gets to the root of the problem.
Thanks so much!
-Dale
Continue reading...
I've recently transitioned from Windows 7 to Windows 10 and have noticed some problems with audio glitches. I've followed all of the advice out there to fix this (driver update, windows update, BIOS update, stopping unnecessary processes/services, etc., etc.) but have not succeeded in getting glitch-free audio.
The evidence suggests that the culprit may be how Windows 10 handles audio device drivers that weren't specifically written for Windows 10. In Windows 7, the OS used a fixed 10ms buffer from which the audio device driver would fill its own buffer. As long as the device buffer wasn't larger than the OS buffer (resulting in buffer underruns or a total failure of the sound device to function at all), audio was glitch free, but, as expected, it included a minimum of 10ms of latency because of the fixed buffer size within the OS. For real-time audio processing where you're trying to eliminate latency wherever possible, 10ms of fixed latency was a problem that Microsoft addressed within Windows 10. Under Windows 10, the OS allows the audio device driver to specify the desired buffer size up to a maximum of 10ms. Because Windows 10 eliminates the fixed 10ms audio buffer on the OS side, audio latency can be significantly reduced... if the audio device driver knows how to inform the OS of the desired buffer size.
Now, you would think that Windows 10 could still work with legacy audio drivers if, unless told otherwise, it still assumed a 10ms buffer from which the audio device driver could draw its data. Unfortunately, this does not appear to be the case. I have evidence from two specific cases of attempting to use legacy audio drivers with Windows 10 that suggest the OS is not defaulting to a 10ms buffer.
Example 1: Using Synchronous Audio Router (SAR, an app that allows creation of synthetic ASIO audio input/output devices) to capture Windows Sound Mapper output to reroute it to Ableton, through a VST Finite Impulse Response correction filter, and finally through a Roland Octa-capture to speakers. Under Windows 7, this configuration worked flawlessly with ASIO buffer sizes up to 10ms. Under Windows 10, however, this configuration is not stable (results in continuous buffer underruns) beyond an ASIO buffer size of 5.3ms. The Roland Octa-capture needs an ASIO buffer size of at least 8ms to guarantee glitch-free audio, so the 5.3ms that Windows 10 is forcing upon this configuration makes glitch-free audio impossible.
Example 2: Once again, using SAR to capture Windows Sound Mapper output to reroute it via ASIO4ALL to Ableton, through a VST Finite Impulse Response correction filter, and finally through ASIO4ALL to a laptop internal sound device to internal speakers. Just like with example 1, under Windows 7, this configuration provided glitch-free audio with ASIO buffer sizes up to 10ms. Under Windows 10, however, this configuration is not stable (results in continuous buffer underruns) beyond an ASIO buffer size of 1.3ms. Just as with example 1, a 1.3ms buffer within ASIO4ALL is not long enough to guarantee glitch-free audio.
In both of these cases, the actual ASIO drivers (Roland Octa-capture and ASIO4ALL) are legacy drivers that weren't specifically written for Windows 10 and its new "latency-reducing" audio buffering architecture. Instead of defaulting to the legacy 10ms buffer to accommodate the legacy drivers, however, the evidence suggests Windows 10 is still attempting to employ "latency-reducing" buffer sizes that simply aren't appropriate for the legacy drivers.
Finally - the question: Is there any way to force Windows 10 into "legacy audio mode" where it always uses a 10ms buffer, just like Windows 7 did?
I apologize for being so verbose; it just seems that there is a metric ton of advice online for correcting Windows 10 audio problems and none of the advice gets to the root of the problem.
Thanks so much!
-Dale
Continue reading...