Directsound vs WASAPI vs ASIO Options

[Midiman]

After reading many of the posts on this forum, it appears that there is no evidence that ASIO or WASAPI sounds any better than Directsound on Windows 7. I understand the importance of ABX testing - and I am interested in a way to do a double blind test of such claims. Let me first of all state that I cannot hear any difference. For me to be objective, I have used the same player for all tests - foobar2000 - and simply changed the Output from Directsound to WASAPI to ASIO. However, to do this, I have to stop the playback, change the driver, and then restart playback. This means comparions are doubtful at best due to the delay involved and my knowledge of what driver I selected.

How have others conducted such tests? Please note I am not trying to prove a difference exists - just to find a way to do the test properly. The ABX plugin for foobar2000 has been very useful to compare different WAV files. Obviously, one cannot use it for this sort of experiment.

Thank you for any suggestions or pointers to where this has been proved one way or another.

[[JAZ]]

@item: The possibility to have an unaltered, bit-for-bit output with either of those three methods exists, and in such case, neither ABX, nor guesswork and supppositions play any role.

In the event that they aren't bit-perfect, the election of one over another is most of the times about performance/usability than about quality. (Expecting very low latency, wanting to play multiple streams at the same time, sampling rates and bit depths supported, etc..).

There are only two scenarios worth mentioning about Directsound or MME failing:
The old "AC97"-compatible hardware/drivers that used to resample with low quality 44.1Khz signals to 48Khz, and the newer "MME waveout recording resampling with low quality, even when unneeded" in Vista and 7 (which was patched some time ago).

That's as much about quality that any Windows user should care when using the different audio APIs (Other than when requiring bit-for-bit signals)

[item]

@item: The possibility to have an unaltered, bit-for-bit output with either of those three methods exists, and in such case, neither ABX, nor guesswork and supppositions play any role.

In the event that they aren't bit-perfect, the election of one over another is most of the times about performance/usability than about quality. (Expecting very low latency, wanting to play multiple streams at the same time, sampling rates and bit depths supported, etc..).

There are only two scenarios worth mentioning about Directsound or MME failing:
The old "AC97"-compatible hardware/drivers that used to resample with low quality 44.1Khz signals to 48Khz, and the newer "MME waveout recording resampling with low quality, even when unneeded" in Vista and 7 (which was patched some time ago).

That's as much about quality that any Windows user should care when using the different audio APIs (Other than when requiring bit-for-bit signals)

The possibility exists that DS can be bit-perfect, but it's less likely. It is likely that if there is a positive difference, it will be on the side of ASIO / WASAPI.

As I said, whether there will be a practical difference in every - indeed any - system depends on many factors that are hard to rule in or out.

This post has been edited by item: Oct 8 2012, 23:12