Analogue output glitches at 44.1, 48kHz with built-in Windows driver

[danielpieczko]

App: xk_316_mc
Config: 2AMi8o8xxxxxx_winbuiltin

Using ASIO4ALL with the built-in Windows driver, there are glitches in the analogue outputs at 44.1 and 48kHz. At 44100, all channels show glitches; at 48000 the glitches were seen on channels 1, 2, 4, 5, 7

The glitches appear to be periodic, with the audio analyzer software detecting approximately proportionally more as the duration increases.

Zip containing a stereo WAV recording of this issue and a USB protocol capture (not captured and recorded simultaneously): sine-glitches-48k.zip (since this was recorded at 48k, the glitches are only on channel 1, not channel 0)

This issue is very similar to #107 which went away when ASIO4ALL was upgraded to the latest version.

[danielpieczko]

There are lots of parts of the recording that look like they have missing samples:

[danielpieczko]

In ASIO4ALL, the buffer was set to the maximum of 2048 samples.

[danielpieczko]

This glitching depends on the version of ASIO4ALL being used.

| Sample frequency | ASIO4ALL v2.15 | ASIO4ALL v2.14 |
|----------------------------------------------------|
|        44100     |   glitches     |    good        |
|        48000     |   glitches     |    good        |
|        88200     |   good         |    good        |
|        96000     |   good         |    good        |
|       176400     |   good         |    glitches    |
|       192000     |   good         |    good        |

[danielpieczko]

The USB data was captures from host to device and converted into a WAV file: this showed glitches on the same channels as the analyzer detected, so this looks to be a host driver issue, rather than a problem with our board or software.

[xross]

At 44100, all channels show glitches

Are the glitches on the channels all at the same time or at different channels a different times?

[danielpieczko]

Are the glitches on the channels all at the same time or at different channels a different times?

They are all at the same time.

[danielpieczko]

This issue has also been seen in I2S loopback testing, where the ramp samples appeared to be divided by 2 at 44100, and 3 at 48000.

Failed: Failure at sample rate 44100
Channel 0: discontinuity (samples 24576, 12291 do not differ by 3 but by -12285) (sample 13144)
Channel 0: discontinuity offset 0: 12291 12294 12297 12300 12303 12306 12309 12312

Sample 24576 / 2 = 12288, and adding the ramp step of 3 gets to the next sample received of 12291 (where 24579 was expected).

Failure at sample rate 48000
Channel 0: discontinuity (samples 18432, 6147 do not differ by 3 but by -12285) (sample 11118)
Channel 0: discontinuity offset 0: 6147 6150 6153 6156 6159 6162 6165 6168

Sample 18432 / 3 = 6144, and plus 3 gives the next sample received of 6147 (where 18435 was expected).

[danielpieczko]

I retried this with the latest ASIO4ALL release v2.16 - the behaviour is unchanged from v2.15.