Firmware q2f10: Hardware Test : Audio: low volume

[earias]

The volume of audio test is more lower than older versions

[Quozl]

Thank you for your problem report.

I have tested Q2F10 and Q2E48 with volume control set to maximum, using a sound level meter at 5mm distance. Both versions produced 81.2 dBA for the sound command (the startup sound), and 68.1 dBA for test /audio sweep.

Please advise your method of measurement, measurement results, firmware versions, and volume level?

[earias]

Q2E45

[Quozl]

Thanks.

I've tested Q2E45 against Q2E48, and the older version yields 82 dBA during test /audio. This is a 14 dBA drop for the Open Firmware audio test only. It does not affect the startup sound, Linux or Sugar.

I've briefly checked for the changes between Q2E45 and Q2E48 that may have caused the lower volume. This period covers development of manufacturing tests for XO-1.5 hardware, and changes to shared code. During these changes there were refinements to the method used for producing sine waves, but no change to the volume settings.

XO-1 is no longer in mass production, so the manufacturing test for audio is not required by OLPC. That leaves use by deployments.

I propose that:

• those using test /audio on XO-1 be taught that the test is less loud after Q2E45, and that this change has no effect and no impact,
• this ticket be closed.

Please respond. In particular I'm interested in why you consider this to be high priority. I must be ignorant of something.

[earias]

ok consider for future versions

[Quozl]

No, we will not consider fixing this for future versions given the information available so far. There is no impact. I shall close the ticket. Thanks for your bug report.

[reuben]

This should be corrected in future versions of all hardware versions. This is both a mfg test and field testing mechanism.

[reuben]

Test using q2e41. Test using q2f11.

There is significant difference in output volume and perhaps even a different sweep being played. This is a regression and should be reverted to the settings used in q2e41.

[reuben]

Any progress here?

[Quozl]

Please test http://dev.laptop.org/~quozl/q2f11je.rom

The calculation of the amplitude of the sine wave was using a single length 32-bit intermediate result, which overflows.

Affects XO-1 and XO-1.5 only. Doesn't affect XO-1.75.

Fixed in svn 2994.

The regression was introduced in svn 2260, source file isin.fth, the last line of set-period which calculates fstep. Changing this code back to what it was before svn 2260 fixes the symptom, but there is a loss of precision due to the order of operations.

(On XO-1 and XO-1.5, the */ operator is converted to >r * r> / by the FCode tokeniser. The */ operator is left as is on XO-1.75 because the audio test is not FCode tokenised, so the intermediate result is a double length 64-bit value. Mitch, perhaps the tokeniser needs a fix?)

[reuben]

Tested okay by earias.

[Quozl]

Thanks. Review changes by Mitch also made. Will be in Q2F12.

[Quozl]

Is in Q2F12.

[Quozl]

Is in Q3C07.