test XO-1.5 audio recording levels

[dsd]

Need to run a side-by-side comparison with XO-1 to check that XO-1.5 microphone settings are adequate and that the volume of recorded sound is similar.

[mikus]

How did you get the software on F11 to perform sound recording? For me, Record-64 doesn't seem to record anything, and Measure-28 doesn't even launch.

[Quozl]

triage. changed from defect to task.

related to #9526. related to #9639. related to #9640.

mikus: see #9526 for a method to record sound suitable for technical testing.

(all: suggestions welcome for test method. tools available to me include a sound pressure meter, a powered stage speaker, an SM58 microphone, several audio sources, and a cathode ray oscilloscope. i do not have an anechoic chamber.)

[mikus]

In os42 the Record activity *does* record audio.

On playback, the volume sounded weak to me.

[triagebot]

changed by irc user Quozl:

[Quozl]

Ticket moved out of 1.5-software-final to 1.5-software-update as a result of a software manufacturing release triage meeting. Per ed, dsd, cjb, reuben, quozl.

[Quozl]

Tested using a recorder (a woodwind musical instrument) and the ALSA utilities arecord and aplay.

On XO-1 build 802 and XO-1.5 build os200.

The laptops were set side by side in a quiet location, equidistant from the sound source. The output volume control on each laptop was raised to maximum using the volume up key on the keyboard.

Commands used in Terminal:

arecord --vumeter=mono --rate 44100 /tmp/test.wav
aplay /tmp/test.wav

There was no obvious difference in the volume meter readings during recording of the sound. (There was a pulse of energy at the start of the XO-1 recording in the silence before the musical instrument was not active.)

There was no audible difference in the recordings when played back.

Conclusion: the signal path from air through to user space via ALSA is clean and at near identical levels between the laptops.

Hardware, driver, kernel, ALSA are all good. Anything we find in the Record activity is probably unrelated.

[cjb]

James, are you happy with the current 1.5 build default recording levels, or do we need to change them?

[Quozl]

Tested using a recorder (a woodwind musical instrument) and the ALSA utilities arecord and aplay.

On XO-1 SKU38 build 802 and XO-1.5 C1 build os119.

The laptops were set side by side on a concrete slab in open air with a light wind, equidistant from the sound source. The default volume setting was used.

A sound level meter was placed between the laptops, also equidistant from the sound source.

Commands used in Terminal:

arecord --vumeter=mono --rate=44100 /tmp/test.wav ^{C aplay /tmp/test.wav}

The musical instrument was played from a distance of about 70cm, at a volume to achieve a display of 90dBA on the sound level meter.

Six test runs were made, using various pitches and cadences. The sound level meter and the laptop display were monitored. Each recording run was played back on one laptop after another.

There was a slight difference between the laptops of about 5% in the arecord volume meter readings.

90 dBA on the sound level meter was represented as 35% level on the XO-1 MP, and 30% level on the XO-1.5 C1.

There was no significant difference in the recordings when played back.

Conclusion: the levels are adequate.

[cjb]

Thanks. Do you know whether gstreamer and Record use the same levels?

[Quozl]

I cannot tell if gstreamer, or therefore Record, use the same levels, but the following test suggests they do.

gstreamer cannot properly record audio on XO-1.5 by default, so a comparison is not as trivial as with arecord. The following pipeline generates incorrect audio on the XO-1.5:

gst-launch alsasrc ! waveenc ! filesink location=/tmp/test.wav

However, the plughw workaround works:

gst-launch alsasrc device=plughw:0,0 ! waveenc ! filesink location=/tmp/test.wav

Eight brief notes, and then a two second burst were played on the musical instrument. The recordings were then analysed using Audacity 1.3.11 on a desktop system.

The XO-1.5 C1 recording suffered from a DC offset (#9851) that had to be removed before rms volume analysis.

The Contrast Analyzer for WCAG 2 compliance was used to measure the rms volume of the two second burst against the two second silence that followed.

model	burst volume	silence volume
XO-1.5 C1	-27.4 dB	-62.4 dB
XO-1	-25.1 dB	-60.0 dB

Conclusion: there is a 2 dB difference; the XO-1 generates a recording that has slightly louder than the XO-1.5.

Risk: there may be significant manufacturing variability that more than compensates for this observation. I've no idea what the manufacturing test criteria are.

[mikus]

Did a test with side-by-side XO-1.5, XO-1, me reciting numbers:

XO-1.5: Custom os118 (with kernel 20100409.1311), Record-66
XO-1: Virgin build 802, Record-59

Using the 'Audio' panel, both XOs recorded what I said. On clicking on the recoding (icon in bottom tray), the audio playback at both XOs had equal volume levels. Since the problem in #9641 is described as "XO-1.5 recording levels", I would say that that VERY_SPECIFIC description of the problem has now been fixed.

---

But there is more to the XO than just doing audio record/playback within the Record Activity. Although both XOs recorded using the 'Video' panel, the XO-1.5 video playback (in Record) had no sound (the XO-1 had both video and sound).

And when the recordings (both from 'Audio' and from 'Video') were put in the Journal, only the XO-1 was able to play them back (both 'Audio' recording and 'Video' recording played o.k. - both launched Browse to output the respective material). On the XO-1.5, the 'Audio' recording gave an error, and the 'Video' recording appeared to be without sound. [I could not tell how well the video information was being shown -- on the XO-1.5 the picture (taken indoors) was so dark that I could not tell what it was supposed to show -- whereas the picture on the XO-1 was much much bigger, and bright enough to show things clearly).

For recording, I expect the XO-1.5 target audience to use the Record activity, rather than CLI commands. Unless the Video/Audio performance of Record on the XO-1.5 matches the performance on the XO-1, I expect XO-1.5 users would be disappointed.

[Quozl]

Mikus, aren't your additional Record activity concerns already covered by other tickets?

[mikus]

Replying to Quozl:

Mikus, aren't your additional Record activity concerns already covered by other tickets?

Yes, they are.

I think of myself as the sort of person who looks at the forest, rather than looking at the trees. While attacking a complex situation may well require solving one problem at a time, my fear is that when that single problem has been tamed, the bean counters will say "enough" and "leave the rest be".

The point of view that I often consider is "How useful is the XO ?" When I see many comments devoted to "audio recording levels", I'm concerned about whether attention is being drawn away from "having video recording match the quality on the XO-1". So I take the opportunity to say that solving "audio recording levels" helps, but does not by itself solve "making the XO useful".

---

In your comment 11, you mention the "DC offset". One of the things that in my mind makes the XO 'useful' is that one can vary the voltage (or is it the resistance?) presented at the microphone input of the XO-1, and thereby use the XO as an *instrument* for recording / remote sensing all sorts of physical phenomena (needing only a simple circuit plugged in to the microphone jack). Again, I fear that measuring decibel levels addresses a particular tree, when there be many kinds of trees in the forest - for instance, think EKG.

[cjb]

Thanks for the testing, let's close this now then.