NXT sound sensor characteristics

[Claude Baumann]

We've worked some hours to do some interesting experiments with the NXT sound
sensor in order to learn more about the sensor characteristics:

Please consult: http://www.convict.lu/Jeunes/RoboticsIntro.htm

... and help discussing.

Claude Baumann

[Brian Davis-3]

In lugnet.robotics, Claude Baumann wrote:

> We've worked some hours to do some interesting experiments with the NXT sound
> sensor in order to learn more about the sensor characteristics:
>
> Please consult: http://www.convict.lu/Jeunes/RoboticsIntro.htm

Very nice. I'll have to dig through it again in detail, but I didn't find
anything interesting at 3 kHz... perhaps because I didn't go that high (oops).
To figure out the frequency response I did it manually: I borrowed a signal
generator and speaker from the university, as well as a sound meter, and then
simple manually cranked up the frequencies on the signal generator, noting what
readings the sound sensor returned in dB and dBA mode, as well as the dB level
displayed on the meter. The summarized results (over a lower range of
frequencies) are here:

http://www.brickshelf.com/cgi-bin/gallery.cgi?i=3348152

My conclusion was the sensor can be used to distinguish fairly pure tones below
roughly 1 kHz (below 1 kHz, there's a nice distinguishable difference between dB
and dBA modes), but not above that. I didn't go up to very high frequencies,
because among other things I was originally trying to get one NXT to distinuish
the tones played by an NXT or RCX, the the sound production at very high
frequencies with the sources is very weak.

More annoying, although I can get an NXT to reliably detect *pure* tones in the
proper range ("is this tone above or below 1 kHz?"), it seemed hopeless for
"normal" sounds that are really a very messy combination of waveforms. In other
words, having one NXT "listen" to another just didn't work out well.

Using the DSP module is a very nice way to do this, by the way. I'll have to dig
through this some more.

--
Brian Davis

[Claude Baumann]

Thanks Brian
>snip.. The summarized results (over a lower range of
> frequencies) are here:
>
> http://www.brickshelf.com/cgi-bin/gallery.cgi?i=3348152
>
> My conclusion was the sensor can be used to distinguish fairly pure tones below
> roughly 1 kHz (below 1 kHz, there's a nice distinguishable difference between dB
> and dBA modes), but not above that..snip

Your excellent graph shows that you tried to maintain the sound pressure
constant. We did the inverse in the last graphs, where we tried to draw the
curve for a constant NXT sensor value.

We also repeated the experiment, and this time, besides the measurements, we
payed attention to the sensation of loudness... and we were amazed to see that
around 3kHz, the sound sometimes has to be louder, sometimes weaker, in order to
maintain the constant 20 NXT-dBA. We sweeped over the range and the value droped
and grew at neighbour frequencies.
This leaves us puzzled. It would be a great thing, if someone verified this.

[Brian Davis-3]

In lugnet.robotics, Claude Baumann wrote:

> ...you tried to maintain the sound pressure constant.

Since I was doing this all manually, it was by far the easiest. I realized that
a better way would be to vary the signal generator level at each frequency, but
often when trying that I simple "maxed out" the reported level from the sound
sensor at some frequency, and wanted to do that as little as possible.

> we were amazed to see that around 3kHz, the sound sometimes
> has to be louder, sometimes weaker, in order to maintain the
> constant 20 NXT-dBA. We sweeped over the range and the value
> droped and grew at neighbour frequencies.

That sounds like resonance. Perhaps in the sensor response, but at 3 kHz the
wavelength is down around 10 cm, meaning a half-wavelength at 5 cm is getting
very close to the physical dimensions of the sensor. Did you check for physical
resonance effects, perhaps within the casework?

> This leaves us puzzled. It would be a great thing, if someone
> verified this.

I'd love to, but don't have the signal generator right now. I'm very tempted to
get something like this:

http://www.pasco.com/featured-products/xplorer-glx/index.cfm

As it would satisfy nearly all of my datalogging needs when they go beyond the
NXT (higher resolution).

--
Brian Davis

[Claude Baumann]

> That sounds like resonance. Perhaps in the sensor response, but at 3 kHz the
> wavelength is down around 10 cm, meaning a half-wavelength at 5 cm is getting
> very close to the physical dimensions of the sensor. Did you check for physical
> resonance effects, perhaps within the casework?

Excellent suggestion!!! We'll open one sensor and try once again.

>...I'd love to, but don't have the signal generator right now. I'm very tempted
>to get something like this:

>http://www.pasco.com/featured-products/xplorer-glx/index.cfm

B.t.w. Vernier also has a remarkable device:

http://www.vernier.com/labquest/

Someone else interested in repeating the experiment? Philo?

Claude

P.S. I added a graph to the page, where the sound pressure is maintained
constant.

[Philippe Hurbain]

> Someone else interested in repeating the experiment? Philo?

I would be interested, but I have no audio equipment (apart from a function
generator)

Very interesting data BTW!

Philo

[Brian Davis-3]

In lugnet.robotics, Claude Baumann wrote:

>> ...I'd love to, but don't have the signal generator right
>> now. I'm very tempted to get something like this:
>
>> http://www.pasco.com/featured-products/xplorer-glx/index.cfm
>
> B.t.w. Vernier also has a remarkable device:
>
> http://www.vernier.com/labquest/

I've looked at that, but honestly got interested in the Pasco product first just
due to who's catalog arrived at my house first. They look almost identical in
functionality except the LabQuest has a stylus and more memory, while the
XplorerGLX has I2C sensors (or seomthing like them; it seems they transmit
digital information back the main unit, while the Vernier sensors are analog and
AtoD is done on the main unit). Is that correct? Do you use a LabQuest and do
you have any more information, reviews, comments etc about it? It's not like
this is a product reviewed by Consumers Reports or something, so I'm finding it
hard to get 1st-hand knowledge.

--
Brian Davis

[Claude Baumann]

> I've looked at that, but honestly got interested in the Pasco product first just
> due to who's catalog arrived at my house first. They look almost identical in
> functionality except the LabQuest has a stylus and more memory, while the
> XplorerGLX has I2C sensors (or seomthing like them; it seems they transmit
> digital information back the main unit, while the Vernier sensors are analog and
> AtoD is done on the main unit). Is that correct? Do you use a LabQuest and do
> you have any more information, reviews, comments etc about it? It's not like
> this is a product reviewed by Consumers Reports or something, so I'm finding it
> hard to get 1st-hand knowledge.

Nop, I have seen it. I will try to get more information about it.