Wind sensor fall response time.

Questions about Modern Device and JeeLabs Sensors
Post Reply
Cabe
Posts: 7
Joined: Sat Feb 25, 2012 8:48 pm

Wind sensor fall response time.

Post by Cabe » Sat Feb 25, 2012 9:29 pm

Hello!

Project background : Midi Ocarina/Flute/Recorder/Breath controller.
Aiming to emulate a 4 hole pendant style initially, more interesting Ocarinas to come later :)

I have been fiddling with the Wind Sensor for the last couple of days and have been impressed with the sensitivity and responsiveness of this little sensor. Or at least I have when the wind-speed has been increasing.

I am seeing lag on the sensor returning to "0" wind of up to a second. I suspect this isn't a problem in its intended environment of meteorological survey but as you can imagine it limits the creativeness of a musical instrument. Notes hang around much longer than they should, and "tonguing" (that tut-tut sound when you say tongue) is only possible to start notes (once again it shows just how good the rise time is.)

I am currently hanging it off the 5v pin from my Arduino (clone from DF Robotics). My first instinct is that it can't supply enough current so it is unable to come back up to temperature quick enough, but who knows it wouldn't be the first time I was wrong :)

Additional Query - Hopefully this sensor will serve my purposes, if so would it be possible to buy the "snap off sensor" part separately? I would love to be able to replace ones that fail due to being in the general electronics destroying environment of hot breath?

Cheers
Gregg
------------
Is this soldering iron still on? OW!......Yeah its still on. ;)

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Sat Feb 25, 2012 10:09 pm

I am currently hanging it off the 5v pin from my Arduino (clone from DF Robotics). My first instinct is that it can't supply enough current so it is unable to come back up to temperature quick enough, but who knows it wouldn't be the first time I was wrong :)
You got this inverted. The sensor is easily able to get hotter - because it is heated. The problem is that it can't get rid of heat with anything but ambient cooling. This goes by the name of thermal inertia. The best anemometers use wires - but they are fragile - but for your app - I think they will work well.

We're working on this - send me some email and I'll send you a beta / proto version of one for testing. They respond almost instantly. In theory the best thermal anemometers have freq responses that top off well over the audio range - so they could potentially be used as microphones - an intriguing possibility but probably useless as anything but a curiosity.
if so would it be possible to buy the "snap off sensor" part separately? I would love to be able to replace ones that fail due to being in the general electronics destroying environment of hot breath?
You can cut off the "fingers" and use three wires to attach them to the electronics part of the sensor.

I'm interested in developing a breath controller too - so I'll be happy to see how you make out. I'll send you a pressure sensor to try if you want. I think you can merge the data from the sensors profitably.


Paul

Cabe
Posts: 7
Joined: Sat Feb 25, 2012 8:48 pm

Re: Wind sensor fall response time.

Post by Cabe » Sun Feb 26, 2012 3:51 am

Paul,

That sounds awesome!

I had looked at the DIY hot-wire anemometer using a small torch bulb, but anyone who has seen a roadie* load a van knows if you can make it solid state, you really should :) I had also noticed the snap off section which is why I was curious about it, for future iterations I didn't really want to buy a whole new sensor if I could just replace small thermistor units.

I have the device doing everything separately (Input Matrix, Breath Sensor, Sending Midi codes, etc) I still need to implement an LCD/Rotary/Menu system for setting up the "instrument" side of things but as that's well known it was a low hanging fruit to deal with.

After that its a "simple" matter of casing it up :)

I will definitely keep this thread up to date as to how I get on with the final stages of the build.

Many Thanks
Gregg.

*I'm a video tech, I load a van like its last orders at the bar :)
------------
Is this soldering iron still on? OW!......Yeah its still on. ;)

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Sun Feb 26, 2012 1:48 pm

Greg
I had looked at the DIY hot-wire anemometer using a small torch bulb, but anyone who has seen a roadie* load a van knows if you can make it solid state, you really should :) I had also noticed the snap off section which is why I was curious about it, for future iterations I didn't really want to buy a whole new sensor if I could just replace small thermistor units.
Actually the (broken) light bulbs are fairly robust - the filament is coiled (in some cases anyway) and supported by flexible leads. I still haven't ruled out going that way for a next gen, but I haven't found a neat, quick way to break the light bulbs. Also the other solution is much lower powered. We'll see how fragile they are - there may be a gnarly materials science problem to solve. The real players in this field (read $300 and up) use platinum wire but it requires some tricky welding equipment to put together, disregarding the ridiculous cost of fine platinum wire.

I have the device doing everything separately (Input Matrix, Breath Sensor, Sending Midi codes, etc) I still need to implement an LCD/Rotary/Menu system for setting up the "instrument" side of things but as that's well known it was a low hanging fruit to deal with.

After that its a "simple" matter of casing it up :)

We've got a laser cutter here - but it might be more of rep rap project. Probably pvc pipe is the first stop.

I will definitely keep this thread up to date as to how I get on with the final stages of the build.

Cabe
Posts: 7
Joined: Sat Feb 25, 2012 8:48 pm

Re: Wind sensor fall response time.

Post by Cabe » Sun Feb 26, 2012 4:14 pm

All excellent points, I did briefly consider stripping a MAF meter from a car as they also use a similar hot wire system. Sadly are are all a little too "robust", as engine bays are some of the nastiest places to shove tech, they all appear to need 12v and have much thicker wires. Also I wasnt putting my lips on anything that might have been a spare part :)

As for the initial query I have performed some actual numerical tests (not just a PWM LED and a MK1 Eyeball...), it seems to hold up my theory, but its about as truly scientific as I can make it without calibrated lab toys to generate/register wind speeds.

http://cabe.fatboylan.net/midi_ocarina/teraterm_out.csv - Raw Data from the Out port.
http://cabe.fatboylan.net/midi_ocarina/teraterm_rv.csv - Raw Data from the RV port. (although there aren't enough samples on that one)
http://cabe.fatboylan.net/midi_ocarina/Wind_Sensor.ods - Spreadsheet with pretty graphs!
Attachments
IMAG0212.jpg
Test "mouthpiece" rig.
(204.12 KiB) Downloaded 19063 times
IMAG0211.jpg
Test "mouthpiece" rig.
(188.58 KiB) Downloaded 19063 times
10sec_tongue.jpg
Graph - 10 seconds with increasing and decreasing tonging force.
(76.56 KiB) Downloaded 19063 times
3sec_SmoothOn.jpg
Graph - 3 seconds smooth breath out.
(63.81 KiB) Downloaded 19063 times
3sec_HardOn.jpg
Graph - 3 seconds fast attacking breath out.
(57.88 KiB) Downloaded 19063 times
------------
Is this soldering iron still on? OW!......Yeah its still on. ;)

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Sun Feb 26, 2012 6:58 pm

Greg,

The data looks good and is about what I expected to see. If you think about the downward curve, it's about the same deal as an RC curve where the the resistor is discharging the cap, and the closer it gets to ground (or whatever discharged is) the slower it goes since the discharge current is based on the voltage differential, which is also dropping. In the wind sensor's case the heat loss is based on the delta T between ambient. For what it's worth - it's easy to mod the sensor to run hotter, which yields smaller output values for a given amount of wind but somewhat peppier response curve as the delta T is larger. If you look at the schematic, there is just a voltage divider that sets the operating current (hence temperature at the thermistor. You could even put a pot or resistor sub box in for the bridge - and just dial up a new value. That's what we did when we designed it. We decided that 25 mA at quiescent seemed like a good trade-off between the various parameters. It's also more immune to ambient temp changes if it's run hotter. The downsides are just more power wasted, and a dampened response curve. If you think about the power that's required to keep it far above ambient say 50 degrees C, the added cooling from a given amount of air movement is going to make less difference in the balance.

Paul

The links on your charts down work for some reason. - They seem to 404

Cabe
Posts: 7
Joined: Sat Feb 25, 2012 8:48 pm

Re: Wind sensor fall response time.

Post by Cabe » Mon Feb 27, 2012 5:56 am

Seems my webspace is playing about at the moment.

http://cabe.multiplay.co.uk/midi_ocarin ... rm_out.csv - Raw Data from the Out port.
http://cabe.multiplay.co.uk/midi_ocarin ... erm_rv.csv - Raw Data from the RV port. (although there aren't enough samples on that one)
http://cabe.multiplay.co.uk/midi_ocarin ... Sensor.ods - Spreadsheet with pretty graphs!

Should work much better, if you want to have a look at the raw recordings.
Thanks for the interesting suggestions on how to increase the speed of the bounce-back, I guess with a higher temp the s-curve on the rising side will be more pronounced?

I think plan A will be to add some signal processing on the sensor data rather than just directly coupling it to the volume (0-127 for the midi values) by using the MidiVolume=(AnalogRead(A0)/8) code I have at the moment.
------------
Is this soldering iron still on? OW!......Yeah its still on. ;)

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Mon Feb 27, 2012 4:44 pm

Greg,

I think signal processing and merging data from both sensors is the way to go too.

Paul

Cabe
Posts: 7
Joined: Sat Feb 25, 2012 8:48 pm

Re: Wind sensor fall response time.

Post by Cabe » Mon Feb 27, 2012 7:08 pm

Still trying to wrap my head around how to ensure it still have that "mouthfeel" of a real instrument, I.E. they provide resistance but they aren't completely sealed.

Considering some kind of Pitot style tube which works on a differential pressure from direct and static pressures in the breath stream. Hummm, I wonder if they do scale model sized ones.....
------------
Is this soldering iron still on? OW!......Yeah its still on. ;)

sethcim
Posts: 3
Joined: Sun Aug 19, 2012 7:45 am

Re: Wind sensor fall response time.

Post by sethcim » Sun Aug 19, 2012 7:55 am

paul wrote: We're working on this - send me some email and I'll send you a beta / proto version of one for testing. They respond almost instantly. In theory the best thermal anemometers have freq responses that top off well over the audio range - so they could potentially be used as microphones - an intriguing possibility but probably useless as anything but a curiosity.
How is the new version progressing? Is it likely to make it to the store soon?

Thanks!

Thot
Posts: 2
Joined: Wed Nov 28, 2012 4:33 pm

Re: Wind sensor fall response time.

Post by Thot » Wed Nov 28, 2012 5:05 pm

Hello Paul and Cabe,
I just purchased the little guy last week.
I too am trying to make a MIDI wind controller with the wind sensor and Arduino, more like a flute than an ocarina but the idea is the same. I power the sensor with the Arduino +5V and ground it on the Arduino and I get the Out signal on A2.

I am also finding that the response is very quick and the resolution quite good but the decay time way too long. If I take the values at face time and turn it into notes, the instruments keeps playing for several seconds after I stop blowing.
I am attaching two curves I obtained, the first by blowing four "pulse" breaths and the second by blowing with a sharp "attack" constant sustained breath and sudden stop (I have the spreadsheeet if of interest but it's probably overkill).

I'd like to know if you have made progress in this respect.
Paul, I'd love to beta-test/purchase your next generation sensor if you have it.
Cabe, have you found a software solution? I am playing with constrain() and map() to go from sensor values to midi velocity but the slow decay curve is proving to be a challange.

Thank you and keep up the good work!
Attachments
Breath Sensor Response.jpg

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Fri Nov 30, 2012 1:18 pm

Thot,

I have a few comments. As I said above the issue with slow response time is that the sensor heats up by force and has to cool off naturally. Here are some possible solutions.

1) Run the sensor at a higher temperature - if you don't mind burning off some more current consumption. Then the sensor is going to cool faster since the sensor is farther above ambient. This can be done by changing the bridge voltage for the reference leg on the schematic. It also helps the sensor be less sensitive to ambient temp so two big wins at the expense of increased power usage.

Try putting a 1M trimpot across R6 (165K) (check the pot first to make sure it's at maximum) then SLOWLY dial it down and you'll see the voltage of the output go up as the op amp has to heat up the sensor more to match the resistance of the sensor. If you do this too much you'll fry the thermistor though, so monitor the input current of the sensor. There is definitely enough current available through the transistor to cook the thermistor.

We'll try this too - maybe we could offer a special "high temp" version of the sensor - I've made some in the past for customers.

2) Use a smaller sensor element - this is something we're working on. Unfortunately some of our research got lost in the move and hasn't surfaced yet. I'm buying more sensor materials.

3) Just solve the problem in software by using the peaks in your graph and as suggested in the Arduino forum.

4) Accept the slow response as ADSR (see electronic music for the uninitiated)

5) Try our MPXV pressure sensor mounted to a mouth piece - with possibly a small hole in the orifice. It responds to pressure not wind but it's all about the same, with the right orifice / setup. You could also merge both sensors.



There are more I'm sure but I'm out of ideas just now - and am out of time.
Hope that helps.

Paul

Thot
Posts: 2
Joined: Wed Nov 28, 2012 4:33 pm

Re: Wind sensor fall response time.

Post by Thot » Fri Nov 30, 2012 3:20 pm

Paul,
thank you for your answer and your suggestions.
I have gotten some good software advice from the Arduino forum folks and that may help when using the sensor as an on/off trigger. I will probably be able to translate sensor values into note velocity for the ascending part of the curve, but for a wind instrument the ADSR curve is not good enough.

I was thinking that running at higher temperature will help (precisely because of the difference with ambient temperature) but I was thinking that I may be able to achieve that by powering it with 9V or 12V?
I am reluctant (because I don't have the skill or knowledge) to "hack" your device (besides, it's a beauty, I would feel sad if I destroy it).
Having said that, I can try the 1M trimpot if I get desperate.
Do I solder the legs of the trimpot to the 2 solder joints across R6?
Where do I monitor the current, across R6 as well?
What value of current should be safe?

I was initially looking at pressure sensors but what I have in mind is more like a flute, where you blow across a tube and do not seal the mouthpiece with your lips (I believe, among other things, that commercial wind controllers use a pressure-type sensor as they do have you blow into a sealed mouthpiece) and I thought a true wind sensor would be Ideal (I still think it is, if I can "straighten" the down curve).

Other Ideas I had was to use one of your barometric pressure sensors (would it work?) or a simple microphone (they make a lot of noise = output) when you blow on them, but they may respond to ambient noise or if you knock on your instrument.
Also considering a straight temperature sensor, but I believe it will not be fast or sensitive enough and will suffer of the same problem, slow passive cooling.

Sorry for the naive questions and thank you for your help.

Keep up the good work!

Luca

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Sun Dec 02, 2012 12:12 pm

Luca,

One thing you could try with a pressure sensor is to use a tube with a constriction - but this is more in the clarinet or brass instrument model and not a flute.

As to soldering to r6 probably the legs a trimpot could be made to work. Just make sure the pot is at max, not min before powering up.

Monitor current in series with the supply pin (meter connects the circuit). It draws about 20 mA currently. (Increases with wind speed)
The thermistor is rated at 100mW so if say 3 volts is across the sensor, at 33 mA you are at the rated max of the thermistor.
Also long-term stability of self heated thermistors is unknown, so things may drift in the future.

Send me email off the forum and I'll send you some beta products to test that may work with your app.

Paul

Cabe
Posts: 7
Joined: Sat Feb 25, 2012 8:48 pm

Re: Wind sensor fall response time.

Post by Cabe » Tue Dec 18, 2012 9:40 pm

Hey all!

All good fun, I do intend to get back to this, I just poked my head in as this is on my list as my "Christmas Project" assuming I actually get any time :)

I will update as and when I make any progress, the high-temp suggestion was my P1.

Once again I need to thank MD for their excellent community and stellar service.
------------
Is this soldering iron still on? OW!......Yeah its still on. ;)

Epen99
Posts: 1
Joined: Mon Sep 09, 2013 11:54 am

Re: Wind sensor fall response time.

Post by Epen99 » Mon Sep 09, 2013 11:58 am

I have in mind is more like a flute, where you blow across a tube and do not seal the mouthpiece with your lips. :) :)


แทงบอลออนไลน์royal1688ดูผลบอลสล็อต

Bill sullivan
Posts: 3
Joined: Thu Feb 20, 2014 11:51 pm

Re: Wind sensor fall response time.

Post by Bill sullivan » Fri Feb 21, 2014 12:43 pm

I am also making a wind sensor arduino device. Does anyone have a simple snipit of code to show how to use the sensor? I'm not sure which output I'm supposed to be connecting too? The temp seems to be working but as others have commented it does have a longer sustain then expected. Are there any examples of how to get usable values from the sensor?

Bill sullivan
Posts: 3
Joined: Thu Feb 20, 2014 11:51 pm

Re: Wind sensor fall response time.

Post by Bill sullivan » Mon Feb 24, 2014 9:13 am

I purchased the wind sensor for a midi instrument I'm working on. when I first took the sensor out of the packaging there was small pin rail to insert into the bottom, while originally testing the sensor I used the pins pushed through the holes into the bread board to connect it to the arduino without soldering the pins to the sensor board. It made it impossible to test the device. I finally soldered the pins to the board and now it works flawlessly. I'm not even noticing the fall off time issue. there is a small ramp down but it's within three cycles and even if that was a problem a simple function to watch for a dramatic fall off would work.

Awesome product thanks, I will post my simple sketch that I used to visualise the sensor.

paul
Site Admin
Posts: 711
Joined: Mon May 12, 2008 4:19 pm

Re: Wind sensor fall response time.

Post by paul » Sat Mar 01, 2014 10:22 am

Glad everyone is working out this issue. I also urge wind sensor builders to try the MPXV - one of my friends built a instrument that works well with that. It's a pressure sensor, so the response if more symmetrical. One would need to use some kind of orifice to generate the required pressure.

Regarding the wind sensor - what is really required is a real hot wire, so that the mass is less and the wire has almost infinite surface ratio to mass, so that the heat disappears almost instantly. I have some working prototypes but the issue is really how to mount, protect, obtain the hot wires. Also as one of the participants in the thread mentioned - it's inconvenient to have to supply 12 volts and a lot of power.

Post Reply