Fast response air temperature measurement

till

It probably easier to treat your test volume as a Helmholtz bottle; the resonance frequency depends on the velocity of sound in air, which is temperature dependent.

You can track the resonance by monitoring the phase relationship between the excitation and the response of your resonant volume. For a high Q resonanace the slope is pretty steep around resonance.

-- Bill Sloman, Nijmegen

still

an

to

Bill, I thought of using a resonator, but I was worried this wouldn't give the 10ms response time. (You've got to get the air into the resonator and then worry about the temperature of the resonator itself.) I was thinking more along the lines of a transducer and detector as posed by John D. above. I have no idea how much of a phase shift you would get for a one degree change in the temperature.

George H.

ill

I was thinking more of running a linear regression, which can be done if the data is anti-logged (I suppose). I haven't really done the math on this. Using a running set of samples, as the thermistor settles, the noise becomes less of an issue. [Obviously a simple average would slow down the response, so the curve fitting is the key.]

The Helmholtz resonance doesn't directly depend on the speed of sound. It's a mass-spring resonance, not an organ pipe resonance. If you believe the ideal gas law, the density dependence should cancel out, leaving the temperature dependence, meaning that it will work as a thermometer.

Unfortunately it needs a closed container, so it doesn't help the OP's problem--he really needs to do it in the open air. It needs something like a pair of piezo elements separated by a known distance, with mounts designed to reduce the conducted acoustic wave, and an XOR phase detector. Filter and square up the received signal, then beat it against the transmit signal, and the PD will give you an output voltage proportional to phase, i.e. to delay, which will go as 1/T.

It's tempting to use a PLL and read out the frequency, which would give the temperature directly, but that'll need a lot more calibration work to get rid of the phase shifts of the transducers.

Cheers

Phil Hobbs

Hi Bill,

I did think of that, also bouncing it off one wall. Unfortunately I don't trust the volume to be sufficiently consistent. Although if I can use something like that to actually *measure* the volume that would be interesting in its own right, will look it up.

Thanks,

Actually it *is* a closed container, but not a very rigid or accurately defined one. (Sorry to dribble out these scraps of information, don't mean to!)

That was the sort of thing I had in mind (but you have thought it through better!)

Thanks,

I would look at using a single self-resonant piezo bender element mounted in a Helmholtz cavity with a small opening to the atmosphere, then just read the frequency off with a counter. This could be really, really inexpensive, in fact it should work using an ordinary piezo beeper.

Whoops. 1/sqrt(T).

Cheers

Phil Hobbs

r to

"Whoops. 1/sqrt(T)."

That sounds better. So at 300K a one degree temperature change will only give you a one part per thousand phase shift? (Did I do that right?) If you want 10 ms ~ 100 Hz, you need a drive freq of 100 kHz?

OK don't quote me on the above. I'm probably missing something.

George H.

still

The thing about different ways of extracting the information is that they don't make the original data any better.

Sometimes a totally impracticable thought experiment makes it easy to see a weak point in a scheme.

-- Bill Sloman, Nijmegen

in

tor to

ou

ch

ing

nts

ge

Opps, cancel that, The sqaure root will only give you a factor of 1/2 in reduced sensitivity. So 1/600 change in frequency for a one degree change at 300K. And the frequency number is only true if you wanted to measure frequency by counting the number of cycles for 10ms. I think the PSD idea of Phil H. will let you meausure changes in times of order one period if your signal is strong enough. If you have to signal average....

Say John you mentioned some container. How big is it? If it's small enough you might be able to lock on to one of the acoustic modes.

George H.

Whatever sensor you go with, you can decrease the system's response time if you can arrange to move the air across it with a fan.

Other than some exotic methods (acoustics, etc), you are actually measuring the temperature of a sensor that has thermal mass and must gain/lose heat to reach equilibrium with the surrounding air. The smaller the temp difference from equilibrium, the slower that natural convection will assist in this transfer.

Get a small (CPU cooler?) fan.

Ah, I see where you got that 100kHz from now! I was assuming looking at the phase all along - there will probably be less than a wavelength between Tx and Rx anyway.

Should be pretty strong with the transducers facing each other. Gap would really need to be Say John you mentioned some container. How big is it? If it's small

It's A few litres, but a bit variable and floppy unfortunately.

ll

Yeah just forget the 100kHz number. 10kHz will be fine. (There are cheap audio speakers and mic's at that frequency too.) 10 mm sounds too close? With a velocity of 300m/s that's only 30 us. and then changes of one part in 600 mean time differences of only 60ns (or so.) I think you'll do better if you can move source and detector further apart. But perhaps there are other constraints.

George H.

This thread should be titled, "Fast air response temperature" measurement, i.e., "FART" meaasurement. ;-P

Cheers! Rich

Pre or post ignition?

Omage sells some non-insultated thermocouples that are 36 gauge wire fro about $5 each.

Thanks Steve. I looked at their site again and found some that I didn't find before.

There is a 0.025mm one with 50ms response in still air. That may be fast enough for what I want (since there should be *some* air movement).