Thermal pile

[yaddatrance]

If you're up for another mammoth post, I'll add my notes on painting a TEC. There's definitely a knack to that, so I wrote down my findings when I worked on it. It's recently posted again to alt.lasers too, cos someone asked there about this.

Definitely up for another post on this! This is interesting.

[The_Doctor]

Ok, here it be... 'Tis big, and out of context, I just saved it whole so I could see what I did long after I forgot.

I removed the damaged volume absorber. It wasn't stuck properly across the
back, just round the rim, which probably accounted for the >5% inaccuracy
across it. I used a carbon steel blade to shave off the glue and other crud
on the thermpile ceramic, cleaned with white spirit then acetone to finish.

I made a black paint out of 3 parts white spirit to 1 part yacht varnish.
(I'd tried matt finish aeroplane kit glue, it was useless. I couldn't get
any butyl acetate as thinner, and other solvents made it coagulate). The
varnish is gloss, but the high viscosity and adhesion allow a heavy loading
of carbon powder (activated charcoal, politely blagged from a brewers
supplier). I can't quantify the amount of powder, as it won't suspend
evenly in the liquid. I treated it as if it was very wet sand, painted on
the ceramic surface of the thermopile, such that vibration could make it
shiny as the particles settled after painting it as flat as I could get it.
I then shook it, keeping it on a plane surface, to even out the thickness
and consistency. I dried it with forced air from a 12V 80mm fan at very
close range. The forced evaporation (and the high-frequency agitation in
the fan airflow) made sure the carbon could not settle in time to allow a
shiny surface to develop in the dried film.

The result is a deep black, almost as good as soot deposition. It looks
like very fine velvet, like that paper which photographers use for extreme
light absorption. Its damage threshold is high, I had to focus 158 mW to
within 0.5 mm diameter before I could get any smoke out of it, and to
within 0.25 mm diameter before any blanching of the surface occurred.
Response time on the thermopile is good (I didn't get numbers for this
though), and the accuracy is good too, about 1% down in sensitivity off-
centre to edge, symmetrical around that centre. I've made a very basic gain
stage with a dual op-amp (LF412), with the first half making a new ground
about 2.5V above negative so I can get accurate through-zero offset tweak,
and headroom to measure up to 30 watts on a voltmeter's 32 volt range. I
don't know how much the carbon coating's varnish binder might vary with
wavelength, but probably very little given that it has very low reflection
now, and I think this system could be good for 1% accuracy providing the
beam is centred on the detector plate and spread just wide enough to be
below the damage threshold.

I'm calibrating it with the Lasercheck that I bought from Rupert Ward.
(Thanks for that, this is extremely useful), and I'm going to make some
PCB's for the circuit. I've designed them, but getting them made up
economically will be more of a challenge. I might make several, they'll be
a useful accurate DC gain stage with terohm input resistance and low
thermal drift, useful for many things, and small, and cheap.

One thing that bothers me about the thermopiles in Scientech heads: They
use 71 couples, instead of 127. Wouldn't they do better to use the 127
couple TEC's? That way they'd have a better coupling from the ceramic
plate, higher accuracy off-centre, and a higher voltage output, and small
differences between couples would average out better as more couples would
be used per unit of beam area. If I make my own thermopile sensors from
scratch, I'll be using the highest couple count and density I can get,
unless someone can convince me I shouldn't.

Sam said that the ceramic is a good heat conductor and evens out the transfer across many couples, so maybe the count isn't too important. I still like the higher count higher density types though, it means a tad less gain to get 1V/W, so the op-amp offset isn't multiplied into as big an error as would otherwise be.

[yaddatrance]

That's very cool information!

[Hemlock Mike]

Doctor --
Damn - your posts keep making me have to think This is the basic info from where we can start our own projects.
Thanks - Mike

[The_Doctor]

This is what I hope.. I like looking at the way problems are solved, and bringing old ideas to new things. Puts a perspective on the real cost, for one thing. If we can make our own meters and lasers, what justifies the thousands of $$$ when we have to buy them pre-built? Just looking at how a $$$ Scientech detector head is constructed is illuminating, there's nothing there that we cannot make with a few dollars and a lathe and a visit to local electronics and hardware shops. I'd prefer to pay for things that really cost because it's damn difficult to do them well, than merely paying top dollar to Keepers of the Sacred Key.

That's a jibe in the general direction of those who guard access to calibration references. At times I think they do this like high priests in a space opera, rather than scientists. It's the only thing that stops us all from making our own cheap valid references. ANYTHING that can erode this process is a good thing, so long as we don't use our new kit to lie with. Not that THAT can work, when anyone with similar skills can show the truth.

[marconi]

Dig that avatar Crow !!!

This thread reminds me of one I started over a year ago..
Im glad to see interest in this direction again.
I agree with you 100 percent.
We should all be able to measure our own devices.
At least for peace of mind.

[Hemlock Mike]

Doctor --
I've been busy lately trying to make money to pay our socialistic taxes here --- However ----

I didn't consider a trim circuit for my first try -- GREAT idea -- I didn't think of it and it will be most needed.

I don't have an LF412 but I found a TL082 in my junk box with very similar specs - Both JFET OAs. Off I go to breadboard this. I'm going to add a range switch into the circuit for some 1:1 and 10:1 control when using fixed input voltage which will limit my "metered" max reading. My "pile" reads 0.10V out with a ~100 mW red LD !!

Mike

[The_Doctor]

Thankyou Marconi. I ought to make more...

Mike, the TL082 should be ideal, I just chose the LF412 as a general purpose high quality 741 replacer. It's particularly nice when you need to play with audio signal cancellation and analog arithmetic, I found it has better linearity and clean cancellation between inputs than many more exotic audio IC's.

I am using a range switching idea too, so here are my ideas for that:

For the 200.0 mV range, send as is to DPM,
for 2.000 V range use 910R and 8K2,
for the 20.00 V range use 100R and 10K.
All resistors metal film 0.6W. Buy ten or more of each value.
Measure and note each resistance for the LOW value for each pairing.
Measure and note each resistance for the HIGH value for each pairing.
For the divide by ten, multiply by nine for the first low value, pick the exact match from the high value set if there is one, if not, try the next low one...
For the divide by hundred, repeat as above, but bear in mind it might be harder to find a match as the values will fall at the edges of the tolerance range. This might be good enough though, no-one will miss a few milliwatts from a multiwatt beam measurement... Besides, any error here will NOT be the most signinficant error in the system, and will still allow well within 1% accuracy to be had.

I found the neatest way to switch them will be a DPDT switch of the On-On-On type, they allow an odd linkage between two terminals per pole that can route an input to one of three outputs. One pole selects the range, the other selects which decimal point lights up on the DPM. It might even be possible to automate this switching in analog switches, but they'd have to have VERY low on resistances, and some finely tuned op-amp comparator switching. I'll settle for a DPM that can take 20V input safely and just display some kind of over-range signal.

Take care with the supply to the DPM and the gain stage! They MUST be separately regulated, or the DPM (if LED type, anyway) will change the detected voltage, in turn changing the DPM... That is nasty vicious cycle that makes a mockery of what could otherwise be a neat system.

[The_Doctor]

Disregard this thread. All my input to it anyway.
I won't elaborate, the results were so error-stricken the thing wasn't worth doing. The arm-in-the-beam method would get you closer to guessing the beam power than my thermopile meter and gain stage.

Buy a decent power meter.

That is all.

[The_Doctor]

Disregard this thread. All my input to it anyway.
I won't elaborate, the results were so error-stricken the thing wasn't worth doing. The arm-in-the-beam method would get you closer to guessing the beam power than my thermopile meter and gain stage.

Buy a decent power meter.

That is all.

Disregard this post.
I'd forgotten this one, it was the result of a dis-spiriting day dealing with a small LED panel meter's load current upsetting the gain stage's voltage reference and causing a nasty feedback error problem. If you run the metering and the gain stage off separate supplies the problem doesn't exist. I was also annoyed by what appeared to be low readings for power of red diodes. My thermopile was actually telling the truth. It led me to realise that my assumptions for wavelength were wrong, the diodes were running at 664 nm or longer, not 658 as hoped, and Laserben recently posted here confirming that diodes can do this:
http://www.photonlexicon.com/forums/...ead.php?t=2722
He found that a red diode laser, strongly driven, was at a measured wavelength of >666nm, which agrees with what I saw. Not sure what the nominal wavelength spec is for Marconi's diodes though, probably 660 nm, so a 658 nm diode could be at >664 when driven hard. I only have my deduction to go on, plus the fact that if I set to 664 on a lasercheck instead of the assumed 658, the reading of the Lasercheck and my thermopile meter agree closely.