Transitor as heater

Agreed. A single sensor/heater combination will work nicely in a thermally insulated (adiabatic) environment. Eventually, the internal temperature will be uniform throughout the assembly thus eliminating the temperature gradient problem that I mentioned. At that point, one can consider the crystal temperature to be the same as the heating transistor junction temperature. Life is great in idealized environments.

However, if there is any temperature gradient between the crystal and the heater, the crystal will NOT be at the heater temperature. When I tried to construct an insulated oven enclosure in a marine HF shore station transceiver, I had problems. There just wasn't enough room for a doubly insulated ovenized oscillator. It was easy enough to tell that there was thermal problem; the aluminum case was fairly warm. Never mind that I created most of the problems myself. I used an AT cut crystal (all I had) instead of the more appropriate SC cut. The fiberglass insulation I used was too dense. Those and other mistakes were eventually fixed.

Before I the problems, I tried to fix the OCXO by rearranging the crystal, heater, and sensor positions. I suspected that it would be a bad idea for the temp sensor to "see" the heater directly. The idea was to regulate the crystal temperature, not the heater temperature. So, I created a sandwich with the crystal in the middle. That change alone almost produced a usable OCXO. Eventually, fixing the other problems (i.e urethane foam not fiberglass, SC cut crystal, double insulated, radiation shield, paper under the aluminum can, aluminum foil sandwich wrap, etc) made it all work correctly.

The down side of the sandwich construction is that the thermal lag is long which requires too much time to stabilize. I "fixed" that by cranking up the maximum power on the heater to the point below where it tries to melt solder or blow a fuse. It was seriously under damped as it would overshoot the target temperature, but it would stabilize much quicker than when slowly approaching the target temperature from a cold start. It also had the side benefit of somewhat accelerating crystal aging, but would also occasionally destroy defective crystals from the thermal shock.

Yep, that can be made to work. I sorta tried it but failed to create a worthwhile improvement. The problem was that I had to add heaters and sensors on all six sides of the package to block the leakage on that side. If one side did not have a heater, it would leak heat from the crystal to the outside, thus recreating the temperature gradient.

Another temperature gradient I found was between the quartz crystal element and the nickel/tin crystal can. I tried to buy UV fused silica or quartz crystal cases, which are IR transparent, but the prices were too much. So, I convinced the vendor to fill the crystal cans with helium, which has a higher thermal conductivity than air. There was no difference with all the other thermal errors that I introduced, but once these were reduced, the helium helped a little with reducing yet another gradient. (I would have used hydrogen, but didn't like the idea of soldering to a potential bomb).

(...)

I forgot to mumble something about another OCXO thermal screwup. I was doing all my testing in a fixture, without the radio. When installed in the radio, there was enough moving air from the power amp cooling fan to cause small frequency excursions when the fan was running. While the cause is quite obvious at this point, it took five experienced engineers and managers a full day to see the obvious cause, and then only after it was pointed out by one of the production techs. Very embarrassing. Rather than trying to move the fan or the OCXO, I concentrated on improving the insulation and adding an air flow deflector, which was sufficient to meet the specs.

But you have time - measure the temperature curve after a heating pulse and correlate that with a thermal model. Using a thermal model is the only way of calculating how much of your heat has reached the target at a given point in time, so you have to do it anyway.

Clifford Heath.

Good idea, since the thread has drifted off into several directions based on the respondents bad guesses as to what you're trying to accomplish, what problem are you trying to solve, and what you have to work with. Context is always helpful (but topic drift is often more interesting).

Now you tell me. Please ignore my previous OCXO horror stories. This is quite different.

I have an idea, but no clue if it can actually be made to work.

Combine the thermal generator and the detector. Use an IR LED to heat the sample. Then, turn it off and use it as an IR photodetector to measure the temperature rise. You already have the LED at cryogenic temperatures, so thermal noise should not be a (major) problem.

Not really. Since the LED will start measuring the temperature the instant its heating function is turned off, you can watch the temperature curve as the heat diffuses through the sample. From the curve, you can predict an end point and don't really need to wait until it gets there.

Maybe. Test the fixture without a sample and subtract the graphs. I'm not sure if the results are useful or meaningful, but if you see a difference, it should be due totally to the effects of the sample.

Think again. If you combine the functions so that one device is switched between a thermal emitter and a thermal detector, you might get something that works.

What servo? If you know how many coulombs you're feeding the LED thermal emitter, the LED efficiency at cryo temps, the radiation pattern, and the area covered, one should be able to calculate how much heat is being delivered to the target. No need for feedback.

Disclaimer: I know nothing about cryogenic techniques and devices, but I'm really good at guessing and don't really mind being totally wrong.

Sounds like you're making it harder than you need?

Use a PTC and drive it with a constant current source that can be set via your desired to obtain the given heat you need.

The feed back would be the applied voltage required to obtain the current verses the temperature you need.

Depending on what you're doing, you could use the tab mounted styles if you are trying heat a specific spot.

We've done such a thing making a make shift vacuum sensor.

Jamie.

Jamie

Yeah, Well I wasn't thinking as far ahead as curve fitting. I was hoping to be just able to wait ~5 time constants or so. I must admit I thought it might be fun to watch the heat leak out of the transistor.

George H.

Hi Jeff, I'm going to cut your response. (because google groups sucks, but is also easy.) And I'm much too lazy to try and trim the whole thing.

The sample and addendum will sit inside another metal can. The outer can will servoed to be at the same temp as the sample. (the can will have a 'weak' thermal link to 77K, a heater and sensor.)

George H.

On Thursday, February 13, 2014 10:26:01 AM UTC-5, George Herold wrote: What the heck is a "transitor" ? :-)

Sniff... I'm crushed but can take it. Besides, I hate reading my own stuff more than once.

My main thing has always been repair and troubleshooting, which is why I'm into such horror stories. Note that my internet domain is LearnByDestroying.com.

Yep. Just remember that hot air rises to the top of the discussion.

I hate solving problems I know nothing about. I guess the best way to find an unknown is to compare it with something that's known in the same setup. In my days of running a print shop, we used a "gray card" which has a guaranteed 18% reflectivity. Whether that works in the IR region can be tested with an IR source and light meter at room temperature. If it works (or is close), cram the card into your LN2 bath, and you should see 82% absorption. You can measure the heat rise of the card with the LED running as an IR photo detector. You can also get gray cards with different reflectivity, which will give you a calibration range. The test sample should fit in between some of the gray card test points.

Don't forget to mask off everything except the area of interest. Also, watch out for nearby object re-radiating absorbed IR from the mask material.

Yep.

I don't understand what that will do, but I'll dig through your other thread and see if I can decode what you're doing.

SITOR is an error correcting teletype system. A tranSITOR transmits SITOR.

The Vbe tells you the temperature. To get a more accurate reading, switch its current by (say) an order of magnitude and measure the amplitude of the Vbe change. Be nasty, make the low current low enough so that the heating can be neglected (make that fixed), and make the high current hi enough to get the heating you desire (adjustable, natch). One part does both and there is NO thermal hysteresis.

That would be a dual like the MMDT2N2222. Rather close spacing.

Bear in mind that there'll be more than one curve. You need to model it as a multi-stage series-R parallel-C circuit. A single stage would get you close, and a second may well go within your measurement error.

that,

and then a wait time for the temperature to settle down. No matter what I use I need good thermal contact between the sample and my heater/ temp sensor, gizmo

measurements.)

use

low mass/ easy to mount/ good themal contact. (yeah I know, pick two of the three :^)

If you are really interested in low mass and good thermal contact being sufficiently primary, you could solder quad transistor dice the thermal interface and wirebond to a PCB. The mounting is hell but it excels on the other parameters.

?-)

ter.

ntially)

g of controlling the current

ransistor as a temp sensor with the c-b shorted. (diode connected transist or.) Could I just push a bunch more current through it for a heater.

g from heater to temp sensor. Do I use relays or analog switches?

Hi Robert, Yeah I've done the current scaling of the Vbe change. This turns out not to work so well for just Vbe, because of the non-idealit y factor which is something near 2 for most diodes. However it works prett y well for a diode connected transistor. (good to 1% or so, well 1% at 300 K is 3 degrees so it's not all that great.)

Arghh.., ya know I just realized that I'd been ignoring the self heating of my temperature sensor. (Bill S. will be scolding me shortly.) .. OK I'm going to go sharpen my pencil. I might have to pulse the current through the sensor.

George H.

(Well I didn't mention to other reason which is that you tend to write *a l ot* :^)

just the addendum.

s.

I started to draw a picture.. but let me just try words.

I've got a sample hanging from a string in vacuum. There are two sources of heat leak. Radiation, Thermal conduction along the wires and support string Self heating of the temp sensor.

Three, three sources of heating.

So to take care of the first two the sample sits inside metal can. The tem perature of the outer can is set to be the same as the temperature of the s ample. Now it won't be perfect because the outer can won't be a perfect is otherm. There will be some radition from those areas that are hotter or co lder than the sample.. but that should be small.

That's the idea anyway. I haven't tried it yet and perhaps it will crash a nd burn because of somethng I've forgotten.

George H.

Assuming the whole thing is small enough. The thermal mass approximation is horrible at short time scales.

Cheers

Phil

Like your pun; SexTronix used the current switching method (i think current ratio was 1:10) to measure temperature with reasonable accuracy over a fair temperature range.

...and one can measure the cooling from service as a heater to get system TC and make appropriate thermal correction.