Diode connected transistor as temp sensor

--
12 years is in the "near future" if you plan to be around forever.

Sounds like a difficult/impossible project, which is what we thrive on bringing to life.

My "research lab" would be happy to help if you can get over the past animosity between us and are willing to pay for the work.

Email me if you're interested.

--
JF

I presume the heatsink tab is the attraction of this transistor? In terms of 'ideal diode' compliance, a low-noise transistor would be better, because at higher currents the design of a power transistor (the TIP32 is a 3A PNP power transistor) includes emitter resistance (which is built-in to the silicon and/or emitter metallization).

An NPN (because N substrate has higher conductance) would be possibly preferable to PNP. The stresses at the tab-transistor interface could cause flexure with temperature, and might argue for a smaller die size to minimize the differential expansion (or, for an exotic package instead of the plastic-and-copper).

The difficulty is having a bunch of pumped-down liquid helium, down to

2K maybe, and the associated temperature instrumentation. A watch-type tuning fork crystal has its parabolic turning point near room temp, and has a very steep tempco slope as it gets cold. It would be interesting to see what happens below around 40K, when material properties can really change.

I actually got some samples of vacuum-packaged tuning fork crystals and was trying to get Jefferson Labs interested, but no dice. I wouldn't go the considerable trouble unless I had a potential customer lined up. It would be a decent grad-student project, maybe.

Yeah, money can pave over a lot of animosity.

John

e e t

The instrument is for use from 77 to 400K. I really like the TO-220 package. I screw the tab to the TeCu sample holder that's also ground, and can measure temperature with one lead. We use two leads to get rid of the lead resistance.

I was trying to figure how much these cost. (The transistors are basically free.) I figure about a day of my time so far. (It took a day to take the data, but I was doing other things at the same time...it's always hard to calculate these things) I've got to write some simple code to generate the calibration data for each diode. It'll take a lot longer for me to figure out how to get a nice print out. We can hope to sell maybe 100 in five years. $10-20 each? It'd be a lot cheaper if I could sell 1,000.

And besides transistors are much more fun. All physics students should see the Ebers-Moll equation at least once.

George H.

age

the

rs.

int

y .

It

n

These things are much too expensive, I wish someone would 'eat their lunch' and sell me liquid nitrogen diodes for $10. (As you say you can go down to 20 K till you 'fall off the edge'.)

I've been trying to design a nice surface mount diode cryo sensor. I'd like something with a ~4-40 size screw hole mount. Saphire would make a nice substrate. (I've got future uses for temp sensors that can't be grounded.) I think I might have to 'tomb stone' the diode, with a flying lead. How much would it cost for a bit of copper, stuck to saphire and good to 77 K? Maybe that's not possible?

(I'm already imagining some saphire washers that will fix me.)

George H.

Wow, do tell!

I interviewed at a frequency standard company and suggested that they cut their quartz at 25 C and use a TEC to hold the temperature there. I figured for most of the users it'd waste a lot less power than 70 C. They said the big problem with quartz was bits of stuff sticking to it, or unsticking from it.

George H.

e...

Yeah, I first played with npn's. But my current source is a simple positve voltage reference, 0.1% resistor, and fet opamp. So I wanted to ground the tab side of the TO-220.. pnp. I had TIP32's lying around.

My 'ideal' diode temperature sensor would certainly have much mass than a TO-220 package. (I'm just thinking about grinding bits off the TO-220 to get the mass down)

A nice metal mounting hole has a lot of attraction in a temperature sensor.

George H.

I've got my own doubts, but I'd imagine that the semi-conductor physics would be okay at 77K.

Differential contraction could well put excessive pressure on the active junctions or rip the some packages apart.

The LM940023 is supplied in a "4-Bump Thin micro SMD Ball Grid Array Package"

which might mean that there was little enough external packaging for the differential contraction to be a non-problem.

Vacuum seals used to be glass-to-metal seals with horrid built in thermal stresses.

-- Bill Sloman, Nijmegen

Hi George

Please look at:

@article{, author =3D "Harro K=FChne", title =3D "Dynamischer temperaturmesser", journal =3D "Elektor", pages =3D "54--58", month =3D "1", year =3D "1993", publisher =3D "Elektor Elektronics", url =3D "

" }

The above temperature measurement accuracy according to the article (in german) - do not need any calibration:

• sensor error

How about channel conductivity on FETs?

The electric (silicon) characteristics between 76 K and 400 K is one thing, the thermal expansion of the copper PCB track and aluminum pins, the thermal expansion between the gold wire and the silicon pad is an other matter.

** A TIP32C has copper pins, no gold wires and a silicon chip mounted on a flat copper tab with some kind of exotic solder.

But I take your basic point - differences in co-efficients of thermal expansion will soon prove fatal to the device when cycling from 50K to 400K

This is way outside the maker's specs.

... Phil

An LM35 outputs 10 mv/deg C. At -60C, its output will be -600 mV and its substrate diode will turn on. And given that it's an LM35, it will likely latch up.

John

ge

he

s.

nt

d 0

True. The LM94023 which I mentioned in the subsequent bit of the post

- which you discarded with an un-marked snip - wouldn't have that particular problem, any more than would the AD590.

I was hoping for for some more detail on how thses parts go bad when operated below the minimum specified temperature, rather than content- less point-scoring.

-- Bill Sloman, Nijmegen

LN2 is easily available, and cheap stainless thermos bottles make good dewars. You could try them and report the results. Maybe publish in RSI.

The substrate diode limit of the LM35 is fact, not "content-less point scoring."

John

=20

Yeah, I hear that. For highest reliability over large temp range, silicon-on-sapphire is good (similar thermal expansions), and I'm thinking the high-end sensor diodes use some customized packaging.

At 10 uA, the bulk resistance and emitter degeneration won't matter much, either. For lower mass, there's TO-202 in my junk drawer, or the glass-packaged Zeners in DO-41 cases (they are for low currents, so no degeneration resistors, but high power so have good thermal coupling to the leadwires). The Zeners would be forward biased, of course. Solder/ braze/weld one lead to a washer and you're good to go.

a

al

0K

I've tortured all kinds of stuff by repeated dunking into LN2. (For $100 I make a call, the next day a truck shows up at the building loading dock. And fills up my 60l dewar.) So far no problems, except with the Nagolen(sp) bottle I was using as a dipper. It got a big crack.

I=92m using power resistors in TO-220 pac=92s as heaters outside their =91spec=92 too. (a heater might be =91happier=92 at lower temperature.)

I thought it was a nice linear curve. If I selected for the higher voltage ones, it looks better than 0.1 K accuracy. I looked up the AD590=92s that Bill S. was suggesting. ~$70 for the 0.5 K accuracy. There were only 8 in stock at digikey so it=92s not a big market.

George H.

"George Herold"

** Piss off you damn fool.