Zenering a big transistor

e:

connected)

s. (LND150)

ar).

he junction?

after

ld do it.

This has interesting measurements of a variety of vendors' 2n3904's:

"Figure 2 also shows why true 2-terminal discrete diodes are not used in temperature sensing applications instead of 3-terminal devices such as the 2N3904.

A discrete 2-terminal diode, ideally, would perform in temperature sensing applications as well as a thermal diode would. However, characterization in the labs determined that discrete 2-terminal diodes typically have an ideality factor much higher (1.2?1.5) than ? ASSUMED of 1.004. This discrepancy (between ? ASSUMED and ? REAL) would cause unacceptable temperature measurement errors at all temperatures."

I just read a number delta Vbe app notes; all recommend diode-connected transistors over diodes.

Cheers, James Arthur

Eh? LM35 IS a transistor used as a temp sensor (with linearizing tricks and a preamplifier). Software is just as good at linearizing, and a transistor makes a very good sensor.

The same equation-of-operation that makes the LM35 massproducible applies to all low-noise transistors (and with some bias precautions, almost all bipolars). I'd recommend against using germanium, though.

"Unacceptable" is arbitrary and judgemental. It's kinda prissy, going around assuming stuff.

I don't care about ideality factor, how well a c-b junction might match some theoretical ideal-diode equation. What matters is how well you calibrate the V-T curve of a device, and how similar later devices might be.

Have a hearing problem? I could type louder.

LM35 IS a transistor used as a temp sensor (with linearizing

It's a modified bandgap reference. And it's specified for temperature sensing accuracy.

Software is just as good at linearizing,

A transistor sensor will not be calibrated.

The LM35 does have some bugs, though.

BTW, LTZ1000 used in the most demanding applications (8.5 digit voltameters) is defined as super-stable and has even a built-in heater. But its application note contains some resistors and even an opamp on the most critical path. How then were they able to achieve that microvolt per 10V accuracy with all that junk in between?

Best regards, Piotr

ote:

e connected)

ers. (LND150)

far).

the junction?

n after

ould do it.

ors

That's a nice paper. Do they say how they get the ideality factor? I've got data (at different currents) for the tip32 and the error is ~3 K if I assume fidelity factor = 1. Changing it to 1.003 (or so) would make the numbers better.

BTW here's a plot/table for tip32 at 10 uA of current.

George H.

rote:

de connected)

ters. (LND150)

o far).

e the junction?

in after

o?

would do it.

tors

se

I sorta agree, If I could find a diode that was repeatable and easily calibrated against some standard that would be fine.

The one nice thing about the 'ideal' behavior is that you can check the calibration (very roughly) by looking at the signal as a function of bias current. That nice.. and also nice physics.

George h.

e:

om

ode connected)

iters. (LND150)

so far).

ge the junction?

ain after

oo?

would do it.

stors

use

Ds

.................. ^ideality.. dang spell checkers. GH

?dl=0

Buy the transistors from the same vendor, no matter how you process the signals. Ideally, but a reel that's a lifetime supply.

Seems to me that you can't avoid some actual temperature measurements, unless you buy an IC or an RTD that has guaranteed specs.

In your production units, are you planning to do the theoretical delta-I/deltaV math, or are you just going to use a constant current and measure the diode voltage drop?

How accurate do you need it to be?

rote:

de connected)

ters. (LND150)

o far).

e the junction?

in after

o?

would do it.

tors

se

If we stop being offended and look at the figures, though, discrete diodes had large and unpredictable variations, while transistors' variations were small and consistent. That's what I got from the same paragraph, anyhow.

I've no quarrel with that. If you're going to individually calibrate the sensor anyhow, any diode will do and ideality doesn't matter.

It seems the advantage of a diode-connected transistor is being repeatable and interchangeable /without/ calibration. Which is what George wants, AIU I.

Cheers, James Arthur

ote:

.com

diode connected)

imiters. (LND150)

d so far).

mage the junction?

again after

too?

es would do it.

sistors

cause

RTDs

ed

That, or use the delta Vbe technique (take the Vbe at two currents, then subtract to get the difference).

That's really the question. That, cost budget, and whether George can calibrate the sensor or not.

Cheers, James Arthur

The LED is a fine solution, simple and rugged. Just don't shine any bright lights on it!

Cheers, James Arthur

Page AN45-7

which I've successively implemented on several custom chips (with embellished mirrors and switching). ...Jim Thompson

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

I'm looking for work... see my website.

Thinking outside the box... producing elegant solutions.

That's backwards. Bandgap references use a balancing trick to null the tempco enirely, they're TERRIBLE temp sensors. But, they ARE transistor-forward-bias-based.

In the LM35 data sheet there's a plot of absolute accuracy limits, that necks down to a tight tolerance... at 25 C. That's what you get when you laser-trim resistors at one and only one temperature. LM35 is manufactured, calibrated, and then you solder it in. A transistor is manufactured, you solder it in, and then you calibrate it.

The only difference is who does the calibration. Either way, the adherence to the theoretical curve allows a one-point calibration.