I use voltage-current characteristic of p-n junction I=Io[exp(eU/kT)-1]. But cannot get T->U right dependence. Right is: if T is increased then U is decreased. But characteristic (above) gives reverse dependence. If diode is excited with constant current then dependence should be approximately -2.2mV/deg. (for silicon diode). I want to have analitical dependence (and sequence of conclusion) for ideal occasion and improvement for real occasion (diode 1N4148 for example). If anybody knows this please help me.

Thank you in advance (and excuse me for my english).

 

There are very low cost precision sensors available that will have far less variance than a production lot of diodes. I have done both and
the need for calibration of individual diodes swamps the cost difference in even modest quantities. And since this is a "typical" spec for diode manufacturers, the next lot may well behave differently. These differences are small but probably sufficient to exceed your error budget. Techniques that depend on the matching of monolithic transistors are better and still
reasonably inexpensive but, transistor arrays are becoming harder to find. The reasons for departure from ideal are many and largely uncontrolled in general diode production. I had a paper on this back in the 70's, I'd have no idea where to find it now. With multi point calibration you might get the accuracy you are looking for.

Regards

cww

--
Free Tools!
Machine Automation Tools (LinuxPLC) Free, Truly Open & Publicly Owned Industrial Automation Software For Linux. mat.sourceforge.net.
Day Job: Heartland Engineering, Automation & ATE for Automotive Rebuilders.
Consultancy: Wide Open Technologies: Moving Business & Automation to Linux.

 

Your observation about the equation you cite is correct. The turn-on voltage, Von, does decrease as temperature increases. Using the simpler version covering just the forward conducting region of the Voltage-Current relatonship, yields:

Id = Io { exp [ (qVd) / (nVt) ] }.

Then, taking the natural log, ln, of each side, and solving for Vd
yields:

Vd = (nVt/q) ln ( Id / Io ).

Substituting known or normally accepted values for q, n, Vt, yields the solution for Von at any temperature, T1. Thus:

Von@T1 - Von@To = 1000 x Kt (T1 - To):

To = reference Temp @ say, 25 deg.C.

T1 = junction temp @ new point in deg.C.

Kt = temperature coefficient, eg, -2.0 mV / deg.C.

Von@T1 = Von@To + [1000 x Kt ( T1-To) ].

Note that the last term [....; ] is negative. Therefore as temperature increases, i.e., T1 is greater than To, Von decreases. Q.E.D!

Regarding the diode you mentioned, I don't have the design parameters immediately available for it, nor its near equal, 1N914, But if in a glass package, the empirical factor, n, may have to be adjusted.

Regards,
Phil Corso, PE
(Boca Raton, FL)

 

My earlier response omitted an essential part of the derivation:

Starting with, Id, current flow in forward direction is:

Id = Io { exp [ (qVd) / (nkT) ], where:

Io = Reverse saturation current.

Vd = Potential across diode.

q = Electron charge.

k = Boltzmann's constant.

T = Absolute temp in deg. K.

n = Empirical scaling factor, based on material construction, etc.

and, letting Vt = kT / q, then:

Id / Io = exp [ Vd / (nVt) ].

Sorry for omission!

Regards,
Phil Corso, PE
(Boca Raton, FL)

 

Better yet, forget deriving a more accurate formula. IF you need to measure some temperature other than your own pcb , just use 2 diodes, don't bother to match them, but have your firmware do a training sequence over temperature range to measure the difference between them. Cycle your product in a temperature chamber between 2 extreme values and measure the
differences between the 2 diodes . You might need a microcontroller with an A/D range of 200mV give or take, depending on your intended temperature
range. You then interpolate (linear or segments) between the 2 extremes, using the known end of scale voltage differences between your 2 diodes . You could build a lookup table with this interpolation . All this can be part of a power up procedure your device performs automatically while being temperature cycled and burned in . With analytic formulas you won't get far. If you want , do a Spice simulation or better yet, look at Spice models for the diodes you were considering . You'll see lots of different phenomena which contribute significantly to the errors , vary with technology, geometry, materials, batch , etc. , very hard to predict .

Best regards,
Matt Tudor , MSEE
http://www.gigahertzelectronics.com

 

Hi,

Hart Scientific discusses precision temperature measurement in great detail at "www.hartscientific.com":http://www.hartscientific.com .

Best::

John G. Boland, president
Strateg!c Method$ Corporation
One Parker Square Suite 408
2525 Kell Boulevard
Wichita Falls, Texas 76308
940.322.9922
940.723.1478 fax

 

Responding to Matt's comments:

Sorry, but I didn't interpret stts' query as a request for improving the cited formula. I thought he wanted proof that Vd is inversely
proportional to T.

Regards,
Phil Corso, PE
(Boca Raton, FL)

 

Yes, I wanted proof that Vd is inversely proportional to T. But I wanted also to use that for getting real temperature measuring and getting maximum allowed accuracy. Therefore thank you. I have understanding that I did not take into account Io that greatly depend on temperature. Also I have learned following method:
to get one transistor(or diode) and excite him with two different but known currents I1 and I2. Then calculate Ube1-Ube2=kT/q*ln(Ik1/Ik2). (there Ube - base-emitter voltage, Ik - collector current) Here Io does not affect and difference Ube1-Ube2 is direct proportional to absolute temperature T. I now examine this scheme in WorkBench and get very good results (relative to theoretical results). Though I do not sure about transistor-model in WorkBench.

Regards,
Evgeny Hrustalev

 

Do a search on the part AD8551 at "http://www.analog.com":http://www.analog.com there is a great example circuit using 1N4148 as a temperature compensating component. with some modification, you can use the circuit to measure tempreture from -55 to 150C. However there is an error on the circuit they have. If you are interested, email me and i can help you at [email protected]

 

under constant diode current, the voltage drop of diode is fairly accurate to -2mV/C. you can set up the current source easily by using a precise opamp, high current gain pnp transistor, a resistor and a reference voltage. here is the connection
1 apply reference voltage to Vin+ of opamp.
2 connect base of pnp i.e.(2N2222) to ooutput of opamp
3 connect Vin- of opamp and resistor terminal a to the emitter of pnp
4 connect terminal b of resistor to supply voltage. note that the tolerance of supply voltage will affect the accuracy.
5 connect a diode from collector to ground and you have form a simple thermometer with analog voltage output

the source current I_d = (V_supply-V_ref)/R_ab
if need further help, email me at [email protected]