temp measure with diode 1N4148 and differential amplifier

[SimonTHK]

hello
https://www.capgo.com/Resources/Temperature/Semiconductor/DiodeTemp.GIF

Basically what I think I need to do is to have a current running through the diode (stable current) , and then measure the voltage over the diode.

Then I will use a differential amplifier to measure the voltage over the diode and amplify it into something usefull.

Is this how to do it?

Can anyone help me figure out what the voltage over the diode will be at a certain degree? It should be liniar if I have a constant current.

Thanks in advance

 

[ericgibbs]

hello
https://www.capgo.com/Resources/Temperature/Semiconductor/DiodeTemp.GIF

Basically what I think I need to do is to have a current running through the diode (stable current) , and then measure the voltage over the diode.

Then I will use a differential amplifier to measure the voltage over the diode and amplify it into something usefull.

Is this how to do it?

Can anyone help me figure out what the voltage over the diode will be at a certain degree? It should be liniar if I have a constant current.

Thanks in advance

hi,
The diode forward voltage drop is about -2mV/Cdeg, look at this simulation for a typical plot.

 

[SimonTHK]

thank you very much

 

[ericgibbs]

thank you very much

hi Simon,
This is one option.

 

[MikeMl]

Why not just use an LM34 or LM35?

 

[indulis]

Just google "diode equation"... from that you can get temperature cofficient .

 

[Vizier87]

Why not just use an LM34 or LM35?

I have to admit, it looks pretty tempting and fun to build another version of a solid-state thermometer other than the LM35. The LM35DZ, being cheapest, has been discontinued from production, btw.

 

[SimonTHK]

I also think it is very interesting with different options. For my examination one of the 12 possible quistions I can get is to talk about temperature measurements. I have to be able to talk about lm35, 1N4148, pt1000 and ntc liniarising.
So I aint gonna build it, but I am gonna learn it and I try to figure out the quistions by reading, but I also have to ask abit.

thank you

 

[Vizier87]

I'll be more interesting to build it.

 

[atferrari]

Si diodes as sensors.

Many years ago I did extensive testing of a bunch of about 20 diodes for temperatures from 0º to 100º. Data was recorded manually in a painfully slow process which I tried to make as precise (!) as posible. Diodes fed with constant current sources.

While I do not doubt that it was imprecise for any reasonable standard I concluded that:

* Voltage versus temperature curve was NOT linear between 0º and 100º.
* The curve varied sensibly from diode to diode (they were supposed to be from the same batch) (??).
* The curve had the widest separation from the straight line (0º to 100º) at 26º or maybe 60º... actual details escape to me because my precious notebook was lost in a port abroad. Cannot say why I took it ashore. Learnt to never bring anything out of my bench again.

All this to say: you will have to calibrate each diode/s you will use.

When I was about to use them in a real design, I run across the LM335 which was MUCH easier to calibrate (you do it once). Maybe you would like to read about it here **broken link removed**

Browse for Politemp 832 under Circuits.

Beaten paths are popular.

Buena suerte.

 

[MRCecil]

Your empirical results testing diodes for linearity highlights the very shortcomings of trying to use them for a reference; the lack of linearity and repeatability even from die of the same lot. Add to that the delta induced from self-heating across a temperature band vs. ambient and one finds the need for calibration for each unit as you discovered. When devices appeared back in the early '70's exploiting the band gap potential of silicon, a whole new horizon appeared for both accuracy and an array of new devices, which led to much more; think time standards and all that implies.

I think that the initiator of this thread got this assignment from his instructor. I believe it is a terrific research project to trace the path of the evolution of one small facet in the field that has helped revolutionize electronics along with the way we have come to live with technology.

 

[aljamri]

I have to be able to talk about lm35, 1N4148, pt1000 and ntc liniarising.
So I aint gonna build it, but I am gonna learn it and I try to figure out the quistions by reading, but I also have to ask abit.

PT 100 is the best among all, It is made of Platinum, and measures 100 Ohm at zero degree( so taht it is named PT-100). It is very linear in ranges of more than -100 to 200 Ohm ( Exact figure vary ) and used in most medical and analytical instruments whenever temperature to be measured. Its resistance rises linearly with 0.038 Ohm every one degree centigrade.

 

[atferrari]

the lack of linearity and repeatability

Why those expresions do elude me when I should use them...?

Add to that the delta induced from self-heating across a temperature band vs. ambient

Yes, by that time, I was already aware and used a quite low current but cannot recall how much it was (certainly much less than 1 mA).

My first reaction, when I learnt that I could use diodes as sensors was: Wow a cheap sensor, really!. The necessity of calibrating them (building a full curve to be exact), one by one, made me to choose something better.

 

[ericgibbs]

Why those expresions do elude me when I should use them...?


Yes, by that time, I was already aware and used a quite low current but cannot recall how much it was (certainly much less than 1 mA).

My first reaction, when I learnt that I could use diodes as sensors was: Wow a cheap sensor, really!. The necessity of calibrating them (building a full curve to be exact), one by one, made me to choose something better.

hi Agustín Tomás,
Its important to consider that many OP's asking for silicon diode temperature sensors are not requiring to measure temperature over a wide range.
Most are for controlling a fan or heater ON/OFF, so the full calibration is not required, only at the On/Off set point temperatures.

One member is using this type of temperature control for the roof solar heated panels and his swimming pool, been working ok for over 2 years.

If you hold the constant current less than 100uA [ some go as low as 10uA] then the self heating is not a problem.

 

[atferrari]

...only at the On/Off set point temperatures.

You are right! I did not think of it that way. So there is hope for Si diodes after all...

 

[Vizier87]

Many years ago I did extensive testing of a bunch of about 20 diodes for temperatures from 0º to 100º. Data was recorded manually in a painfully slow process which I tried to make as precise (!) as posible. Diodes fed with constant current sources.

While I do not doubt that it was imprecise for any reasonable standard I concluded that:

* Voltage versus temperature curve was NOT linear between 0º and 100º.
* The curve varied sensibly from diode to diode (they were supposed to be from the same batch) (??).
* The curve had the widest separation from the straight line (0º to 100º) at 26º or maybe 60º... actual details escape to me because my precious notebook was lost in a port abroad. Cannot say why I took it ashore. Learnt to never bring anything out of my bench again.

All this to say: you will have to calibrate each diode/s you will use.

Regarding the calibration issue, I'd like to know if it is possible if the non-linear output (of the temperature sensors using those diodes) be fed through a electronic network (RLC-op-amp, etc) simulating a transfer function to process and linearize the output? I suppose that's how the circuit like the one designed by **broken link removed** is produced?

 

[ericgibbs]

Regarding the calibration issue, I'd like to know if it is possible if the non-linear output (of the temperature sensors using those diodes) be fed through a electronic network (RLC-op-amp, etc) simulating a transfer function to process and linearize the output? I suppose that's how the circuit like the one designed by **broken link removed** is produced?

hi Viz,
Yes its possible to produce a close fit using short 'chords' to most non linear functions.

 

[Vizier87]

Can you elaborate on that, sir Eric? I was imagining that the designs which linearize the output of circuits like these utilize laplace/fourier/Z transforms in s/z domain and it would be matched with another transfer function to act on the linearization, where some transistor-resistor-capacitor-op-amp networks may produce. Am I correct?

 

[ericgibbs]

Can you elaborate on that, sir Eric? I was imagining that the designs which linearize the output of circuits like these utilize laplace/fourier/Z transforms in s/z domain and it would be matched with another transfer function to act on the linearization, where some transistor-resistor-capacitor-op-amp networks may produce. Am I correct?

hi Viz,
I will try to produce a graphical example to answer your query, post tomorrow if I can...

 

[Vizier87]

Okay.. take your time and take your rest, Eric.