Thermocouple signal amplification

It can be, however it's a particularly unsuitable choice with its large offset voltage and no TCVos specification and limited common-mode input range.

The input common mode range of that part does not include ground, so something like that might be required.

Don't know what circuit you're talking about, but you'd need 2 or 3V negative IIRC, to allow it to work with the input at ground. If you're going to use a crap op-amp you may as well use an LM358 then. But it won't be very stable with temperature.

You /can/. Why would you want to?

You can't afford $3 for a decent op-amp? You can hardly get a cup of coffee for that.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Hello Apprentice_nerd,

Since you are dealing with signal differences in the sub-millivolt range this amp isn't very suitable. It has several volts in offset error. You either need a very precise amp or a regular one with a chopper stabilization (clamping away CD errors and DC drift).

You have to look at the "compliance range" or "common mode range" in the data sheet of the opamp you select. It should clearly state how close inputs and outputs can get to the rails until things become non-linear. Please mind that the DC performance may not be as good when hugging the rails.

What should it do there?

In that case I guess you have to familiarize yourself with "auto-zero", "chopper" or "clamping" techniques to get rid of DC offset errors. The AVR can greatly ease that task because it can act as the "capacitor" by storing and subsequently subtracting the DC offset.

Regards, Joerg

Hello Spehro,

We can still get a mid size (venti?) cappucino for $2.90 ;-)

Regards, Joerg

Interesting. How much amplification do you get? I suppose the source impedance must be low, to avoid noise/offsets from charge coupling from the FET's gate drive. What frequency and inductor values are you using?

--
 Thanks,
    - Win

Hello Joerg

Thanks for your informative answer.

I read in the LM1458 Datasheet that a negative voltage is recommended for better performance. I dont have any negative voltage generator at hand only some 555 and this schematic about producing negative voltage.

Best regards.

Hello Mr Pefhany

Thanks for your time and your valuable suggestions.

I dont have a negative voltage generator only some 555 and this

I thought maybe its possible to further amplify the signal in case the gain from LM1458 is not big enough.

I totally agree that $3 is affordable but at this moment i am only investigating if i can measure temperature even with +/- 10 acurracy.Unortunately I still have a lot of things to learn and i think its better to struggle with minimal budget than to make costly mistakes destroying IC's because of ignorance. Also i live in Greece and I order from digikey every 3 months because Greece is the most expensive country on earth and a cup of coffee is at 3.5 Euro!:)

Best regards and thanks again.

You've given no indication of knowing the properties of thermocouples, and no one else has piped up with it, so here goes.

First, amplification is pointless without first having a well-established reference. The DC voltage you get from a thermocouple is meaningless without a known temperature to reference your reading to.

Second, thermocouples are nonlinear with respect to temperature. You'll need a conversion algorithm more complex than mx+b to get anywhere near a real temperature reading over more than a few 10's of degrees.

Look at

to get a very brief look at reference junctions, and at for a quick look at linearization.

Google can be a big help: google "thermocouple reference junction" and "thermocouple linearization" to get some real information.

As others have said, _after_ you've got the reference junction out of the way, you still need an auto-zero amplifier to get an accurate voltage. Then you can use your AVR to get a temperature.

Are you sure you want to get into this? Diodes and thermistors are much easier to deal with.

John Perry

acurracy

in

SENSOR

The biggest problem with using non-specialized circuits for thermocouple voltage amplification is the lack of "cold junction compensation". Suppose you have a copper-constantan thermocuople. If you connect the constantan thermocouple wire to your circuit via a copper wire, you have formed a new "phantom" thermocouple in series with the "real" thermocouple. This "phantom' thermocoule generates a voltage in series with the voltage from the "real" thermocuople. The phantom thermocuple voltage is a function of its temperature. Thermocuple interface amplifiers have built in voltage generators (cold junction compensation) that takes care of this problem. . .Historical note: In the "old days", for laboratory work, the cold junctions were immersed in an ice water bath to ensure that they produced a known "phantom" voltage in series with the "real" thermocuple voltage, so that it could be subtracted out to get the correct net voltage. . Regards, Jon

I am for the moment doing some experiments on DC-amplification for low impedance sources. I use a DMOS switch( 5 ohm), an inductor and a transistor. I kind of buck-converter, i.e. the small dc gives a current through the inductor which is suddenly stopped, and the voltage spike is analyzed. The spike is proportional to the source (thermo-)voltage. The outcome of this experiment is in short: Better offset stability than the best op-amps. (About 10-50 nanovolts for hours) Low cost components, but a few too many maybe, are needed. But if you can use the functionality of the microcontroller, the number of components would be reduced.

/Sven Wilhelmsson

[...]

You might be able to find useful parts in old computer monitors, pc power supplies, printers, modems, telephone answering machines, clock radios, tv's, xerox copiers, and other things that people usually throw away. Usually the older designs have more usable parts.

For example, answering machines and monitors may contain op amps such as LM358. Some modems may have TL074's. Look for ones that combine a modem with a soundcard.

A monitor may have useful parts, such as high voltage MOSFET's like the IRF634 and IRF640, potentiometers, electrolytic caps, voltage regulators such as 7812, 7809, and 7805, and high voltage transistors.

Usually people are happy to let you have stuff they normally throw away. Just let the word get around that you like these things, and pretty soon you will be swamped with parts:)

Mike Monett

Also, when you start stripping old electronics, you need to find out what the part numbers mean. Here's some info on decoding semiconductor part numbers and finding data on the parts:

"Decoding Valve, Transistor and CRT Numbers"

"Transistor Specifications Database"

"Transistor Database"

"Transistor Cross-Reference TCG/NTE/ECG To JEDEC and Japanese"

"BC Transistor Cross-Reference"

Here's a good source for datasheets. It allows partial searches, so you don't have to know the full part number:

Here's another one in case the above doesn't have what you are looking for:

Mike Monett

Yes, I thought it might be something like that. ;-) You may not get enough negative voltage using a bipolar 555 at 5V with two diode drops. Consider using a CMOS 555 or driving a CMOS buffer with the

555. Or consider a 7660. You'll have to worry about exceeding the allowable negative input voltage on the AVR (typically something like

-300mV).

Okay, here's a chance to put pen to paper. Given your low accuracy requirements, does the open-loop gain of the LM1458 really matter?

Calculate the gain of the closed loop amplifier given a finite gain op-amp, and substitute the minimum and typical open-loop gains of the

1458. Simple algebra, and you'll get a better feel for what's important. Say you want the output of the op-amp to swing from 0 to 2.5V (it won't go much higher reliable with +5/-3V supplies, IIRC) with a 60mV input, the closed-loop gain would be about 40.

Is that for that mud-like stuff that most of the rest of the world calls "Turkish coffee"?

No problem.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

[snip]

A thermocouple is perhaps the most difficult temperature sensor to learn on, and no benefit unless you actually need to measure temperatures above about 150C. That is especially true when you mention a uP and single rail in the same sentence.

If the range is within -20C to +120C you might find it easier to get going with (say) a 1% thermistor.

--
Tony Williams.

That's not really a problem in this case, he's got a micro and there's absolutely no reason to use the old-fashioned analog methods for CJC. In fact if you want to design either a universal T/C front end or a very accurate one (with delinearization of the CJC signal) it's like totally the wrong way to go about it.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Amplification in step one, the inductor, is around 100, which is the integration time (1 ms) over the spikewidth (10 us). Amplification step two is the transistor. Maximum 50. The overall amplification is uncertain but reasonably stable, but it is clear that it must be stabilized using feedback.

Yes, the method is pointless unless source impedance is low, say below 10 ohm. If you have Z_source > 100 Ohm, using an OP is a better way. One of the best low-Z OP is LT1028, and it works best at about 200 Ohm.

In fact, the main idea is to use an inductor to transform the low source impedance (The inductor can of course have a secondary winding, a transformer), so that it matches the optimal Z for the transistor. This would be in the range 100-1000 Ohm.

In my experiments I use a frequency of 1kHz. But this is not so important. The important thing is to make use of all the integration time: L/R > T where T is integration time and R is the resistance of the switch including source and coil. In my case R=5 Ohm and L>5 mH.

Noise/offsets from charge coupling: This is dominant but fairly stable. I try to balance charge coupling somewhat. Yes, this might be the limiting factor besides thermal noise. I'm not sure about the noise mechanisms here. In switched capacitor devices we have kTC-noise, but what is its equivalent in "switched inductor devices" ?. Any ideas?

In any case, I'm almost convinced that this technique is better than the best OP-amplifiers for sources below one Ohm.

Finally, it is important to keep things thermally clamped, and to be careful with powerburning near the sensitive area. Any solderpoint is a thermo voltage sorce.

/Sven Wilhelmsson

If he's planning on measuring the temperature in, say, a ceramic kiln then that part would have a rather brief career. ;-)

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

And the LM335 could probably be used as for CJ compensation.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Why reinvent the wheel?

Analog Devices makes several temp sensor chips. Look up the AD22100 to use the chip as a sensor with an output voltage proportional to temp. Also the AD594 or AD595 if you need to use a thermocouple as a sensor. It is a lot easier than trying to build your own.

Hello Apprentice_nerd,

That would be a charge pump. You can do that but the 470uF caps seem a bit large for the little current you need.

I would look for a single supply opamp which can go to GND. Less hassle.

Regards, Joerg