Hello, I'm trying to build a temperature logger for very high temperatures (1500F-1900F). I'd like to build an amplifier circuit for a type K to convert the 0-2000F into 0 to 2V for my data logger. This seems like a very simple project for an op amp with a low pass filter. But when I look on line I find that analog devices makes a thermocouple amplifier for like $30. So, what am I missing?
I am planning on using a single supply op amp (LM324) with a gain of
200 and a first order low pass filter with a corner frequency of
30Hz. Does anyone have experience with this? Approximately how much error should I expect from this circuit?
The LM324 is a very poor choice of amplifier; it has a high input off- set voltage and the temperature coefficient of input offset voltage is on the high side. Moreover, the bipolar input transistors effectively rectify any high frequency noise on the input (all frequencies above the - low- bandwidth of the LM324). It takes some noise to drive the input into non-linearity - the "rectification" becomes progressively less efficiently as the noise swings less than 52mV - but it can still be a problem with much smaller swings.
FET-input amplifiers are an order of magnitude or two less sensitive to high frequency noise.
If your first order low pass filter keeps on rolling off steadily up into the radio spectrum (which is unlikely - most capacitors eventually look like resistors or inductors at sufficiently high frequencies) you wouldn't need to worry about this, but in practice it is usually a problem that needs to be dealt with.
The Analog devices AD595 circuit includes cold junction compensation and laser-trimmed non-linearity corrections
formatting link
so you get quite a lot more bang for your $30.
The application note doesn't seem to mention the noise-rectification problem, but it probably wasn't written by anybody who had used the circuit in the field.
The Linear Technology application note on thermocouple measurements AN-28 was written by Jim Williams, who is pretty good, but it still only mentions noise in passing
An accuracy specification. All else will follow from that. The 10-cent LM324 is not a very good low-level amplifier, but it's not ridiculous in this application either, if your requirements are very loose.
The typical DC drift is 7uV/K, 30uV/K maximum so the zero drift due to op-amp TCVos will be around 0.2~0.8 °F/°F. You need a gain of around
40 to cover the 0-2000F range you mention-- 200 is too high.
Add onto that the error due to whatever you are using as a cold-junction sensor, gain error, and whatever error exists in your calculations for thermocouple linearization and cold-junction compensation (to do the best job possible you need two roughly 10th order polynomials- one for forward linearization and one for reverse). Older (and new lower performance) designs such as the AD595 tend to use a linear approximation for one of the polynomials, which works okay in your type of situation if the cold junction is not too far from room temperature and mostly on the + side.
Also there will typically be a small error due to bias current change vs. sensor resistance (which also changes due to temperature).
You should be able to 'typically' get within 10-20F over a narrow ambient range, with care, a decent cold-junction sensor, and good calibration. A better op-amp will easily cut that in half, but your cold junction compensation will begin to dominate. I'm not a fan of expensive chips like the '595 that require you to closely thermally couple a DIP to a terminal strip.
Your easiest (and best for a one-off) solution is to go buy a thermocouple transmitter that will provide galvanic isolation, filtering, cold-junction compensation and linearization for perhaps $200 off the shelf- less than an hour or two of engineering.
Thanks for your advice. I really like the LM324 because it has a single supply voltage, even though it's not the best of op amps. Dual supply voltages are fine in a lab, but in field instruments they are a pain. There has to be a good way of taking a single 9V battery and creating a pos and neg supply. Anyway, I'll probably go with the AD595 due to the fact that I need about 6 of these. And I can make the whole package, including data collection, for less than $400.
I have not played with them seriously, but you may have a look at Texas Instruments' TLE2426CLPE3, which provides a virtual ground with quite low output impedance.
- ---------------------------------)----------------------- potcore | |+ This is a sine wave generator | === 10u | --- |___ k diode a____|_____ - 12 V Original hand painted ASCII
Thermocouples always come in pairs (a complete electric circuit has a hot junction and a cold junction), so you'll need to digitize both the loop voltage and the cold junction temperature, OR use a functional module that sums in a cold-junction compensation and maybe does some linearizing of the voltage/temperature curve. The 'thermocouple amplifier' is the full functional module, and saves you from dedicating two datalogger channels.
You can do your own temperature compensation with a diode-connected transistor, if there's any kind of regulated power available... but your calculations will get somewhat messy, and calibration is up to you. And, it DOES take some kind of reference source, not just op amps.
There have been a few more single supply op amps developed since the LM324 hit the market.
The Linear Technology LT1006 is almost as old, and has much better off- set and drift specifications. Linear Technology has introduced a lot of other precision single supply op amps since then; the LTC1050 and its successors might be worth a look, and the Analog Devices equivalents are also interesting.
Beside the poor choice of an OpAmp, I'd skip the OpAmp altogether and have a sufficiently good ADC to measure the thermocouple. A K type does in the order of 40uV per degree Celsius, making in the order of 25uV per Fahrenheit, and the 1500 F thus give in the order of
40mV, a useable input for a standard 20bit ADC. Some may even have selectable amplifier stages built in.
Once the circuit is running the PNP's eb junction is back biased most of the time. 9 turns will have 9V peak =3D 18V p-p on it so one turn has 2Vp-p.
It is running as a class-C1 stage. During saturation, the hfe of the transistor goes nearly to zero. A lot of the base drive current is taken up during that time.
I am sort of surprised it doesn't squeg. The 10uF gives a time constant way longer than the ring up time of the tuned circuit.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.