Chopper stabilised op-amps - is there a mystery to the noise?

use the

es

It may be easy, but it isn't super accurate. Super accurate would be Ac exc itation - or at least reversing DC.

The voltage drop across a nominally 100R Pt sensor at 1mA is 100mV, and if you want super accuracy you have to be able to correct for any thermocouple voltages present. Reversing the direction of the current through the senso r allows you to that easily, generally using the same sensor, resistor and ADC, if your power supplies allow it.

My 1996 per on our millidegree thermostat does mention this, but we could o nly have done it (at the time) if we'd had access to a -15V rail, which we didn't.

Bill Sloman wrote

The self heating aspect can be sorted by turning on the current only for a reading.

Lots of instruments do that. I bought a cheap chinese one on ebay. It wasn't accurate so I went to measure the sensor current and... there wasn't any :)

Measuring DC accurately must be easier than measuring AC, I would think. Or perhaps you are referring to slow switching, using four analog switches to reverse the sensor, so you take just the two readings. I saw some appnotes on AC excitation and synchronous detection but it didn't sound all that accurate.

To eliminate PT100 self heating "totally", with even a tiny package in free air, you need to go down to something like 100uA and then you are measuring just 10mV which is hard work, with noise, etc.

Exactly; we share the multiplexed delta-sigma ADC among three RTDs and some other stuff. RTD current duty cycle is low and self-heating is minimal anyhow. The error is dominated by our initial calibration accuracy and drift in the Susumu resistor and the thin-film RTD itself. Platinum resistance changes about 3800 PPM/K, and we measure to roughly 10 PPM, so the temperature error is microscopic, way better than any thermocouple can claim. Our RTD measurements are fun but, in real life, gross overkill.

I got a cheap dual-channel thermocouple meter from Amazon. It is very good. Delta-t offset between two thermocouples is zero. Good for heat sink tests.

Absolutely. A cheap delta-sigma ADC makes ratiometric measurements to PPMs. Intelligent PCB layout avoids any stupidities.

Or perhaps you are referring to slow switching, using four

A 1206 thinfilm RTD can have some thermal copper pours on the PCB, on the end caps, to reduce self-heating by a lot. 1 mA and 100 ohms and

excitation - or at least reversing DC.

if you want super accuracy you have to be able to correct for any thermoco uple voltages present. Reversing the direction of the current through the s ensor allows you to that easily, generally using the same sensor, resistor and ADC, if your power supplies allow it.

Depends on the thermal time constant of of the resistance sensor.

If John Larkin had done the job properly he'd be able to say how much the s ensor warms up while he was reading it. "Minimal" means that he doesn't kno w.

So you don't calibrate all that accurately, and don't spend enough on your low drift reference resistor.

National standards labs use AC bridges to measure the resistance of their p latinum sensors

Larsen N T 1968 Rev. Sci. Instrum. 39 1?12

They tend to use ratio transformers to establish the relationship between t he resistance of the sensor and the resistance of the reference resistor, w hich gets you down to Johnson noise limited accuracy.

You could play around to get that a bit lower. if you tried hare enough.

Sloman A W 1978 J. Phys. E: Sci. Instrum. 11 967?8

It's been done.

Sloman A.W., Buggs P., Molloy J., and Stewart D. ?A microcontroller

-based driver to stabilise the temperature of an optical stage to 1mK in th e range 4C to 38C, using a Peltier heat pump and a thermistor sensor? ? Measurement Science and Technology, 7 1653-64 (1996)

Admittedly we used an interchangeably thermistor rather than a platinum res istance sensor (which is a whole lot less sensitive).

It's less accurate than a Blumlein bridge built with properly wound ratio t ransformer.

It's a terrible title - what was meant was that the transformers involved w ere completely shielded.

The Brian Kibble listed as one of the authors is the same Brian Kibble for whom the Kibble balance is named.

You clearly didn't understand it properly.

It doesn't eliminate it - just makes it smaller.

But not zero. And you need to know the thermal time constant before you can work out how much self-heating you've actually got.

Of course, once you've done that properly, you can subtract it out of the r esult you present to the customer.

Thanks John, that's a good one to know about!

When they say the long-term stability is 0.1%, that 0.1% of the 100 ohms, right? Which with the 3850ppm/K sensitivity is 1/3.85 - 0.26 degrees, right? I see the B indicates class F0.3, or 0.3C initial accuracy, best offered for this part, right? Together that's 0.56 C.

No, I think "minimal" means minimal, which is well below the 0.3C rating of the part. If one has a precision RTD, they can can pull out all the stops for measurement, but this $2.05 part (D-K, 100 pieces) is not in that class. Moreover, with modest care, thermocouple voltage effects won't be encountered onboard the PCB.

The Vishay data says that the thermal time constant in air is seconds; we digitize in milliseconds. We do add big copper pours on the PCBs to reduce theta and add thermal mass. Our ref junction accuracy is way overkill for processing thermocouples.

One interesting use for these surface-mount RTDs is as a simultaneous heater and sensor, to explore the thermal effects of soldering parts to PC boards.

I did the math once to turn that into K/W. Can't find it just now. But the effects of the added copper are pretty stark, theta and tau. I recall the free-air (just some tiny long wires) theta being something like 200 K/W. On a PC board it's much less.

We also hang thermistors here and there on boards, to snoop local thermals. My new class-D power amp will be thermally instrumented, so we can play with air flow.

I design electronics. Sloman whines.

John Larkin claims to design electronics. I point out how he could have done it better. He doesn't like that - what he wants is uncritical admiration, which he's poorly equipped to attract.

That's one meaning you can attach to the world "minimal".

John Larkin hasn't been that specific, and clearly hasn't bothered to work out how much his sensor does warm up. Quite a few people make that particul ar mistake.

Sloman A. W. ?Comment on ?A versatile thermoelectric temper ature controller with 10 mK reproducibility and 100 mK absolute accuracy? ?? [Rev. Sci. Instrum. 80, 126107 (2009)] ?, Review of Scientif ic Instruments 82, 27101 - 027101-2 (2011).

In that particular case it vitiated their claim of 100mK absolute accuracy

- they had 200mK of self-heating.

A printed circuit board carries elements that dissipate power, so it's got thermal gradients. There a lots of different metals in the components - cop per tracks, solder joints, aluminium metalisation on the integrated circuit s.

It takes more than modest care to eliminate thermocouple voltages - more a renegotiation of the laws of physics.

A $2.05 part isn't all that cheap, and with proper care could deliver very accurate measurements. The metallurgical quality of the platinum film may b e a bit dubious - if it's coefficient of thermal expansion doesn't exactly match the substrate you can get thermal stress which can change the observe d coefficient o of the resistance against temperature term - but Vishay are n't the only source for that kind of sensor.