RTD help?

We use RTDs as reference junction sensors for thermocouples. You can't use a thermocouple for that!

Si diodes work down below 2K. Below about 20K they stop behaving like diodes and start acting like thermistors. The signals down there are huge.

John

Well, i figured that with a resistive bridge, i better use an instrumentation opamp. So...i found the TI INA114 and lo and behold, on the datasheet page

11 figure 8 is an RTD temperature measurement circuit. Uses the current i had in mind, namely 100uA via current sources - not resistors. Used SPICE on that idea, and it is close enough to perfect to be more accurate (in theory) than i need; error seems to be zero at 0C and creeps up to an astounding 0.016C at 185C! Like with that massive error, aint nobody can complain!

Almost all "garden variety" electronic devices start to fail near

160C, and most are dead by 185C and virtually all dead at 200C and irrecoverably dead in the 220-240C region. Well, i figured that with a resistive bridge, i better use an instrumentation opamp. So...i found the TI INA114 and lo and behold, on the datasheet page 11 figure 8 is an RTD temperature measurement circuit. Uses the current i had in mind, namely 100uA via current sources - not resistors. Used SPICE on that idea, and it is close enough to perfect to be more accurate (in theory) than i need; error seems to be zero at 0C and creeps up to an astounding 0.016C at 185C! Like with that massive error, aint nobody can complain!

• Balancing the bridge with a resistor equal to the RTD value at a given temperature works nice as seen with SPICE. The bug is that a mere 0.1% resistor firstly is not available for the temps i have in mind, but if they were, the INaccuracy achieved if one is subjected to the +/-0.1% gives an error of about +/-0.5C which is juuust tolerable if truly repeatable. So i do a bit of "trimming" of resistors to get what i want.

Shows that one can drive the p*ss out of them! And that some heatsinking can go a loong ways. Going to follow Figure 8 page 11 of the INA114 datasheet (100uA current sources).

Run those diodes towards 200C and report back...

For a 100R platinum resistance thermometer it is 1mA. At a guess, a Pt-1000 would take 316uA.

Taking the Pt-100 at 1mA as the standard, 100uW.

Larsen's classic paper on micro-degree thermostats includes an expersion for the noise-limited sensitivity of a resistance sensor

Larsen N T 1968 Rev. Sci. Instrum. vol. pages 39 1=9612

which depends on the power dissipated in the sensor - more power means less noise, but more self-heating. It you excite the bridge from a stable voltage source the self-heating ought to be stable, predictable and measureable.

You can linearise a platinum reistance sensor with a little bit of postive feed-back, allowing the sensing current/brige drive voltage to increase with increasing temperature. Honeywell did it for years.

-- Bill Sloman, Nijmegen

Yes, single digit mw range or less was my idea. Going to use 10uA current sources.

..no..Harley hog.

I'd go for more drive, generally, preferably pulsed. 100 uA will only make 10 millivolts, about 40 uV/degC, so microvolts will start to matter. Thermocouple effects, RF pickup, and amp drift become hazards down there.

John

000 f ,

r?

Serious temperature measurement with resistance thermometers uses an AC-excited Blumlein bridge with a ratio-transformer setup to balance the bridge.

Larsen's classic paper on his microdegree thermostat

Larsen N T Rev. Sci. Instrum. 39 1=9612 (1968)

spelled it out in detail. I didn't think much of his use of a narrow band filter before the demodulator to reduce the noise on the demodulated signal - excessively complicated for the benefit it gave - but the paper is comprehensive.

You can use "reversing DC" to get rid of the thermocouple effects - Crystal Semiconductor used to sell as sigma-delta A/D converter set up to handlke exactly this job, and the "no-latency" Linear Technologies

24-bit A/D converter

lends itself to the same kind of use, but that isn't all that compatible with with a ratio transformer bridge, as the excitation frequencies become a bit low.

Ratio-transformers - wound with loosely twisted bundles of identical wires - offer ratios that are accurate and stable to one part in 10^7, which makes them a good deal more accurate and stable than any resistor pair you can buy.

The catch with AC-excited transformer bridges is that you need to balance the reactive impedances as well as the resistive impedances. It make sense to excite them with pure sinewaves so that you can balance everything at one frequency.

You can still amplify and demodulate the out-of-balance signal and digitise the detected output with sufficnet accuracy - over a relatively small range - without throwing away the accuracy of your basic ratio-transformer, but it does make life more complicated.

-- Bill Sloman, Nijmegen

That's overkill for an RTD measurement. And nowadays, what with Susumu resistors and chopamps and 24-bit delta-sigma ADCs, the AC thing is less appealing. It has its own problems, like phase shift, not to mention the complexity.

John

I don't think any of those get rid of thermocouple voltages though, do they?

--

John Devereux

You can use an analog mux and a delta-sigma ADC, and measure the RTD voltage with and without the excitation applied, and that takes out any tc voltages or DC offsets.

All you need is a reasonably stable (but not accurate) voltage reference, one good thinfilm resistor, a cmos mux, and a differential-input delta-sigma ADC. Measure various things and do the math.

John

f e

ore

I've had problems using current a source, the impedance seen by each wire to the rtd is so different there's very little common mode rejection

-Lasse

OK, still AC excitation then :)

Simpler than sinusoidal or reversal though - thanks!

--

John Devereux

At toasty 300K+ temperatures anyhoo.

a 1000

to

0

n of

tie,

? ,

s,

s

Which is why I went on to mention reversing DC, in a section which you repressed with an unmarked snip.

Your Susumui resistors aren't competitive with a properly wound ratio transformer and no better than your used to be able to buy from Vishay.

Here's the rest of what I wrote

------

You can use "reversing DC" to get rid of the thermocouple effects - Crystal Semiconductor used to sell as sigma-delta A/D converter set up to handlke exactly this job, and the "no-latency" Linear Technologies

24-bit A/D converter

lends itself to the same kind of use, but that isn't all that compatible with with a ratio transformer bridge, as the excitation frequencies become a bit low.

Ratio-transformers - wound with loosely twisted bundles of identical wires - offer ratios that are accurate and stable to one part in 10^7, which makes them a good deal more accurate and stable than any resistor pair you can buy.

The catch with AC-excited transformer bridges is that you need to balance the reactive impedances as well as the resistive impedances. It make sense to excite them with pure sinewaves so that you can balance everything at one frequency.

You can still amplify and demodulate the out-of-balance signal and digitise the detected output with sufficnet accuracy - over a relatively small range - without throwing away the accuracy of your basic ratio-transformer, but it does make life more complicated.

------ Not only do you fail to notice what I post on electronics, but you also go out of your way to ignore the interesting bits when you do see it.

-- Bill Sloman, Nijmegen

y s

Crystal Semiconductors did sell the CS5520 20-bit bridge transducer A/ D converter which supported low frequency,=91reversing DC=92, AC-bridge excitation, which provided an even simpler way of getting rid of offsets - it did the math.

does allow for self-offset calibration, which may do the same job

-- Bill Sloman, Nijmegen

Interesting..but that feedback would have to have a parabolic function - not linear..