Interfacing with RTDs

My application is a temperature controller that uses a surface temperature sensor and Triac to adjust a heating element using a uC using PID control. Although I only need about a +/- .5 degree accuracy, I'm aiming for a temperature range of 0 to 200 degrees C which seems to be comfortably out of the range of most thermistor solutions but seems to be within the capabilities of a standard RTD.

I'm struggling to implement an RTD in my design as the interfacing solutions I have seen require a lot of opamps and a precise reference voltage. Here's specifically the one I'm talking about from this microchip application note:

formatting link

My design uses a transformerless power supply that allows me to detect zero crossing as seen in this microchip application note:

formatting link

I feel like my range and accuracy requirements are modest, do I really need to add regulation for a reference voltage, a constant current source, + and

- voltage sources for op amps in order to use an RTD??? Since I'll be taking my temperature readings in phase with the voltage fluctuation of the power supply, I wonder if regulation for a reference voltage is really necessary. Is there a smoother way to interface with an RTD given my requirements?

--------------------------------------- Posted through

formatting link

Reply to
segwaypirate
Loading thread data ...

If you understand the principles behind the working of the RTD, you ought to be able to make a current source that's proportional to your ADC reference. That'll make the readings ratiometric. Then use rail to rail op-amps. That'll get you down to only needing one supply.

--
Tim Wescott
Wescott Design Services
 Click to see the full signature
Reply to
Tim Wescott

e
.

You might want a combination of thermistor (low temp) and thermocouple (high temp).

ro

What are you measuring? Frequency or voltage?

ed

nd

he

No, just a stable voltage to the uC.

Reply to
linnix

e
.

chip

ro

ed

nd

he

a lot depends on whether you can calibrate the system, or do you expect it to meet accuracy first time it is power up?

will the interface circuit experience temperature extremes too?

the .01% error that article claims requires a lot of constraints, I suspect

Reply to
steve

No op amps needed. A delta sigma ADC w/ differential ref and signal inputs (e.g. LTC2433), a stable reference resistor, and another resistor (stability not important) to lower the overall current. I would change the zener to 8 volts and add a regulator on the 5V, but you might not need it.

5V | crappy 10k | +----- A | RTD 1k nominal @ 0C (up to 1.76k @ 200C) | +----- B | 4.99k good tempco (and must be 2x the highest RTD value) | +----- C | ground

ADC signal is A-B ADC ref is B-C

Bob

Reply to
Bob

Often interfacing to an RTD is done differentially using 2 current sources, one driving the RTD and the second driving an on-board resistor. The difference is measured by means of a difference or instrumentation amp. This allows you to improve the dynamic range of the measurement.

To understand, look at a TI dedicated part like the XTR105. Then try and take the same approach.

For a current source try the LM334 but watch for its thermal performance and use the compensation circuit recommended. Linear makes a better and more expensive part- the number doesn't come to hand at the moment.

--------------------------------------- Posted through

formatting link

Reply to
antedeluvian

There's actually a really nice RTD measurement circuit as an app note at the back of the AD7793 datasheet.

--
Rob Gaddi, Highland Technology
Email address is currently out of order
Reply to
Rob Gaddi

Thanks for all the responses.

To answer some questions:

I'm detecting the zero-crossing only to control the mains phase with a triac I'm not measuring the voltage or frequency.

Only the surface temperature sensor and its wiring will be exposed to the

200 C but the lead length will be less than 2 or 3 feet.

Some calibration is ok at least for now.

Bob,

Thanks for your help. I know its pretty normal to use high resolution ADC's that have a differential input but in my range of 0 to 200 I can't help but think that the 10 bit ADC on board my uC can get me an acceptable result. Shouldn't I be able to use the configuration posted by Bob using two channels of that ADC taking the difference in software?

If anyone is aware of an NTC surface temperature assembly that can operate at 200 C, that would obviously be the better option.

--------------------------------------- Posted through

formatting link

Reply to
segwaypirate

Good luck. In my experience, a "10-bit A/D" on a uController is usually good for about 8 bits of actual, usable resolution.

--
Grant Edwards               grant.b.edwards        Yow! In Newark the
                                  at               laundromats are open 24
 Click to see the full signature
Reply to
Grant Edwards

n
t

ble

Bit 9 should be OK with good PCB layouts. Bit 10 would need properly designed uC with separate power rails. As they think the world is all digital, not all uC supports separate analog power rails.

Reply to
linnix

Modest? Well, 0.5 degree C accuracy implies better than 0.2% total system accuracy in resistance terms, so the error budget for each component of the error is not very large. In fact, the sensor error ALONE for a DIN class A sensor is more than 0.5 degree at 200°C, and B, C grade devices are considerably worse. Maybe you're not actually serious about that 'accuracy' or can trim it later.

As others have said, the reference voltage would normally be designed to cancel out, but you need to do ADC to higher resolution than otherwise and the reference shouldn't change much between readings. Often we're also concerned about leadwire resistance compensation.

Having the sensor and signal conditioning circuitry common with one side of the mains may not be the best idea, although it's certainly do-able in principle. Depends in part if you have to meet some degree of safety and if the controlled surface needs to be safe to touch.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
 Click to see the full signature
Reply to
Spehro Pefhany

Sure, just do all the signal conditioning up front with good components well trimmed and deliver to the micro (say) a clean filtered 0-4V signal for 0-200°C and you may be able to cut corners on the ADC. It's a trade-off. Decent op-amps and precision resistors instead of some more expensive silicon.

Depends on what you find acceptable in performance too.

Maybe, and you might have to add dithering and digital filtering. Assuming a PT-100 DIN alpha = 0.385 standard RTD operated at 1mA nominal, you need an order of magnitude higher than the resolution that a PIC 10-bit ADC offers, so op-amps.

Actually, that's not obvious at all.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
 Click to see the full signature
Reply to
Spehro Pefhany

I'm curious about the type of application that needs 0.5 degree accuracy over that temperature range. Many ovens, heaters, and other process control devices may need repeatability on that order, but absolute accuracy is more questionable. Unless the heating system has infinitely high thermal conductivity and the thermal mass is very high, getting the process to actually stabilize at a temperature within 0.5 degrees seems a difficult chore. Placement of the sensor and tuning the PID loop may be difficult if the temperature is to stabilize in less than an hour or two.

Now, if you can get 0.5 degree resolution and repeatability, then you have a calibration and curve-fitting problem. Whether that is suitable for the problem will depend on the calibration cost and the number of units to be built.

Mark Borgerson

Reply to
Mark Borgerson

You have already lost the game with so many analog components to go wrong.

The way to go with modern components is to have a high-resolution ADC and one precision resistor feeding the RTD from the reference source of the ADC. You'll lose most of the resolution, but there is enough left to get reasonable results. To not lose precision in the calculation, you need at least 32 bit arithmetic carefully tuned to avoid overflows and underflows, but it is doable.

A different story is that the thermal resistances and time constants will play havoc with the control to the attempted precision.

--
Tauno Voipio,
been there, done that ...
 Click to see the full signature
Reply to
Tauno Voipio

That 2 to 3 feet could eat your entire error budget w/o 4-wire or thick enough RTD wires.

Work out the error budget and report back, please.

Why would that be better? Look, the LT2433 is $2, a 0.1% 25ppm/C resistor about $0.20, and put together they take far less than 1 square inch of board. What is it that you're trying to optimize?

Bob

Reply to
Bob

The application is a heated oil bath.

Tauno Voipio,

I'm not sure what you mean by "You have already lost the game with so many analog components to go wrong."

The configuration posted before by Bob sounds about like what you described:

5V | crappy 10k | +----- A | RTD 1k nominal @ 0C (up to 1.76k @ 200C) | +----- B | 4.99k good tempco (and must be 2x the highest RTD value) | +----- C | ground

ADC signal is A-B ADC ref is B-C

Perhaps this is a stupid question but why is it good practice to use a resistor at least 2x the highest RTD value? Wouldn't putting it exactly at the highest RTD value give the greatest possible resolution to the ADC?

Bob are you suggesting I need four wire for a distance of 2 or 3 feet? I'm a little surprised as I wouldn't expect much difference.

The reason I guessed that a conventional NTC surface assembly would be preferable to using an RTD is that since my range isn't that large a NTC could do as good of a job with much less and cheaper parts without using a separate ADC. Also, thermistors are just generally cheaper themselves.

--------------------------------------- Posted through

formatting link

Reply to
segwaypirate

Using RTD's and similar devices for Cryogenic temperature ranges, I have adopted the four-wire measurement system. The RTD (or other element) is configured into a wheatstone bridge arrangement. Your electronics needs to supply a stable voltage source to feed two opposite corners of the bridge. You then measure the differential voltage from the other two corners into your chosen ADC. It tends to be a quieter, easier measurement to make.

There are plenty of references for Wheatstone Bridges and four wire measurement techniques on line.

--
********************************************************************
Paul E. Bennett...............
 Click to see the full signature
Reply to
Paul E. Bennett

2x is specific to the LTC2433. Look at the data sheet. I'm suggesting you do a ratiometric conversion and the signal FS of the LTC2433 is +-ref/2 IIRC. A different ADC may need something else.

I'm trying to prod you into working out the errors for yourself. A Pt RTD 100 has a defined resistance for a given temperature. If you add 0.2 ohms of total lead resistance, you've already failed your 0.5C accuracy spec.

They're cheaper because they aren't as accurate. Or. more precisely: they aren't as interchangable. Despite your original post asking for

0.5C accuracy to 200C, you actually seem rather unconcerned with accuracy, so a thermistor may be fine.

Good luck, Bob

Reply to
Bob

Second the use of an AD77xx A/D and datasheet design. Worked good for a lazy fool like me back when I didn't have the time to do a proper design...

Reply to
Jim Stewart

Well, here you go, knock yourself out.

formatting link

Say ~32K at 0C and ~65 ohms at 200C. Good luck with a 10-bit ADC.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
 Click to see the full signature
Reply to
Spehro Pefhany

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.