PT1000 circuit

Hi all,

Need your advice....

I'm building a prototype of a bench-top instrument. I need to measure absol ute temperature using a RTD pt1000 sensor with short lead wires (

Reply to
jmariano
Loading thread data ...

You can build a signal conditioning stage with a single op-amp.

Put a series resistor from your sensor to a positive reference. Something like 250uA is appropriate.

Build a positive gain stage with appropriate output for your ADC (zero

the output fed back to the non-inverting input of the op-amp. Select theat resistor (should be hunreds of K ohms) so that the output

(mid-scale)

The resistor linearizes the non-linear response of the sensor, which is slightly more than a constant current.

The resistor values interact, so it's a PITA to change ranges, but this is all do-able with ordinary parts from an online supplier like Digikey.

Algarve eh? I'd like to pop over and show you on a whiteboard.

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

Hi Spehro

Thanks for your input.

You mean voltage reference? How about current excitation? In this particula r case I don't get the pros and cons of using one or the other... On the other hand NS suggest using a 2 ampop configuration to built a float ing current source (AN1559) but LM334 claims to be floating. So, why should I built one if I can use one of the shelf?

th a bit of

Linearize the transfer function of the sensor or the voltage divider equati on? I will be using the Callender-Van Dusen equation(s) on the microcontrol ler to get T from R so I don't understand the need to linearize the sensor output.

Please do that, you'll be very welcome here. It will be a pressure to talk to a fellow engineer. No such a thing were, only golf courses, beautiful be aches and girls in bikinis. Very boring.... :-)

Regards

Reply to
jmariano

Den onsdag den 14. maj 2014 13.45.17 UTC+2 skrev jmariano:

lar case I don't get the pros and cons of using one or the other...

ating current source (AN1559) but LM334 claims to be floating. So, why shou ld I built one if I can use one of the shelf?

with a bit of

tion? I will be using the Callender-Van Dusen equation(s) on the microcontr oller to get T from R so I don't understand the need to linearize the senso r output.

k to a fellow engineer. No such a thing were, only golf courses, beautiful beaches and girls in bikinis. Very boring.... :-)

if you drive it with a current source you get a voltage that is linear with resistance thus (almost) linear with temperature

if you drive it with a voltage source and a resistor your voltage is no lon ger linear with resistance, you can of course fix it in software but you'll nee d more bits in the ADC

-Lasse

Reply to
Lasse Langwadt Christensen

You can make it really linear by using a negative resistance, iirc about

-25 times the 0 C resistance of the sensor for -150C to +500C. (The exact optimum depends a bit on the range you want it to operate over, but it's easy to find the right number.)

Also make sure to use a reference, resistors, and ADC with a low enough tempco--the RTD only moves 3500 ppm/K, so if your resistors are 200 ppm, you'll be pretty sensitive to the temperature of the instrument.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Have a look at you gmail mailbox.

--

-TV
Reply to
Tauno Voipio

If you don't mind doing some math, put the RTD in series with a good resistor, the pair across a reasonably stable power supply. Measure the voltage across the RTD and the voltage across the resistor (or the supply voltage, if that's easier), and divide to compute the RTD resistance, then do the CVD equation. The measurement is ratiomatric on the reference resistor. Use a Susumu 0.05% thinfilm maybe. This works great with a dual-channel, diff-input delta-sigma ADC.

In our climate, you get to really admire a girl in a well-fitting parka.

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

One way uses a few more parts (a current source and a separate amplifier) but is a bit simpler to calculate and will have fewer weird values of resistors. So you have a current source (transistor, op-amp, reference, say) and an amplifier.

The other way uses only one op-amp and feeds back the output to maintain either a constant current, or a bit more than that, which linearizes an RTD pretty well, and you can probably avoid the equations to the accuracy of the RTD.

The current source is effectively one reference and the ADC usually has another, so any drift in one relative to the other results in errors. I like converting the reference to a current and bouncing it off the positive rail with a precision current mirror, which is a dual op-amp, two precision resistors (can be the same value or a matched pair) and two transistors. More parts, but it can be very precise. I would not use an LM334.

If you want to use the equations you can use a constant current. The analog linearization was common back in the day when the equations were difficult to do in a microcontroller. The RTD is almost linear itself when excited with a constant current so there is no appreciable downside to doing the linearization in the micro (no loss of resolution as there would be with a really nonlinear sensor such as a thermistor over a wide temperature range).

Reply to
Spehro Pefhany

icular case I don't get the pros and cons of using one or the other...

floating current source (AN1559) but LM334 claims to be floating. So, why should I built one if I can use one of the shelf?

The LM334 is good, but not quite up to PT100/Pt1000 bridge standards.

o

is slightly more than a constant current.

quation? I will be using the Callender-Van Dusen equation(s) on the microco ntroller to get T from R so I don't understand the need to linearize the se nsor output.

It's all in the hardware, which is cheaper and quicker and harder for the p rogrammer to mess up.

talk to a fellow engineer. No such a thing here, only golf courses, beautif ul beaches and girls in bikinis. Very boring.... :-)

with resistance thus (almost) linear with temperature

longer linear with resistance, you can of course fix it in software but yo u'll need more bits in the ADC.

All perfectly correct. I worked out the linearising trick for myself back i n 1979, only to have some half-wit dump it because it was "positive feedbac k" and thus "made the circuit unstable", rather neglecting that it was gain of 1.04, and it takes a positive feedback of 2 to get instability.

Honeywell put the same scheme into their Pt100 sensing modules a few years later. It doesn't do a perfect job of linearisation, but it's pretty good.

I ended up with a four op amp circuit - which my boss hated - so I had to d raw the one, two and three op amp versions, showing how the extra op amps c ould be progressively cheaper and let you get away with progressively cheap er passive components.

One of the cuter aspects of my circuit was the use of a four 10k (IIRR) tig ht tolerance metal film resistors on a common substrate. They tracked to ab out 5ppm/C, which helped. That didn't make it into production either.

--
Bill Sloman, Sydney
Reply to
Bill Sloman

Hmmm, wouldn't it make sense to use a ratiometric measurement to eliminate the error of *both* references? Add a low tempco resistance to the constant current path between the PT1000 and ground. Use this voltage as your reference for the ADC. Then the only temperature related errors are the reference resistance and any other parts used to get this voltage to an appropriate level for the ADC.

--

Rick
Reply to
rickman

Yes, you can do that, but it can be challenging to keep the noise down when you attach an antenna to the reference. I've used that method for relatively low performance setups.

The method I suggest is also dependent only on one absolute resistor value and resistor ratios.

Best regards Spehro Pefhany

Reply to
Spehro Pefhany

Thin film resistor arrays do track pretty well.

The one time I used the negative-resistance trick, I used a ratiometric A/D, one resistor in series with the RTD, and hung the positive-FB op amp off to one side, i.e. I didn't use it for the measurement, just the correction. That made me 25 times less sensitive to its offset. The negative resistance also cancelled the top resistor in the bridge.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Den onsdag den 14. maj 2014 17.02.59 UTC+2 skrev John Larkin:

ular case I don't get the pros and cons of using one or the other...

oating current source (AN1559) but LM334 claims to be floating. So, why sho uld I built one if I can use one of the shelf?

with a bit of

ation? I will be using the Callender-Van Dusen equation(s) on the microcont roller to get T from R so I don't understand the need to linearize the sens or output.

stor,

oss the

on. The

gma

yes it works well, but you need quite a few more bits that 12 if you want t o do the linearizing in software and get 0.1'C

-Lasse

Reply to
Lasse Langwadt Christensen

20 bits (e.g. LT2420) is more than enough.

0.1 C will be better than Class A sensing element at its best (0.15 C at

0 C).
--

Tauno Voipio
Reply to
Tauno Voipio

By antenna, do you mean the sensor?

--

Rick
Reply to
rickman

and the wire to the sensor

I prefer a resistor to both ends to keep things symmetric.

I've seen one design with an instrumentation amp and a current source feeding the sensor begin very sensitive to noise, with one wire a high impedance current source the other hard ground the common mode rejection sucked

-Lasse

Reply to
Lasse Langwadt Christensen

Something like this maybe.

formatting link

The upper voltage can also go into the ADC reference, instead of CH1,which has its good and bad points. The usual delta-sigma ADC can be programmed to have a notch at 50 or 60 Hz, which keeps the noise down.

The V+ reference needn't be stable; the measurement is ratiometric on Rref.

--
John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

Yup, can do that, and put a low(er) pass filter on Rref voltage.

Don't do much four-wire, usually three wire sensors.

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

Three wires? So a drive, upper voltage sense, and then ground. (one sensor end is tied to a *hard* ground.) I've done that, but didn't know it was common.

If the OP has limited bits, then why not a bridge.. centered in the middle of his temperature range.

George H.

Reply to
George Herold

Yes, and two or three of the wires have current going through them. If the wires are matched in resistance you can make the leadwire resistance cancel 100%. It's still not as good as a Kelvin connection, but plenty good enough for most control systems.

I think of the ADC as the missing half of the bridge.

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.