Opamp offset question...

[Klaus Vestergaard Kragelund]

On 29/11/2021 02.52, Anthony William Sloman wrote:

On Monday, November 29, 2021 at 9:46:07 AM UTC+11, Chris Jones wrote:
On 29/11/2021 08:35, Klaus Kragelund wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential
coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inherited .

Problem is the large offset voltage of the opamp is amplified producing
large errors

We are contemplating production calibration, but I am worried that the
offset isn\'t stable after the calibration has been done

In literature the offset comes from mismatch of the long tailed pair.
Is that expected to be stable, so a calibration done in production also
cancels out after 10 years operation?

By the way, my suggestion is to ditch the opamp and feed the signal
directly into the ADC. All the opamp errors disappears then

I just need a big sample cap to reduce charge injection problems from
ADC channel switching and sample/hold effects

If the offset has a temperature dependence then you would not be able to
take it out without doing calibrations at multiple temperatures. If the
opamp and the RTD are not always at the same temperature then it may be
impossible to fix by calibration.

Can you swap the opamp for a type that has less offset? There are ones
that would have the same pinout.

https://www.analog.com/media/en/technical-documentation/data-sheets/LT1013-LT1014.pdf

The LT1013 is quite nice, but not all that cheap. You can run a Pt1000 bridge with more volts across the platinum sensor than you can with a nominally one hundred ohm sensor - those are normally rated for 1mA through the sensor, which is only 100mV and 0.3% of that per degree Celcius is only 30uV per degree.

At that level you start seeing thermocouple voltages in the copper to invar to alumium-on-silicon junctions you\'ve got around the op amp, and you rapidily start thinking that the complexities of AC bridge excitation might be worth the effort.

The LT1013 is good. It was preferred when I did space stuff. But it\'s
too expensive

 

[Klaus Vestergaard Kragelund]

On 29/11/2021 08.12, Jasen Betts wrote:

On 2021-11-28, Klaus Kragelund <klauskvik@hotmail.com> wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

Can you drive it with AC instead?

That is not possible I am affraid

 

[Klaus Vestergaard Kragelund]

On 29/11/2021 10.52, Tauno Voipio wrote:

We jettisoned the amplifier and used a minimal component
count solution: A 20 bit sigma-delta ADC and a precision
10k resistor to the ADC reference and the sensor between
the input and ground.

This will lose 90% of the range of the converter, but we
still have more than 16 bits of resolution left. With
suitable microprocessor linearization, it is more than
needed for the sensor tolerances in a range of say, -50C
to +200C.

We have a 12bit ADC, so raw performance is not good: We do use
oversampling, so that can get the performance back on track

 

[Klaus Vestergaard Kragelund]

On 29/11/2021 15.04, Spehro Pefhany wrote:

On Sun, 28 Nov 2021 22:35:11 +0100, Klaus Kragelund
klauskvik@hotmail.com> wrote:

Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

We are contemplating production calibration, but I am worried that the offset isn\'t stable after the calibration has been done

In litterature the offset comes from mismatch of the long tailed pair. Is that expected to be stable, so a calibration done in production also cancels out after 10 years operation?

By the way, my suggestion is to ditch the opamp and feed the signal directly into the ADC. All the opamp errors disappears then

I just need a big sample cap to reduce charge injection problems from ADC channel switching and sample/hold effects

Regards

Klaus

LMV358 is not a precision op-amp- its a general purpose low-voltage
part with Vos as much as +/-9mV. TCVos is not guaranteed and is
\'typically\' 5uV/°C. You can translate that into degrees error from
whatever circuit you are using.

Not sure about your comment about the resistor, you need at least one
precision resistor somewhere or you won\'t get a voltage as a function
of the sensor resistance. Maybe you\'ve got a bridge circuit given your
\"differential coupling\" comment- in which case all the resistors
affect the accuracy in general and the zero in particular.

If there is 2V across the sensor then the output at the sensor is
about 7.7mV/°C ,The Vos of the LMV358 is a bit more than 1°C error
worst-case, if there is 2V across the sensor. If you have a bridge
configuration a 1% error in one of the resistors represents several
degrees C error (20mV or so). So unless you\'re using precision (like
0.1% resistors) even the LMV358 is not your main issue.

With the series resistor, the excitation should be taken from the ADC
reference voltage, obviously, so the ADC reading is ratiometric. Using
the supply voltage as a reference for both can cause noise issues.

Good comments

About the ADC reference it is taken from the SMPS with post LC filter.
So there is a little difference between the reference voltage ripple and
the voltage fed to the pullup resistor for the PT1000

One option to get better performance is to sync the PT1000 sampling to a
time slice when the microcontroller has steady current consumption, but
that makes the SW more complicated and susceptible if changes are done
in other places in the SW

There are advantages and disadvantages to haveing an op-amp in there,
I would probably stay with it in most situations because it allows a
nice low pass filter. Sounds like your ADC has no PGA or buffer
amplifier.

You can get op-amps that are pin-compatible and have offset voltages
in the microvolts and TCVos in low tens of nV/°C. An old-fashioned
non-zerodrift precision type may be more resistant to EMI, though.

That said, you can probably count on the change in offset being within
100uV at the same temperature far off into the future, even with the
crappy LMV358. If the voltage across the sensor is 2V, that\'s one or
two hundredths of a °C. If the Pt1000 sees much in the way of
temperature swing or mechanical stress that\'s in the wash.

Yeah. problem is that we can have wide temperature range, so I am afraid
the calibration is not valid

 

[Lasse Langwadt Christensen]

tirsdag den 30. november 2021 kl. 01.20.00 UTC+1 skrev Klaus Kragelund:

On 29/11/2021 08.12, Jasen Betts wrote:
On 2021-11-28, Klaus Kragelund <klau...@hotmail.com> wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

Can you drive it with AC instead?


That is not possible I am affraid

could you toggle between two currents ?

 

[Phil Hobbs]

Klaus Vestergaard Kragelund wrote:

On 28/11/2021 23.23, Jeroen Belleman wrote:
On 2021-11-28 22:44, Klaus Kragelund wrote:
28.11.21 22:35, Klaus Kragelund   wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential
coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified
producing large errors

We are contemplating production calibration, but I am worried that
the offset isn\'t stable after the calibration has been done

In litterature the offset comes from mismatch of the long tailed
pair. Is that expected to be stable, so a calibration done in
production also cancels out after 10 years operation?

By the way, my suggestion is to ditch the opamp and feed the signal
directly into the ADC. All the opamp errors disappears then

I just need a big sample cap to reduce charge injection problems
from ADC channel switching and sample/hold effects

Regards

Klaus

I have been doing some tests. Quite odd, but hitting the circuit with
a hot airflow of 60 Degrees creates 20mV offset on the output of the
opamp. The specs define temperature drift of 5uV/K, so something
weird is going on...

--
Klaus

The offset and its drift are specified at the input.

Yes. The differential gain is 6 times, so even for a 60 degrees
temperature change, the output would only change by 2mV (60*5uV*6)

Probably some horrible temperature gradient times a huge thermocouple
coefficient. Silicon has huge TC coeffs versus any metal, for instance.

It should be much smaller in an isothermal test.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[John Larkin]

On Tue, 30 Nov 2021 01:12:55 +0100, Klaus Vestergaard Kragelund
<klauskvik@hotmail.com> wrote:

On 28/11/2021 23.26, jlarkin@highlandsniptechnology.com wrote:
On Sun, 28 Nov 2021 22:44:27 +0100, Klaus Kragelund
klauskvik@hotmail.com> wrote:

28.11.21 22:35, Klaus Kragelund wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

We are contemplating production calibration, but I am worried that the offset isn\'t stable after the calibration has been done

In litterature the offset comes from mismatch of the long tailed pair. Is that expected to be stable, so a calibration done in production also cancels out after 10 years operation?

By the way, my suggestion is to ditch the opamp and feed the signal directly into the ADC. All the opamp errors disappears then

I just need a big sample cap to reduce charge injection problems from ADC channel switching and sample/hold effects

Regards

Klaus

I have been doing some tests. Quite odd, but hitting the circuit with a hot airflow of 60 Degrees creates 20mV offset on the output of the opamp. The specs define temperature drift of 5uV/K, so something weird is going on...


Got a schematic?


It\'s quite close to this one:

https://i.stack.imgur.com/8uwas.png

Except the R1 and R3 is replaced by a fixed voltage reference. Gain is
about 6 of the diff amp. Output feeds into an ADC with RC filter to deal
with charge injection/sample capacitor

Are you seeing resistor tempcos maybe? An RTD is about 4000 ppm/k, so
you\'d need some very good resistors to keep their error contribution
down. If the pullup is 1K, that kills half of the gain too.

The pullup is 3kohm, and the amplifier is used to get full use of the
ADC range (0 to 3.3V)

All resistor are 0.1%/5ppm, so it does account for some error, but does
not explain it all

Would your RTD make enough voltage to drive the ADC directly? Too much
excitation current could self-heat the RTD.

My point about this circuit is to shift to a direct input into the ADC,
then only the pullup has temco. I would then use a pullup with the same
resistance as the PT1000 (1kohm), to maximize the dynamic range. Use a
FET to turn on the pullup to prevent self-heating of the PT1000

If you can turn off the pullup, maybe you can autozero. Turn it off
and measure offset.

What about the RTD+resistor voltage reference? Is it the same as the
ADC ref?

Yes, for ratiometric measurement

Can you auto-zero?

No, it\'s a simple circuit so not an option

I would expect opamp offset to be pretty stable over time, so you
could cal it out.

We like to do RTDs like this:

https://www.dropbox.com/s/kcb705scwytfzty/RTD-ADC.jpg?raw=1

The measurement is purely ratiometric on the Susumu thinfilm resistor.



--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

 

[Lasse Langwadt Christensen]

tirsdag den 30. november 2021 kl. 02.22.04 UTC+1 skrev John Larkin:

On Tue, 30 Nov 2021 01:12:55 +0100, Klaus Vestergaard Kragelund
klau...@hotmail.com> wrote:

On 28/11/2021 23.26, jla...@highlandsniptechnology.com wrote:
On Sun, 28 Nov 2021 22:44:27 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

28.11.21 22:35, Klaus Kragelund wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

We are contemplating production calibration, but I am worried that the offset isn\'t stable after the calibration has been done

In litterature the offset comes from mismatch of the long tailed pair. Is that expected to be stable, so a calibration done in production also cancels out after 10 years operation?

By the way, my suggestion is to ditch the opamp and feed the signal directly into the ADC. All the opamp errors disappears then

I just need a big sample cap to reduce charge injection problems from ADC channel switching and sample/hold effects

Regards

Klaus

I have been doing some tests. Quite odd, but hitting the circuit with a hot airflow of 60 Degrees creates 20mV offset on the output of the opamp. The specs define temperature drift of 5uV/K, so something weird is going on...


Got a schematic?


It\'s quite close to this one:

https://i.stack.imgur.com/8uwas.png

Except the R1 and R3 is replaced by a fixed voltage reference. Gain is
about 6 of the diff amp. Output feeds into an ADC with RC filter to deal
with charge injection/sample capacitor

Are you seeing resistor tempcos maybe? An RTD is about 4000 ppm/k, so
you\'d need some very good resistors to keep their error contribution
down. If the pullup is 1K, that kills half of the gain too.

The pullup is 3kohm, and the amplifier is used to get full use of the
ADC range (0 to 3.3V)

All resistor are 0.1%/5ppm, so it does account for some error, but does
not explain it all

Would your RTD make enough voltage to drive the ADC directly? Too much
excitation current could self-heat the RTD.

My point about this circuit is to shift to a direct input into the ADC,
then only the pullup has temco. I would then use a pullup with the same
resistance as the PT1000 (1kohm), to maximize the dynamic range. Use a
FET to turn on the pullup to prevent self-heating of the PT1000
If you can turn off the pullup, maybe you can autozero. Turn it off
and measure offset.

What about the RTD+resistor voltage reference? Is it the same as the
ADC ref?

Yes, for ratiometric measurement

Can you auto-zero?

No, it\'s a simple circuit so not an option

I would expect opamp offset to be pretty stable over time, so you
could cal it out.





We like to do RTDs like this:

https://www.dropbox.com/s/kcb705scwytfzty/RTD-ADC.jpg?raw=1

The measurement is purely ratiometric on the Susumu thinfilm resistor.

an extra resistor on the ground side puts the differential voltage in
the middle of the input range, and makes the noise \"impedance\" of
the two signal wires similar

 

[Spehro Pefhany]

On Tue, 30 Nov 2021 01:28:01 +0100, Klaus Vestergaard Kragelund
<klauskvik@hotmail.com> wrote:

Yeah. problem is that we can have wide temperature range, so I am afraid
the calibration is not valid

Not just calibration of the electronics - those cheap sensors have
significant hysteresis.

--
Best regards,
Spehro Pefhany

 

[Klaus Vestergaard Kragelund]

On 30/11/2021 01.35, Lasse Langwadt Christensen wrote:

tirsdag den 30. november 2021 kl. 01.20.00 UTC+1 skrev Klaus Kragelund:
On 29/11/2021 08.12, Jasen Betts wrote:
On 2021-11-28, Klaus Kragelund <klau...@hotmail.com> wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

Can you drive it with AC instead?


That is not possible I am affraid

could you toggle between two currents ?

Yes, could be. So you are thinking like syncronous detection? Shifting
the signal to a higher frequency avoiding DC errors

 

[Lasse Langwadt Christensen]

onsdag den 1. december 2021 kl. 00.12.39 UTC+1 skrev Klaus Kragelund:

On 30/11/2021 01.35, Lasse Langwadt Christensen wrote:
tirsdag den 30. november 2021 kl. 01.20.00 UTC+1 skrev Klaus Kragelund:
On 29/11/2021 08.12, Jasen Betts wrote:
On 2021-11-28, Klaus Kragelund <klau...@hotmail.com> wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

Can you drive it with AC instead?


That is not possible I am affraid

could you toggle between two currents ?

Yes, could be. So you are thinking like syncronous detection? Shifting
the signal to a higher frequency avoiding DC errors

something along the time lines of

two currents I1 and I2

I1 * R + offset = V1 , I2 * R + offset = V2

(I2 * R + offset)-(I1 * R + offset) = I2*R-I1*R + offset- offset => V2-V1 = (I2-I1)*R

R = (V2-V1)/(I2-I1)

 

[Klaus Vestergaard Kragelund]

On 01/12/2021 17.16, Lasse Langwadt Christensen wrote:

onsdag den 1. december 2021 kl. 00.12.39 UTC+1 skrev Klaus Kragelund:
On 30/11/2021 01.35, Lasse Langwadt Christensen wrote:
tirsdag den 30. november 2021 kl. 01.20.00 UTC+1 skrev Klaus Kragelund:
On 29/11/2021 08.12, Jasen Betts wrote:
On 2021-11-28, Klaus Kragelund <klau...@hotmail.com> wrote:
Hi

I have a PT1000 circuit where a LMV358 is used in a differential coupling to feed it to an ADC

PT1000 is pull up with a resistor

I would NOT do it like that, this is a design I have inhireted

Problem is the large offset voltage of the opamp is amplified producing large errors

Can you drive it with AC instead?


That is not possible I am affraid

could you toggle between two currents ?

Yes, could be. So you are thinking like syncronous detection? Shifting
the signal to a higher frequency avoiding DC errors

something along the time lines of

two currents I1 and I2

I1 * R + offset = V1 , I2 * R + offset = V2

(I2 * R + offset)-(I1 * R + offset) = I2*R-I1*R + offset- offset => V2-V1 = (I2-I1)*R

R = (V2-V1)/(I2-I1)

That\'s nice, so in effect the dynamic resistance is found between two
load lines