thermocouple input, 0..20mA output

[Eur van Andel]

Hi

I needed a circuit that converted a thermocouple (40 uV/K) to a 0..20mA input
of a PLC. Normal opamps can't touch the rails and I didn't like creating a
negative voltage. All I had was 24V DC. Here is my circuit:
(fixed pitch)

+24V DC +24V DC
| |
| |
+--------------------+
| | | |
| |\| 1k | |
+-|-\ ___ |/ |
| >--|___|--| |
/---------|+/ |> | 0.8 Ohm current output: 0..20mA
* |/| | | ___
\---------------+ +---+--|___|+->|-->|-->|--o--+
thermocouple | | | |
0..16mV | +--------------------+ .-. inside PLC:
0..400C | | | 250 Ohm to GND
| | |
| '-'
| |
| |
=== ===
GND GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de


The thermocouple generates about 16mv @ 400C (400*40uV), there is no
cold-junction compensation (with a working point of 300C, I don't need one).
The opamp tries to match the 16mv difference and raises its output. The NPN
boosts the current (if your opamp can't source 20mA) until 20mA*0.8Ohm = 16mV
over the resistor, which is fed back into the negative opamp input.
Most opamp outputs and inputs can't come near the rails and since it is current
output only, I added some diodes between the current measuring ressitor and the
PLC input. That way the thermocouple and the opamp in- & outputs are raised far
enough from the bottom rail.

The 1k is a currrent limiting resistor to protect the NPN transistor.
Measure the polarity of the thermocouple while holding it in a flame.

 

[James Meyer]

Most opamp outputs and inputs can't come near the rails and since it is current
output only, I added some diodes between the current measuring ressitor and the
PLC input. That way the thermocouple and the opamp in- & outputs are raised far
enough from the bottom rail.

Wouldn't a single resistor work just as well as a string of diodes?

Jim

 

[Winfield Hill]

Eur van Andel wrote...

I needed a circuit that converted a thermocouple (40 uV/K) to a 0..20mA
input of a PLC. Normal opamps can't touch the rails and I didn't like
creating a negative voltage. All I had was 24V DC. Here is my circuit:

[ see edited version below ]

The thermocouple generates about 16mv @ 400C (400*40uV), there is no
cold-junction compensation (with a working point of 300C, I don't need
one). The opamp tries to match the 16mv difference and raises its output.
The NPN boosts the current (if your opamp can't source 20mA) until
20mA*0.8Ohm = 16mV over the resistor, which is fed back into the negative
opamp input. Most opamp outputs and inputs can't come near the rails and
since it is current output only, I added some diodes between the current
measuring ressitor and the PLC input. That way the thermocouple and the
opamp in- & outputs are raised far enough from the bottom rail.

The 1k is a currrent limiting resistor to protect the NPN transistor.
Measure the polarity of the thermocouple while holding it in a flame.

A pretty nice circuit. I've taken the liberty to make a few changes.

All single-supply opamps, and most low-cost chopper opamps, etc., work
down to the negative rail, so the diodes aren't necessary. I replaced
the 0.8 ohms with a more common 1.0 ohms, and made up for the change by
adding a little gain with the opamp. This also provides a possibility
for different temperature ranges, or gain calibration for various types
of thermocouples.

.. ,-----------+----------- +12 to +28V DC
.. | |
.. ,-------2k49---------,
.. | |\| select | |
.. +--|-\ |/ |
.. | | >-- 1k --| |
.. o--+-2k----|+/ |\> | current output:
.. / | | |/| | | 0..20mA
.. * 470k 10k0 | '---+-- 1R00 --+----o
.. \ | | | LT1013 | |
.. o--+----+-------------------------------' .-. inside PLC:
.. thermocouple | | | 250 ohms
.. 0..16mV | | | to GND
.. 0..400C | '-'
.. | |
.. === ===
.. GND GND

A resistor on the opamp's + input cancels voltage offset errors from
the input bias currents and also protects the opamp in case of static
discharge into the thermocouple wiring. Adding a pair of back-to-back
diodes from the opamp's + input to the other thermocouple lead would
provide further protection. The 470k input resistor prevents a high
output current in the event of a disconnected thermocouple (but watch
out, because the unit will falsely indicate +175C or so).

 

[Spehro Pefhany]

Eur van Andel wrote...

I needed a circuit that converted a thermocouple (40 uV/K) to a 0..20mA
input of a PLC. Normal opamps can't touch the rails and I didn't like
creating a negative voltage. All I had was 24V DC. Here is my circuit:

[ see edited version below ]

The thermocouple generates about 16mv @ 400C (400*40uV), there is no
cold-junction compensation (with a working point of 300C, I don't need
one). The opamp tries to match the 16mv difference and raises its output.
The NPN boosts the current (if your opamp can't source 20mA) until
20mA*0.8Ohm = 16mV over the resistor, which is fed back into the negative
opamp input. Most opamp outputs and inputs can't come near the rails and
since it is current output only, I added some diodes between the current
measuring ressitor and the PLC input. That way the thermocouple and the
opamp in- & outputs are raised far enough from the bottom rail.

The 1k is a currrent limiting resistor to protect the NPN transistor.
Measure the polarity of the thermocouple while holding it in a flame.

A pretty nice circuit. I've taken the liberty to make a few changes.

All single-supply opamps, and most low-cost chopper opamps, etc., work
down to the negative rail, so the diodes aren't necessary. I replaced
the 0.8 ohms with a more common 1.0 ohms, and made up for the change by
adding a little gain with the opamp. This also provides a possibility
for different temperature ranges, or gain calibration for various types
of thermocouples.

. ,-----------+----------- +12 to +28V DC
. | |
. ,-------2k49---------,
. | |\| select | |
. +--|-\ |/ |
. | | >-- 1k --| |
. o--+-2k----|+/ |\> | current output:
. / | | |/| | | 0..20mA
. * 470k 10k0 | '---+-- 1R00 --+----o
. \ | | | LT1013 | |
. o--+----+-------------------------------' .-. inside PLC:
. thermocouple | | | 250 ohms
. 0..16mV | | | to GND
. 0..400C | '-'
. | |
. === ===
. GND GND

A resistor on the opamp's + input cancels voltage offset errors from
the input bias currents and also protects the opamp in case of static
discharge into the thermocouple wiring. Adding a pair of back-to-back
diodes from the opamp's + input to the other thermocouple lead would
provide further protection. The 470k input resistor prevents a high
output current in the event of a disconnected thermocouple (but watch
out, because the unit will falsely indicate +175C or so).

Usually we want an instrument such as this (for temperatures above
room temperature) to fail at somewhat about the 20mA range when the
T/C opens so that heat source will turn *off*. For a general purpose
instrument, it's thus necessary to add another bit of circuitry to
limit the current to 21-25mA or so, as well as putting a bit of bias
current through the sensor (perhaps the op-amp Ib if it's the right
polarity and about the right value, but I usually like to control it).
If it's a one-off hack for a specific application, some series
resistance can be added on the output that limits the current,
assuming the loop supply voltage is known and regulated, and assuming
the load resistance is known. Obviously without cold-junction
compensation, RF filtering, and output polarity protection, this
circuit is very limited in application anyhow.

Also, note that all these circuits will behave rather badly with
grounded-junction thermcoouples (which are otherwise generally
preferred). Adding isolation while maintaining loop-power is
non-trivial.


Best regards,
Spehro Pefhany

 

[Rich Grise]

Hi

I needed a circuit that converted a thermocouple (40 uV/K) to a 0..20mA
input of a PLC. Normal opamps can't touch the rails and I didn't like
creating a negative voltage. All I had was 24V DC. Here is my circuit:
(fixed pitch)

If you're going into a PLC, you probably want 4-20 mA where 4 is the lowest
possible value. Typically, in 4-20 mA control loops, 0 means a fault of some
kind.

Have Fun!
Rich

 

[Winfield Hill]

Spehro Pefhany wrote...

Usually we want an instrument such as this (for temperatures above
room temperature) to fail at somewhat about the 20mA range when the
T/C opens so that heat source will turn *off*. For a general purpose
instrument, it's thus necessary to add another bit of circuitry to
limit the current to 21-25mA or so, as well as putting a bit of bias
current through the sensor (perhaps the op-amp Ib if it's the right
polarity and about the right value, but I usually like to control it).
If it's a one-off hack for a specific application, some series
resistance can be added on the output that limits the current,
assuming the loop supply voltage is known and regulated, and assuming
the load resistance is known. Obviously without cold-junction
compensation, RF filtering, and output polarity protection, this
circuit is very limited in application anyhow.

Also, note that all these circuits will behave rather badly with
grounded-junction thermcoouples (which are otherwise generally
preferred). Adding isolation while maintaining loop-power is
non-trivial.

"behave rather badly" = fail completely! It seems clear this
circuit is entirely too simple for most serious applications.

Didn't we have a thread on isolated-sensor 4-20mA loop circuits
a few years ago?

 

[Tony Williams]

Also, note that all these circuits will behave rather badly with
grounded-junction thermcoouples (which are otherwise generally
preferred). Adding isolation while maintaining loop-power is
non-trivial.

I liked (aka made money with) the flying capacitor
with thermocouples. Auto-zero the amplifier on the
cold junction whilst the cap is over on the t/c.
Solid state switching if the CMV is known to be low
(eg, just to avoid any earth loop), or use a relay
for high CMV and/or extreme CMRR.

 

[Spehro Pefhany]

Spehro Pefhany wrote...

"behave rather badly" = fail completely! It seems clear this
circuit is entirely too simple for most serious applications.

Didn't we have a thread on isolated-sensor 4-20mA loop circuits
a few years ago?

Not specifically, that I recall, though we've talked about some
elements of the problem.

Best regards,
Spehro Pefhany

 

[Fred Bloggs]

Also, note that all these circuits will behave rather badly with
grounded-junction thermcoouples (which are otherwise generally
preferred). Adding isolation while maintaining loop-power is
non-trivial.

The OP mentioned that the junction is in a flame- so probably not grounded.

 

[Spehro Pefhany]

The OP mentioned that the junction is in a flame- so probably not grounded.

That's not necessarily true. It's common to TIG weld the junction into
the end of a metal protection tube or to TIG weld a mineral-insulated
swaged T/C junction together with the outer shell, which is in turn
grounded by the mounting bracket or compression fitting. Insulating
the junction electrically from the protection tube (if present) slows
down the response, costs more, and tends to be less reliable.

Not so much of a factor in a relatively low temperature system such as
this one, but I've gotten burned by DC leakage in high temperature
insulation affecting accuracy. Best to keep relatively high DC
voltages off the T/C if possible- AC can be filtered out easily.

Best regards,
Spehro Pefhany

 

[Eur van Andel]

[ see edited version below ]

The thermocouple generates about 16mv @ 400C (400*40uV), there is no
cold-junction compensation (with a working point of 300C, I don't need
one). The opamp tries to match the 16mv difference and raises its output.
The NPN boosts the current (if your opamp can't source 20mA) until
20mA*0.8Ohm = 16mV over the resistor, which is fed back into the negative
opamp input. Most opamp outputs and inputs can't come near the rails and
since it is current output only, I added some diodes between the current
measuring ressitor and the PLC input. That way the thermocouple and the
opamp in- & outputs are raised far enough from the bottom rail.

A pretty nice circuit. I've taken the liberty to make a few changes.

You're welcome.

All single-supply opamps, and most low-cost chopper opamps, etc., work
down to the negative rail, so the diodes aren't necessary. I replaced
the 0.8 ohms with a more common 1.0 ohms, and made up for the change by
adding a little gain with the opamp. This also provides a possibility
for different temperature ranges, or gain calibration for various types
of thermocouples.

. ,-----------+----------- +12 to +28V DC
. | |
. ,-------2k49---------,
. | |\| select | |
. +--|-\ |/ |
. | | >-- 1k --| |
. o--+-2k----|+/ |\> | current output:
. / | | |/| | | 0..20mA
. * 470k 10k0 | '---+-- 1R00 --+----o
. \ | | | LT1013 | |
. o--+----+-------------------------------' .-. inside PLC:
. thermocouple | | | 250 ohms
. 0..16mV | | | to GND
. 0..400C | '-'
. | |
. === ===
. GND GND

A resistor on the opamp's + input cancels voltage offset errors from
the input bias currents and also protects the opamp in case of static
discharge into the thermocouple wiring. Adding a pair of back-to-back
diodes from the opamp's + input to the other thermocouple lead would
provide further protection.

You are very right. The circuit (probably the opamp) has died already in the
industrial environment where it was used. The 2 kW heater is switched by
turning on a toroid transformer, which generates huge EMF pulses. The
optocouple leads are isolated and coaxial at the sensor end, but after 20" is
is a simple twin wire, non-twisted, about 3' long.

I'll post the next version here too.

 

[Eur van Andel]

If you're going into a PLC, you probably want 4-20 mA where 4 is the lowest
possible value. Typically, in 4-20 mA control loops, 0 means a fault of some
kind.

This PLC can do 0..20mA as well, which makes it easier for me.

http://www.unitronics.com/pdf/m90_91/m90-19-b1a.pdf

If I had to do it, I'd use a PIC. I don't like PLCs.

 

[Eur van Andel]

That was the testing method.

The thermocouple is in a heated Aluminum block. It is a coaxial type K juction,
isolated in a stainless (316) sleeve.

Farnell 424-8326 exactly this one
Conrad 120757 similar, not the same

 

[Spehro Pefhany]

That was the testing method.

The thermocouple is in a heated Aluminum block. It is a coaxial type K juction,
isolated in a stainless (316) sleeve.

Farnell 424-8326 exactly this one

That mineral insulated one is isolated-junction. Made by Labfacility
in the UK. I wonder if the same guy still runs the place- anyone
happen to know? A big jolly fellow, can't recall his name atm.

Best regards,
Spehro Pefhany

 

[Eur van Andel]

You're welcome.

It works, too.

http://www.fiwihex.nl/couple_amp/

Index of /couple_amp
Parent Directory
couple_amp_0-20ma.brd
couple_amp_0-20ma.sch
couple_amp_brd.png
couple_amp_sch.png
p1010062.jpg

SCH and BRD file in Eagle format. (www.cadsoft.de) For these board sizes, Eagle
is Freeware. Also PNG's of board and schematic and the circuit on my bench,
with ugly 10k pot mod for the 2k2, which doesn't work?

2k2 gives a zero current of about 2.7 mA.

I forgot the 2k as well :-|

 

[Spehro Pefhany]

SCH and BRD file in Eagle format. (www.cadsoft.de) For these board sizes, Eagle
is Freeware. Also PNG's of board and schematic and the circuit on my bench,
with ugly 10k pot mod for the 2k2, which doesn't work?

2k2 gives a zero current of about 2.7 mA.

I forgot the 2k as well :-|

Could you have picked a worse op-amp for use with a thermocouple?

Best regards,
Spehro Pefhany