Current measuring resistor calibration

[Petkovic]

Hi, I need to measure current in the range of 1mA to 10A.
The precision I need is 0.1% of the current that I'm measuring i.e for
10A current i need 10mA resolution for 1A I need 1mA resolution and so
forth. Potential across resistors are fed to an Opamp that has offset
and gain trimming networks.
The current is changing within the range limits and I need to measure
current continuously and send the value to computer.
I have built a prototype but I have some problems that I hope you can help.

Because the range is too broad to be measured accurately with only one
resistor, I have used these resistors:

-----------------------------------
0.1 Ohm -> 1A to 10A
1 Ohm -> 100mA to 1A
10 Ohm -> 1mA to 100mA
-----------------------------------

Now Problems that I have :

1- When I try to calibrate each resistor's gain an offset it causes a
gap with other resistors e.g. when current is increasing from 50mA to
200mA and I have to change current measurement resistor from 10 Ohm to 1
Ohm at point of change I read two different currents for 100mA i.e. 10
Ohm resistor reads it 100mA but 1 Ohm resistor reads it 115mA.

2- When current is high (>1A), current measurement resistor becomes hot
and after a while current that it reads change e.g. first it reads 2.16A
and after 10 minutes, it reads the same current 2.65A.

3- If I want to mass produce this, is it an easier way to calibrate
offset and gain of each resistor instead of manually change trim
networks of Opamp?

Thanks in Advance.

 

[Petkovic]

Actually, you could do it with just one resistor assuming that the
current is DC. With a chopper op-amp and careful design, you can get
down to a micro volt. You have to beware of the metals involved.

-----------------------------------------------------------------------
I need 0.1uV resolution if I want to measure 1mA with 0.1 Ohm Resistor
with 0.1% accuracy. I never could measure lower than 10uV in practice
and that only happens if I use every trick that I know. for me 100uV
sounds more practical. So if you excuse me I can't do this unless you
teach me practically how.


-----------------------------------------------------------------------

Look very carefully at how you have things wired. It could be that
the wiring resistance is getting into your measurement.

I assume you didn't let the drop in the switches get into it.

---------------------------------------------------------------------------
I have used very wide tracks to ensure there is no IR drops in tracks
and also in switches.

but even in lower currents this problem exists. I mean 100mA is not much
current for a short(a few Cm) 200 mil track. It should not cause this
problem.
---------------------------------------------------------------------------

 

[Petkovic]

The first resistor dissipates 10 watts max, and the next one
dissipates 1 watt. They'll get hot, and the heat will change their
values and generate thermoelectric potentials. Both will need 4-wire
connections to avoid copper errors, especially the first one.

Use lower-value 4-wire shunts to keep the errors down, and low-drift
diffamps with lots of gain. 10 to 50 mV max drop is reasonable. Do the
zero and span calibrations in the computer.

One good trick is to make provisions for shorting the diffamp input
occasionally and auto-zeroing in software. Then you can tolerate a
cheaper, driftier diffamp. Software is free in production.

As noted elsewhere, when you switch resistors, make sure the
resistance of the switch contacts isn't part of the measurement
resistance. In other words, switch all 4 wires to each shunt.

------------------------------------------------------------------------------
Do you think it is possible to produce a board that can produce very
accurate currents for this purpose ?
I mean how can I make sure that my reference current is let's say
exactly 100mA. I mean (100.0mA). Is there an alternative to check it
with a very accurate galvanometer?
I can produce very accurate voltages using high precision reference ICs
but what about current?
------------------------------------------------------------------------------

 

[Petkovic]

Sure, depending on what exactly is your purpose. That's not clearly
defined here.


Fluke makes bench DVMs that are very accurate on current, like 0.02%
to 10 amps. Use one of them as a tracable cal thing.

--------------------------------------------------------------------------
John, My purpose is to automatically calibrate those current measurement
resistors but for that I need to either know the very exact value of
each resistor or pass a known(reference) current through them and
measure their voltages.
Yes there are accurate measuring instruments but problem is they also
have to be calibrated. If I remember correctly Fluke guarantees its
calibration for just 2 years.
Is it possible to produce a reference current without such instruments?
--------------------------------------------------------------------------

 

[Petkovic]

John,
[Larkin]

My purpose is to automatically calibrate those current measurement
resistors but for that I need to either know the very exact value of
each resistor or pass a known(reference) current through them and
measure their voltages.

---
I don't mean to be unkind, but you don't seem to have grasped the
fundamental fact that if the resistance of your "reference" resistor(s)
changes when current through them changes, then you are lost.

Consider: (View in Courier)


+--[Vin]--[Rs]-+
| |
[R1] [R2]
| |
+-----[RL]-----+
| |
+-----[VL]-----+


Where Vin is the voltage source feeding the circuit,
Rs is its source resistance,
R1 and R2 are the resistances of the wires connecting Vin to the load,
RL is the load resistance, and
VL is the voltage appearing across the load,

then if the resistance of RL changes (because of its temperature or
voltage coefficient of resistance, or both, or whatever) when Vin, Rs,
R1, and R2 are perfectly stable, VL will also change and won't truly
reflect the current in RL.

------------------------------------------------------------------------
John Fields,
The most usual way of measuring currents is to use a resistor as a
current to voltage converter. Yes all resistors have TC,... but this is
another issue. For most precise measurements one should uses a resistor
with lowest possible TC. what you mentioned is a general issue and not
addresses my specific problem. Yes TC is important and you can suppose I
make all my effort to lower that effect as mush as I can.

I think I was not clear enough in previous post so I say it again.

I want to calibrate current measuring resistors -> From calibrate I mean
to know exactly what voltage a specific current produces across that
resistor. If I know this coefficient I can convert the measured voltage
to a valid current value in computer and hopefully then there is no gap
when I change from one resistor to another.
what I said in previous post was that I have two different ways. I said

EITHER:

1- I should learn the exact value of the resistor-> if I know that
then I can calculate a valid value for current : V/R.

OR:

2- I should produce a REFERENCE CURRENT not a reference voltage
If a reference current passes through a resistor tht I don't know its
exact value then I measure voltage and can find the exact value of R
(V/I) and from exact value or R I can have a valid current value(See #1).

------------------------------------------------------------------------

---
Yes, of course:

http://en.wikipedia.org/wiki/Bandgap_voltage_reference

But once you build it what are you going to use to make sure it's
working properly?

-------------------------------------------------------------------------
I should produce a ***REFERENCE CURRENT*** not a reference voltage.
I explicitly mentioned there are many voltage reference ICs. what I want
is either to measure the very exact value of a resistor or produce a
***REFERENCE CURRENT***. And please note that reference currents that I
need are 100mA , 1A, 10A not a few milliamp where there are some
reference current ICs for them.
-------------------------------------------------------------------------

 

[Petkovic]

If you want to be legit, anything that you sell as electrically
accurate has to be tracable to a recognized standard, like NIST (in
the USA) or equivalent. A Fluke that is in cal is tracable; after the
cal interval has expired, it's not.

You can also buy lab standard resistors - very expensive - but they
need to be periodically certified, too.

No instrument can calibrate itself, at least unless it relies on some
fundamental quantum measurement. A primary-standard current source
would cost as much as a thousand Flukes.

Why not buy a Fluke and have it cal'd every two years? Use that to
calibrate whatever it is that you're making.

Since I don't know what you're trying to do, that's all I can say.

----------------------------------------------------------------------
John Larkin,
what you said makes sense to me. Thank you.
My only purpose is to calibrate current measuring resistors.
I want to calibrate with least effort.
----------------------------------------------------------------------

 

[Petkovic]

Better is to go the whole hog and use an alternating current bridge;
this lets you use transformer widings to define your bridge ratio's -
a Blumlein bridge.

National standards laboratories use such bridges to compare and
calibrate standard resistors,with one another and with quantum Hall
effect primary standards of resistance.

Rayner and Kibble's "Coaxial Alternating Current Bridges" discusses
this approach at some length

http://www.amazon.co.uk/Coaxial-Alternating-Current-Bridges-Kibble/dp/0852743890

The book was published in 1984, so it doesn't cover quantum Hall
effect standards of resistance. Rayner and Kibble both worked at the
British National Physical Laboratory (NPL), and they pretty much
reported on what they were doing there then - the work is still going
on

http://www.npl.co.uk/electromagneti...oducts-and-services/coaxial-ac-bridge-systems

The book is out of print at the moment but you can find it in
university libraries - for a long time the NPL would sell you a copy
printed on demand, but they've now stopped doing that.

One of the approaches used to calibrate a resistor was to use a
"quadrature" bridge to compare the resistor to the impedance of a
known capacitance - eventually derived from a Thompson-Lampard
calculable capacitor of around 0.15pF - at a well-defined frequency.

http://www.measurement.gov.au/index...&objectID=D90A574B-BCD6-81AC-15616BB69A99D113


http://en.wikipedia.org/wiki/Four-terminal_sensing

--

----------------------------------------------------------------------------------------
Hi Bill, That was very interesting. I couldn't find enough information
Online on how to build a calculable capacitor , so maybe I can find a
proper book later but I think if I realize that this method is really
practical and applicable, definitely it is very useful for me.
----------------------------------------------------------------------------------------

 

[Petkovic]

A micro can assist resistor switching and calibrate linear output. It
can also monitor temperature changes for compensations. In
productions, you can build lookup tables while calibrating and
programming the micro. The so called "ASIC" are probably pre-
programmed micro.

-------------------------------------------------------------------------------
Hi Linnix, That was also a very interesting idea. In theory what you say
completely makes sense because by creating a look up table one can take
it into account the effect of Tc and also it can minimize non-linearity
related to CMRR of opamp. does anyone knows if this method of creating a
Look up table for calibrating resistors is actually used in any
commercial product?
-------------------------------------------------------------------------------

 

[Petkovic]

It's preposterous. Buy a decent DVM that's certified calibrated.

---------------------------------------------------------------------------------------------
John, I have more than 5 decent DVMs. calibration of all of them have
expired of course. Meanwhile I have experienced different reading even
with calibrated DVMs, one from Fluke and another from Hioki both very
decent DVMs.
---------------------------------------------------------------------------------------------

 

[Petkovic]

We are selling to end-user for a profit, so I guess it's a commerical
product. The main problem we are facing is variations in components
accuracy, with or without temperature changes. With a One-Time-
Programmable micro, we build unique lookup table for each unit. Piror
to programming the chip, we isolate the I/Os from the external circuit
and test it with an external tester. Since we are only interested in
linearizing an RC feed op-amp circuit, we can easily construct an
inverted lookup table indexed by operating temperature. The lookup
table is part of the micro's runtime code adjusted by current
temperature.

-----------------------------------------------------------------------------------
Thanks linnix, your input was very useful for me. may I ask if you also
had to overlap current values read with one resistor on other resistors?
as I mentioned earlier our most annoying problem is that when current
range changes i.e. current measuring resistor changes, they produce
different values for the same current. we calibrate them of course but
if we calibrate lower end to overlap on previous range then upper end
values are invalid and if we calibrate upper end to overlap on next
range then lower end values are invalid and produce gaps when range
changes.maybe producing a look up table also can make this problem to go
away.
-------------------------------------------------------------------------------------

 

[Petkovic]

http://books.google.com.au/books?id...teDiDQ&sa=X&oi=book_result&ct=result&resnum=7

It should be a definite over-kill for you application. Vishay will
sell you four terminal resistors with a temperature coefficient of
less than 1ppm/C for of the order of $100 each - or they would the
last time I looked into it - and you can compare these with the
resistors you want to calibrate to better than 1ppm with a "coaxial"
AC bridge.

The decade ratio transformers you'd use to build the bridge can be
constructed to give stable and predictable ratios at the one part in
ten million level without using a machine shop. Setting up the metal
screening boxes to totally enclose the components - that's what
"coaxial" means in this context - does require competent metal bashers
with some kind of gear for folding thin sheet metal and soldering up
long seams, but it is a lot less like hard work than putting together
a Thompsonp-Lampard calculable cross-capacitor, which is a symmetrical
arrangement of at least four parallel circular metal rods, typically
centimeter diameter rods a few hundred milimetres long.

-------------------------------------------------------------------------
Hi Bill, your posts are very useful for me, Thanks a lot.
Yes, I also think Vishay's resistors are more practical for short term
at least.
In this thread, Linnix, suggested that I can produce a look up table for
each resistor to make it more linear.

As I explained in the original post, I have 3 different ranges and I
want to read the same value for a current when range changes.
Till now using DVMs I could not do that.

My calibration procedure for each range is:
1-When there is no current passing through resistor, I read voltage(v1)
across the resistor(It includes Opamp's Offset voltage).
2- I pass a current(i1) near the maximum range of that resistor and read
voltage(v2) across the resistor(It includes Opamp's Gain Error).
3- Now to read a current, I read voltage(V) across that resistor and
calculate current(I) from this
I= (V - v1) * (v2/i1)

Do you also think that Linnix's suggestion of producing a look up table
can solve the problem of reading different values for the same current
when range changes?

Thank you in advance.
-------------------------------------------------------------------------

 

[Petkovic]

------------------------------------------------------------------------------
John,
you said
"The right way is to start with resistors/shunts which have a temperature
coefficient of resistance which will allow you the precision you require
over the temperature range which they will experience."

I have used 0.1% metal film resistors and there was this problem. while
I haven't used precise 4 terminal resistors that suggested elsewhere but
I think 0.1% metal film resistors should be considered low Tc.
--------------------------------------------------------------------------------

 

[Petkovic]

That's not what I was referring to, what I meant was if you'd tried the
constant current circuit I posted earlier.

But now I'm totally confused!

What is the opamp used for and can you post a schematic of your setup
somewhere, please?

If you have access to binary newsgroups,
alt.binaries.schematics.electronic would work.

--------------------------------------------------------------------------------------



John,

this is my setup.
I have used OP27.

-----------------------------. .-----------.
| | | | | | |
=== \ o \ o \ o | | |
PGND \ \ \ | | |\| |
\. \. \. | '-----|-\ | ___ ___
o o o | |

--o--|___|---|---------|___|-|

| | | '-----------|+/ |
|
| | | |/| | |\|
|
.-. .-. .-. |---|-\
|
| | | | | | | >-
------|
| | | | | | |-------- | |-----|+/
'-' '-' '-' | | | |/|
| | | | |\| | |
| | | |---|-\ | ___ |
\ o \ o \ o | >- | -|___|- |
\ \ \ .---|+/ |
\. \. \. | |/| .-.
o o o | | |
| | | | | |
'-----------------------------' '-'
|
|
===
GND
(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)


--------------------------------------------------------------------------------------

 

[Petkovic]

--------------------------------------------------------------------------
Ok, John,
There are 3 resistors that only one of them is connected via 2 switches
and three Op27 are configured
as a Differential amplifier.
Nothing complex, really. I hope it is clear enough.
--------------------------------------------------------------------------

 

[Petkovic]

----------------
Ok, John,
There are 3 resistors that only one of them is connected via 2 switches
and three Op27 are configured
as a Differential amplifier.
Nothing complex, really. I hope it is clear enough.
----------------

I think it is, but let's go slowly here in order to eke out a proper
solution for your problem.

I've taken the liberty of redrawing your circuit in order to make it
easier to visualize and to remove any existing ambiguities, and this is
what I think you have:


.VIN>--O--------O--------O------|+\
. |S1A |S2A |S3A | >-+-[R4]-+---+
. O-> | O-> | O-> | +-|-/U1 | | | +V
. | | | | | [R5] | |
. [R1] [R2] [R3] +------+ | +-|+\
. | | | GND | >--+
. O-> | O-> | O-> | +-|-/ U3 |
. |S1B |S2B |S3B | | |
.GND>--O--------O--------O------|+\ | -V |
. | >-+-[R6]-----+--[R7--+
. +-|-/U2 |
. | |
. +------+

Am I right?

 

[Petkovic]

this is my setup.
I have used OP27.

-----------------------------. .-----------.
| | | | | | |
=== \ o \ o \ o | | |
PGND \ \ \ | | |\| |
\. \. \. | '-----|-\ | ___ ___
o o o | >--o--|___|---|---------|___|-|
| | | '-----------|+/ |
|
| | | |/| | |\|
|
.-. .-. .-. |---|-\
|
| | | | | | | >-
------|
| | | | | | |-------- | |-----|+/
'-' '-' '-' | | | |/|
| | | | |\| | |
| | | |---|-\ | ___ |
\ o \ o \ o | >- | -|___|- |
\ \ \ .---|+/ |
\. \. \. | |/| .-.
o o o | | |
| | | | | |
'-----------------------------' '-'
|
|
===
GND
(created by AACircuit v1.28.6 beta 04/19/05www.tech-chat.de)

The switches look like they are not used right. I will attempt a
little
ascii art:

I+ ----------+--------
! !
[sw] [sw]
! !
! -------------------[SW]------+-------
!/ ! !
[R] ! !
!\ ! ! Op-amps
! ------------------[SW]------!---- go here
! ! ---------[SW]------ !
! !/ !
! [R] !
! !\ !
! ! ----------[sw]----------+--
I- ----------+--------

--------------------------------------------------------------------------------------
MooseFET,
The switches are double pole actually (DPST) and when a switch turns on,
it places a resistor in current path. At any time only one resistor is
in current path. Differential amplifier reads voltage of common ends of
switches.
--------------------------------------------------------------------------------------

 

[Petkovic]

It looks to me like you're measuring the switch resistance as well as
the resistor resistance. That may add big errors.

Amplifier common-mode rejection ratio will need to be very good, too.

--------------------------------------------------------------------------------------
John,
you are right switch resistance are in the current path, but they are
negligible I assume(<10 milli Ohm).
yes CMRR is very important, I use 1% metal film resistors.
The resistor that is connected to ground actually is a 1% metal film
resistor + bourns 25 turn Trimpot to adjust ratio of diff amp's resistor.
actually in another attempt to solve this problem I tried to lower
maximum voltage of each range from 1V to 100mV(down scaled each resistor
by a factor of 10 and added a x10 gain) to have less CMRR problem. It
improved the problem to some degree but problem still exists.
--------------------------------------------------------------------------------------

 

[Petkovic]

It costs money to have your secondary standard calibrated against a
primary standard. Make the service a monopoly, and the sole supplier
can charge what they like.


He hasn't told us why he hasn't sent them back for recalibration.

As I've mentioned, he does seem to have already gone down that route.
Telling him to do it again isn't useful advice.


And that's you excuse for telling him to do what he has done before,
which didn't seem to work?



They should be good to one part in ten million, if properly built.


You use them to set up an AC-excited resistance bridge (a Blumlein
bridge) which allows you to compare a wide range of resistors with a
single secondary standard resistor with fairly spectacular precision -
national standards laboratories have been doing this for many years,
and publish at length on the subject.


Too true. Since you posted this message, you've got into the
discussion between Petkovic, John Fields and MooseFET, which does make
it clear that Petkovic doesn't seem to understand the idea of using a
bridge to compare two resistors, which happens to be the gadget he
should be trying to build and calibrate.

---------------------------------------------------------------------------
Bill,
I will build a Blumlein bridge to find out exact value of my resistors
and it solves problem of calibrating one range but as it is evident in
my original post, I also need that when I change current measuring
resistors, I read the same value in different current ranges.
As I've mentioned I've used 0.1% resistors but this problem persists, so
in addition of knowing exact value of my resistors I have to find a way
to eliminate this problem.
---------------------------------------------------------------------------

 

[Petkovic]

This is a clumsy way of comparing two resistors. The classic circuit
for comparing two resistors is a direct-current excited Wheatstone
bridge, where you use a potentiometer to balance the bridge and
establish the ratio of the resistors being compared.

As John Larkin pointed out earlier in this thread, it is difficult to
get real precison out of a DC excited bridge because the imbalance
signal can be influenced by thermocouple voltages around the circuit
and off-sets in any op amps you use to realise the circuit. As John
went on to point out these can be corrected out by eliminating or
reversing the DC voltage across the bridge from time to time.

If you go over to an AC-excited bridge you can use transformer
windings in two arms of the bridge - the Blumlein bridge - and replace
the potentiometer with one or more ratio transformers, which are
rather easier to build than a precision potentiometer, and can be a
good deal more precise. You do need a stable AC source to drive the
bridge and in-phase and quadrature AC detectors to measure the out-of-
balance signal, but all this has been standard practice for many years
now.

-----------------------------------------------------------------------------
Bill,

The circuit I've shown is not intended to compare 3 resistors. It is not
intended
to measure resistance and it is not intended to replace a Blumlein
bridge. The sole purpose of this circuit is to measure current
continuously from 1mA-10A. For measuring current accurately, I need to
know the exact value of resistors and that needs another setup and as
you indicated probably a Blumlein bridge.
-----------------------------------------------------------------------------

 

[Rich Grise]

I am not bound by your absurd definitions of what I am.

There are lots of cal labs around. But I wouldn't use one that can't
trace their measurement chain to a national standard. What's the point
of having a free market in the definition of a volt?

It's worse than that - if you buy a bushel of fruit from a vendor, what
guarantee is there that you're both using the same bushel?

Well, we've all known for some time that Sloman's an idiot.

Cheers!
Rich