Searching \ for '[EE] thermocouples and muxes' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/index.htm?key=thermocouples+muxes
Search entire site for: 'thermocouples and muxes'.

Exact match. Not showing close matches.
PICList Thread
'[EE] thermocouples and muxes'
2006\11\20@223723 by Bob J.

picon face
I have a Max6675 thermocouple interface chip that I have had working with an
18F with no problem.  I've added two muxes (CD74HC4067) which are eight
channels each, so that I can multiplex the 6675 to 16 thermocouples.  Once I
threw the muxes into the mix, not unexpectedly the thermocouples read higher
temperatures at room temperature.  I suspect the "on" resistances of the
muxes (70 ohms) are having an effect.  Through some trial and error I came
up with a correction factor that I subtract from the 12-bit result of the
6675 which has moved the resultant temperature in line.  However, I
presently don't have a thermocouple calibrator to test across the
temperature range I am going to use the circuit in.  I've tested the
thermocouple readings by dipping the thermocouple in a cup of ice water and
a cup of boiling water, and the results are correct at those temperatures.
So my question is: will the additional impedance of the muxes have a linear
effect on the voltage produced by the thermocouples (which, by themselves
aren't linear) or should I expect the circuit to behave in a non-linear
fashion which would require a correction table that I will have to build?

Regards,
Bob

2006\11\21@073605 by Gerhard Fiedler

picon face
Bob J. wrote:

> I have a Max6675 thermocouple interface chip that I have had working with
> an 18F with no problem.  I've added two muxes (CD74HC4067) which are
> eight channels each, so that I can multiplex the 6675 to 16
> thermocouples.  

I'd probably amplify (and output impedance-reduce) the thermocouples before
muxing them (if you want to mux them). Bring the 30 mV or whatever is your
input voltage range up to a few volts, and muxing becomes a lot easier. (Or
use one MAX6675 per input... That's what they're made for :)

The other thing to consider with multiple thermocouples is that depending
on the application they may be at different potentials. Thermocouples are
sometimes electrically connected to something in the process. So your
inputs may have to be able to cope with that (ie. be differential).

{Quote hidden}

It doesn't sound like it was a linear resistor divider that you're seeing.
The MAX6675 claims 60k input impedance; how could a switch resistance of
70R could make a difference of a few degrees at room temperature -- and
increase the reading? So it looks more like something non-linear. And it
might very well be that it's dependent on the specific devices (ie. not the
same on all boards you make). In any case, introducing a correction factor
before exactly understanding the effect doesn't seem a prudent thing to do
:)

Don't forget that you're dealing with uV at the input. There's all kinds of
effects that can be in the same range.

Gerhard

2006\11\21@095031 by Bob J.

picon face
On 11/21/06, Gerhard Fiedler <spam_OUTlistsTakeThisOuTspamconnectionbrazil.com> wrote:
>
> Bob J. wrote:
>
> > I have a Max6675 thermocouple interface chip that I have had working
> with
> > an 18F with no problem.  I've added two muxes (CD74HC4067) which are
> > eight channels each, so that I can multiplex the 6675 to 16
> > thermocouples.
>
> I'd probably amplify (and output impedance-reduce) the thermocouples
> before
> muxing them (if you want to mux them). Bring the 30 mV or whatever is your
> input voltage range up to a few volts, and muxing becomes a lot easier.
> (Or
> use one MAX6675 per input... That's what they're made for :)


The reason I want to avoid having sixteen 6675's is cost.  Sixteen of them
would cost around $96.  On the other hand when I get the mux to work
properly it should cost around $8 in parts.  Obviously a very significant
cost savings.

In one of the app notes that Analog has for its AD594 thermocouple
interfacing device, they show a mux on the thermocouple input.  The mux that
they use in the app note has a much higher on resistance and no
amplification is done before the AD594.  I'm beginning to wonder if my
problem has to do with the thermal connection between the thermocouple
terminals, the mux, and the 6675.  Its all on a breadboard right now which
is likely not ideal in that sense.

The other thing to consider with multiple thermocouples is that depending
> on the application they may be at different potentials. Thermocouples are
> sometimes electrically connected to something in the process. So your
> inputs may have to be able to cope with that (ie. be differential).


You're likely correct on that, being that thermocouples generate only a few
millivolts at high temperatures.

{Quote hidden}

Sure, I agree, and that's what I'm trying to find out :)  But after doing
some brainstorming this morning, I need to eliminate any isothermal
differences between the 6675 and the cold-junction (the terminals the probes
are connected to) and make sure the mux has a good thermal connection to the
6675.


> Don't forget that you're dealing with uV at the input. There's all kinds
> of
> effects that can be in the same range.
>
> Gerhard
>
> -

2006\11\21@095113 by Geo

flavicon
face
Sent from correct address this time - sorry Dan...

On 20 Nov 2006, at 22:37, Bob J. wrote:

> I have a Max6675 thermocouple interface chip that I have had working
> with an 18F with no problem.  I've added two muxes (CD74HC4067) which
> are eight channels each, so that I can multiplex the 6675 to 16
> thermocouples.  Once I threw the muxes into the mix, not unexpectedly
> the thermocouples read higher temperatures at room temperature.  I
> suspect the "on" resistances of the muxes (70 ohms) are having an
> effect.
You could try a thermocouple direct to the ADC with a 100 ohm
resistor in series to see if your thery is correct.
I do not see why in-line resistance should have any effect if the
themocouples (with about zero source impedance) are feeding into
a high impedance, low-offset measuring device (ADC). I suspect it
is more likely to be a voltage effect - which may be temperature
dependent. You could test for this by warming and freezing the
multiplexors. In commercial data logging equipment that I used to
maintain, the switching was done by QA selected reed relays
and they had to be encased in insulating foam to prevent thermal
offsets.


George Smith

2006\11\21@105016 by Mark E. Skeels

picon face
Hi, Bob,

I'm currently mpx'ing 8 thermocouples through 4066's. I've also got
ESD/overvoltage protection devices and series resistors before the mpx's
and RC networks after.

They feed into a MSP430F2013 which has an onboard Delta Sigma ADC with
diff inputs.

I've not encountered this offset problem as far as I know, but I did
have to slow down the mpx to avoid voltage settling problems due to
impedance effects. I'm getting suitable accuracy for my application so
far; within 1 degree of my Fluke portable temp meter given similar
mechanical transfer arrangements. It seems to me that the input
impedance is great enough on my MSP that the small resistance your
talking about would be essentially swamped out. Other resistances in my
circuit may have affected software calibration values, but they did not
seem to affect final performance in my case.  I had a chance to
encounter a similar problem too, since my first version did not have the
protection, and of course, I kept frying  4066's.

More thorough testing I have not done yet. I have to say though, that I
probably don't need the kidn fo accuracy you're looking for; I'm happy
with +-2-3 deg C.

Mark

Bob J. wrote:

{Quote hidden}

>>-

2006\11\21@124549 by Spehro Pefhany

picon face
At 10:37 PM 11/20/2006, you wrote:
>I have a Max6675 thermocouple interface chip that I have had working with an
>18F with no problem.  I've added two muxes (CD74HC4067) which are eight
>channels each, so that I can multiplex the 6675 to 16 thermocouples.  Once I
>threw the muxes into the mix, not unexpectedly the thermocouples read higher
>temperatures at room temperature.  I suspect the "on" resistances of the
>muxes (70 ohms) are having an effect.  Through some trial and error I came
>up with a correction factor that I subtract from the 12-bit result of the
>6675 which has moved the resultant temperature in line.  However, I
>presently don't have a thermocouple calibrator to test across the
>temperature range I am going to use the circuit in.  I've tested the
>thermocouple readings by dipping the thermocouple in a cup of ice water and
>a cup of boiling water, and the results are correct at those temperatures.
>So my question is: will the additional impedance of the muxes have a linear
>effect on the voltage produced by the thermocouples (which, by themselves
>aren't linear) or should I expect the circuit to behave in a non-linear
>fashion which would require a correction table that I will have to build?
>
>Regards,
>Bob

Looks like this part has quite a bit of current flowing out of the T+ input--
there's 330K nominal resistance to the internal reference (whatever voltage
that is). So, it will read high, with an offset dependent on the particular
mux unit you have and on the temperature. There will also be a span error
in the other direction. If we assume it's 2.5V, then you've got 7.6uA flowing
out of the input, so you'd have around 13°C error, nominally, with a 70 ohm
resistance on the input and it will vary with temperature etc.

Is your actual error that high? I try to design thermocouple instrumentation
to be within spec with at least a 100 ohm lead resistance on the sensor.

Best regards,

Spehro Pefhany --"it's the network..."            "The Journey is the reward"
speffspamKILLspaminterlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com
->>Test equipment, parts OLED displys http://search.ebay.com/_W0QQsassZspeff


2006\11\21@150121 by Bob J.

picon face
On 11/21/06, Spehro Pefhany <.....speffKILLspamspam.....interlog.com> wrote:
>
> Looks like this part has quite a bit of current flowing out of the T+
> input--
> there's 330K nominal resistance to the internal reference (whatever
> voltage
> that is). So, it will read high, with an offset dependent on the
> particular
> mux unit you have and on the temperature. There will also be a span error
> in the other direction. If we assume it's 2.5V, then you've got 7.6uAflowing
> out of the input, so you'd have around 13°C error, nominally, with a 70
> ohm
> resistance on the input and it will vary with temperature etc.
>
> Is your actual error that high? I try to design thermocouple
> instrumentation
> to be within spec with at least a 100 ohm lead resistance on the sensor.


Good catch Sphero!  I did not realize that an offset voltage was coming from
the T+ lead until after reviewing the block diagram of the 6675.  I display
the thermocouple reading in °F so the 13°C you mention would be 55°F, and
the indicated temperature after adding the muxes with no correction factor
the temperature was somewhere in the neighborhood of 120°F with the
thermocouples at room temperature.  So this makes sense, even if it is a
ballpark number as you suggest with span error being 2.5V.  I will
definitely look into this and as George suggested, will look into
temperature cycling the muxes.

Thanks for the help gents!

Regards,
Bob

2006\11\22@055051 by Gerhard Fiedler

picon face
Bob J. wrote:

> Good catch Sphero!  I did not realize that an offset voltage was coming from
> the T+ lead until after reviewing the block diagram of the 6675.  

This is the problem with such integrated devices. Traditionally, you'd have
the thermocouple break detection (and maybe some amplification) before the
mux. With this all-in-one device, you can't easily separate this from the
rest.

Using it with 16 thermocouple inputs, it seems you're working against the
design of this device (unless you use 16 of them). Use a traditional ESD
protection and break detection on each input, add possibly a differential
amplifier for each input (alternatively only one after the mux), mux that
into an ADC, add a simple cold junction temperature sensor and do the
compensation and linearisation in firmware. This sounds like a lot of
stuff, but it's probably easier than trying to multiplex such an active
input that wasn't designed to be multiplexed.

Gerhard

2006\11\22@063643 by Alan B. Pearce

face picon face
> Good catch Sphero!  I did not realize that an offset voltage was coming
> from the T+ lead until after reviewing the block diagram of the 6675.

Although you are currently using a Maxim part, you might like to look
through the Linear Technology datasheets and application notes.

AN28 - very good looking application note which discusses accuracy and
       sources of error, along with methods for doing high accuracy
       isolation of thermocouples. Based around the LT1025 compensator
       chip, but has wider application.

DN302 - Design Note discussing robust thermocouple interfaces built around
       the LTC2053 instrumentation amp. Again using an LT1025 compensator
       chip.

LT1013/4 - Dual/Quad precision op-amp with a number of circuits on the
       datasheet for using a single quad op-amp with one section as the
       cold junction compensation reference, and the other three sections
       to sense 3 thermocouples before using a mux. Also has circuits for
       hot wire anemometer and liquid flow sense (I should have remembered
       these for the recent discussion where someone wanted airflow
       measurement)and strain gauge signal conditioner.

2006\11\22@152000 by Bob J.

picon face
On 11/22/06, Gerhard Fiedler <EraseMElistsspam_OUTspamTakeThisOuTconnectionbrazil.com> wrote:
>
>
> Using it with 16 thermocouple inputs, it seems you're working against the
> design of this device (unless you use 16 of them). Use a traditional ESD
> protection and break detection on each input, add possibly a differential
> amplifier for each input (alternatively only one after the mux), mux that
> into an ADC, add a simple cold junction temperature sensor and do the
> compensation and linearisation in firmware. This sounds like a lot of
> stuff, but it's probably easier than trying to multiplex such an active
> input that wasn't designed to be multiplexed.
>
> Gerhard


I am leaning on going down this path of doing all of the linearization and
cold-junction compensation myself, depending on how my testing goes with the
current circuit.  As soon as I can get my hands on a thermocouple calibrator
so I can see if the circuit I designed has decent accuracy with the Max6675,
I may stick with it.  However I am not fond of Maxim's small order policies
nor am I fond of being dependent on a propriatary chip.  So we'll see.

Thanks again for the help gents.

Regards,
Bob

2006\11\22@192912 by cdb

flavicon
face
Wow, that'll do for further investigation for the flow meter.

Thanks

Colin

:: LT1013/4 - Dual/Quad precision op-amp with a number of circuits on
:: the
::
::        datasheet for using a single quad op-amp with one section
:: as the
::        cold junction compensation reference, and the other three
:: sections
::        to sense 3 thermocouples before using a mux. Also has
:: circuits for
::        hot wire anemometer and liquid flow sense (I should have
:: remembered
::        these for the recent discussion where someone wanted airflow
::        measurement)and strain gauge signal conditioner.

--
cdb, colinspamspam_OUTbtech-online.co.uk on 23/11/2006

Web presence: http://www.btech-online.co.uk  

Hosted by:  http://www.1and1.co.uk/?k_id=7988359

Do No Harm.

We must never do evil that good may come of it: William Penn 1693

.



--
No virus found in this outgoing message.
Checked by AVG.
Version: 7.5.431 / Virus Database: 268.14.13/546 - Release Date: 11/22/2006 9:01 AM


2006\11\23@040126 by Alan B. Pearce

face picon face
>Wow, that'll do for further investigation for the flow meter.

Not only that, but you should be able to get 2 free samples out of them.
They seem to be reasonably liberal in giving samples in much the same way as
Microchip do. I have had a number of samples of switchmode regulator chips
from them.

More... (looser matching)
- Last day of these posts
- In 2006 , 2007 only
- Today
- New search...