multiplexor Inputs

[Richard]

I need to come up with some ideas to take a Thermocouple chip like the AD595
and have 16 type K- Inputs going to it. I would imagine it would be best to
use 4 - 8 channel multiplexors, using 1 pair for each 8 channels, in order
to switch the + and - of each thermocouple.

I have not dealt with multiplexors yet, so do you all have any
recommendations on which would be the best ones to start playing with. I
also need someway to filter noise at the each channel before the
multiplexor, to eliminate one open channel from injecting noise into other
valid channels. Any ideas on this one? Simple caps or some pull downs?

Thanks for any help.

 

[Richard]

One of the muxes I'm looking at is the Intersil 8x2 DG407DY. Using two of
those would switch the pos and Neg of 16 channels.
Any comments on these? Wow, that AD595 is pretty expensive. Any
alternatives that work well?

Thanks

 

[Ban]

I need to come up with some ideas to take a Thermocouple chip like
the AD595 and have 16 type K- Inputs going to it. I would imagine it
would be
best to use 4 - 8 channel multiplexors, using 1 pair for each 8
channels,
in order to switch the + and - of each thermocouple.

I have not dealt with multiplexors yet, so do you all have any
recommendations on which would be the best ones to start playing
with. I also need someway to filter noise at the each channel before the
multiplexor, to eliminate one open channel from injecting noise into
other valid channels. Any ideas on this one? Simple caps or some pull
downs?
Thanks for any help.

Richard,
I'm afraid you will need one compensation chip for each channel and also the
required filter afterwards, and only then mux, because a mux in front will
corrupt the signal. The charge injection will probably saturate the amp and
the long time constants of the filter will need very long settling times.
If you make up some kind of amplification/filtering all connections have to
be at the same temperature ( that is why that chip is ceramic and
expensive).
But depending on your required accuracy you could thow away precision and
have an instrumentation amp up front followed by a filter and you could
subtract some temp dependent ice-point correction, that you have amplified
identically, or probably leave that out as well. Who cares about +/-20°C
then? So you save the AD595 as well. Go figure out.

 

[MK]

Richard,
I'm afraid you will need one compensation chip for each channel and also the
required filter afterwards, and only then mux, because a mux in front will
corrupt the signal. The charge injection will probably saturate the amp and
the long time constants of the filter will need very long settling times.
If you make up some kind of amplification/filtering all connections have to
be at the same temperature ( that is why that chip is ceramic and
expensive).
But depending on your required accuracy you could thow away precision and
have an instrumentation amp up front followed by a filter and you could
subtract some temp dependent ice-point correction, that you have amplified
identically, or probably leave that out as well. Who cares about +/-20°C
then? So you save the AD595 as well. Go figure out.

You don't really need one compensation chip per channel - you need to
measure the temperature of the cold (local) junctions, so if you can keep
them the same within the limits of accuracy you want to work with then one
sensor will do. Most small multi input instruments work like this.

It is possible to use passive filtering, protection etc and a multiplexer
and this can work well. You can get better performance by using
buffer/filter amplifiers before the mpx. If you do it right you can use your
buffers to get rid of common mode noise, filters to get rid of serial mode
noise and then use a cheap modern mpx running off a 5V supply. Several
modern ADCs or micros have built in temperature sensors and you can (with
care) use one of these for the cold junction compensation.

Linear technology have the LTC2480 with built in sensor and an app. note
showing how to use it for a single channel thermocouple system. (The package
is not hand built prototype friendly - I know I've just soldered 5 of them
!).

(I'm running a test on this now and the reference thermometer in the dry
well calibrator says 70.02C and my PC reading out from the prototype says
69.9C. I'm happy because that was calibrated at 25C. I know from tests on
other prototypes that the effect of ambient temperature changes in the range
5 to 40C is less than 0.5C.)

www.mkesc.c.uk


Michael Kellett

 

[[email protected]]

Hi, it's not really feasable to switch the thermocouples directly. A
better aproach is to have each thermocouple connected to its own
amplifier/cold jinc compensator/filter input stage, then multiplex the
output voltages. You dont need any exotic chips, even a 741 will do at
a pinch and a diode can be used for cold junc comp.

 

[Spehro Pefhany]

Hi, it's not really feasable to switch the thermocouples directly.

Ach, ye could hae fooled me. What makes ye think ye cannae do it?

better aproach is to have each thermocouple connected to its own
amplifier/cold jinc compensator/filter input stage, then multiplex the
output voltages.

....grounded junctions...

You dont need any exotic chips, even a 741 will do at
a pinch and a diode can be used for cold junc comp.

If you don't care a whit about accuracy and are using only the basest
of base-metal thermocouples.


Best regards,
Spehro Pefhany

 

[Jon]

As others have mentioned, the problem with using a mux upstream of the
AD595 is cold junction compensation errors. One way around this, if
you don't need extreme accuracy is to use a temperature sensor, such as
the AD590 to measure the cold junction temperature (assuming that it is
roughly the same for all 16 phusical thermocuople interface points).
You can then replace the AD595 with an ordinary OP Amp or
Instrumentation amplifier. You can do your own cold junction
compensation in software, using the known thermocuple characteristics.
I have succesfully used this technique for measuring high temperatures,
where the variation of the cold junction temperatures among
thermocouples was negligible. One way to ensure that the temperature
variation is small is to place a large ground plane under the
thermocouple connections. The ground plane will tend to equalize the
cold junction temperatures. You can place the cold junction
temperature sensor on or above the ground plane.

 

[Richard]

I had planned on using Phoenix Contact connectors all in one row and have
the muxes and AD595's right next to them pretty close. I would think the
temp would be very, very close this way, however I could be more wrong than
I think.

The large ground plane sounds like a great idea.

Most of the temps measured would be in the 200 to 1200 Degree F Range if
that make a difference. Accuracy of 1 - 2 Degrees F is plenty good for this
application.

I have seen on the Net where some state to add 10k resistors to ground for
each lead on each pair right before the mux to reduce noise, and also to
help reduce the antenna effect from one open channel from affecting other
channels. Does this make sense?

Richard

 

[Spehro Pefhany]

I had planned on using Phoenix Contact connectors all in one row and have
the muxes and AD595's right next to them pretty close. I would think the
temp would be very, very close this way, however I could be more wrong than
I think.

The large ground plane sounds like a great idea.

Most of the temps measured would be in the 200 to 1200 Degree F Range if
that make a difference. Accuracy of 1 - 2 Degrees F is plenty good for this
application.

That can be quite challenging to achieve in practice especially if you
don't want to allow a 1 hour warmup or have a housing to contend with.
A 10°F difference can arise fairly readily, depending on what else is
around.

I have seen on the Net where some state to add 10k resistors to ground for
each lead on each pair right before the mux to reduce noise, and also to
help reduce the antenna effect from one open channel from affecting other
channels. Does this make sense?

Richard

Not to "reduce noise", but it might be important where you have
multiple instrumentation amplifiers- to establish a ground within the
common-mode range. Most T/C's are low enough resistance that 10K will
only affect the reading by a small amount. Commercial instruments will
usually operate within specs for source impedance of 100 or even 1000
ohms maximum, so the effective shunt 20K could add significant error.


Best regards,
Spehro Pefhany

 

[Richard]

Thanks for all the input. Sure gives me more to think about.

Richard