MAX31855 temperature error (reads 25% low)

[David_AVD]

I'm using a MAX31855 (K-type) thermocouple chip and am having issues with the returned readings being approx 25% low.  eg. It reads 77°C when the probe is actually 100°C (boiling water).

I've plugged the same thermocouple into my Fluke DMM adapter and it reads the correct 100°C.  The percentage error seems to be fairly constant over a range of temperatures.

I did have issues with erratic (random) readings before I added more PSU decoupling.  I already have the recommended 10n cap across T+ and T- and am using a K-type socket.

Any ideas?  There has to be something very simple (silly) going on.

[SArepairman]

can it be backwards?

[David_AVD]

The thermocouple?  No, I triple checked the connections.  I did try backwards to see what happens and you just get a negative reading (as expected).

[senso]

If it was backwards it would read lower, I mean, as the temperature goes up in the sensor tip, the chip will read lower and lower, been there done that, was really confused thinking that the thermo-couple was a chinese special

[retrolefty]

I've always found this Omega Engineering data page useful in identifing the various thermocouple types (J,K,etc) and their polarity, it's not always intuitive.

http://www.omega.com/techref/colorcodes.html

Also are you wiring the TC all the way directly to the MAX31855 where the cold reference junction compensation is applied? If you transition to copper wire prior to the MAX chip and their is a temp difference between the transition point and the MAX chip that can introduce a measurement compensation error.

[David_AVD]

Yeah, I actually meant "negative going reading" (with increasing temperature) when I said negative above.   

[CrabxCore]

Is it possible you need the ferrite beads like in the eval board schematic?

[David_AVD]

Here's the relevant portion of the PCB:

[David_AVD]

Is it possible you need the ferrite beads like in the eval board schematic?

I had not seen an eval board for it, it should be easy to cut the Vcc track and insert a bead as a test.

[SArepairman]

can you splice that connector in circuit with your fluke meter to see if it is good?

like thermocouple wire - silver solder - connector- thermocouple wire. I had weird stuff like this happen when I ordered a K connector on ebay and got a J or something like that.

[SeanB]

Your cold junction compensation is off? What does it read when short circuited with no TC in the socket, and when the TC is at room temerature along with rom temperature from another thermometer.

[CrabxCore]

Here is the link to the data sheet of the eval board. http://datasheets.maximintegrated.com/en/ds/MAX31855EVKIT.pdf

If that fixes the offset problem, I will need to add that to my board i'm working on before I order it. Winging it and learning as you go gets expensive! 

[plesa]

Is it possible you need the ferrite beads like in the eval board schematic?

I had not seen an eval board for it, it should be easy to cut the Vcc track and insert a bead as a test.

shematic of Eval kit
http://datasheets.maximintegrated.com/en/ds/MAX31855EVKIT.pdf

Are you 100% sure that you're using proper MAX31855K type chip?

[David_AVD]

Your cold junction compensation is off? What does it read when short circuited with no TC in the socket, and when the TC is at room temerature along with rom temperature from another thermometer.

My take on it is that the MAX chip does the cold junction compensation and applies it to the thermocouple portion of the digital reading.  You can read the comp reading separately of course, but as far as I know you don't have to do anything with it.

Hmmm... if I disconnect the TC and short the input, the reading is 26°C which would be about right for my office at the moment.

EDIT:  Adding the bead to the Vcc (before the cap on the MAX) made no difference.

[han]

could you measure the voltage across the IC input,
and calculate the value using K type chart, and compare with the reading from the IC.
1. If the voltage reading match the actual temp and chart, then something wrong with the IC
2. If the voltage reading didn't match with the actual temp and chart, then probably something wrong with the soldering/ metal type on the cold junction.

[CrabxCore]

Your cold junction compensation is off? What does it read when short circuited with no TC in the socket, and when the TC is at room temerature along with rom temperature from another thermometer.

My take on it is that the MAX chip does the cold junction compensation and applies it to the thermocouple portion of the digital reading.  You can read the comp reading separately of course, but as far as I know you don't have to do anything with it.

Hmmm... if I disconnect the TC and short the input, the reading is 26°C which would be about right for my office at the moment.

EDIT:  Adding the bead to the Vcc (before the cap on the MAX) made no difference.

On the eval board the bead is on the t+ and t- trace. I'm pretty new to all this so not sure if it matters.

[gman4925]

It might be a J type chip.
100 deg is around 4127.6 uV on a K Thermocouple.
4127.6 uV on a J chip is around 71 deg.

Also check what the measured voltage is at the chip T- and T+ pins is.
And check what the error bits show.

[gman4925]

Or T would be around 79 deg.

[David_AVD]

I just tried another K-type probe.  Since it had spade terminals on it, I soldered them to the back of the existing K-type socket.  With that one I get a reading of 80°C in boiling water.  Grrr...

[SeanB]

Time for the right Max chip then. Yours is either the wrong type of has accidentally been programmed to a different thermocouple type.

[David_AVD]

Just checked the chip (again) and it's marked as a M31855K type.

I read approx 3.4mV at the MAX T+ and T- pins which seems to correspond to 80°C.

[gman4925]

Make sure T- is not grounded.
Try measuring the TC voltage without it connected to anything.
Post schematic.

[ejeffrey]

With a hot junction at 100 C and a cold junction at 20 C, the thermo-electric voltage should be around 3.4 mV.  So your reading is correct for a K-type thermocouple in boiling water if your meter is at "standard" room temperature and there are no serous temperature gradients except near the tip of the thermocouple.  That pretty much narrows the suspicion down to the chip.  Either the cold junction compensation is wrong or the chip is mislabeled or otherwise has the wrong calibration dataset loaded.

[David_AVD]

It really does look like the chip is the wrong curve.

I bought the chips from Digikey; 2 pieces originally then another 2 a month later.  The two prototypes have one from each purchase, but they are probably the same batch.

Unfortunately neither RS or Farnell have stock, so can't easily get another chip to see if that's the issue.   

I'll see if I can get one somewhere else and go from there.

[ve7xen]

Sample from Maxim?