4
\$\begingroup\$

I have designed a board with four MAX6675 thermocouple digitisers to measure temperatures at four different positions in a device. The board is connected to a Raspberry Pi, which reads the data via SPI. Unfortunately, the measured values are very different at room temperature (e.g. U8 got 16°C, U7 got 19°C, U6 got 30°C and U5 got 25°C).

This is the relevant part of the PCB: enter image description here

  • Blue: T+ of Thermocouple to MAX6675
  • Grey: T- of Thermocouple Connector to MAX6675 (via GND)
  • Green: 3.3 V
  • Purple: CLK and MISO for SPI, connected to Pi
  • White: CS via U4
  • Red: GND
  • C6-C9: 0.1 µF ±10% 25V Ceramic Capacitor 0805
  • U5-U8: MAX 6675 ISA
  • U4 is used for Chipselect: 74HC139D

We have ruled out a problem in the software by testing it with another board with the same interface. We also excluded the type k thermocouples by swapping them. The measured temperature jumps back to the same wrong value with swapped sensor at room temperature.

Are there any faults on the board that may be causing the problem? On the backside of the board there are no traces above the purple trace.

Edit: The Cable Length are all the same and around 75cm.

\$\endgroup\$
13
  • 1
    \$\begingroup\$ The Cable Length are all the same and around 75 cm.U8 got 16°C, U7 got 19°C, U6 got 30°C and U5 got 25°C \$\endgroup\$
    – dthal
    Oct 12, 2022 at 9:21
  • 2
    \$\begingroup\$ This may be important depending on the answer :-) . 1. The distance from thermocouple gnd to IC gnd varies in each case, and shares the copper with other current generating 'stuff'. This may relate to 2. From the available layout it APPEARS as if the ground is physically discontinuous between the four ICs. Is in fact there an essentially direct ground line between the pin-1 of all ICs or does it wander hither and yon bwteen them? || You are dealing with 10.5 uV per LSB. 1 mV of ground difference is about 8 bits. || When dealing with low voltage conversions extreme care needs to be taken. \$\endgroup\$
    – Russell McMahon ♦
    Oct 12, 2022 at 9:54
  • 2
    \$\begingroup\$ I imagine (and I may imagine wrongly) that the ideal would be to bring the thermocouple ground to the IC ground directly and "star" connect it to any other ground at the IC pin. If there are inter IC or across track voltages these should then be irrelevant. \$\endgroup\$
    – Russell McMahon ♦
    Oct 12, 2022 at 9:55
  • 2
    \$\begingroup\$ Maxim says "Make the [T-] ground connection as close to the GND pin as possible." I don't know how close they mean. I note that high-reading U5 and U6 have longer, thinner power, but it doesn't look bad. Is there anything which might differentially heat up the four chips? Have you tried shorting T+ and T- at the chips, to simulate Δt = 0? This should give you each chip's ambient. Have you tried swapping two individual chips, to see if it's board or chip? \$\endgroup\$
    – jonathanjo
    Oct 12, 2022 at 11:20
  • 2
    \$\begingroup\$ Can you short the thermocouple connections with copper wires and report back the resulting Ta/Tj readings? \$\endgroup\$
    – Spehro Pefhany
    Oct 12, 2022 at 11:26

1 Answer 1

Reset to default
1
\$\begingroup\$

From comments:

Q. Maxim says "Make the [T-] ground connection as close to the GND pin as possible." I don't know how close they mean. Have you tried shorting T+ and T- at the chips, to simulate Δt = 0? This should give you each chip's ambient.

A. I tested by holding T- directly on the pin on the MAX6675. It looks like that is actually the problem. I have now adjusted the layout and reordered. I will report if that solved the problem.

I note that the MAX6755 has a replacement part MAX 31855. The main difference appears to be that this latter chip says do not connect T- to GND, because there is an internal connection enabled during conversions. Datasheet says "strongly recommended to add a 10nF ceramic surface-mount differential capacitor, placed across the T+ and T- pins".

Adafruit has circuit diagrams and PCB layouts for both chips which might be helpful to compare with yours.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.