Experiment: TCN75A Temp sensor investigation

I covered the TCN75A i2C sensor a few weeks ago in a simple test to practice using I2C. I had never used I2C before and I knew it was time to stop avoiding it and jump in if I was going to use some of the fun chips and modules. Shortly after posting my results and code I jumped into @tymkrs IRC (visit their site if you’re curious) and Victor ( @vicatcu ) must have at least caught the article subject as he asked me if I noticed the chip runs a little warm (I think he mentioned roughly a deg C?). I had also noticed that oddity myself but I ignored it as I was only concerned with the actual communication. Victor  ( @vicatcu ) posed the question: what would happen if I had a vented circuit board? Would that cure the chip from running hot?

Well that was half-brilliant wasn’t it? I had to know! Does under chip venting on the PCB effect temperature readings? I decided to purchase more sensors and SOIC break out boards as well as make sure my Fluke 51 was calibrated.

Upon receiving the SOIC breakout boards I took one and carefully used a Dremel to mill out the package outline. I had started to design my own PCB  but I abandoned that because I didn’t want to wait for fabrication. I mounted the new sensor on it’s vented PCB then plugged the sensors right next to each other on breadboard and put the temperature probe wedged between the boards. Obviously factors like mounting the sensor near power transistors or near your power supply are always going to have some effect on your readings, and in some cases maybe that’s what your application is meant for. I wanted to take those factors out of the equation. My workshop heating and A/C was off for over a day when I did this test… no windows where open and the chips had been prepped for about 1/2 hour on the board while I wrote the code to test them out. I decided to read data straight from the I2C bus with some confirmation on RS-232 telemetry. I did this in case I introduced any math problem in my conversion in code. I also stuck with Celsius over Fahrenheit because that was the native output of the sensor.

SOIC breakout board with vent cut into it
SOIC breakout board with vent cut into it
The setup of the initial test.
The setup of the initial test.

After downloading code and checking everything out my Fluke 51 read 25.8 deg C, the un-vented TCN75A read 26.06 deg C and the vented TCN75A read 25.81 deg C. Over time the un-vented sensor would measure up to 0.400 deg C away from the other two. It’s not 1 deg C but it is something.

The TCN75A boards are back to back with the K type probe between them. To the bottom right you see the I2C pull up resistors with the logic analyzer tied to them. The PIC16F1509 is just to the right.

Temp sensor telemetry
Temp sensor telemetry; don’t mind the calculator. I was confirming a hexadecimal I2C address.

Test #2, a muffin fan…. same set up. No real change; roughly the same results.

Test #3! What about a change in temperature? I grabbed a cardboard box, put my test circuit and a bowl of ice with some water in the bottom of it, with a muffin fan blowing into the bowl and covered the whole thing in with saran wrap. Not the most insulated thing but I was only looking for a couple degrees. The test proved interesting… both sensors had some lag, and the venting only offered some minor help. It caught up first but they still have a fair amount of lag over the K type probe. In the graph provided I measured at even intervals over 15 minutes. The box settled out at the end (not enough insulation I imagine) the temperature flattened out so I dropped that data.

The Fluke 51 with K type thermocouple vs the normally mounted and modified mounting TCN75A temp sensors.
The Fluke 51 with K type thermocouple vs the normally mounted and modified mounting TCN75A temp sensors.

I can live with the results. I think if I were going to design a board where I wanted as close to an accurate temperature as I could get with this part I would definitely have a milled vent. In hind-sight I would make the vent so it is along the chip axis and protrudes past the edges of the chip. Thanks again to Victor for the great idea!


  • Konrad
    April 8, 2014 - 9:15 pm | Permalink

    I know that the TN measures the temp from the bottom of the ic. ?
    I have a TN in a thermometer kit but its mounted on the pcb and enclosed so the reading is not exactly accurate as it tends to read internal pcb temp instead of external air temp hence I am wanting to change its placement.
    I was considering mounting the bottom of the ic on a copper heatsink and exposing the heatsink to the surrounding air for accurate measurement. would mounting the TN on a copper heatsink work just as well as on a vented soic pcb?

    • Chas
      April 9, 2014 - 3:53 am | Permalink

      I wouldn’t know with out trying it of course but I would think that the heat sink would work well in keeping the inside of the case close to the same temperature as the ambient air outside of the case. I think if you could work the heat sink in contact with the TCN part you’d probably be better off than venting the board. Good luck, I’d like to see how that got accomplished if you end up going that route.

  • Konrad
    April 22, 2014 - 4:19 pm | Permalink

    Hi Mate,
    further to my last message I tested the chip with the heatsink and without. Instead of drilling the pcb I reverse mounted the chip so that the bottom faces up as the sensor measures from the bottom. I found that the measurement is more accurate with no heasink.
    here is a pic of reverse mount:

  • Chas
    April 22, 2014 - 4:23 pm | Permalink

    Konrad — thanks for the addition and the photo of your mounting job. Interesting .. I wonder if it would be easy just to fab a small board with the whole package minus leads milled so you could easily reflow the TCN75A upside down like shown. I might consider trying this just to see how it turns out. Definitely interesting!

  • Leave a Reply