TMP006: Crazy response to temperatures

This is somewhat weird...

I have a TMP006 wired up on a double-sided FR4 PCB. This board was manufactured in-house as a quick-turn prototype, has 1oz copper on a 1.6mm laminate, and no solder mask. This is the same board for which I requested design advice in a previous thread. There IS a metal pad under the centre of the chip (grounded), but it doesn't have an insulating solder mask over it. The IC was soldered with paste flux (which has been removed with an ultrasonic cleaner and electronics-grade Isopropyl Alcohol) and a hot-air soldering station (preheater and hot-air jet).

The Gerber files for the board are in my 'files' area.

The TMP006's Ambient sensor seems to be behaving itself, and is reading 25C in my fairly warm office / lab (that'll teach me to leave the heating on). Thing is, the IR sensor is misbehaving quite badly. If I leave the sensor in 'free air' (actually mounted in a Panavise), I get these values in the registers:

  R[0] (Vobject) = 0xFBE1
  R[1] (Tambient) = 0x0C5C

Putting this into the TMP006 spreadsheet results in an ambient temperature of 24.718750 Celsius... and an object temperature of 2.295038 Celsius...

OK, so what happens if I put my hand about an inch above the sensor?

  R[0] (Vobject) = 0xFA82
  R[1] (Tambient) = 0x0C6C

Putting this into the spreadsheet gets me an ambient temperature of 24.843750 Celsius... and an object temperature of -8.337225 Celsius. Now, I've been accused of having cold hands, but if they were that cold I doubt I'd be typing this message!

So, here's the question -- what's going on? Is it possible that the exposed copper pad is shorting out the IR sensor? Would I be better off removing it if I can't place an insulating solder mask over it?



  • Hello Philip,

    It's very unlikely that the exposed copper pad is shorting out the IR sensor. The assembly error that would have caused that would likely also have shorted adjacent balls of the BGA together, preventing any sort of functionality.

    The issue is that your copper area under the center of the chip is not covered with solder mask. The sensor is looking straight down at this area on the board and is 3x more sensitive in this direction than looking up. When the sensor sees either the exposed copper or any solder that's flowed to this area during reflow (since there's no solder mask to stop it), it sees a shiny metallic surface with poor IR emissivity and the potential to scatter any IR that does manage to get radiated from the area. The recommended layout requires solder mask to be present in this area for the proper thermal behavior to be achieved and correct object temperature measurements to occur. I apologize that I didn't catch this when reviewing your layout.

    A way to test this without having to spin your PCB is to apply an overcoat of solder mask to this area, for example, by using an overcoat pen such as the Circuit Works CW3300G

    Best regards,

    Ian Williams
    Linear Applications Engineer
    Precision Analog and Sensing Products 

  • In reply to Ian Williams:

    See, I'm not convinced CW3300G would do much good. It's only a conformal coating, and isn't likely to survive the soldering process (the maximum rated temperature is 255F when cured), or the ultrasonic cleaning process (acetone removes it completely, and I'll wager that isopropyl at least weakens it).

    The only other thing I can think of being worth a try is Kapton tape... or possibly dry-film solder mask (which I may well be able to get hold of...)

    Also, I noticed an issue with the datasheet - some of the "shifted hex code" values for the Tambient register examples are not correct. How does one go about filing a bug report for a datasheet? :)

  • In reply to Philip Pemberton:

    Well, I managed to sneak two boards onto the edge of a panel, with solder mask this time. You're absolutely right -- the solder mask does make a lot of difference. As in, the sensor is now producing sane values. It clips to -INF or NaN if the object temperature goes too high (putting a soldering iron tip in front of the sensor is generally enough to do that) but it works a treat for just about anything within the specified temperature range in the datasheet.

    The boards were etched in house using Mega Electronics Microtrak pre-coated 1.6mm double-sided FR4 laminate, with Harwin track pins to link the top and bottom layers, and dry-film solder mask (also from Mega). This combination worked a treat!


  • In reply to Philip Pemberton:

    Hello Philip,

    I'm glad to hear that you got the TMP006 working!  


    Ian Williams