• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 31 subscribers
  • Views 884 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMP006 mystery

David Luff
Prodigy 50 points
Other Parts Discussed in Thread: TMP006

Hi,

We've got a PCB, which we've got mounted a TMP006. We've prototyped and tested and its all working really well, very happy with the results.

We then moved to a short production run and the TMP006s all failed. On investigation the problem is that as soon as we use a protective coating around the TMP006 it seems to kill it. Note that we had masked the top of the sensor and sprayed on the conformal coating in the normal way. The top sensing surface was clean and unmarked.

Since then we've done lots of experiments and whatever product we use to seal the TMP006 onto the board we either get crazy readings from the TMP006, or no output at all, its just "dead".

Under a microscope we can see that there is a square in the middle of the sensor, and we assume that this needs to be free from contamination, but we cannot find anything about the requirements for this.

The problem for our product is that the PCB is not sealed and is used externally, and hence we need to add on some protection from the moisture.

We read on the datasheets that you have to keep the underneath clear of flux, and we've done this. Before we add a coating around the ship it is functioning perfectly. We've tried some very viscous sealants around the chip and are certain that it's not going underneath - the sealants are just providing a deal from the sides of the IC down to the PCB.

Can you offer any explanation or advice on what we should do? At the moment the only option to us is to add a mechanical seal over the chip with a germanium window, but the germanium window is really expensive and would require a PCB redesign. In this case it would be far cheaper to use a different temperature measurement method.

Does the "mystery" square at the center of the underside of the IC need to be vented to the atmosphere for it to work?

Is this mystery square very sensitive to some particular out gassing from sealants etc?

I've been working with ICs for my whole career, never come across anything like it! Please help us!!!

  • 0
    TI__Expert 3370 points

    Hi David,

    The "mystery Square" on the underside is the thermopile. It is sensitive to any heat source, including conduction. It sounds like at some point in the process someting is contacting the thermopile and creating a thermal short.

    The TMP006 Layout and Assembly guide (SBOU108) section 5.4 gives guidance on cleaning and section 5.5 basically says no underfill should be used or it will caue a thermal short. The behavior you are seeng is consistent with a thermal short.

    I would recommend to look at the cleaning procedure used and verify the underfill is removed. Also, ensure that no material is on the top of the device; the common overcoats are not IR transmissive.

    The alternative is to build a metal shield around the TMP007 with a window or aperture in it. If an aperture then mask during the overcoat process.

    If a window is used, a germanium window is not the only option; there are several cost effective plastics such as polyethylene and polypropylene (use thin windows, <0.5mm). Also, there are specialty materials such as the IRx series from Fresnel Technologies. Also, a silicon window will work for most applications in the MW and LW IR.

    I will check with our PCB experts and pass on any additional suggestions.

    Regards,

    Werner Metz

     

  • 0 in reply to Werner Metz
    Prodigy 50 points

    Morning,

    Well we can do some more experimentation, but to be honest I'm not at all convinced:

    - before adding any protection to the PCB the sensors are working fine, that is very clear
    - we tried a number of protection materials including epoxy, a few varnishes, silicon sealants, super glue, loctite, some general purpose and some sealants targeted at electronic with low fume etc. No matter what we used in ALL cases the sensor failed.
    - In some cases the "protection materials" we applied were quite fluid and I can believe with surface tension they COULD be "pulled" under the chip between the solder balls. Others were very vicious (like the sealants) and it is absolutely not plausible that this was "pulled under" to the central square.
    - with the application of some protection materials we were getting erratic readings, which would look like we shorted the thermophile. For other cases the sensor was completely DEAD and outputting nothing.

    So I don't believe that "underfilling" is the complete issue - it appears to us that the underside must be very sensitive to outgassing of any sealants that are used. However it is confusing.

    If none of this will work, then the best option is the metal cans soldered down and with a window - we can investigate this. If this is really the best way to handle it then it seems to me that Ti should supply the sensor in a suitable package like this that can be used directly - I think that ALL commercial PCBs are typically treated with a conformal type finish. To supply a ship that cannot be handled in this way is nuts isn't it?

    With regards the cover material, again we can do some more experiments but we have tried thin polypropylene and polyethylene from a few sources and in both cases the signal was effectively "killed" have you really tried these materials on the office, or are you just going from the theory? We've not tried silicon, so we can look at that.

    Thanks, Dave

  • 0 in reply to David Luff
    TI__Intellectual 910 points

    David,

    The sensor should continue to output something even if there is contamination.   Are you able to read the local temperature at all?     If the part is compleatly dead, it sounds like there was some sort of physical stress that may have damaged the device or broken the solder connection.     In the coating or cleaning process, is there anything that may be causing excessive mechanial or thermal stress on the part?    As this is a chip scale package, it is more fragile than plastic coated packages. 

    -Daniel

     

  • 0 in reply to DMAR
    TI__Expert 3370 points

    Hi David,

    Daniel is correct, you should be able to read local temperature if assembly process is not damaging the die. The undercoat issue will affect only the thermopile. WCSP is widely used by TI and its customers. If the symptom is indeed dead devices with no response to I2C and current draw outside the expected limits, then issues other than undercoat are occurring.

    I consulted with our PCB experts and they suggested to try the following.

    "We have used a syringe  and normal red RTV to deposit  a barrier around the bottom Of TMP006 and TMP007 to keep this area clear, of not only flux but any compound that will thermally short the sensor.  Only one side shown in these pictures but would need The bead all the way around.  Then you could coat the device with conformal coat  but Would need to keep the top side of the die open.  I’m sure the Conformal coats out there are not IR transparent ."

    See attached file.

    4743.RTVSolution.pptx

    As regards windows, yes we have successfully used a variety of windows. However, note that the transmission is very dependent on window thickness.

    I have attached an excerpt from the forthcoming TMP007 Layout and Assembly Guide concerning estimating window performance from a first order model for selected materials. Note that for polyethylene typically a thickness of 2mm (80 mils) or less is recommended, preferably ,30 mills.(0.75mm).

    4048.Estimating IR Performance of a Window.pdf

    Regards,

    Werner Metz

  • 0 in reply to Werner Metz
    Prodigy 50 points

    That is a very full and clear answer - its greatly appreciated.

    I'll do some more experiments over the next few days and get back to you.

    Kind regards!