• State TI Thinks Resolved
  • Locked Locked
  • Replies 13 replies
  • Answers 1 answer
  • Subscribers 84 subscribers
  • Views 9038 views
  • Users 0 members are here
Support feedback
Options
Options

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.

piccolo internal temperature sensor

Max Sonnaillon
Intellectual 600 points

Hello,

I designed my system using 28035PNT (105 deg C) DSPs, the internal temperature measurements were very reasonable in about 20 different units I tested.

When I went to production (28035PNS - 125degC), the internal temperature sensor is completely useless, it has a +/- 20deg  tolerance and they are very noisy.

My question is: why I am having such bad measurements from the internal temperature sensor? is it a known silicon bug that the temperature sensor is useless?

Thanks for your answer.

Max

  • 0
    TI__Expert 8500 points

    Max,

    There is no known bug on the internal temp sensor making it worthless. You should be able to get a pretty accurate reading of the silicon temperature. Note that this is the junction temperature that is being measured, not the ambient air temperature. The two will likely be different with the junction temperature being higher. Take a look at the errata for the 28035. There is an ADC erratum on the first sample of every sequence. If you are measuring the temperature as its own conversion sequence and only measuring it once you could be getting an erroneous value.

    Regards,
    Dave

     

  • 0 in reply to David Foley
    Intellectual 600 points

    Dave,

    Thank you for your answer. I am aware of the bug on the first sample, just in case I ignore the first conversion.

    I was successfully using the temperature measurement in the first lot of DSPs (they were the 105 deg C version). But in my first production build, with a different DSP lot (125 deg C) but the same hardware, I can guarantee that the temperature measurements are useless. They are very noisy and have a huge random offset. Could you help me if I send you the date code? I need to know if I can trust the temperature measurement in the future.

    I posted another issue with the ADC regarding the ACQPS, could you please take a look at that? I am pretty sure that this issue is a bug in the ADC, I have seen it in all the DSPs I have tried and also other people have posted similar issues.

    Thanks,

    Max

  • 0 in reply to Max Sonnaillon
    TI__Expert 8500 points

    Max,

    Please go ahead and post the date code from the top of the devices.

    Regards,
    Dave Foley

     

  • 0 in reply to David Foley
    Intellectual 600 points

    Dave,

    The code that I see under the part number TMS320F28035PNS is:

    C-08AGN9W G4

    Thanks,

    Max

  • 0 in reply to David Foley
    Intellectual 600 points

    Any news Dave?

    Thanks,

    Max

  • 0 in reply to Max Sonnaillon
    TI__Expert 8500 points

    We do see anything aberrant with that lot trace code. Not sure what the issue is. If you would like you can send one back to TI and we can take a closer look at it.

    Regards,
    Dave Foley

     

  • 0 in reply to David Foley
    Prodigy 90 points

    I'm using TMS320F28027PTT for some time now for production and observed following:

    The ones with the code CA11DD47T  showing significantly higher level of temperature then any other code, include ones with code 0 (which were a bit bugsy).

    I'm quite sure that the actual temperature should be around 32degC, because it used on the automatic tester with very good repeatability of the results. I also double checked it looking through Fluke T20 thermal imaging camera and I don't see any difference on the surface temperature in very stable load conditions. I've done it dozens of times already.

    My setup is:

    1) 2.5V ref voltage

    2) I'm using classic offset/slope recalculation formulae from the ADC datasheet, but first recalculate ADC measurement from 3.3V ref to 2.5ref.

    3) I'm using internal 1.8V LDO

    4) I don't see any noise due to software filtering and using all errata's suggestions

    Possibilities are:

    a) TI mfg calibrated it badly on the temperature deviation

    b) at some point ref voltage changed from 3.3V to 3.1V or whereabouts. That will give about the same deviation I'm observing

    c) Some hot spot on the die which is impossible to see outside even after an hour of the stable work?

    Please comment on my observation. For the moment data from that sensor is not very reliable.

     

  • 0 in reply to Igor Abramov56564
    TI__Expert 8500 points

    Igor,

    I'm confused. Point 1 says you are using 2.5v ref voltage which makes me think you are running the ADC in external reference mode with 2.5v on the VREFHI pin. Option b proposes that the ref voltage changed from 3.3v to 3.1v. Can you explain what you mean by that?

    If you have multiple devices failing please send one back to us with a thorough description attached so we can take a look at it and try to determine the validity of option a.

    Regards,
    Dave Foley

     

  • 0 in reply to David Foley
    Intellectual 600 points

    Dave,

    I never sent you the IC to TI to be analyzed, but I tell you that I don't use the internal temperature sensor any more. Some lots have good temperature measurements but some others are bad or very bad, so the measurements are no reliable.

    I'm happy with the ADC performance for the other channels, but the internal temperature sensor is useless. The tolerance is too high. Maybe it uses too few bits to represent a big change in temperature so it would need a perfect ADC to produce a reasonable measurement in degrees.

    The on-chip ADC has another issue with the sampling window that I posted in the forum, but other than that it works fine.

    Regards,

    Max

  • 0 in reply to David Foley
    Prodigy 90 points

    David,

    yes, I'm using 2.5V exterhal ref.

    option b is my speculation on what TI have changed in their calibration process. if instead of 3.3V, as stated in datasheet, TI are using 3.1V, my numbers (from some of MCUs I've used) does add up. It's too much f the difference to blame it on just tolerance, also it corresponds to the different lot numbers.

  • 0 in reply to Igor Abramov56564
    TI__Expert 8500 points

    Igor/Max,

    Can you two each send a failing part back in to TI? We don't see any issues with the temp sensor, even at 2.5v VREFHI. We'd like to get to the bottom of this and understand if we have a device issue of if there is a nuance in the way we use the temp sensor vs. you. If you can spare it, sending a good device back as a reference may help us. We can use one of ours as a reference device, but it always helps to have one the customer says is good.

    Regards,
    Dave Foley

     

  • 0 in reply to David Foley
    Prodigy 90 points

    My MCU are soldered to my product.

    1) I can try to unsolder it, risking a bit of mechanical/thermal damage

    2) Give to you my PCB with TI chip on it + JTAG connection schematic

    Please provide the address, or should I give to the local TI rep?

    regards

    Igor

  • 0 in reply to Igor Abramov56564
    TI__Expert 8500 points

    Igor,

    Number 1 would be our choice. We don't really have equipment to desolder customer's boards and then there is the question of what to do with the board after we desolder it. We do have equipment to clean and straighten the leads, so as long as you don't break the pins off or totally mangle them we should be able to correct them.

    I'm attaching a form with instructions. Typically we like to go through the local TI reps so that we have a contact point. In this case I think the Forum will suffice. If we need to pull in the local rep later we can. Please reference my name and this thread in your problem description on the form.

    Regards,
    Dave Foley

    TI_customer_return_form_FLASH.doc