• State TI Thinks Resolved
  • Locked Locked
  • Replies 16 replies
  • Answers 3 answers
  • Subscribers 31 subscribers
  • Views 2716 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.

LM76: step in error curve @ -45degC

Mike Hubert
Prodigy 80 points
Part Number: LM76
Other Parts Discussed in Thread: LM95172-Q1

Hello,

 

We use the LM76 in some automatic test equipment. It is calibrated by curve fitting using a reference thermometer accurate to 0.02 deg C – target resulting accuracy of +/- 0.25 deg C. Temperature range -70 to +100 deg C. 20 temperature points in the calibration process (every 10 degrees +100 to -30 deg C, then every 5 degrees from -40 to -70 deg C). Process takes about 15 hours as great measures are taken to ensure temperature stability prior to taking measurements.

 

There is a recurring phenomenon whereby the error curve has a step around -40 to 45, the trend of which is maintained as temperature gets colder. In repeated tests, the magnitude of this step varies (0.05 to 0.2 deg C), sometimes barely visible, but more often than not it is there. This step has significant impact on the resulting curve fitting. The repeatability factor is also of concern obviously.

 

 

Looking into a number of possible causes, but wanted your opinion please.

 

We are aware of the following note in the datasheet: “Accuracy of the LM76CHM-5 will degrade 0.08°C for a ±1% variation in +VS from the nominal value.” Once we rule out any issues with the reference temperature probe used, we will perform some measurements on VCC. But this is an unlikely cause as the 5V supply is external to the thermal chamber, at room temperature. A power glitch could however still be related to connectors in the system.

 

Here are some sample datecodes, not sure how recent our stock is on this part:

 

M8RA

MSAB

 

Your help would be much appreciated!

 

Thanks,

 

Michel

  • 0
    TI__Expert 4905 points

    Hi Michel,

    Thank you for your inquiry.

    Was there an attachment of some sort? (there a big blank spot in your reply that is why I ask.)

    Can you please attach your schematic and the error transfer function you are seeing?

    We did not observe such a step function,

    Sources of the error that you may want to pursue:

    1. Investigate thermal stress caused by your PCB material on the device
    2. Consider using another part the LM95172-Q1 as it comes in a ceramic package thus package stress a cold temperatures would be less than the LM76 plastic package. We also have newer smaller parts that are 0.5C accurate over a wider band that may be of interest. (Let me know and I'll make a list)
    3. Check your temperature chamber stability - make sure the reference probe is placed really close to the part. (also keep in mind that the reference probe has more thermal mass than the sensor so the sensor may be picking up chamber fluctuation that the probe does not see)

    Take care,

  • 0
    TI__Expert 4905 points
    Hi Michel,

    I just saw the curve of the error. It looks to me like you have two discontinuities or jumps there's another one at hot 70 to 80C. So this point to a temperature chamber issue more than anything.

    Take care,
  • 0 in reply to Emmy Denton
    Prodigy 80 points

    Hi Emmy,

    Thank you for your response.

    In reference to your 3 points:

    (1) Our PCB is made of polyamide, therefore I do not think PCB stress is an issue.

    (2) We will definitely keep the LM95172-Q1 in mind, however it is not a drop-in replacement (package + communication protocol).

    (3) We may actually have an issue with the reference probe stability (which is not the chamber's probe), will look into that.

    ---

    I proceeded to some AC measurements on VCC = 5.0V, see attached document, which includes a schematic excerpt.

    This 160mV dip is definitely in excess of the +/- 1% variation in VCC ("Accuracy of the LM76CHM-5 will degrade 0.08°C for a ±1% variation in +VS from the nominal value."). However it appears as though actual temperature conversion occurs at a fixed rate, independently of temperature readout, and therefore may have a stable VCC when doing temperature conversion, although perhaps there are no guarantees on this.

    Would be surprised that this would only have an impact at -45C however. And you are right yes, sometimes there is also a step on the hot side, although much less consistently.

    Should I be concerned about this 160mV?

    LM76 - AC measurements on VCC.docx

  • 0 in reply to Mike Hubert
    TI__Expert 4905 points
    Hi Mike,

    When does this jump in the power supply happen? Can you control the LM76 conversion so that you avoid this? Perhaps use the one-shot mode?

    You are trying to get such accuracy and repeatability that I would be concerned about this.

    If you are concerned about the inductance and C5 you may want to replace C25 with a 0.1uF. If the caps are surface mount you could just parallel a 0.1 on top of the 0.01 to see if it works (I'm not saying go into production this way just for experimentation purposes). Make C1 bigger? What's happening on the other side of L1?

    Thank you for using TI products and your patience this is a hard one to solve!

    Take care,
  • 0 in reply to Emmy Denton
    Prodigy 80 points

    Hello Emmy,

    Thank you for your response.

    I haven't analyzed in great detail but all I know is the VCC dip corresponds to the 1Hz poll rate I had going during that test. I may indeed try adjusting the capacitance.

    What do you mean by one-shot mode? Doesn't come up in the datasheet I have ('SNIS109D –MAY 2004–REVISED NOVEMBER 2010').

    Thanks,

    M|ichel

  • 0 in reply to Mike Hubert
    TI__Expert 4905 points

    Hi Mike,

    Sorry I meant shutdown mode (got my parts confused). When it comes out of shutdown it starts to do a conversion you would wait 1 sec and then read the part. Actually also every time you read the part it starts a new conversion so you could do a double read. First read would start a conversion (you would not use the data) and then the second read 1 sec. later would access the conversion data.

    Take care,

  • 0 in reply to Emmy Denton
    Prodigy 80 points

    Hi Emmy,

    Still working on this.

    Would it be possible to please look up the datecodes I provided earlier, it is possible that we are using old chips, possibly as old as 2002, at best from 2011. Has there been a lot of work on the die since its initial release?

    Here are those datecodes:

    M8RA
    MSAB

    Thanks,

    Michel

  • 0 in reply to Mike Hubert
    Prodigy 80 points
    In case you need this and it hasn't been provided previously, here is the specific part number: LM76CHM-5

    Thanks!
  • 0 in reply to Mike Hubert
    TI__Expert 4905 points
    Hi Michel,

    Do these parts have a TI logo or a National Semiconductor logo?

    Thanks!
  • 0 in reply to Emmy Denton
    Prodigy 80 points
    Hi Emmy,

    They have the National Semi logo.

    Thank you for your continued support!

    Michel
  • 0 in reply to Mike Hubert
    TI__Expert 4905 points
    Hi Michel,

    BTW, there haven't been any changes to the part that would effect it's performance. I think if the parts were damaged you would see another response. Perhaps they wouldn't function at all or over-heat. You know that the LM76 is only operates down to -55C and that it's specified for accuracy down to -20C.

    I'm trying to find out about the date code for you - it may take some time. My first attempt say that the first two letters are the year and month but the coding does not make sense. Thank you for your patience and designing with TI products!

    Take care,
  • 0 in reply to Emmy Denton
    Prodigy 80 points

    Hello Emmy,

    Yes we are aware the part is only rated to -55C, but the step starts at -45C. And as I understand it, generally there is no issue with using parts outside their rated range, just that they have not been characterized / tested by manufacturers. Do you have any specific information to suggest that we shouldn't be using the part below -55?

    Regarding the accuracy specified to -20C, the discussed problems are occurring during our calibration process to improve that aspect.

    Thanks for continuing to look into the date codes. The ones I provided are definitely accurate, looked under the microscope.

    BTW, do you think the fact that the ICs are conformally coated with HumiSeal could have anything to do with it? Perhaps locking in some humidity?

    Thanks,

    Michel

  • 0 in reply to Mike Hubert
    TI__Expert 4905 points
    Hi Michel,
    Is R15 indeed 0 ohms? If it's not change it. The datasheet specifies the average current - the peak can be in the mA range so series resistance with VDD is not recommended.

    Take care,
  • 0 in reply to Emmy Denton
    Prodigy 80 points
    Hi Emmy,

    R15 is indeed 0 Ohms.

    I ordered in some new parts from Digikey (as you may recall, our current inventory may be anywhere from 2002 to 2011, but likely 2011). And they have the National logos with the date codes which seemed strange to you.

    Unfortunately what I received from Digikey also has the National logo, I had hoped / assumed I would be getting Texas-logo parts. I will see with them if they have any such parts. Otherwise, if I request samples from ti.com, do you know if they are guaranteed to be Texas logo?

    The date codes on the Digikey parts do not have the issue you were mentioning earlier (re: “I'm trying to find out about the date code for you - it may take some time. My first attempt say that the first two letters are the year and month but the coding does not make sense.”).

    Digikey marking (National logo): 34DC

    So these parts would be from 2003?? Trying these isn’t exactly accomplishing what I wanted, re: trying the latest ICs. But I will give it a shot.

    Thanks,

    Michel
  • 0 in reply to Mike Hubert
    TI__Expert 4905 points

    Hi Michel,

    thank you for your patience!

    Take care,

  • 0 in reply to Mike Hubert
    TI__Expert 4905 points

    Hi Michel,

    Here is the decoding of the date codes:

    34DC is April 2013

    M8RA is 48th week in 2009 or 2006

    MSAB is 6th week in 2010 or 2007

    I really don't think what you are seeing is a date code problem. If it is then it can come back and you may be stuck in the future with only getting product with the non-monotonic behavior as no changes to the part have been made. Since the part is within spec I can't comment on the non-monotonic behavior. I still really think it has to do with your setup because of the architecture of the LM76 should not show such behavior. When we have encountered such behavior we have found it be our lab setup. That being said we do not specify that the part is monotonic.

    Take care,