• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 3 answers
  • Subscribers 20 subscribers
  • Views 1077 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

TPL5010: Clock timing is not what I expected

Steve Carkner
Intellectual 370 points
Part Number: TPL5010
Other Parts Discussed in Thread: TPL5000,

Support Path: /Product/Development and troubleshooting/

It would be REALLY nice if TI put out a spreadsheet or other calculator for this product. Even better would be a simple table of the values Rext = this time...

I have gone through the calculations and am pretty baffled by the quantization as it seems to indicate massive (+/- 6 second) steps in the setting 3 zone (10 - 100 seconds).

IN MY APPLICATION: I am using a 13.0K 0.1% accurate resistor, I was expecting a 15.0 second timing pulse, but instead I am getting 15.5 seconds.

Where did I go wrong?

T 15 Seconds
a b c
Use 0.1972 -19.345 692.1201
Rext 13000 Ohms
Rd 130
Tadc 6 Seconds
Err 60 %
  • 0
    TI__Mastermind 32575 points
    I believe the quantization error can be large since 1 s to 7200 s time interval range is divided by 1641 intervals, per section 8.5.4.

    Alan
  • 0 in reply to Alan O
    Intellectual 370 points

    I believe that the part designer specifically choose the four distinct ranges in order to reduce quantization error. With 1641 total across 5 ranges, this allows about  328 choices per range, so the 10-100 second range would therefore be broken up into perhaps 0.275s per segment. But there is absolutely NO information given about this. It would be INCREDIBLY helpful if a table of values was provided, or if TI put out a spreadsheet calculator.

    But to be clear - I don't actually care about how the product works internally, what I am really trying to figure out is what time period I should expect from a 13.0K resistor because the calculations seem to indicate 15.0 seconds, but I get 15.5 seconds. I also want to know how tight a tolerance I need on that resistor to ensure that when I produce a few million products, they all end up selecting the SAME timebase value.

    The Datasheet provides an example which is wrong in many ways:

    Example given says Rext is between 56.960K and 57.900K - Note however that Equation 1 produces a result that is in OHMS, so you would expect the other calculations to also use ohms

    It also says that using these values with Equation 4 produces values of 586.85s and 611.3s,

    Error#1 - Equation 4 has an INT in front of it, so it cannot ever return decimal places and therefore it is questionable what the part actually would produce (does it always round down?) You simply can't get a value of 586.85 seconds out of Equation 4 as it is written.

    Error#2 - Equation 5 says that Rd is equal to the INT of Rext/100, which would mean plugging the values of 569 and 579 respectively (ie. INT(56960 ohms/100) into Equation #4. However, this results in time values of 56 seconds, clearly wrong. Instead if you plug the OHMS value of Rext into Equation 4 (ignoring equation 5) then you at least get the times shown in the example. So what is the purpose of Equation 5?

    But this still fails to answer the fundemental questions of:

    WHAT TIME DOES THE PART ACTUALLY PRODUCE if I use a PERFECT, Ideal 0.000000001% tolerance resistor?

    and

    WHAT VALUE OF RESISTOR TOLERANCE WILL ENSURE THAT THE SAME TIME-VALUE IS SELECTED EVERY TIME, FOR EVERY PRODUCT?

    Doesn't seem like it should be this hard to say "what resistor value and tolerance do I need to create a 15 second pulse".

  • 0 in reply to Steve Carkner
    TI__Mastermind 32575 points
    Unfortunately I'm not very familiar with this device beyond the datasheet. I'm more knowledgeable on the LMK/CDC clock devices. I'm not aware of any calculator or tool for this either.
    But have you tried using a trim pot to determine the resistor value that yields the desired time interval? And adjust up and down to determine the sensitivity to the resistor variation/ tolerance?

    Alan
  • 0 in reply to Alan O
    Intellectual 370 points
    Sadly using a resistor to empirically measure the range on a few devices would not provide the proof required to satisfy our internal ISO processes whereby component value calculations must be used to prove operation prior to heading into full high-quantity production.
  • 0 in reply to Steve Carkner
    TI__Mastermind 32575 points

    For what it is worth, we did emperical testing on a typical EVM at nominal conditions and got the following result.  Only a 13k resistor with 0.1% tolarance could meet this.

    What is your end-equipment/application and estimated run rate?

    Rext (kΩ)
    T (s)
    12.84
    14.52
    12.91
    14.84
    12.95
    14.84
    12.97
    15.16
    13
    15.16
    13.02
    15.16
    13.05
    15.16
    13.06
    15.48
    13.13
    15.48

    Regards,
    Alan

  • 0 in reply to Alan O
    Intellectual 370 points

    The datasheet has a parameter called the "time interval setting accuracy" given as 0.6% typical, which I assume to mean the accuracy the resistance is read. Therefore a 13K resistor could be read as 12.92 to 13.08K.

    The 3% step between each time selection coupled with the ability of the part itself to read the resistance likely means that I would have to assume 14.84 or 15.48 seconds could also be selected (which is consistent with my findings that my first prototypes are giving a 15.47 second pulse using a 13.0 0.1% resistor.

    It is certainly an unusual part, it has an oscillator accuracy of 0.5% (which is fundementally what I need), but an inability to be set to anything better than 3%. Ooops.

    Production calibration could be done, but would require waiting 15 seconds for the first pulse to complete, that's a lifetime in high volume production. Also, no guarantee that the next time the part boots-up it won't jump a time step.

    Application: This is for a piece of soldier or police-officer worn equipment, so volumes on successful product launch would exceed 100K, probably would exceed 1M pcs.

    We are running the product on a CR2032 cell and are trying to get 7 years of battery life. Central processor is an MSP430, we are NOT using Vlo or a watch crystal because it draws too much power (1-2uA) compared with the phenomenal 0.05uA of this part.

    Is there perhaps a factory-set option where we could buy this part fixed permanently at 15.16 seconds (or some other value). 

    If you are producing a fixed value product for another customer, that may work too. I don't actually care about the exact value value as I can easily adjust my software to suit anything from 5 seconds to around 30 seconds. But i do need about 1% time accuracy over a wide operating temperature range and voltage range, this part just barely meets that criteria (for timebase accuracy, not including the setting-problems above).

    I really appreciate the results you have provided, they do explain a lot of what I am seeing.

  • 0 in reply to Steve Carkner
    TI__Mastermind 32575 points
    I'm not aware of a factory-set option. Were you aware of the TPL5000 which has 3 logic inputs to select from 8 settings between 1 s and 64 s?

    Alan
  • 0 in reply to Alan O
    Intellectual 370 points
    That's an excellent suggestion. I originally looked at the TPL5010 due to to the low 100ppm/C drift compared to 400ppm/C for the TPL5000. However, given the issues with the TPL5010, I will certainly revisit the TPL5000 as it should be possible to calibrate out any drift issues.

    Thanks for the suggestion!