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.

  • Resolved

OPA827: Vos drift over temperature

Prodigy 120 points

Replies: 9

Views: 119

Part Number: OPA827

Hello,

I have some questions about the Vos drift over temperature.

In the initial description, I have found that typical value is 0.5uV/°C:

but in the following table, typical value becomes 0.1uV/°C:

Why are they different?

I have usually found zero mean Gaussian distribution in other OpAmp datasheets, but this is not the case. Looking at distribution of fig.7, it seems to be like a Gaussian distribution with a mean different from zero:

Pheraphs, is the 0.5uV/°C considered as mean and the 0.1uV/°C value as the standard deviation?

But if that's the case, why is 0.1uV/°C value only taken into account in the following example (an extract from precision labs 2.1 solutions) using the same OPA827?

  

In case of distribution with mean different from zero,as in fig.7,  I would expect total Vos variation over temperature to depend on both mean and standard deviation (and not only on standard deviation):

+/- Vos(@25°C)   plus  +/- (mean value of Vos drift over temperature * deltaT)  plus  +/- (standard deviation of Vos drift over temperature * deltaT)

That comes back to the common form:

+/- Vos(@25°C)   plus  +/- (standard deviation of Vos drift over temperature * deltaT)

in case of mean equal to zero.

I apologize for the lenghty email, but I would like to build a solid background in this matter.

Thanks and best regards,

Massimo

  • Hello Luchi,

    The OPA827 datasheet has been around for 14 or 15 years and has undergone a few revisions. I suspect that some inconsistencies have creep in during the revisions.

    The datasheet Electrical Characteristics table specifications are what TI assures in terms of performance. You ask, "Perhaps, is the 0.5uV/°C considered as mean and the 0.1uV/°C value as the standard deviation?" For TI's Precision Amplifier products the typical "TYP" number of 0.1 uV/°C is the one standard deviation number (one sigma), and 2 uV/°C maximum "MAX" is a 5 or 6 sigma number based on what was selected during the original characterization. I am not sure why the 0.5 uV/°C typical drift number is stated in the datasheet Description section, but it may be a carry over from the original preliminary datasheet before the spec's were completely established. The Figure 7 graph isn't consistent with the Electrical Characteristics table either, but it may be that it represents a sample of devices at the time of characterization and different from what ended up in the table.

    The key is to use Electrical Characteristics table numbers. The calculations you show are correct for the Typical offset and drift numbers. I would perform the calculations for the Maximum numbers as well so you have an idea how high the input referred voltage offset will be at a particular temperature of interest. Do realize that the voltage offset might be higher at some point between the temperature extremes. It just depends how the voltage offset and drift behave for a particular device.

    Regards, Thomas

    Precision Amplifiers Applications Engineering 

  • In reply to Thomas Kuehl:

    Morning Tom et al, 

    I would say there have been some changes on this device and datasheet, I happened to have a 2009 revision in my directories, 

    opa827.pdf

    1. The front page typ matches spec table typ there at +/-1.5uV/C but with no maximum, 

    2. The histogram has two issues, 1st taking the absolute value which I think is odd for a bipolar physical spec, 2nd clearly some outliers that might suggest a design change somewhere along the road to reduce that maximum. Could also have been a packaging flow change to reduce die stress. 

    Michael Steffes

  • In reply to Thomas Kuehl:

    Hello Thomas,

    Thanks for the information.

    Now things are more clear and I will consider the value in the table as the typical one. I agree with you about the importance of "maximum" calculations to estimate the worst case. I know that the max value should be a tested value(OpAmp having higher values are discarded during  production tests),but the thing that is not clear to me is: if we assume it is a 6-sigma value and sigma is 0.1uV/°C, why is it equal to 2uV/°C instead of 0.6uV/°C for OPA827?

    In other words,if six sigma is 0.6 uV/°C, I would surmise it is quite impossible to find a part with a value over 0.6uV/°C so I can't understand why a 2uV/°C test should be done.

    Thank you in advance for your help.

    Regards,

    Massimo

  • In reply to Michael Steffes:

    Hello Michael,

    thank you for the additional details. I think that the answer to my question is a mix between changes in both documentation (as also highlighted by Thomas) and the device itself, as suggested by you. As regards your observation "taking the absolute value which I think is odd for a bipolar physical spec",  I didn't worry about it since I thought (but please correct me if I am wrong) that it depended on what definition was used (TI precision labs part 2.1):

    1)Without absolute value:

    2)Or with absolute value:

    Sometimes, in fact, I find datasheet with bipolar values or only with positive ones.

    Regards,

    Massimo

  • In reply to Massimo_Luchi:

    Most offset drift specs are centered bipolar, showing bipolar plots emphasizes if the mean is shifted more clearly than absolute value. Also, folks occasionally assume those absolute value plots mean a unipolar drift, which is misleading. Except for these parts, most device datasheets show the actual bipolar distribution - for instance, the OPA837 shows this, 

    Michael Steffes

  • In reply to Michael Steffes:

    Also Massimo, 

    Just to be clear, there are very few op amps that have a tested screen on offset drift. Two exceptions are the JFET input

    OPA656 and OPA657 that do have final test offset drift screen implemented. 

    The OPA837 parts were screened at probe for tighter than final limits, but not tested after packaging where we expect some shift in drift distribution - the drift outliers are not packaged, so the distribution I sent is a little more reliable than most in terms of not exceeding the stated +/-2uV/C max. 

    Michael Steffes

  • In reply to Michael Steffes:

    Hello Michael,

    thank you very much for your explanations.

    I think that I was confused by these two slides regarding OPA827, where max level is considered to be tested:

    As regards OPA837 Vos drift, notes in datasheet state: "(B)Not tested in production, limits set by characterization and simulation" and "Maximum drift specifications are set by min/max sample packaged test data using a wafer-level screened drift. Min/Max drift is not specified by final automated test equipment (ATE) nor by QA sample testing."

    So, comparing the two notes with your previous message, why is OPA837 considered not tested in datasheet? Isn't the screening at wafer level, to discard outliers before packaging, enough (despite a shift that may occur)?

    Here Vos drift is a specific case,but I think it may be useful for me to understand datasheet terms and assumptions in a more deep and general way.

    Thank you

    Regards,

    Massimo

  • In reply to Massimo_Luchi:

    Well Massimo,

    If you are not testing at outgoing final test, the actual drift in the population is not guaranteed. The untested limits on drift in a part like the OPA837 is more reliable given the wafer level screen, but not 100% for sure. Even the OPA656 is a bit vague in that the delta T for that test is not the full range. once you look at the full temp range, the linearity assumption for a small range is a bit of a stretch. There is quite a range of methodology across all parts in the industry, and rarely described clearly for this somewhat fundamental spec. Kind of surprising actually. A relatively new nuance came up in recent HSP releases. While the offset limit at room temp is a tested spec, we came to appreciate that ATE room temp is actually more like 22 to 32C, so we had to widen those test limits to accomodate drift issues over an this poorly controlled ATE environment. All parts have this issue, few consider it. 

    Tough issues, but parts are expected to show a max drift number for rough comparisons (and competitive positioning). 

    Michael Steffes

  • In reply to Michael Steffes:

    Hello Michael,

    your experience has given me a new insight on these topics. I really appreciate your help.

    Thank you for your time!

    Regards,

    Massimo

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.