• State TI Thinks Resolved
  • Locked Locked
  • Replies 10 replies
  • Answers 1 answer
  • Subscribers 50 subscribers
  • Views 981 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.

OPA2673: OPA2673 non-inverting input bias current shift

Koen Verdijck
Prodigy 50 points
Part Number: OPA2673

We recently shipped over 250 products including the OPA2673 from Belgium into Spain by road transport (Barcelona, then further to 23 different sites). Once arrived, on all sites, 90% of the products showed degraded performance. After analysis, we found that the degradation is due to a dramatic increase in bias current on the noninverting input: while we normally measure a few microamperes (consistent with datasheet), we now measured tens of microamperes, with some devices up to 55 microamperes at 25°C, while the spec says maximum 25 microamperes. We have no idea what could have caused this shift in input bias current. All units were duly packed in anti-static bags. Some units stayed in our Belgium warehouse and are not affected. Any idea of what could cause such a bias shift over the course of a few months time?

  • 0
    TI__Genius 15975 points

    Hi Koen,

    Were these products subjected to any kind of mechanical/electrical stress during transport? It is possible to shift the OPA2673 device (or any op-amp) DC parameters if the device is subjected to some form of mechanical/electrical stress.

    Also, the dramatic increase in the bias current only shows up for the non-inverting input, or its visible for the inverting input as well? Is it possible for you to measure supply current of the OPA2673 as well to see whether it is out of limits?

    I am not sure if these parts were soldered or not, but it might be good if you can attach your OPA2673 schematic in-order to help debug further.

    Best Regards,
    Rohit

  • 0 in reply to Rohit Bhat
    Prodigy 50 points

    Hi Rohit,

    Many thanks for your quick reply.

    On the damaged device, the current measured on the non-inverting input almost equals the current on the inverting input (~60µA). Supply current is difficult to measure, but when measuring the temperature at the top of the package, a temperature rise of 3.5°C is measured, with respect to the PCB temperature. This is consistent with non-damaged devices, so supply current should be similar.

    The OPA2673 is soldered to a PCB, about 220 x 60mm, which itself is mounted on top of a heatsink with the same dimensions. Mechanical stress is therefore quite unlikely. Electrical stress is also unlikely since the units were never powered during transport. Attached the schematics of the circuit for your information.

    We were also thinking about radiation (x-ray, gamma, beta). However, the returned units showed no increased radiation levels.

    Best regards,

    Koen

    OPA2673 Schematics.pdf

  • 0 in reply to Koen Verdijck
    TI__Genius 15975 points
    Hi Koen,

    Looking at the OPA2673 schematic, the non-inverting inputs are biased to a 2.5 - 3V DC voltage on a single 10V supply which seems to be very close to the negative rail or GND. The datasheet specifies +/- 2.5V headroom from either supplies at the non-inverting input at 25'C room temperature. So, have you made sure that this 2.5V - 3V bias voltage is not drifted to less than 2.5V, or say 2V? Also, have you tried increasing this bias voltage to 5V DC voltage or mid-supply and seen any improvement in performance?

    When you mention 90% of products showed degraded performance, what parameter is actually affected by the increase in the input bias current? Are you noticing any distorted output waveform or oscillations at the output?

    I also noticed that the R3 & R4, or essentially the feedback resistor (Rf) in your circuit is set too low which could result in some frequency response peaking and affect stability of the circuit. So, would it be possible for you to make the R3 & R4 as 511 ohms and similarly change the R5 as 340-ohms to see whether the performance is improved.

    Best Regards,
    Rohit
  • 0 in reply to Rohit Bhat
    Prodigy 50 points

    Hi Rohit,

    Thank you for your answer.

    Maybe a bit more explanation about the actual application: the OPA2673 is actually used as a preamplifier and drives the gate of a mosfet transistor pair, providing the bias gate voltage of the transistor at the same time. This is why a voltage of 2.5-3VDC is required. So it is not possible to increase this voltage at will as this would set the transistor to another working point which is not what we want.

    The problem is that due to the increased bias current at the non-inverting input, the voltage drop over R25 and R28 becomes too high (up to 0.5V), which causes a major shift downward of the transistor operation point.

    We did not observe any oscillations or higher distortion levels.

    At the moment, we are implementing a solution where we tune up VB to compensate for the voltage drop. However, there is no guarantee that this is a stable solution since we do not know why the OPA2673 input bias current has increased. During this campaign, we see that about one third of the units need a significant rise of the VB voltage (300mV or more), while about 50% require less than 100mV increase. Units with more or less VB compensation seem randomly distributed among delivered products.

    I will test your suggestion to increase R3 & R4, but I guess this will have no effect on the noninverting input bias current.

    Best regards,

    Koen

  • 0 in reply to Koen Verdijck
    Guru 140200 points
    Hi Koen,

    can you add a small negative supply voltage? Eventually made from a charge pump?

    Kai
  • 0 in reply to Koen Verdijck
    TI__Genius 15975 points
    Hi Koen,

    One thing you can certainly try and see is reducing the R25 and R28 from 9.1 k to 1k or 2k, and see whether the voltage drop across these resistors reduces. I am not sure if there is any reason for these resistors to be in the k-ohms range as the non-inverting input impedance is ~1.5 Mohms. It should be possible to scale down these resistors even further below 1k, if that helps reduce the voltage drop.

    You might also be able to extend the negative input voltage headroom by slightly offsetting the negative supply below GND, as Kai mentioned. However, I am not sure whether this is feasible now given that you have already built the boards.

    Best Regards,
    Rohit
  • 0 in reply to Koen Verdijck
    Prodigy 50 points

    Hi Rohit,

    I hope you are doing fine.

    This week, we have set-up some units in our lab in similar conditions as in Spain, in an attempt to simulate the problem.
    We took four new units from our warehouse, and had them in our lab for observation while they were simply powered-on.
    The result is that after one week, one module showed a severe increase in bias current:38 microamperes, while the three others show only a few microamperes, as was initially the case for all four.
    It appears that over the course of a couple of days the OPA2673 non-inverting input bias current shifts beyond published limits only while powered-on.
    This confirms the behaviour we see in Spain. Therefore, we can exclude any causes due to transport, storage and handling after the units were shipped.
    We could not observe this during our initial tests as the units were then only powered for a couple of hours.
    When we now measured several units that were produced years ago, and have been powered for a long time, we do not measure increased input bias currents.
    We therefore start to suspect that we have a bad batch of OPA2673 components, or mishandling during assembly of the boards. We contacted our contract manufacturer for clarification.
    All the OPA2673 devices used are taken from a single production batch: identificaton number 201731MYWG.
    With this new information could you please think about what could be the cause. Did you experience this behaviour before? Did you have other issues with components from this batch?
    Would it be useful if we send you some damaged devices for your analysis?

    Best regards,
    Koen Verdijck.

  • 0 in reply to Koen Verdijck
    TI__Genius 15975 points
    Hi Koen,

    Thanks for the production batch identification number, and let me see if I can find out the test results that would indicate anything unusual.

    Something you may already be aware of is that it is TI policy to always test the production parts before shipping them. If you check the Electrical characteristics table in the OPA2673 datasheet, the test level A indicates 100% tested at 25'C and the non-inverting input bias current is certainly categorized as test level A. So, I am not sure whether it is possible to find the OPA2673 test data with non-inverting input bias (Ib+) current beyond the +/-25uA test limits.

    For returning the damaged units, would it be possible for you to do the return through a local TI FAE (if units directly bought from TI) or through the distributor (if units bought from the distributor)? I think this just ensures safe handling and proper channel of communication for the damaged units. Once we have the units, it is possible to determine the exact root cause of the failure.

    Best Regards,
    Rohit
  • 0 in reply to Koen Verdijck
    Guru 140200 points
    Hi Koen,

    I don't think that the chips are bad. I think it has to do with the common mode input voltage. Input bias currents are specified in the datasheet at a common mode input voltage of 0V, which is in the middle of the +/-6V supply. Translated into your application this would be a common mode input voltage of 5V. But your common mode input voltage is only 2.5...3V. This can result in a different input bias current.

    Kai
  • 0 in reply to kai klaas69
    Prodigy 50 points

    Hi Rohit and Kai,

    Many thanks for your valuable inputs.

    We would very much like to know the root cause that resulted in a shift over time of the non-inverting input bias current.

    We take this up with our contract manufacturer, who will send some devices back through the normal supply chain for further investigation.

    I will keep you posted about the results.

    Best regards,

    Koen.