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.

OPA2314 in constant current control circuit latch up

Other Parts Discussed in Thread: OPA2314, OPA2340, OPA340, STRIKE, OPA313, OPA314, OPA344, OPA343, OPA342

The drawing below shows an OPA2314 in a constant current control circuit.

When the +5V supply is turned on, the DAC puts out 0.07V (0.7V and 0.8V are typos), and the OA drives the FET gate with 5V, since the 1.6V supply is not on yet.

When the VCC (1.6V) supply turns on, the OA output does not stop driving 5V even though the difference between -IN and +IN is much more than the OPA2314 input offset voltage worst case of +3.5mV.

This behavior is observed on about 10% of the units built.

The +5V supply is shared by the DAC and OPA2314 OA.

Other clues.

+ If a small amount of noise is injected in the +IN input after VCC is on, the OPA2314 starts behaving.

+ If the VCC supply voltage is increased so that the Vds voltage is greater than 200mV, the OPA2314 behaves normally.

thanks for any insights you might have.

  • Additional test information.
    + no impact by removing 4700pF cap or replacing it with 1000pF, 2200pF.
    + changing 100 Ohm resistor to 1330 Ohms had no impact.
    + 1 out of 3 devices gets better when the circuit is at 85C.
    + testing at -5C shows no changes, e.g. a good part remains good and a bad part remains bad.
  • More test results.
    + loading the OpAmp output with a 1.3K resistor to ground helped some parts, but not all. The others railed to 4.2V and stayed there.
  • David,

    OPA2314 linear output swing does NOT extend within 70mV above the negative supply as the application requires. The specified linear swing is defined in the open-loop gain, AOL, conditions of the datasheet to be 200mV above the negative rail – see in red below.  What this mean is that even though the output may swing closer than 200mV, it is a non-liner operation where Vos max spec will NOT be met.

  • Thank you Marek.

    It is a disappointing revelation that the test condition is actually the linear operation limit.

    What good is the RRIO if it’s non-linear within 200mV to the rails?

    Is there a way to lightly load the output to keep it linear?

    Thanks
  • David,

    Even though linear op amp input common- mode voltage range can be truly rail-to-rail, the output can never reach either of the rails.  Lightening the load of OPA2314 will not help much because It's already very light. If you really need to drive the output to within 70mV of the negative rail, you may have to pick another op amp like OPA2340. However, you could instead increase the shunt resistor to 0.3ohm.

  • Marek,

    Can you explain why increasing the current sense resistor from 0.1 Ohms to 0.3 Ohms would resolve the problem with the output being within 200mV of the rails?

    Thanks,

    David

  • Hi David,

    Marek is out of the office on international travel so I'm going to provide some help on this one.  Let me get the circuit into TINA SPICE perform some analysis and then get back with you.

  • Thank you Collin,
    It would be great if this could be simulated, but that could be challenging since the problem is in <5% of the product. We have tested "good" and "bad" devices from -20 to 100C. The problem does not appear to be temperature dependent. It is not random - "good" devices remain "good", etc.
    Marek stated that the Vos and Aol specifications do not cover when the output is at the rails.
    Perhaps you can explain why?
    Thanks,
    David
  • Hi David,

    Somewhat as expected simulations don't reproduce this issue.  The simulations do confirm one of our initial worries that the circuit has the opportunity to produce very large currents which may damage the load or other circuitry in the condition where the load, and/or power supply voltage are removed and then reapplied.  The results are shown at the bottom of this message. 

    The initial analysis of this circuit was a little off as the op amp output does not need to swing to the negative rail when the input is a 70mV.  This is because to create the desired 0.07V / 0.1Ohms = 700mA load current the op amp must drive the gate to a voltage near the 2V threshold voltage. 

    Marek stated that the Vos and Aol specifications do not cover when the output is at the rails.  Perhaps you can explain why?

    This behavior is typical of all op amps and is specified in the open-loop gain (Aol) specification table.  Basically as the output approaches the rail the output stage devices begin to get saturated (or cutoff) and as a result the Aol specification, which is defined as ΔVout / ΔVos, begins to degrade.  The open-loop gain and feedback factor (β) determine the loop-gain (Aol*β) which directly relates to the linearity of the circuit.  Once the output is fully saturated the virtual short between the inputs is broken and the offset voltage can be very high.  Different op amps will have different linear swing-to-rail limitations based on the internal design, for instance the OPA340 which Marek mentioned can swing linearly within 5mV of the negative rail with a light 100k load. 

    Here's a few questions:

    1.) Is there something that limits the current in the condition where the op amp output is forcing the gate to +5V after the MOSFET drain voltage is reapplied? 

    2.)  What level of current flows during the condition above?  What voltage does the inverting input go to during this condition?

    3.)  Could you take scope captures showing the output of the op amp, voltage across the 100mV resistor, and IN+ voltages when the power supply is reconnected?

    4.)  Going back to your first post, you mentioned:  "+ If the VCC supply voltage is increased so that the Vds voltage is greater than 200mV, the OPA2314 behaves normally."  Could you elaborate on this test a little more?  With the 1.6V supply shown in the image, the Vds voltage should be well above the 200mV mentioned in your test.

    The circuit in an active dc state along with simulated transient results are shown below.  SW1 is opened at 750us and then closed again at 2ms, this behavior repeats.  As expected when the 1.6V supply (or load) is removed the output of the op amp saturates to the positive rail.  At 2ms when the supply is reconnected there is a large burst of current limited only by the source impedance of the supply, on resistance of the MOSFET and the 100mOhm shunt resistor.  The current spike will occur until the overload recovery time of the op amp is exceeded and the 1k+4.7nF feedback network can discharge to the proper level.  If there is not already one in the system, you may consider adding a series BJT based current limiting circuit to prevent this condition from occurring.

  • Hi Colin,

    The load can best be described as a 1.3V voltage drop in series with 0.5 Ohms. The supply voltage is adjusted so that there is ~50mV across the FET when the OPA2314 IN+ is set.  With that, the maximum current increase when the OPA is in overdrive is approximately 50mV/0.6 Ohms or 83mA.

    With the load explained, i think some of your questions are answered.

    Do you still want me to provide some scope pictures?

    Thanks for explaining the Aol and Vos connection.

    david

  • Hi Colin,

    Is there any possibility that the Vos behavior we're seeing resulted from an EOS/ESD stress to the part?
    Could an EOS/ESD related degradation have the same "signature"?

    Thanks,
    Dave

  • Hi David,

    We would still like to the see the scope captures if it's not too inconvenient.

    Thanks for the information on your load.  We may have misinterpreted what you described, but we still aren't sure why large currents don't flow when the power-supply is re-connected with the op amp output driving the FET gate to +5V, but what you described may help diagnose this issue.  If you reconnect the supply/load and only 83mA flows through the 0.1Ohm shunt resistor, the IN- pin of the op amp is still only at 8.3mV while the IN+  pin is about 70mV.  In this condition the op -amp will continue driving the output into the positive rail because the IN+ pin is greater than the IN- pin.  I think we would be able to see this behavior in the scope shots once you collect them, or if we've misunderstood the situation the scope shots may clear things up.

    It's difficult to state exactly how an op amp will behave after being exposed to ESD/EOS damage.  If large amounts of EOS energy are allowed to flow then catastrophic damage will occur which will result in complete device failure.  ESD can be a little trickier to diagnose depending on how severe the ESD strike was relative to the built-in ESD protection inside the devices. 

  • Hi,
    Sorry, I wasn't clear. The 83mA is the extra current available that adds to the ~700mA load current when the Op Amp is driving the gate at +5V.
    The current is limited by the voltage across the FET going to almost 0.
    david
  • Colin,

    Another question from our reliability folks.

    Since not all OPA2314 devices behave in a manner that presents this problem, is there a test or combination of tests that we could do to identify a part as being likely to exhibit the behavior?
    For example, would measuring the maximum +/-Vos under the condition where the output is allowed at the rails be a meaningful and repeatable test?

    Thanks,
    David
  • Thanks for the additional information.  In the condition where the output remains stuck at +5V, the IN+ voltage is roughly 70mV and the IN- voltage is 783mA*0.1Ohms = 78.3mA, resulting in an 8.3mV differential voltage between the inputs. 

    It's sounding like Marek's original suggestion regarding the degraded Aol is likely the cause of the issue.  If the degradation of Aol near the +5V rail causes the offset voltage to increase to >8.3mV then the IN+ voltage will look higher than the IN- voltage and the op amp output will stay stuck to the +5V rail.  Disturbing the system with an input transient will cause the output to move off of the positive rail and the offset voltage will reduce back to the expected datasheet levels allowing the loop to regulate again.  Increasing the FET VDS voltage increases the current flow and a larger differential voltage is produced between the inputs which overcomes the increased offset voltage, again allowing the loop to regulate.

    It may not be an acceptable condition in your system, but if more current was allowed to flow through the FET after the load/supply is reconnected then the differential voltage between the inputs would increase which would likely overcome the shifted offset voltage, as you saw when increasing the VDS voltage. 

    Otherwise, you may consider using the OPA313.  It is a very similar product to the OPA314 but with an open-loop gain specification that is valid to within 0.05V of the positive rails with a light load.  I agree that when the output saturates to the positive rail (within 5mV typically) the Aol and therefore offset voltages won't be specified, but because the discrepancy between the specification limit and the expected operating range is much reduced compared with the OPA314.  If you're interested, you can order some free samples from the link below: 

     

     

  • Colin,

    Your description of Op Amp behavior is in line with my observations during the "stuck" condition and how it is impacted by more load current or a disturbance on the IN+ input.

    By lowering the FET Vds we have been able to place a limit on how much current can flow in the load.

    With respect to the OPA313 you touched upon the concern that we may not be eliminating the problem, just lowering its probability.

    Most parts seem to have no problem. Is that because they don't saturate as much to the rails?

    Thanks,
    David
  • Hi David,

    Let me discuss this with some of my colleagues to see if we can come up with suggestions for a test that will help you screen for products which may display this behavior.  

    There's a few things in play regarding why some devices perform this way while others do not.  First, if the change in input offset voltage as the output approaches the rail is negative, then when the load is reconnected the IN- pin will always still seem higher than the IN+ pin and as a result the output will move off of the positive rail to try to reestablish the virtual short between the inputs.  Similarly, if the change in offset is positive but  is < 8.3mV then when the load is reconnected the IN- pin will be at a higher voltage than the IN+ pin and the output will move off the rail to regulate the output.  Unfortunately as noted earlier, the typical/maximum Aol specifications are not provided for this region so we don't have any information on what a typical device may do versus a device that exhibits larger amounts of this behavior.  We do think the OPA313 will perform better in this situation, but I understand your concerns since we won't be able to guarantee the performance in that region.  

    A device like the OPA340 is fully specified for output voltages within 5mV from the rail and we're pretty confident you wouldn't see the same issues with this device.  Other similar products with different bandwidths, offset voltages, and prices are the OPA342, OPA343, OPA344 

  • Also, I remembered one of my former colleagues published a nice blog post on offset voltage and open-loop gain that you may find interesting:
    e2e.ti.com/.../offset-voltage-and-open-loop-gain-they-re-cousins
  • Hi Colin,

    Thanks for the blog reference. I had seen a similar treatment in a book. It explains the macro level level.

    I have been able to take the waveforms you requested. Hopefully they are self explanatory. Thanks, David

    Control loop Start Up Problem waveforms.pdf 

  • Hi David,

    We took a look and would like you to ship us some of the units you're having trouble with for us to analyze.  Do you have ~5 units you can identify that show this issue along with a similar number of units that do not?  We'll take a look at the device parametric specifications to see if there's anything obvious that would help you look for units that may exhibit this behavior in your product.  I'll send you a private message with an address and some instructions.

  • Yes, i can provide 2 of each. They are all in a DFN package, so i need to go outside to have them removed safely.
    Thanks,
    David