• State Resolved
  • Locked Locked
  • Replies 10 replies
  • Subscribers 47 subscribers
  • Views 1019 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.

OPA378 odd behavior

Tim Sobering
Intellectual 860 points
Other Parts Discussed in Thread: TPL0102-100, OPA378, OPA330, OPA335, OPA376, OPA188, OPA333, OPA2180

I am using an OPA378 as a buffer on a reference.  It is configured as a unity gain buffer, with an RC filter (1MEG series, 15nF shunt) before the noninverting input, and followed by a connection to pins 3 and 4 (high terminals) of a TPL0102-100 (digital pot).  The low terminals are tied to ground so this looks like a 50K load.

V+ of the OPA378 is just shy of 3.3V.  V- is at ground (measured).  The input to the RC filter should be 500mV but in circuit is 478mV.  The NI input of the OPA378 is at 436mV so there is a 42mV drop across the 1MEG resistor (well beyond what the bias current would normally cause).  The output of the OPA378 is at 497.6mV even though the output is tied to the inverting input...no gain resistors. 

I have looked with a scope and all of the waveforms are solid DC...no oscillations.

If I remove the input RC filter and inject a DC signal from a reference directly into the NI input, the amplifier works perfectly.  I have used RC filters on the NI input of opamps before without a problem.  The application required a very low-noise reference so I would prefer not to remove the RC filters. I'm well within the CM spec for the part.

Any idea as to what is causing the problem?

Tim

  • 0
    Intellectual 860 points

    FWIW...exact same circuit (on a different design/different PCB) using an OPA330 running at 2.7V/GND exhibits the same behavior. 

    Also, removing the cap in the RC filter doesn't fix the problem. I can "hide" the problem by reducing the value of the series resistor.

    Tim

  • 0
    TI__Guru 68661 points

    Tim,

    It would be really helpful to see the actual schematic of your application instead of its description - especially when it comes to the digital pot connection, loading, etc.

    Having said that, OPA378 and OPA330 are chopper op amps where the input voltage offset is actively corrected by the internal circuitry by continuous swapping of the input terminals at fc=350kHz and 150kHz, respectively, clock frequency - for that reason, placing the large input resistor and cap at the non-inverting input will drastically unbalance the impedance seen at each of the input terminals (at fc frequency) preventing the internal chopping action of the amplifiers from working properly.

    I would suggest to move the RC filter to the output of OPA378 as long as it is possible from the loading point of view - what is the load (other than 50k pot) at the output of OPA378?  If this is not possible, I would recommend to use a non-chopper type of amplifier like OPA376 or OPA335.

  • 0 in reply to Marek Lis
    Intellectual 860 points

    Attached is a pdf of the schematic. 

    I suspected the chopper as being the issue until I replicated the problem using an OPA330 which is not chopper stabilized.

    8308.OPA378 Schematic.pdf

  • 0 in reply to Marek Lis
    Intellectual 860 points

    Unfortunately neither of the parts you suggested will work as the OPA335 is not available in an SC-70 package and the OPA376 has way too high a quiescent current.  I do appreciate your suggestions and I suspect you are correct about the chopper, but I'm skeptical as to that being the whole problem as the OPA330 has the same issue (although it is "auto-calibration" so perhaps you are spot on if that is similar internally). 

    Any other ideas would be appreciated as I have 380 boards on the bench waiting for rework. :(  While I can use a higher quiescent current in this application, the second application requires micropower and so whatever I use needs a quiescent current less than about 50uA/amplifier

    Tim

  • 0 in reply to Tim Sobering
    TI__Guru 68661 points

    Tim,

    OPA330 as well as OPA333, OPA378, OPA2180, OPA188 are all chopper amplifiers.  Your best bet would be to use instead OPA376 - it is a linear amplifier with the input voltage offset comparable to the chopper amplifiers (max 25uV) and 0.1-10Hz input voltage noise of 0.8uVpp.

  • 0 in reply to Marek Lis
    Intellectual 860 points

    Oops...I guess you already told me the OPA330 was a chopper...I missed it in reading your response.  Sorry.

    OK...I'll solve this by removing the filters for now.  Once everything is installed I would have 41 OPA376's/system and I cannot afford the additional current.

    Thanks!

    Tim

  • 0 in reply to Tim Sobering
    Guru 21975 points

    Tim,

    Something doesn't seem quite right here and I'd like to look at this further. Could you please contact me at thesignal@list.ti.com.

    Regards, Bruce.

  • 0 in reply to Tim Sobering
    TI__Guru 68661 points

    Tim,

    I re-read your description of the problem below and it occurred to me that your issue might be related to the loading effect of the multi-meter you use.  Most hand-held meters have 10Mohm input impedance so if you measure the voltage on the non-inverting input (with respect to ground) it would result in a voltage divider between the 1Mohm R24 input resistor and 10Mohm impedance of the meter; this would explained why 0.478V on the input of R24 results in 0.436V on the right side of the resistor (pin 1): 0.478V*10M/11M=~0.436V.  Likewise, this would explain why the output of OPA378 is back at 0.4976V - your must have unloaded the input by removing the multi-meter from pin 1.

    You do not specify the value of the R36 resistor but if the multi-meter loading is the cause of the problem I would expect R36 to be ~46k so 0.4976V on the left side of R36 becomes 0.4953V on the right side (pin 5 of U27):  0.4976*10M/(10M+46k) = 0.4953V

    I would suggest to try to measure the voltage across the R24 (not with respect to ground) to verify that there is no 42mV drop across it - this would confirm that the multi-meter loading and not OPA378 is the cause of the error you see.

  • 0 in reply to Marek Lis
    Intellectual 860 points

    Marek,

    Odds are you are correct.  The meter I'm using has a 1 Gohm input impedance, however, it has to be set that way in the menus every time it is powered up.  I usually have it set to Hi-Z but if power dropped out the meter would reset it.  Guess what...thunderstorms last night and I recall hearing my UPS beep.

    I'll check in the morning.  Cross your fingers.  Possibly this is all a measurement problem.

    Nice catch!

    Tim

  • 0 in reply to Marek Lis
    Intellectual 860 points

    Marek,

    I verified the DMM input Z this morning.  You were correct...it was 10Mohm.  I reset it to 1Gohm and the "problem" went away.  I'm somewhat embarrassed that this bit me. I'll let Bruce know you solved it.

    You know you have been in this business too long when you start remaking the mistakes you made when you were young. :)

    Thanks again,

    Tim