• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 47 subscribers
  • Views 1857 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.

Photodiode Transimpedance Amplifier

Andy Stangeland
Prodigy 30 points
Other Parts Discussed in Thread: OPA657, OPA847, OPA354, THS4631

I need to measure current from a photodiode that has 80pF of capacitance.  The diode's current can be as low as 0.1uA.  I only care about static brightness of the light, not an AC signal.  Because my signal is DC, I don't think "noise" really affects my application, so I'm fine with a two stage design (although less components / less cost is always preferred).

The light is only turned on for 10 microseconds to be measured.  So I need to settle in roughly 5us so I still have time to take A/D readings.  I don't have a bandwidth requirement, just this settling time requirement.  My understanding is that a 5 microsecond settling time will require an effective bandwidth between 600kHz and 2MHz.

To obtain a 100mV output would require a megaohm feedback resistor.  According to equation 4 of AN-1803 to get to 600kHz would require a GBWP of 180MHz, and 2MHz would require a GBWP of 2000MHz.  Using a smaller feedback resistor might allow a slower opamp, but I worry about how small of output voltage is really measurable.

I'm still trying to understand some of the basics of transimpedance amplifier design.  Obviously there are some very fast opamps out there, but there are other parameters to consider: input bias current, input offset current/voltage, voltage/current noise versus sqrt(frequency).  I'm trying to understand how these factors will impact accuracy of the opamp.  How can I calculate how much these different parameters matter?

I was looking at another post, JFET + transimpedance amplifier, where Julien TAIEB was trying to measure a 0.1uA signal with an OPA847IDR opamp.  However this part lists -19uA typical Input Bias Current and 0.1uA typical Input Offset Current.  I guess I thought that either one of these input current errors would make it impossible to measure a signal as small as 0.1uA

  • 0
    TI__Expert 5400 points

    Andy,

    A couple of things here:

    First, the settling time of the response will be determined by both adequate bandwidth as you mention (rise/fall time), and the stability of the amplifier (ringing).  I have attached an example using the OPA657 that achieves 650ns 1% settling time with 1.4MHz bandwidth (at your specified gain of 1Mohm).  Notice, however, that without appropriate compensation, excessive ringing causes significantly longer settling time.

    To implement 1Mohm transimpedance gain in a single stage, a FET input is appropriate for its low bias currents, as you mention.  Though faster opamps are available, such as the OPA847, the combination of higher bias currents and a large feedback resistor cause the output offset voltage to quickly get out of hand.  Per the OPA657 datasheet, the worst case offset due to bias currents would be 20uV (20pA * 1Mohm).  Additionally, since the circuit is in a DC voltage gain of 1, the input offset voltage translates directly to the output.  For the OPA657 this is 1.8mV worst case at room temperature.

    Determining noise is a bit more complex, and will depend on multiple factors.  For the case you describe, there will be a tradeoff between the noise integration limit (frequency of the system), and the minimum bandwidth that you determine is necessary to adequately sample your signal.  For a more complete analysis, I would point you to the Application Note Noise Analysis of FET Transimpedance Amplifiers.  Since you are only interested low frequency content, averaging multiple samples could also reduce the impact of noise.

    2273.OPA657 TIA 1MHz.TSC

  • 0 in reply to Bart Stiller
    Prodigy 30 points

    Thanks for the explanation of offset and bias.  With R5 and C3 matching the impedance, can I ignore current bias and just use current offset as my error term?

    Is there any way to overcome more of this error?  In the "JFET + transimpedance amplifier" forum discussion, how would you use the OPA847 to measure the 0.1uA signal using a part with 0.1uA of current offset error?

    The OPA657 looks like a great opamp, that would work well for my application.  However it is a bit more expensive than I was hoping for.  So I'm just trying to understand if there is anything I'm missing here that would allow me to use a lower performance opamp.

  • 0 in reply to Andy Stangeland
    TI__Expert 5400 points

    Recall that input offset current is just the difference between the inverting and noninverting bias currents.  So in this case, there are three total sources of DC voltage offset error at the output.  First is the input voltage offset, that is seen directly at the output.  The second is ((Ibias-) + (Ioffset/2)) * Rf.  The third is ((Ibias+) - (Ioffset/2)) * R5.  Since Rf and R5 are equal, the Rf * Ibias terms cancel, leaving Voffset + (Ioffset * Rf).  Ultimately the voltage offset is the dominant term.

    The THS4631 and OPA354 are similar amplifiers that you might consider, though with lower bandwidth, but could be viable options here since you're looking at relatively low frequencies.

    Regarding the OPA847 situation, the offset current is a DC error, so it is possible to have an AC signal on top of that of equal magnitude.  They are just both subject to the same amplification through Rf, so it gets problematic pretty quickly at the output.