• TI thinks resolved

OPA847: High gain in 0-30 MHz range

Part Number: OPA847


I have assembled the transimpedance amplifier with OPA847 for balanced homodyne detection. Feedback resistance 4.3 kOm, feedback capacitance(with parasitic capacitance of feedback resistor) 0.4 pF. The scheme is:

To simulate frequency response I used the next model:

According to TINA the square of gain-modulus versus frequency for my scheme is shown here:

However the measured gain differs from calculations in the frequency range 0-30 MHz. Red line shows approximation line from my theoretical model.

What is the reason for the discrepancy? What is missing in calculations?

I used Hamamatsu S5972 photodiodes with reverse voltage 13 V.

  • Hi Arsen,

    The discrepancy seems to be at the lower frequency end where the chances of correlating measured response to the simulated response are very high. Would it be possible for you to plot the measured and simulated response in the same y-axis [such as dB]? Also, it would be better if the frequency axis is plotted in the more recognizable MHz unit same as simulation.

    Also, are you using the same value Rf resistor in your measurement as the simulation?

    Best Regards,
  • In reply to Rohit Bhat:


    Thank you for response!

    Yes, I'm using  the same value of Rf=4,3 kOm and Cf=0.3 pF (+parasitic capacitance of Rf (resistor 0805) ~0.1 pF).

    The normalized gain for Rf=4.3 kOm shown below:

    High gain in 0-30 MHz only appears for low Rf (I tested it for Rf=4,3 kOm and 2 kOm).

    For example, on next  figure shown normalized gain for Rf=12kOm, Cf=0.2 pF.


    In this case no gain peaking  in 0-30 MHz range.

  • In reply to Arsen Kuzhamuratov:

    Hi Arsen,

    Would it be possible for you to describe the setup that is being used to measure the ac response? Are you using the photo-diode while doing the ac-measurement or directly injecting the ac-signal at the inverting pin? Are you calibrating out for any intensity noise, or responsivity errors in the ac frequency range?

    It looks like the ripple in the frequency response is ~1dB which somewhat resembles a Chebyshev filter response with a ripple zero. It is possible that the ripple could be coming in from un-calibrated cables. You could also be adding for unaccounted trace parasitic (cap/inductance) if the photo-diodes are quite a distance away from the inverting input pin of the OPA847.

    The other possibility is the roll-off of the photo-diode frequency response when using in back-to-back configuration as you have shown. It is possible that the photo-diode BW is being limited in this configuration which is causing the overall frequency response to dip before the OPA847 device roll-off. Is the 13V reverse voltage for each of the back-to-back photo-diodes or is it for the combined across both of them? It might also be worth just using one photo-diode and measuring the frequency response to make sure the frequency response dip is not caused due to limited BW of the photo-diode.

    Best Regards,
  • In reply to Rohit Bhat:

    Hi Rohit,
     The setup is shown here (taken from arXiv.org:1111.4012.).:

    "After having built the HD we proceed to test its performance. To that end, we place it in the experimental set-up shown in Fig. 5. Our LO is obtained from a modelocked Coherent MIRA 900 Ti:Sapphire laser producing 1.8 ps pulses with a repetition rate of 76 MHz, central wavelength of 791 nm. The HD beam splitter configuration is implemented by using a half wave plate and a polarizing beam splitter. The two beams coming out of the beam splitter are focused into the photodiodes. The HD is balanced by first adjusting the waveplate in order to equalize the responses of the photodiodes, compensating for a possible difference in their quantum efficiencies, thereby minimizing the spurious signal at the local oscillator repetition rate. A further crucial step in balancing the HD is to equalize the path lengths of the two beams entering the photodiodes by using a XY translation stage on which the HD is mounted. This is necessary because the pulses, if arriving at the photodiodes at different times, can lead to subtraction pulse having a bipolar shape. Even after these alignment steps, the subtraction may not be perfect due to different capacitances of the two photodiodes." 

    After  this procedure we get output signal ( light noise) with white spectrum thereby we can catch frequency response.

    Electronic scheme shown in next figure:

    I suggest high gain in 0-30 MHz doesn't result from parasitic cap/inducance . The 13V voltage was applied for each photodiodes. Hamamatsu S5972 bandwidth is 500 MHz at 10 V, so it didn't limit frequency response. I put 50 Om resistor on the ground before output BNC to  exclude cable  influence on frequency response.

  • In reply to Arsen Kuzhamuratov:

    Hi Arsen,

    When looking at the frequency response of the TIA, I think the normalization needs to happen with respect to the gain at the low frequency end. This is because the chances of matching the trans-impedance (Tz) gain at lower frequencies is higher between simulation and measurement. If I then look at the frequency response, you seem to have a gain droop between 10 MHz to 40 MHz.

    Looking at your setup, you seem to put a 50-ohm resistor to ground before the output BNC on your board. We do not normally recommend to put a 50-ohm resistor from the output to ground because it loads the op-amp directly when you connect the output BNC to any electrical equipment. You seem to have connected the output BNC to the digital oscilloscope as well as the spectrum analyzer which presents a combined load of 16.67 ohms at the OPA847 output (if I assume 50-ohms input for both the oscilloscope and the analyzer).  The 16.67 ohms load at the output is a pretty heavy load and might be responsible for causing the gain droop between 10 MHz to 40 MHz. In reality, you need a series 50-ohms from OPA847 output to the BNC such that the output is isolated from the test equipment.

    I also tried replicating your setup in my bench here for an Rf = 4.3k and assumed PD cap of 6.8pF. Attached is the measurement result that matches somewhat with the simulation and shows no gain droop in the frequency response between 10-40 MHz. 

    Best Regards,


  • In reply to Rohit Bhat:

    Rohit, sorry for my late reply. Thank you. I have to think more about the output circuit. It might be one of the reasons. At least I should understand how output circuit after OPA847 influence input signal.