• State TI Thinks Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 48 subscribers
  • Views 1364 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.

Hobbs Cascode Transimpedance fails simulation, what's wrong?

Helmut Forren
Mastermind 6805 points
Other Parts Discussed in Thread: TINA-TI

At http://www.electrooptical.net/www/frontends/frontends.pdf Philip C. D. Hobbs suggests using a Cascode Transimpedance Amplifier.  I have attempted to simulate his Figure 9 circuit in TINA-TI.  However, the simulation doesn't work at all.  Surely he's right and I'm wrong.  But what did I do wrong?  Simulation attached.

Specifically, consider the Rbias from the cascode emitter to negative supply.  My input is a square wave ranging from 0 to 1pA (chopped light).  If Rbias is absent, I get a square wave out.  But if I insert an Rbias that's less than some huge G value, the output falls to just noise.  Also, if subsequently bring my negative supply up closer to ground, such as -0.6V supply, the output starts working again.  It seems like the circuit quits working as soon as there's a full diode drop across the base-emitter such that it starts conducting.  I think that's the whole point, however.  So I don't know what's wrong with my simulation.  Also, it may be that a sine wave signal gets through, but the square wave doesn't.  

Note that my absent or larger Rbias results didn't reproduce just now.  I believe that's because it was on a simplified cascode circuit, not the one attached.  Anyway, for the one attached, it still behaves as described when I reduce the negative supply.  At 0V, I get a good square wave out.  At -1V, I get a very noise square wave.  At -2V out, I don't see the square wave anymore, just the noise.  This noise is very slow, like less than 10 times my 10Hz chopped input signal.

Note that I'm simulating using MPSA18 in both transistor locations, although perhaps Q1 should be a different model.  But I don't have a different model installed that I can find.

So what am I doing wrong?

Thanks very much.

2234.OPA637 PRE-AMP (cascode fig 9, rev 3).TSC

  • 0
    Guru 21975 points

    Helmut,

    It appears that the expected operation of Q1 is not achieved because the U1 is not providing a virtual ground (0V) potential at Q1's collector. The emitter current set by the bias resistor, Rbias2, must flow in the collector. If this current is too large, Rbias1 and Rf cannot supply enough current to create balance at the inverting input node of U1.

    If you decrease the value of Rbias1 and/or Rf it will allow U1 to achieve linear operation and the inverting input of U1 will again assume a near 0V potential.

    Regards--  Bruce

  • 0 in reply to Bruce Trump
    Mastermind 6805 points

    Bruce,

    Your advice makes sense, but I can't get it to work in the simulation.  Did you simulate or just decide from general knowledge?

    I tried Rbias1 values of 2M,200k,20k,2k,200,20,2 and none worked.  I added a voltage monitor at the virtual ground, negative terminal of U1.   For values of Rbias1 of 2k and lower, that virtual ground is above 13V.  Not ground at all, of course.  For 20k it's 54mV.  Much closer.  For 200k and larger, it's -656mV and lower.  Futher.  So 20k plus or minus something gets closest to ground.

    But the Pre-Amp-1 output never looks right.  For Rbias=20k, I get the image further below.  I shouldn't have to fine tune the value of Rbias.  Is it because Q1 as MPSA18 is the wrong type, the wrong gain, or something?

    Then when I do some calcs... AM1 is about 7.18uA.  So Rbias2 of 2M should see 14.36V, putting Q1 emitter at -0.64V.  Sounds reasonable.  Rbias1 from 15V to vitual ground, and 7.18uA suggests 2.089M.  Hmmm... That's very close to Dobb's original 2M.  But 2M didn't work.

    I'm confused...

    Ah ha!  My photodiode model was I believe wrong.  Subsequent to this I had that D1 diode in the model.  Looking at currents just now, I saw AM2 was not equal to AM1.  The photodiode model must be carrying something.  I realize a few days ago, I took the diode out of the model.  It shouldn't be there.  I just removed the diode, leaving Rbias1 at 2M, and my PRE-Amp-1-Out is now working.  For 1pAp-p input, I'm getting about 327uVp-p out.  That gain is about 327 kV/A, close to the 301k Rf.  

    I think I can more forward from here now.  You didn't provide the answer, but you caused me to find it.  Thanks very much.  Second working graph attached.

  • 0 in reply to Helmut Forren
    Guru 21975 points

    Helmut,

    Looks like you are on track now. I just made a DC simulation and could see that your operating points were out of whack. Then could see the problem from inspection. I did not notice the problem with the photodiode model. You have the polarity of the diode backwards. See this blog for the proper polarity and other discussions.

    Illuminating Photodiodes ;-)  the basics of photodiodes

    You can add a diode with proper polarity and it should behave as an actual photodiode. Of course, the capacitance and other parasitic components of the diode will be added to the model.

    Regards-- Bruce

  • 0 in reply to Bruce Trump
    Mastermind 6805 points

    Thx.  Yep, I was up on the diode polarity, as well as I the parasitics.   I don't find an ideal diode in Tina, so in a different thread of simulations, I just removed the diode.

    In addition, it turns out this circuit uses the PIN in biased mode rather than zero biased mode.  So the diode cap becomes consequential.  The cascode isolates it.  Then by the way I don't get full Rf gain.  I suspect this is because a biased PIN might be, what, faster?  I don't need hi speed signaling.  It's better for the PIN to be in zero biased mode.  I believe both the dark current and capacitance are less of a factor then.  And my other simulation with regular transimpedance amplifier gets full Rf gain.

    Bottom line, this cascode arrangement might be a good idea for hi speed signaling, but not for low (dark) sensing.  I'm dropping it because my other thread for transimpedance amp is now working great in sim.

    -Helmut