Connect with us

Transimpedance amplifier (TIA) oscillating at above 1 GHz. Need Help

Discussion in 'Electronic Design' started by [email protected], May 14, 2013.

Scroll to continue with content
  1. Guest

    Hey Folks,

    I need to design a broadband TIA for photo current amplification.
    I am using a LMH6629.
    I simulated the circuit in Tina-TI an it is working pretty well for a feedback resistor of 330Ohm and a feedback cap of 4pf and a signal of 10MHz.
    However the simulation is showing oscillation if I increase the cap to something above 20pf. In my opinion it should be stable for higher caps and only start oscillating if the cap is to small.

    However, I developed the board and now it is oscillating at 1.2 GHz with a perfect sine, even without a signal applied!
    This oscillation is nearly independent of the feedback cap. I tried form 1pf up to 10pf.
    Only if the cap is missing totally (still around 2pf from resistor) the oscillation is getting worth.
    So, I think it has nothing to do with the feedback cap but some other issue.

    I really have no ideas left..


    Julian Arnold
  2. Guest

    Am Dienstag, 14. Mai 2013 10:25:29 UTC+2 schrieb :

    Actually it is so ugly because it was my first approach to add a custom part to my layout editor ;)

    How do you calculate the noise gain of 2?
    Without the feedback cap and the diode it should be 1, shouldn`t it?

    I am using a pin diode with 1.2pf of capacitance at 3.3V reverse bias.
    I thought the more input capacitance the more unstable the circuit is so I tried to keep it small..

  3. Am Dienstag, 14. Mai 2013 10:25:29 UTC+2 schrieb Julian Arnold:
    Thanks a lot for the responses!

    Unfortunately I need to build up the circuit for a university project whereI need to transmit an OOK modulated signal at around 100 MHz...

    The resistor across the diode is not soldered to the board.
    I think I should remove it from the schematic.
    It was placed there to remove biasing voltage if the diode would not be connected to VCC, which is at 5V.

    So if the stability of the circuit depends on the noise gain, which needs to be above 10, than wouldn´t it be impossible to get it stable without a signal applied to the diode? Because without the resistor across the diode the noise gain should always be 1 without applying a signal.

  4. Am Dienstag, 14. Mai 2013 10:25:29 UTC+2 schrieb Julian Arnold:
    Just made a few measurements with different caps and no optical signal applied:
    R_f = 1,2k
    C_d = 3pf => V_o: Sine wave with 440mVp-p at 418MHz

    R_f = 1,2k
    C_d = 3,9pf => 448mVp-p at 477MHz

    R_f = 1,2k
    C_d = 1pf => 190mVp-p at 1,2GHz
  5. So the PD is across pad 1 and pad 2? And R5 1.2k is not in the
    (I was wondering what R5 was doing.)

    I must admit I'm also a bit confused by Phils advise that you need
    more C on the input. But I've never used a opamp in a TIA that wasn't
    unity gain stable.

    Your circuit looks very similar to one in the LMH6629 data sheet...
    but there they do have a 10 pF PD on the input.

    Finally I assume that you've got some nice bypass caps on the power

    Please let us know when (and how) you get this working.
    I might have learned something today :^)

    George H.
  6. Yes, the Resistor across the diode is not uses, sorry for that.
    The schematic actually is:

    What I am confused about is, that the datasheet tells, that at least a gainof 10 is required for a stable operation. What gain does this refer to? Ifit is the noise gain, than a stable operation could never be possible without a signal applied because than the noise gain is 0.
    Furthermore, if I use no feedback cap, than the TIA is nearly oscillating from the upper to the lower rail. With the feedback cap it is only oscillating with an amplitude of around 70mV. These 70mV even stay the same if I reduce the feedback resistor form 1,2kOhm to 330Ohm.
    When I press my finger onto the feedback network, the amplitude is only increasing slightly.

    Maybe Phil can take away some of our confusion tomorrow;)


    Am Dienstag, 14. Mai 2013 10:25:29 UTC+2 schrieb Julian Arnold:
  7. tm

    tm Guest

    Don't you have that backwards? The capacitance decreases with increasing
    reverse bias.
  8. Geesh, don't be too hard on him Tim. I was wondering myself about the
    circuit operation.

    To the OP, So I also don't see why it's enough to keep peak noise gain
    (~Cin/Cfb) at ten. But that is the place the circuit wants to
    So try a 10pF input C.

    The other way to make it happen, is to reverese bias the PD (with a 9V
    battery perhaps) into a resistor (maybe 50 or 100 ohms) and then use
    the 'screaming' opamp (non-inverting) at a gain of 10-20 to 'gain up'
    the resistor voltage.

    I've done that with little PD's and mini-circuits amps.

    (Tim's certainly correct about reverse biasing the PD to whatever it
    can take.. if you want speed.)

    George H.
  9. Hmm, Hi Tim, I was reading his post differently.
    (Your reading is more accurate.)
    But reading between the lines, I thought he was asking about the noise
    gain as a function of frequency. Which is only one at low

    George H.
  10. Tim Williams

    Tim Williams Guest


  11. Jamie

    Jamie Guest

    You're just lucky! :)

  12. Am Dienstag, 14. Mai 2013 10:25:29 UTC+2 schrieb Julian Arnold:
    Sorry for the questions about basic stuff.
    I am still studying and have not heard so much about OpAmps so far,
    but should not the gain always be a function of frequency?

    Furthermore Tim said:
    "The open-loop gain is determined by the amplifier characteristics and the
    feedback network you put onto it"
    but is not the open-loop gain only determined by the amplifier and can be approximated by a single pole response?
  13. Sure I was just suggesting this as a means to test the stability...
    not a permanent solution.
    I agree! But I guess there is some price to pay for using an opamp
    that is not stable at unity gain.
    Grin, sure more noise... The TIA with a 1k feedback R will have the
    johnson noise of the 1k R plus the opamp voltage noise (at gain 10).
    If he was to use 100 ohm to ground and a gain of ten after (for the
    same output signal level) that will have the equivalent noise of a 10k
    ohm resistor... but lots of times the noise doesn’t matter that much.
    Especially if he’s got a big hairy laser with lotsa photons. Then
    (ignoring any laser noise) the shot noise can swamp the johnson
    noise. For the TIA case I know the ‘break even’ point is when there
    is 50mV of “photocurrent” (I = 50mV / R_feedback). So at a guess
    maybe 500mV of signal for the R to ground with x10 gain.

    George H.
  14. Hi Tim, just re-reading your post.
    Noise gain is pretty much entrenched in the PD literature.
    I don't really like the term either. I know it caused me hours of
    head banging when I first started to study PD's and TIA's. But let's
    just stick with it for consistencies sake.

    I’m not sure why you are concerned that we equate stability with
    (But you probably just have a deeper understanding than moi.)
    So here’s a question. If I take the same opamp out of the TIA and
    just put resistors around it to set the gain. Then it will oscillate
    unless the gain is greater than 10. Now if I was to try and roll off
    the gain at high frequency, by putting a small cap across the feedback
    R. Would that cause it to oscillate?
    (I’ve only used minimum gain of 10 opamps a few times... and rolling
    off the gain was *not* what I wanted.)

    George H.
  15. Am Dienstag, 14. Mai 2013 10:25:29 UTC+2 schrieb Julian Arnold:
    Good question, George!
  16. Frank Miles

    Frank Miles Guest

    Yes - if that the capacitive reactance is low enough (Zf/Zi < Amin)*.
    By inserting a zero** - a resistor in series with this feedback capacitor
    to keep Zf/Zi > Amin you should be stable, though of course you won't be
    killing the HF gain as much as you might have wanted. The other classic
    approach is to add capacitance from the inverting node to ground to
    increase the noise gain even with the added feedback capacitance. Of course
    there may be other sources of instability, e.g. capacitive loading.

    * -evaluated in the vicinity of f=GBW/|Zf/Zi| to GBW.
    **- a frequency domain zero, of course.
  17. Thanks Frank, I'll see if I can dig up a linear chip with a minimum
    gain requirement and play in LT spice. So I should be thinking about
    (at least) another pole in the open loop transfer function. (That
    looks to really complicate the math.) Hey maybe Julian can run 'his'
    screaming opamp in the TI spice simulator and post some pics. (You'll
    have to calculate 'right' C to put accross the feedback R.)

    George H.
  18. For anyone following along at home, here's an LTspice file (with
    LT1127 opamp)
    If you change C1 from 10 pF to 1 pF it will stop with the
    oscillations. I don't know what to make of the 10k ohm * 10 pF = 1e-7
    second time constant.

    George H.

    Version 4
    SHEET 1 896 680
    WIRE 192 80 176 80
    WIRE 304 80 272 80
    WIRE 176 144 176 80
    WIRE -256 208 -256 176
    WIRE -128 208 -128 176
    WIRE 304 208 304 80
    WIRE 304 208 112 208
    WIRE 400 208 304 208
    WIRE 304 224 304 208
    WIRE 400 256 400 208
    WIRE 176 288 176 272
    WIRE 112 304 112 208
    WIRE 144 304 112 304
    WIRE 304 320 304 304
    WIRE 304 320 208 320
    WIRE 400 320 304 320
    WIRE 480 320 400 320
    WIRE 576 320 560 320
    WIRE -256 336 -256 288
    WIRE -128 336 -128 288
    WIRE 80 336 -32 336
    WIRE 144 336 80 336
    WIRE -32 352 -32 336
    WIRE 176 368 176 352
    WIRE 576 400 576 320
    WIRE -32 448 -32 432
    FLAG -128 336 0
    FLAG 176 272 +V
    FLAG -128 176 +V
    FLAG 176 368 -V
    FLAG -256 336 0
    FLAG -256 176 -V
    FLAG -32 448 0
    FLAG 176 144 0
    FLAG 80 336 IN+
    FLAG 576 400 0
    SYMBOL voltage -128 192 R0
    SYMATTR InstName V1
    SYMATTR Value 15
    SYMBOL voltage -256 192 R0
    SYMATTR InstName V2
    SYMATTR Value -15
    SYMBOL opamps\\LT1127 176 256 R0
    SYMATTR InstName U1
    SYMBOL res 288 208 R0
    SYMATTR InstName R3
    SYMATTR Value 10k
    SYMBOL res 288 64 R90
    WINDOW 0 0 56 VBottom 2
    WINDOW 3 32 56 VTop 2
    SYMATTR InstName R6
    SYMATTR Value 2k
    SYMBOL res 576 304 R90
    WINDOW 0 0 56 VBottom 2
    WINDOW 3 32 56 VTop 2
    SYMATTR InstName R7
    SYMATTR Value 10K
    SYMBOL voltage -32 336 R0
    WINDOW 123 24 124 Left 2
    WINDOW 39 0 0 Left 2
    SYMATTR InstName V4
    SYMATTR Value SINE(0 10m 1meg)
    SYMATTR Value2 AC 10m
    SYMBOL cap 384 256 R0
    SYMATTR InstName C1
    SYMATTR Value 10pf
    TEXT 408 424 Left 2 !.tran 30u
  19. Am Donnerstag, 16. Mai 2013 02:47:56 UTC+2 schrieb George Herold:
    Okay, so here are my simulation results.
    At first the configuration that, I am think, I schoul have on my board:

    C_in = 7pf R_f = 330Ohm C_f = 4pf

    Next with increased C_f (It is still hard to believe, that it is really starting to oscillate with increased feedback cap, but it definitively is...).

    C_in = 7pf R_f = 330Ohm C_f = 15pf

    Than, as suggested, with increased C_in to increase the closed-loop? gain(still a bit confused about the gain thing...).

    C_in = 7pf R_f = 330Ohm C_f = 15pf

    I will try to increase the C_in on my board tomorrow and report if it is helping.
    I can not believe, that my current C_in should be bigger than 10pf and so there is a difference between simulation, where oscillation starts at around15pf, and reality. So maybe there is also something else wrong...

  20. Cool, Everyone should do their own measurements.

    So I don't recommend the 10pF as a permanent solution... just to see
    if it makes the oscillations stop.

    I'd try the opamp as X10 buffer,
    or get a bigger photodiode,
    or find a different opamp,
    There's a fast Fet one from AD.

    George H.
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day