Connect with us

Photodiode Transimpedance Amplifier

Discussion in 'Electronic Design' started by Nomen Nescio, Aug 27, 2003.

Scroll to continue with content
  1. Nomen Nescio

    Nomen Nescio Guest

    Hi. I have to design (or more accurately, copy) a photodiode amplifier
    for one of our IR detectors, but since my knowledge of electronics is quite
    poor I thought it'd be a good idea to post my questions here. I just need
    suggestions about possible improvements to this circuit.

    The situation is this: we have a very sensitive InSb infrared detector
    (photovoltaic) in our lab. It is basically a 300 pF photodiode and it works
    without bias voltage. The amplifier circuit is a transimpedance amplifier
    built around an OP27, with a 10M feedback resistor. There is however one
    difference in this circuit (actually, two) with respect to the usual plain
    transimpedance design: the first difference is that the output current of
    the photodiode doesn't go directly to the opamp input (or summing junction,
    if you prefer to call it that), but to the gate of a prior JFET
    amplification stage: just a JFET with the drain connected to the power
    source (+9V, through an 8k resistor) and the source to the input of the
    opamp (with a derivation to -9V through a 75k resistor) . The feedback loop
    encloses also this JFET (that is, it is tied to its gate. not to its
    source), so one can imagine that we have just a big FET-input opamp
    (FET+OP27) working as a transimpedance amplifier.

    The second difference is (in my opinion) a smart one: since InSb
    detectors work usually cooled with N2, the photodiode itself is mounted
    into a dewar. Whoever designed the system decided to use that fact to his
    advantage by placing some of the electronics into that dewar (in a cold
    plate touching the inner wall, not submerged in N2) to reduce the noise. So
    both the FET and the feedback resistor are working cooled, near liquid-N2
    temperature. That is probably the point of using an external FET as input
    stage instead of just a FET-input opamp in the feedback loop: cooling the
    FET helps to reduce the noise a lot (and in fact this detector is way more
    sensitive than regular, commercial InSb detectors). This design probably
    gets you the best of both worlds regarding the opamp and the JFET:
    FET-input opamps (if one were to use one instead of the OP27) have a higher
    voltage noise that BJTs like the OP27, but lower current noise. Working
    with a cooled FET allows you to get the benefit of the low current noise of
    the FET, plus the improved performance due to the cooling. At the same time
    one still uses a BJT opamp after the FET, where current is not a problem
    anymore, and benefits from its low voltage noise (that is at least how I
    understand the design, I hope I am not very wrong. Please correct me as
    soon as you find any error in my interpretation of these advantages and

    The FET is a dual package: 2N6483. The other FET inside the can is
    used to apply a bias -by applying a manually adjustable voltage to its
    gate- to the noninverting input of the OP27, so that instead of tying it to
    ground one has an adjustable DC bias. Due to the initial FET stage the
    voltage at the inverting input (summing junction) in conditions of zero
    photodiode signal is not zero, but something around 1 Volt, and having a
    10M feedback resistor it is easy to saturate the opamp if one doesn't
    adjust the bias properly to match both inputs, so this adjustment is really

    The frequency requirements of this circuit are relaxed: we just need a
    few tens of kilohertzs (no more than 50, tops).

    That is more or less the description of the circuit... Right now it is
    fried -long, sad story-, so I have to build another one and of course the
    easiest approach is to shamelessly copy the original. I am planning on
    using the same FET (seems to be ok, and anyway replacing it is really
    involved, it is into the optical head), but was wondering whether I could
    still gain a little more of S/N ratio by using a different opamp, or
    introducing any other improvements somebody might suggest... Perhaps the
    main source of noise in the current setup is not the opamp but some other
    (Johnson noise?), in which case replacing the opamp wouldn't buy me
    anything. Unfortunately I don't know enough about all this stuff to be able
    to figure out by myself how to perform a noise calculation on this system.

    Anyway... That is the story. I'd love to hear suggestions or advice from
    anybody to help me improve the circuit (or just to tell me that it isn't
    worth trying!). Thanks in advance!

  2. I used to work with similar devices in a medical device.
    Beyond finding low noise opamps I found that the selection of the feedback
    resistor was important. I don't remember which was best Carbon Comp, Metal
    Film, etc.
    Even bigger was microphonics. Microphonics are small mechanical vibrations
    caused by fans, motors, solenoids, things hitting their stops, etc. These
    vibrations can't be felt by hand but if you have a cap in parallel with the
    feedback resistor it can cause a lot of noise.
    If you consider a cap with a charge on it and think of it as two plates, if
    the plates change their distance it changes the capacitance. The charge on
    the cap doesn't change so the voltage has to change. This is how the
    mechanical noise turns into electrical noise.
    You can see if this is a problem by flicking your circuit with your finger.
    Our circuit originally had a mica cap (really bad). Changing to a ceramic
    cap was better then to a ceramic monolithic was better again.
    The best was a glass capacitor. It has NO microphonics but they are
    I also found that mounting the components with extra lead height helped (we
    had thru-hole) by having the leads absorb the vibration.
    Just changing the capacitor got us 18 dB increase in SNR.

    Hope it helps and good luck,
    Colin Jackson
  3. Uwe Bonnes

    Uwe Bonnes Guest

    : Hi. I have to design (or more accurately, copy) a photodiode amplifier
    : source), so one can imagine that we have just a big FET-input opamp
    : (FET+OP27) working as a transimpedance amplifier.

    : The second difference is (in my opinion) a smart one: since InSb
    Look at Linear Technology's datasheets for the Lt6200/LT6020. They show the
    circuit you mention.

  4. Daniel Haude

    Daniel Haude Guest

    On Wed, 27 Aug 2003 19:40:01 +0200 (CEST),
    From your very detailed description of the circuit it looks as if it's
    already pretty well-designed, so you're obviously not going to get any
    advice such as, "just swap the OP27 for this-and-that and it'll become
    much better". Before you start experimenting, however, you need to take
    measurements (noise, freq response) so you can see if you actually gain

    One thing you might try while you're at it: I've successfully run whole
    opamps at LN2 temperature (I think it was an OPA111 but I'm not sure).
    That was just a quick test to see if it worked at all, so I can't give you
    any details. Might not necessarily improve your performance, but would
    make the design simpler. And get a good feedback resistor like stuff that
    ELTEC will happily sell you for an arm and a leg.

    BTW, I couldn't find any data on your dual JFET. Who manufactures it? Are
    there online data sheets available? With large source capacitances like in
    your case, input voltage noise becomes an issue (as has been discussed
    here many times).

    There's an application note by Burr-Brown (now TI), dealing with this and
    pretty much all other aspects of transimpedance amplifiers:

  5. Daniel Haude wrote...
    I'll send you an email data sheet. The 2N6483 was offered
    by NSC, Harris and Calogic. InterFET makes similar parts.

    With a Ciss of only 20pF, it wasn't the lowest e_n noise
    JFET on the block, the curves showing 4nV/rt-Hz at 1kHz.
    It was a monolithic IC, intended for low offset voltage.

    We featured the 2N6483 and LM394, both dual transistors,
    in our JFET - BJT matchup, see AoE page 441. (The result
    was a split decision.)

    - Win
  6. Daniel Haude

    Daniel Haude Guest

    On 28 Aug 2003 07:38:27 -0700,
    ....which reminds me of the $250 worth of IF3602s (650pF, .35nV/rz-Hz)
    sitting in front of me, calling my name, wanting to be played
    with. Ho-hum, need to do other things first...

  7. Daniel Haude wrote...
    Why don't you send some to us, we'll put 'em to good use.

    - Win
  8. Daniel Haude

    Daniel Haude Guest

    On 29 Aug 2003 06:10:44 -0700,
    That's a very sensible suggestion, Win, considering that
    you're very correct in assuming that the $250 figure
    (actually it's more like $300, all things included) was
    forced on me by Interfet's minimum-order policy. But all the
    bang I got for those bucks was five (in digits: 5) pieces
    which I'm going to defend like a mother bear would defend her
    young. That's 19 million dollars per nanosaarland of silicon
    real estate!

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