Connect with us

More about Photodiode Transimpedance Amplifier

Discussion in 'Electronic Design' started by Deepak K Gupta, Sep 11, 2003.

Scroll to continue with content
  1. Hi,

    I'm trying to understand (and if possibe improve) an existing Photodiode
    Transimpedance Amplifier I have which is design to measure the small
    fluctuation over a dc level of light. Seems that all cares one find in
    litrature about the resistances, capacitances, pcb design are mainly taken
    care.

    A low noise opamp (~3nV/rt-hz) is used with 300MOhms feedback resistance.
    However the circuit is different from the usual opamp Photo-Trans-Ampli
    one finds in application-notes and most litratures. It also a low
    noise (~1.2nV/rt-hz) JFET in the input. Actually output of biased
    photodiode goes to the gate of JFET. Source of Jfet is grounded and Drain
    is connected to non-inverting input to opamp and also to +15V supply
    throgh a 1.2K resistance. Feedback loop also enclose this Jfet, means the
    300M feedback resiatance is connected between gate of jfet and output of
    Opamp. The inverting end of opamp is effectvely connected to a +5V supply
    throogh a network of Rs and Cs to provide a virtual biasing to JFET.

    I wish to undertand how the use of fet in with and opamp help to improve
    the noise ? It there any litrature where I can see to learn more about
    this circuit and it's noise analysis? Can one really operate fet like
    this--forcing a virtual source-drain biasing ? Love to hear any suggestion
    to improve this circuit, or that "it can't be done!".

    Thanks !

    -deepak
     
  2. Guest

    Yup... your circuit is remarkably similar to the one I have been working
    with, I posted a few questions about it a couple of weeks ago. If your news
    server doesn't have the thread any more you will be able to find it doing a
    google groups search, the title was "Photodiode transimpedance amplifier",
    dated august 27th. Some of the people who answered gave very useful
    suggestions and directed me to further reading. A good place to start -for
    the noise calculation- would of course be the Art of Electronics. There is
    also this Burr-Brown app note on transimpedance amplifiers, suggested by
    Daniel Haude: http://www-s.ti.com/sc/psheets/sboa060/sboa060.pdf.

    My knowledge of electronics is rather limited, so take this with a graint
    of salt, by my understanding of the circuit, and especially of the reason
    for using a FET+opamp combination instead of just an opamp is this:
    photodiodes are basically current sources, and you want to receive and
    amplify that signal -a current-, typically by using a transimpedance
    amplifier to obtain a proportional voltage. Of course you want to do this
    without modifying your signal current, so that your voltage faithfully
    represents your signal current, and is not "contaminated" by the
    measurement itself. Since the amount of current generated by photodiodes
    can be rather low in weak signal conditions, the way to measure that
    current without modifying it is by using a very high impedance to receive
    it, like that of a FET or FET-input opamp (instead of a BJT operational
    amplifier). A BJT opamp has much higher bias current and will more likely
    "contaminate" the measurement in low current conditions.

    A second important consideration is the current and voltage noise of your
    opamp. Ideally both have to be low, but in transimpedance circuits the
    value of the feedback resistor can be quite large, megaohms is not
    unusual... that means that current noise of the opamp, converted to voltage
    noise by this large feedback resistor, tends to be the dominant factor...
    so you want an opamp with a very low current noise... now, that's a
    FET-input opamp, which also matches nicely the requirement of the above
    paragraph. Not surprisingly because both things are related. Unfortunately,
    FET-opamps tend to have a little higher voltage noise than BJT-opamps.

    So it is clear that if you want to use just a single opamp a FET-input
    opamp is normally the best choice. There are quite a few transimpedance
    amplifiers around based on FET opamps. Then, why using an independent FET
    plus an opamp? Well... I am not quite certain, but here the designer is
    combining the low current noise and high impedance of a FET and the low
    voltage noise of a BJT-opamp, which I suppose is the one you have in your
    circuit... Think of it as the guy making his own FET-input opamp. Still, I
    guess that one should be able to get similar performance with a single
    low-noise FET-opamp, so there might also be other considerations, perhaps
    the FET-opamp with the specs he wanted didn't exist... I really can't say
    more. For example, in my case the reason is (my circuit is VERY similar to
    yours) that the JFET is mounted together with the photodiode into a dewar
    and works at liquid N2 temperature, thus reducing the noise even further.
    This gives a final noise figure which is quite a bit better than using a
    single (not cooled) FET opamp.

    I hope this helps you, but as I said, this is my interpretation and I am
    far from being an expert on this, don't take this stuff as gospel.
    Hopefully some other people with more expertise will jump in and further
    clarify the matter.

    Good luck!
     
  3. wrote...
    Clearly you guys are enjoying yourselves with transimpedance amps.
    I'll just add a few comments that may be useful. A big issue with
    low-noise current-sensing circuits is current noise, so of course
    you use a FET opamp and take care of the feedback resistor's Johnson
    noise. But at higher frequencies the opamp's voltage-noise is also
    important, due to a current-noise contribution I call e_n - C noise.
    This noise current is given by i_n = w e_n Cin, and it's proportional
    to frequency, opamp voltage noise, and the total input capacitance.
    This e_n - C noise becomes the predominate noise above a frequency
    given by, f = (4kT/Rf)^1/2 / 2pi e_n Cin, if I have worked the math
    correctly. This is the dominate noise source for many applications.

    For example, consider a photodiode amplifier used with a large-area
    PD detector, like UDT's PIN10D. This has 1000pF of capacitance, so
    we'll reverse bias it to -10V reducing the capacitance to only 300pF.
    With 10V reverse bias it has a dark current of 2nA (typ), and we'll
    use Rf = 10M in our amplifier for 1 uA sensitivity at 10V full scale.
    We'll use a TL082A FET opamp, which has e_n = 18nV at 1kHz. In this
    case e_n - C noise will dominate at frequencies above 1.18kHz. Wow!

    Sure, you can use a more suitable opamp, like Analog Devices AD743,
    with e_n = 2.9nV, but this only raises the breakpoint to 7.3kHz.

    One other observation: most noise sources we worry about are white,
    i.e. their noise density is flat with frequency, and the rms noise
    increases mildly by the square-root of the bandwidth. But not so
    with e_n - C noise. For frequencies above the frequency breakpoint
    I mentioned the noise density rises with frequency. This means the
    rms noise increases by the bandwidth raised to the 3/2 power.

    That's a real killer!

    Thanks,
    - Win
     
  4. Thanks Winfield! You hit right at our problem.

    We also think, strongly, that we are seeing, what you called -- e_n-C
    noise. In our case this starts taking over around 100kHz. Let me give you
    few numbers about our circuit.

    We are using 300M feedback resistance (Rf). The input FET we are using
    have e-noise, e_n, 1.2nV/rt-hz and input capacitance 7.2-8pf. We are
    using the Advance Photonix's small photodiode that is having a capacitance
    of ~2pf at 15V biasing voltage. So according to your formula...
    frequency,f turn out to be ~100KHz. This what we are noticing. A noise
    increase with frequency above 100kHz.

    We want to improve this circuit for new larger photodiodes with
    capacitance of around 10pf. Moreover we also wish to increase the
    frequency response of the circuit (ideally wanted to exceed 1Mhz!). So I'm
    looking for the design/idea to eliminate (or atleast minimize) this e_n-C
    noise. Any Suggestion ?

    Probably you noticed (from my first posting) that our circuit not only
    have a fet-opamp (and it is AD743) but also an external fet along with
    this fet-opamp. I believe that the analysis for e_n-C noise is applicable
    for both the cases, i.e., opamp as well as external fet with opamp.
    If we use an opamp only circuit (say using AD743; e_n=2.3nV/rt-Hz, i_n=
    6.9fA/rt-hz, C_input=18-20pf), we will get i_n-C noise 2.4 times more than
    out fet circuit. Moreover, this will start showing up from ~50kHz compared
    to present 100khz. At least in this case external fet with opamp circuit
    shows better response.

    We do cool at LN3 temperature to reduce noise but at higher frequencies, we
    think, this e_n-C noise start dominating and accedes acceptable limits. So
    I'm still looking for details analysis/literature which can explain me that
    how the presence of an external fet improves the noise behaviors of a
    circuit compared to an only opamp circuit. Any idea/circuit for improving
    our existing circuit to reach our attempt for bigger photodiodes and
    higher frequency response is welcome.

    keep commenting please. Thanks!

    -deepak
     
  5. Anonymous

    Anonymous Guest

    We do cool at LN3 temperature to reduce noise but at higher frequencies, we
    I may be mistaken, but thinking about the advantages of having a FET input
    stage... wouldn't that eliminate (well, ok, let's say reduce) the problem
    of current noise in the operational amplifier? In a "normal" transimpedance
    configuration (no FET), when the operational amplifier senses variations of
    the current at its summing junction (originated by its own current noise)
    it tries to compensate them by changing its output accordingly. This change
    of current output, that goes to the summing junction through the feedback
    resistor, it is reflected in your output signal as a change of the voltage
    output of the opamp. This is how opamp current noise affects your output,
    right?

    Now, if you have a FET into the feedback loop, that FET is already
    amplifying the current that comes from your photodiode, as well as the
    current that comes through the fedback resistor. That means that the
    operational amplifier doesn't need to change its output so much to
    "compensate" the current noise that it is perceiving at the summing
    junction, because the FET is also amplifying this feedback current: if with
    your FET you are amplifying -for example- a photodiode signal of 1 microamp
    to 100 microamps, then you are reducing the effect of amplifier current
    noise by a factor of 100. You are basically dividing your total gain (given
    by the feedback resistor) in two stages, so that the system is now less
    sensitive to the current noise of the opamp (second stage) because that
    stage has less gain than in a single-opamp configuration with the same
    feedback resistor.

    Advantages of this amplifier scheme (that is, supposing that my reasoning
    isn't fundamentally flawed)... well, I suppose this would allow you to use
    a BJT opamp with very low voltage noise, lower than that of a FET opamp,
    and still not be seriously affected by its current noise. Reducing that
    voltage noise would buy you some more margin with that nasty e_n-C noise
    Win mentions in his post. Of course for this to be effective you also need
    a very low noise FET (but at the same time you also want a low
    capacitance... it is a matter of compromise, I guess). This scheme seems to
    make unnecessary the use of a FET opamp, since you already have a FET
    input.
     
  6. bombkbro

    bombkbro

    1
    0
    Feb 22, 2011
    Can you help me about caculating (f_3dB)bandwidth when there is a discrete JFET ?
    Thanks!
    Duy
     
    Last edited: Feb 22, 2011
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

-