Connect with us

Suggestions for analog switch to select feedback resistor in TIA

Discussion in 'Electronic Design' started by Kevin, Jun 25, 2005.

Scroll to continue with content
  1. Kevin

    Kevin Guest


    I need to sense the output from an InGaAs photodetector with a BW of
    about 1MHz and a dynamic range of almost 60dB optical (~500pA to 250uA
    photo current). In the past for low bandwidth applications I have used
    a log amp approach (discrete or more recently monolithic solutions)
    very effectively.

    The best approach I can think of is to use a linear Trans-impedance
    amplifier with switched feedback components to give 2 or maybe three

    One problem I am having is that the analog switches I have found all
    have very high capacitances in the off state both between in and out
    and to ground. The capacitance to ground is especially bad for the
    noise performance as it appears in parallel with the photo-detector.
    The photodetector has 5-10pF capacitance (5pF with 5V bias, 10pF with
    0V) plus the input capacitance of an opamp such as the OPA657 and
    strays gives a total of around 20pF.

    For analog switches the old 4053 series looks to be better than most of
    the new ones and even it has >8pF capacitance to ground as far as I can
    determine from the data sheet.

    Another alternative would be to use a discrete JFET or MOSFET although
    I haven't been able to find the various capacitances when used as a

    Does anybody have any suggestions for analog switches or other ways to
    solve the problem?

    I expect that I will have to make a compromise on the low end because
    of noise - the goal is to have similar performance to the existing
    system that only requires low bandwidth sensing.


  2. Joerg

    Joerg Guest

    Hello Kevin,

    Not sure what you are trying to do but AD has lots of nice logamps. That
    and a low noise current to voltage converter might do.

    Anyway, if you are looking for a very low capacitance switch there is a
    nice quad array: SD5400. But it ain't cheap, several Dollars. For some
    reason the Vishay site doesn't seem to find it today but that site seems
    to have problems at times.

    Regards, Joerg
  3. Joerg

    Joerg Guest

  4. Pooh Bear

    Pooh Bear Guest

    I use discrete jfets for audio switching. J174... and J111... family. Also
    in SMD - use SST prefix instead. Vishay Siliconix make them and I think
    they're second sourced.

  5. Kevin

    Kevin Guest


    Thanks for the pointer to the SD5400, that looks promising.

    We are using a scanning mirror to sequentially reflect a number of
    light beams over a single photo-detector to measure the optical power
    in the beams. From the scanning rate we want to use the bandwidth
    required is in the region of 1-2MHz, and we are looking for a 60dB
    dynamic range.

    I have used the Analog Devices AD8305 log amp on a similar product
    where there was a photodetector per channel (no scanning) and they
    worked well but their bandwidth at 10nA photocurrent is less than
    100KHz. In general a log amp has trouble maintaining a wide bandwidth
    over a large dynamic range as the compensation has to be selected to
    get stability at high currents which results in over compensation at
    low currents. I haven't yet looked into the more complex log-amp
    comfigurations that can avoid some of this problem.

    I may also be able to use an integrating approach to measure the energy
    in the light beam but even there I have had trouble in that the charge
    injection from the best analog switches tends to be in the 1pC range
    and I am trying to measure something a couple of orders of magnitude
    less (10nA for 1us).


  6. Joerg

    Joerg Guest

    Hello Kevin,
    That sounds very feasible. Think about a digital solution as well. To
    obtain a true 60dB dynamic range look at the effective number of bits
    (ENOB) versus frequency and subtract one or so to compensate for
    possible marketing optimism ;-)
    I was in a similar pickle in the 90's and had to cascade two logamps
    because I needed at least 70dB at a good bandwidth. I believe it was the
    AD640. Getting the whole thing stable is no small feat and you need to
    master RF layout pretty well for that. But performance was stellar. The
    only tear in the beer is that these chips now cost about $30 a pop.
    The SD5400 contains four switches on one die, meaning their capacitances
    match very well. The trick is to use one switch as a "dummy" just to
    compensate for charge injection.

    Another switch method that I have used several times is a quad-diode
    set, typically very fast Schottkys. I usually transformer-couple the
    drive signal but that isn't a big deal. If you use a matched quad or one
    of those "quad diode pills" the charge injection almost becomes extinct.

    Regards, Joerg
  7. James Meyer

    James Meyer Guest

    A "T" type feedback topology might help. Switching the lower value
    element in the vertical leg of the "T" should allow for much higher parasitic
    capacitance than trying to switch a classical single feedback resistor.

  8. Fred Bartoli

    Fred Bartoli Guest

    I once used a good trick for ultra low charge injection. You can notice that
    for some switches the charge injection sign reverses for some value of VDD -
    Then you have some charge calibration phase in your process and servo the
    switch power supply accordingnly.
    A bit heavy on the stuff but ultra low charge injection guaranteed, no aging
    and no temperature dependance.
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