High gain photoconductor amplifier design questions

Discussion in 'Analog & Mixed-Signal Design' started by remyxomatosis, Apr 7, 2018.

  1. remyxomatosis

    Thread Starter New Member

    Apr 7, 2018
    3
    0
    Hi all,

    I am not an EE by training so please bare with me. I designed a photoconductor amplifier with two stages with a gain of 100x each. We want to measure signals that are typically around 1 muV from a modulated infrared light source (using a chopper wheel, maximum 1kHz).
    Here is the circuit I came up with:
    schematic-phip-amp.png
    The bias voltage is provided by an external SMU, the op amps are battery powered. The schematic omits the bypass caps for both, the SMU and amps are bypassed by 0.1 and 10 nF ceramic caps. Currently, it is built using a perfboard and using dip sockets for the ICs.

    The circuit works and we can detect the modulated signals both on a scope and using a lock in amplifier. For this particular application it is "good enough", however in the process of learning about these circuits I got interested and I would like to improve the design so I have a number of questions:

    • There is a significant amount of noise. It seems to be random (I cannot see any characteristic frequencies other then my modulation using the FFT on the scope). Here is a link to a video of the scope readout. https://giphy.com/gifs/fdVrmojnIF4WE7PwGn As you can see, the readout "jitters" and I wonder what causes this. The scope is triggered by the external reference of the chopper, so it seems that the signal changes with time and does not always arrive at the scope at the exact time relative to the reference. What causes this? And is there any way of fixing this?
    • I chose to use the ALS1722 since it is described in the specs as usable for "a broad range of precision applications requiring extremely low input signal power". Is this a good choice? Are there better op amps for this purpose?
    • As I mentioned, the circuit is currently put together on a perfboard, inside a metal project enclosure for shielding. I currently use a bare 22 gauge wire soldered to the backside of the board as ground rail. How much, if any, improvement can I expect by actually creating a PCB with a dedicated ground plane and smd components? Is it worth using guarding techniques on the opamp input if making a custom PCB? Would an additional EMI shield on the board help?
    I would greatly appreciate any input on improving this design. This is for a custom measurement system we are building at university, if it works well it will be used for a while. We have budget for this project so cost (within reason) is no issue, going through one or two iterations of a custom PCB is definitely OK.

    Thank you all!
     
  2. ericgibbs

    Moderator

    Jan 29, 2010
    6,183
    1,171
    hi rem'
    Welcome to AAC.
    Would I be correct in assuming that your local mains supply is 60Hz.?
    E
     
  3. ian field

    AAC Fanatic!

    Oct 27, 2012
    6,252
    1,131
    My memory is getting a bit vacant - but transimpedance amplifier springs to mind for some reason.

    Around the 80s/90s, discrete component IR receivers were common in front panels on TVs etc. There are schematics archives and some chance of finding the IRRC amplifier schematics.
     
  4. PeteHL

    Member

    Dec 17, 2014
    200
    8
    In the schematic diagram anyway the op-amp input terminals are backwards. Connections to the non-inverting (+) and inverting (-) input terminals of each of the two op-amps need to be reversed if what you had in mind is the non-inverting amplifier configuration.
     
  5. sghioto

    Active Member

    Dec 31, 2017
    530
    73
  6. PeteHL

    Member

    Dec 17, 2014
    200
    8
    If the two op-amps were configured as shown in the diagram of the first post of this thread, would the circuit work in some fashion?

    Edit: Well! Live and learn. This is my first time seeing an inverting amplifier configured this way.
     
    Last edited: Apr 8, 2018
  7. Bordodynov

    Well-Known Member

    May 20, 2015
    1,979
    597
    The photoresistance (LDR) is a very slow device.One kilohertz is very fast for its.Look at the bottom picture:
    draft8.png
     
    OBW0549 likes this.
  8. OBW0549

    Distinguished Member

    Mar 2, 2015
    2,585
    2,216
    That's a good point. Furthermore, the lower the light level the slower a LDR responds. A silicon photodiode probably would be a much better sensor and probably wouldn't require such high gain in the amplifier.
     
  9. ian field

    AAC Fanatic!

    Oct 27, 2012
    6,252
    1,131
    AFAICR: a PD has very low photocurrent and needs lots of gain - photo transistors need less gain but are slower than PDs. Photo darlingtons are much slower, but most likely faster than LDRs.
     
  10. MisterBill2

    Well-Known Member

    Jan 23, 2018
    1,704
    305
    If the circuit is actually working then it is only the drawing that is wrong, since it would not work as shown.
     
  11. PeteHL

    Member

    Dec 17, 2014
    200
    8
    In a simulation that I did, following the inverting amplifier configuration of the circuit as shown, voltage gain by a single op-amp of a 100 mV input sine wave equaled 100. The output was a sine wave.
     
  12. remyxomatosis

    Thread Starter New Member

    Apr 7, 2018
    3
    0
    Hi all,

    Thank you for the fast responses. I need to clarify a couple things:

    I made an error in sketching the circuit. As was pointed out, the op amp inputs are mistakenly switched in the schematic. The actual device is wired up correctly as a non-inverting amplifier with a gain of 100x per stage.

    Another thing: I put a photoresistor in the schematic, since it is the closest device I could think of to what we are actually using. We are studying new devices to detect IR light, so it is actually the photoactive component that is the DUT. These devices are based on meta-materials absorbers for the infrared, and a reasonable good model for them are photoresistors (think PbSe photocells for 3-5 micron wavelengths.) If anyone is interested in these new devices, I am happy to elaborate.

    Speed is not a issue, the circuit was tested up to 3kHz which is all we need.

    So again, the circuit works, I am looking for some input regarding the noise issue I am seeing (as demonstrated in the linked .gif) as well as wether I can expect improvements if we built this circuit on a custom PCB.

    Thank you!
     
  13. OBW0549

    Distinguished Member

    Mar 2, 2015
    2,585
    2,216
    I can't determine the amplitude of the random noise you're seeing (mainly at the top of each cycle of the waveform) in the linked .gif, so it's hard to say.

    With an input noise voltage density of 26 nV/√Hz at 1 kHz, the ALD1722 is not the lowest-noise op amp available; over a 3 kHz bandwidth and with a circuit gain of 10,000 I would expect to see roughly 15 mV of noise on the output. An LT1113 dual JFET input op amp has about 4.5 nV/√Hz at 1 kHz, so substituting one of those for the ALD1722 should cut the noise down by roughly a factor of 5.

    As for building the circuit on a PCB, that almost always improves things some; how much, is hard to say.
     
  14. sghioto

    Active Member

    Dec 31, 2017
    530
    73
    I would certainly try and use the photovoltaic circuit I posted in #5. With your circuit you have DC current through the diode which causes noise.
    You can use the same opamp as in the circuit below.
    SG
    EEE  photodiode amp 2 stage.png
     
  15. ebp

    Well-Known Member

    Feb 8, 2018
    2,215
    774
    "please bare with me"
    AAC permits nudity?

    What is muV? milli-microvolt = nanovolt?

    There is a great big giant Johnson/Nyquist noise generator at R4 though it is shunted through C1 by R21 and the detector (both of the latter are also noise generators).

    Here's a bit of useful info on noise in op amp circuits. Many others can be found on the web.
    http://www.analog.com/media/en/training-seminars/tutorials/MT-047.pdf

    Every connection in a circuit makes a thermocouple and potentials can be pretty high depending on the materials involved. Many cancel each other, but that depends on the cancelling junctions being isothermal. When you are dealing with microvolt signals it can be an issue especially at DC or very low frequency. Waving air over the circuit with a piece of cardboard or the like will often show if there might be a problem.

    ===
    Be careful what type of capacitors you use in your signal path. Most ceramic dielectrics and many plastic film types exhibit considerable dielectric absorption that can make quite a mess of analog signals. C0G (still often called NPO) ceramics are generally pretty good but get quite expensive for higher values. For through-hole, polypropylene caps are probably the best compromise of performance and price, but can't be had in SMT because the melting point of PP is too low. Many ceramic caps are also non-linear (large voltage coefficient of capacitance) and some are quite remarkably microphonic, which can make for some "interesting" effects.
     
  16. KeepItSimpleStupid

    Distinguished Member

    Mar 4, 2014
    2,918
    510
    Your best bet is a DSP Lock-in Amplifier. In reality, it can return the Fourier coefficients for any harmonic. The chopper will introduce a trapezoidal waveform. The book I had with the equations in it has been stolen, but the usual way is to just find the fudge factor.

    You generate a clock when you know your signal is present and feed that into the amplifier as well as the chopped signal.

    I actually built a 2/4-terminal I-V converter with 4 ranges from 100 mA, 10 mA, 1 mA and 0.1 mA with +-10 V full scale. It was capable of biasing +=10 V. It had +-50 mA of suppression available if the bias was within +-5V

    The only thing I didn't get quite right was DC nulling which isn't important for the lock-in measurement. Two 5 digit system voltmeters could read the buffered voltage and the DC current.

    4-terminal means that the devices were contacted with Kelvin probes. Some devices has 25 mA of DC output along with a monochomatic chopped light signal input.

    So there was 2T/4T; ZC/ZC (Zero check/Zero correct); Voc (Voltage:eek:pen circuit) and Suppression.

    Missed problem I wasn;t alllowed to fix, was the D/A used for zero input did not output ZERO at zero out.
    In any even the DC offset was about 40 pA as designed and zero when removed by the Lock-in.
     
  17. remyxomatosis

    Thread Starter New Member

    Apr 7, 2018
    3
    0
    I do not need to build a DSP lock-in, even though this might be fun, I can just use an external lock-in amp.
    A couple follow up questions:

    In case of the TIA approach: How would I best deal with AC coupling this? The current signal rides on a substantial dark current, that could easily rail the amp at high gain settings.

    @ebp: how much can I reduce the return path to ground resistance (R4)? I know I need to have it for the circuit to work, but I'm unsure how how to choose the R value.
     
  18. OBW0549

    Distinguished Member

    Mar 2, 2015
    2,585
    2,216
    Rather than trying to AC couple the TIA input, it might be better to null out the DC dark current. The data sheet for the LT1793 op amp shows an example of this approach:

    Untitled.png
     
  19. KeepItSimpleStupid

    Distinguished Member

    Mar 4, 2014
    2,918
    510
    The standard lock-in assumes a sine wave input. It works, but the fudge factor is different. That's what we used prior to the DSP lock-in.

    You have both a DC gain, the TIA gain and an AC gain (in the lock-in). If the output of the TIA swamps the lock-in (Above the CM or NM range range), you can capacitively couple at the output of the TIA. That's what we did.

    For the TIA I used a circuit in the LT1010 application notes where the LT1010 is placed within the feedback loop to isolate capacitance,
    I also needed +-100 mA. (nominally 25 mA)

    Suppression, never actually ended up being used. @OBW0549 That's a cool circuit. It would not work in my case, because the device was a solar cell with a very low intensity chopped light source. It would need to be bi-polar. That's why I went with suppression. A current source, but we never had to use it.

    In the latest generation, I had a differential amplifier with a 400 M ohm resistor to attempt to cancel Ib. There are likely better approaches, but I wanted 0 input with a disconnected lead.

    The TIA requires a CAP across the FB resistor.

    With a discrete dual FET front end, I think you can eliminate the problem nearly completely, but I haven't found any good design notes.

    In your case, you have a filter with R4. That creates problems. I needed Ib to go somewhere, so I picked a 400 M resistor instead of trying to design a dual FET differential amplifier front end.
     
Loading...