• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 47 subscribers
  • Views 1268 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMP7721 inverting trans-impedance amplifier and guarding

k1mgy
Expert 1680 points
Other Parts Discussed in Thread: LMP7721

Inheriting an existing design that is very high impedance high gain fempto-amp current to voltage converter (5G Ohm feedback resistor and 1M ohm from sensor to the inverting input.  I am new to the guard trace concept, so needing a bit of tutorial here...  Now, please be gentle...


In the current configuration the non-inverting input is connected to signal ground.  So is the "guard" trace.  And the sensor and amplifier package is "shielded" to this same "ground".  (You can now imagine the noise on ground, plus what I think is the effective loss of the guard trace).


Questions:

1. without lifting the non-inverting input up through a 10K to "ground", is it legal then to simply join signal (DC) ground to the non-inverting input, and also to "guard"?

2. if we are to create an effective guard in the non-inverting transimpedance mode, can a stable-enough bias be had to keep the output at 0 with no signal in?


3. The current design operates the amplifier with the V- connected to DC ground.  I suppose this is reasonable...

Reason for these questions: I've yet to test, but reports are that the noise (and leakage) is apparently high enough to make signal differentiation impossible.

  • 0
    TI__Guru 52296 points
    73's K1MGY,

    If the non-inverting (+) input is tied to ground, then the reference level is common (GND).

    To better understand the guard concept, read the latest LMP7721 datasheet, section 8.1.1. I re-wrote it last year to explain guards a little better. Also see the "Related Documentation" section 11.2.1 - particularity the femtoampre article and the Keithley handbook (section 2).

    The idea of the guard is that it driven to the average signal (reference) level. If the reference was at 2V above ground, then any leakage resistance from the input noise to common would see 2V across it. 2V/1G=2nA. If you surround the input with a similar 2V potential, the voltage difference between the two is then zero and no current flows across the leakage resistance. 0V/1G = 0A. Any leakage or noise from the outside gets shunted by the guard.

    In the trans impedance application you describe, the guard should be at the same potential as the non-inverting input, which in your case is "GND".

    To be clear - there are three items here...

    1. "Ground", or frame or case ground. "Green" ground. Usually noisy.

    2. The "Signal Common", or the return from the sensor. Should be clean.

    3. The "guard", which surrounds the input and "bootstraps" the input path to reduce leakage.

    So for your design, signal common = Guard. You do not really need a separate guard as the common will suffice.

    Common and GND should NOT be the same. Frame/chassis ground can have ground currents that may swamp your measurements. You want a separate signal ground and frame ground with them tied at one point. Typical "star" grounding technique.

    #3 is a good question. Normally, having the reference at V- (0V) is bad because the bias currents are highest near the rails. Normally the lowest bias current is where the inputs are at mid-supply. But the 7721 internally bootstraps the ESD structures, so the bias current is still fairly low at 0V. It's "okay"...but not great. You really want to stay at least 500mV away from the rails.

    Measurements at the femtoamp level are noisy. The feedback resistor is generating more noise than your expected signal. You have to control the bandwidth of the stage to get the best S/N ratio. See Section 2.6.5 in the Keithley handbook.

    Hope this helps...

    Regards,
  • 0 in reply to Paul Grohe
    Expert 1680 points

    Paul,


    When I saw your name I nearly fell out of (or off) the chair... having just today viewed a video in which you and the remarkable Bob Pease showed off that amazing copper-clad test fixture, demonstrating a sensitivity in the sub-femtoAmpere range.  I've been drinking from a firehose lately and am fortunate to find yet another hydrant.  So shall now return to re-read the data sheet and check out the other material as I need to get my grounds in order.

    This stuff is really cool.  I'm excited to be working on a neat application and hope to stabilize the front end with a re-design that brings the amplifier a lot closer to the sensor. 

    More questions to follow and would be ever so grateful for your thoughts.  I can't discuss the specifics of the project here, but if you can take my call, would love to bat it around.

    /mark richards

    K1MGY

  • 0 in reply to Paul Grohe
    Expert 1680 points
    Paul, further work is moving us in the direction of an integrating amplifier configuration. There are two issues we need to tackle in the short term:

    1. selection of capacitor
    2. method of reset that will limit the leakage resistance in the feedback path

    On (2) I hear there's a device that provides a nice FET interface with inconsequential resistance in the off state. I just don't have enough information to find it.

    Any thoughts on these? From anyone?

    We're looking to improve S/N ratio and better stabilize the front end.

    /m