Connect with us

Grounding and Circuit Noise

Discussion in 'General Electronics Discussion' started by AdamZ, Jul 3, 2016.

Scroll to continue with content
  1. AdamZ

    AdamZ

    18
    1
    Mar 17, 2016
    Attached is a simplified circuit. The circuit accepts a voltage command at AO1 (analog output 1) and outputs a current at AI1 (analog input 1) via DAQ card PCI-6230 (NRSE configuration). All grounds (1-5) are connected to the same point. The computer is grounded to this point as well, which is connected to a very silent ground to earth.

    I am wondering if my grounding scheme gets me the best possible noise level, regardless of the load. I do not want to modify the rest of the circuit (i.e. add capacitors or swap op amps).

    Other details:
    1. AI1 voltage is in mV range (+/-1.5 mV at 125 kHz). I'd like to get to <+/-0.3 mV at 125 kHz.
    2. When I pull out AO1 from the DAQ panel, the op amps drive to the rails
    3. I tried floating the circuit (disconnecting grounds 1-5) and connecting AI1 to a 10 kOhm bias resistor to ground, but this drove the op amps to the rails.
     

    Attached Files:

  2. CDRIVE

    CDRIVE Hauling 10' pipe on a Trek Shift3

    4,960
    648
    May 8, 2012
    I don't see anything odd about GND in your circuit but I must admit I'm lost with your + & - supply to OpAmp C. Unless OpAmp A is a separate chip from OpAmp C this can't be done because a single chip shares the same supply pins. On that note I've never seen an OpAmp circuit that divides the supply voltage like that. Those voltage divider resistors are quite high and will prevent OpAmp C from supplying anything close to its rated output current. Or is that what you're trying to do?

    Chris
     
  3. AdamZ

    AdamZ

    18
    1
    Mar 17, 2016
    The op amps have different max supply voltages: A is 35 V and C is 18 V. That's why I used the resistor divider.
     
  4. CDRIVE

    CDRIVE Hauling 10' pipe on a Trek Shift3

    4,960
    648
    May 8, 2012
    Got that but I think a voltage regulator would be a better choice.

    When dealing with extraneous noise pickup it's hard to beat differential inputs with high common mode rejection. That's how EKG machines reject much higher ambient noise while still measuring our bodies (minuscule) impulse voltages. Common OpAmps can be configured as common mode amplifiers but instrumentation amplifiers perform best.

    Chris
     
    AdamZ likes this.
  5. CDRIVE

    CDRIVE Hauling 10' pipe on a Trek Shift3

    4,960
    648
    May 8, 2012
    I forgot to mention that Max specs means just that. It's never a good idea to operate any component at its maximum limits. ;)

    Chris
     
  6. Harald Kapp

    Harald Kapp Moderator Moderator

    10,346
    2,250
    Nov 17, 2011
    Definitely. A resistive divider's output voltage canges with current drawn by the opamps which results in a fluctuating power supply. While an opamp can suppress this partly (the parameter is PSRR) it's never good to rely on this. Use a stabilized supply.

    Definitely. When you operat a component at max specs any slight excess of the relative parameter (here. supply voltage) may result in damage to the component. It is good practise to stay within typical operating parameters. In your circuit +-12 V is a good supply which can be used for both amplifiers.

    Another flaw I see with your circuit is the load being conencted from the output of amplifier 1 to the input of amplifier 2. The input of amplifier 2 is high impedance, so is the feednback (5 MΩ). There is no way for the load current to return to ground.

    AI1 is not a controlled current output. The amplifer is conected as a transimpedance amplifier, the output being a voltage proportional to input current. The output current will vary depending on the load resistance.
     
  7. hevans1944

    hevans1944 Hop - AC8NS

    4,535
    2,112
    Jun 21, 2012
    AdamZ likes this.
  8. CDRIVE

    CDRIVE Hauling 10' pipe on a Trek Shift3

    4,960
    648
    May 8, 2012
    Thanks Hop, but it's too heavy for my retired mind. I don't understand why this (same) topic was started.

    Chris
     
  9. hevans1944

    hevans1944 Hop - AC8NS

    4,535
    2,112
    Jun 21, 2012
    Well, clearly @AdamZ is still having problems. His is a difficult task, if you read and understand the paper cited in the previous thread. This is basic R&D done at the molecular level. Adam has lots of spiffy tools, such as electron microscopes and a National Instruments data acquisition system, but the project has "some assembly required" to make the measurements he wants to make. Apparently he did not follow our previous advice to power the op-amps from separate voltage regulators, provide by-pass capacitors soldered close to the op-amp power terminals, to ditch his voltage-divider approach to lowering op-amp power supply voltages, and place all the sensitive circuitry inside a Faraday shield. From what I "know" about this experiment, I would be looking into using a modulated excitation source and a lock-in amplifier to recover a small signal buried in noise.

    Of course, not being there, I am in no position to say what will and will not work, but I do know that taking care of the basics first usually allows discovery of a much clearer path to an eventual solution to his "noise" problem. Adam needs to find a good electronics tech, or maybe even an engineer, with hands-on experience building and troubleshooting high-impedance, low-noise, op-amp circuits. Grounding, ground loops, and noise become huge problems when nanoampere currents and megohm impedances are part of the picture.

    Another mistake I have seen is a scientist trying to faithfully reproduce the work of another working in the same or a closely allied field without fully understanding the experimental protocol and knowing what is important and what is not important to obtaining previously published results. I once tried to help a chemistry professor who insisted that the power supply for his experiment had to be exactly the same identical obsolete and unobtainable model as the one used in the paper whose results he was trying to duplicate. Life is too short to argue with stubbornness like this that bears no relationship to reality. I wasn't able to help with his project, so after I explained the situation to my supervisor, he let me drop it to work on something more productive.

    That might be the appropriate path to take here since Adam has stated, "I do not want to modify the rest of the circuit (i.e. add capacitors or swap op amps)." That kind of unwillingness to let go of a failed design says a lot about Adam's approach to science, and naive approach to electronics in general. It is so very easy to get committed to a failing path and then try to add patches here and there in an attempt to rescue a poor design.

    I must admit, however, that I have succumbed to the same thing in my career after "investing" lots of time and OPM (other people's money) trying to resurrect a design that was DOA. My first civilian job in 1967 as a technician fell into that category. I spent months and thousands of our tax dollars "putting lip-stick on a pig" to no avail. It was a pig when I started, and it remained a pig. The original contractor went bankrupt trying to "save" their pig. I don't think Adam's problem is anywhere near that magnitude, but sometimes you just have to abandon things that don't work, and start over from scratch hoping to find something that does work, if you ever expect to get any good (publishable) results. If that means new op-amps and more capacitors, so be it, but the experimental paradigm should also be investigated to see what trouble (if any) other researchers have encountered. Electronics Point (a mainly hobbyist forum) may not be the best place to do that.

    Hop
     
    Harald Kapp likes this.
  10. CDRIVE

    CDRIVE Hauling 10' pipe on a Trek Shift3

    4,960
    648
    May 8, 2012
    Uh, Jeff Goldblum (Jurassic Park) said something like that! :D

    Chris
     
    hevans1944 likes this.
  11. AdamZ

    AdamZ

    18
    1
    Mar 17, 2016
    You're right, a voltage regulator would be better, but I'm trying to avoid ordering more parts.
     
  12. AdamZ

    AdamZ

    18
    1
    Mar 17, 2016
    Since time is extremely limited, I have to prioritize tasks. The device is working perfectly from one perspective--it is able to produce the intended "product" with very high success rate From an electrical engineering perspective, it is a failure: wrong op amps, missing capacitors, etc. etc. From another perspective, it is a significant improvement over an existing design.

    I found the noise to be more than acceptable by just incorporating 2-3 of your design suggestions. I thought now that I could reduce the noise a little bit more--a mere aesthetic improvement--by fixing the grounding. Simultaneously, I would learn a little bit more about electrical engineering.

    BTW I am using a a Faraday shield.
     
  13. hevans1944

    hevans1944 Hop - AC8NS

    4,535
    2,112
    Jun 21, 2012
    So, you are not trying to put lip-stick on a pig! More like "tuning" the fine points of an acceptable design that has, perhaps, some room for improvement. Said improvement being part of the next phase when someone starts throwing more money at it. Been there, done that, sometimes was awarded a tee-shirt for my efforts, along with everyone else on the team. Life goes on. Next project, please.

    Tweaking an otherwise acceptable circuit to reduce the "noise" is a time-honored procedure. Sometimes it even works, depending mostly on the "noise" source. I doubt the voltage-divider supply for the op-amp that produces a signal proportional to current is the problem. A quick check with an AC-coupled oscilloscope at high sensitivity should verify that. If the same noise is observed on the power supply rails as appears on the op-amp output, by-pass capacitors can be solder-tacked in across the grounded divider resistors to see if any improvement results.

    As far as voltage regulation of the op-amp supply is concerned, a pair of zener diodes could also replace the grounded divider resistors. Again, I doubt this is the source of any remaining "noise" problem. Op-amps, by design, do a pretty good job of rejecting variations in their power supply voltage, as @Harald Kapp mentioned in his post #6 in suggesting you check the PSRR (Power Supply Rejection Ratio) spec in the op-amp datasheet.

    If you really suspect the "noise" is related to the power supply, an easy check is to substitute batteries for mains-operated power supplies. Batteries have very little "noise" in their outputs and are isolated, so "ground loops" cannot occur. That doesn't mean you cannot induce noise in the batteries with poor construction procedures, like using un-shielded wire, not insulating the batteries from the laboratory environment with Teflon pads, and not placing the batteries in their own Faraday shielding box. All of these are extreme measures that are usually not required, but in some instances I have found them to be absolutely necessary. It is just one step toward finding where the real "noise" source is.

    Sometime, even if you find the "noise" source, it is impossible to eliminate it because it is a component of the physics of the device generating the noise. There are sophisticated ways around the problem, but all that I am aware of trade bandwidth (response time) for signal-to-noise improvement. Lock-in amplifiers are a typical example. A lock-in or boxcar integrator integrates (averages) a synchronous signal buried in noise such that the noise integrates to zero but the signal doesn't. The down side is it takes time to build up the signal in the integrator output, so less bandwidth for changing signals is the result.

    Best wishes for success, @AdamZ! Just realize that sometimes "gud enuf" is good enough and it's time to go on to other things. Done is better than perfect.

    Hop
     
    AdamZ likes this.
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

-