• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 49 subscribers
  • Views 2507 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.

How to find source and range of gain independent output voltage offset in 2-Amp instrumentation amplifier using LMH6643, OPA2830, OPA2301

Douglas Stierman
Prodigy 90 points
Other Parts Discussed in Thread: LMH6643, OPA2830, OPA2301, LMH6619, TINA-TI, VCA821

I've designed and built a 2-Amp instrumentation amplifier like the one below:

My primary candidates for the amplifier are the LMH6643, OPA2830, and OPA2301. Currently,  I have selectable gains of 200, 300, and 800 via 1K resistors at R1 and ~200/300/800K at R2 along with a 10K trimmer at Rg. I'm using this IA to measure a voltage drop across a very small resistance (10's Milliohms) with 50-100mA current flowing through that resistance, and the whole thing is only a couple ohms above ground as far as common mode voltage is concerned.

What I am seeing is a wide variance between multiple samples of the same part. We can stick with the LMH6643 for this discussion, but I see the same with the OPA parts as well. Some of the parts have the "zero" level(when diff input is 0V) above ground and some other parts are not showing some of the smaller amplitude signals that my "Golden Amp" can. The golden amp is a sample that has the zero level very close to ground and is able to produce amplification of the smaller differential inputs created by the system.

The LMH6643 is supposed to be good for common mode input voltages down to Vs - 0.5V. But I thought I would try using a negative supply to see what happens. I moved just the negative supply pin (Pin 4) to an adjustable -1.2V to -3.3V supply hoping that it would "pull" the signal out of the ground for the "Low" reading amplifier, but instead it "lowered the water" revealing the zero input  level at a voltage below ground. Here are plots of the three samples with the top being the golden sample the middle being the low-biased sample and the bottom being the high-biased sample. The left side plots show single rail operation (4.5V and Ground) and the right side is the same circuit with a -1.2V rail. Note that all other elements of the circuit are the same, and resistor R2 remains connected to the system's ground node for all three. These offsets are independent of gain, at least for gains of 200,300, and 800 as observed.

What parameters found in the data sheet would allow me to determine the full range of possible offsets so that I can attempt to design in some null-offset circuitry. I was hoping to use this part in a single supply design, but the existence of parts with the hidden negative offset would be an issue as the intended signal would get masked in the single supply implementation. Also this first diff to single ended stage gets fed into second single ended stage where the output offset error will get much larger, possibly even saturating the second stage. I'm hoping to at least be able to find a single supply part that does not have any negative output offset as I can then possibly create a dynamic null offset stage using a differencing stage with a sample and hold capacitor. Any help would be appreciated!

  • 0
    TI__Mastermind 22825 points

    Hello Douglas,

    I think what your experiment / data shows is that the op amp input offset voltage can have either polarity (pos or neg). When you apply the negative supply to the VEE pin, you see that some devices outputs swing below ground, whilst single supply operation had run into the VEE (ground rail) and had not swung below ground because of output voltage clipping. To me, this is not unusual!

    Apart from selecting a part with a smaller offset voltage (which could still be positive or negative voltage), or adding some kind of offset nulling to your circuit, I am not sure what else can be done. Since you have large gain (800x maximum), you need very tight offset voltages. Depending on your speed / bandwidth needs, you may also need a fairly fast part so that your closed loop gain is not compromised.

    Here is another high speed dual device (LMH6619) with improved offset voltage (+/-0.75mV vs. ~+/-3.4mV) relative to the LMH6643:

    Here is one possible way to compensate for LMH6643 offset voltage by injecting some adjustable current into the inverting input of the 1st op amp. I've intentionally added 5mV offset to the 1st stage (V_offset) and corrected for it at the output by the proper choice of "V_correction" and R3:

    Here is the TINA-TI simulation file:

    /cfs-file/__key/communityserver-discussions-components-files/14/6131.LMH6643-Instrumentation-Amplifier-E2E-2_5F00_17_5F00_15.TSC

    For reference, and to investigate effect of component variations, here is your input to output relationship:

    Vo = Vin_diff (1 + 2R2 / RG + R2/ R1)

    Vo_1st_stage = (Vin+) x (-R1/RG) + (Vin-) x (1+ R1/R2 + R1/RG)

    Hope this helps your situation.

    Regards,

    Hooman

  • 0
    Prodigy 90 points
    Hello Hooman,

    Thanks for the reply! I am familiar with some of the parameters that affect the output offset such as Input Offset Voltage and Input Bias Current. However, I was under the impression that those will cause a gain dependent output offset since they affect the differential input voltage of the Op-Amp. These output offsets I am seeing appear to largely be gain independent. Are there parameters that can be used to predict the range of gain-independant output error as well?

    Since I would like to keep this circuit as a single supply design, I have been thinking about ways to bias the input so that the maximum specified negative offset would be brought to ground. This would create a circuit with only positive offsets which I would then feed into a sample and hold stage to subtract out the error before feeding to the second amplification stage, which is going to be a log-converter to expand the dynamic range of the amplifier. Another issue here is that the production nature of this circuit will not allow for manual trimming/calibration, so whatever countermeasures are put in for offset correction need to be automatic in the circuit design(Or possibly able to be done via D/A with the micro-controller).

    Is there an application note that outlines the gain independent portion of Voso? I can find a lot of analysis on the gain dependent components caused by Input Offset and Input Bias Current, but I'm not seeing much on intrinsic output offsets. I'll try my circuit again with gain=1(if these amps are unity gain stable, that is), to determine how much of this offset error is gain independent.

    Thanks so much for your help so far!

    Doug.
  • 0 in reply to Douglas Stierman
    TI__Mastermind 22825 points

    Hi Doug,

    I'm not very familiar with any offset / bias error terms which are independent of the closed loop gain / resistor values referenced to the output. I'd recommend that as you vary the gain and look for the effect at the output, you'd experiment only with VEE tied to a negative bias. The reason is that single supply testing might "fool" you in that output could be railed against VEE and it might give the impression of gain / resistor-value independence. HOPEFULLY SOMEBODY ON THIS FORUM HAS IDEAS ON WHAT OTHER ERROR TERMS COULD BE HURTING YOU IN YOUR CIRCUIT. I'll go ahead and post your question also onto the Precision E2E site as well to see if we can get some help from those folks.

    You could probably also use TINA-TI to investigate various error sources' effects at the output. For example, if you suspect input bias current as a culprit, you'd add additional shunt current source at the input(s) and see the AC Simulation "gain" at the IA output.

    Regards,

    Hooman

  • 0 in reply to Hooman Hashemi
    TI__Mastermind 22825 points

    Hi Doug,

    Pete (from Precision group) has made some comments which may be useful:

    http://e2e.ti.com/support/amplifiers/precision_amplifiers/f/14/t/402591

    Regards,

    Hooman

  • 0 in reply to Hooman Hashemi
    TI__Genius 12631 points

    Hello Doug,

     Any variation in the input common-mode of the circuit will also cause an offset in the circuit. Most of our parts measure offset at mid-common mode. As the common-mode moves towards the rails that offset voltage would increase. For example a part with 60dB CMRR (1mV/V) @ dc and a 1mV offset at mid-common mode can have an offset of 2mV at +1V common mode and 3mV at +3V common mode (the offset change can actually be in either direction). The same thing can happen with varying the supply voltage as well, i.e 5V supplies will have a different offset voltage compared to 10V supplies (if the part supports that). This power supply effect is usually a lot smaller.

    Have you looked into using the VCA821 for your application? The VCA821 works as a differential to single-ended amplifier and has high input impedance as well. You will need split supplies though.

    -Samir