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.

INA118: Vos, Vdiff and Vout: does the input Offset Voltage depend upon Vdiff ? Is 'Vout vs. Vdiff' linear when Vdiff is close to Vos ?

Part Number: INA118
Other Parts Discussed in Thread: INA290, TINA-TI, INA190

Dear E2E Staff,

in using the INA118 to measure the current flowing into a sensing resistors, we noticed that the Vos effect on the current read-out does change with Vdiff at the input of the INA. The overall effect is compatible with a Vos which depends from Vdiff. The Vos vs. Vdiff modulation is evident for Vdiff values close to the Vos values published on the data sheet. As a consequence, we concluded that a minimum Vdiff is needed at the input of the INA in order to guarantee linear operations. Such Vdiff(min) seems to be of the order of the  Vos(typ) published on the datasheet.

 I did not find any comment in the data sheet and/or in specific app notes on such effects. Notably, I have found this comment in the INA290 data sheet:

The linear range of the output stage is limited to how close the output voltage can approach ground under zero-input conditions. The zero current output voltage of the INAx290 is very small, with a maximum of GND + 25 mV.Make sure to apply a sense voltage of (25 mV / Gain) or greater to keep the INAx290 output in the linear region of operation.

I am wondering if something like that does apply to the INA118 and, more in general, to all instrumentation amplifiers. In our application, we have to measure current in the sub-uA range, so such behavior comes into action. It would be very useful  to have  a plot of Vos vs. Vdiff and/or any plot showing the Vdiff vs. Vout  trend for very low Vdiff values, in the range of tens of Vos, where the classic gain equation seems to fail.

Thank you so much for your help and support.

Best,

Alberto

  • Ciao Alberto Relaxed

    Can you post a schematic?

    Kai

  • Dear Kai,

    nice to read you again!

    Please find attached the TINA-TI schematic:

    4152.current_sensing_with_INA118.TSC

    I need to monitor the current flowing into the sensor. The V1 generator supplies from -1V to -4V, this is important for CM considerations.

    I calibrated the current read-out by using a Keithley SMU. My circuit reads ALWAYS more than than the Keithley. I guess this is due to the Vos of the INA. However, such 30nA offset does change with the current: at very low current (<100nA), the offset is about 10nA. For higher current, the offset goes from 30 to 40nA. It looks like the Vos does change with the Vdiff at the INA input.

    I am not sure I am using the best approach to measure, on the high side,  currents in the range 0-1uA and 0-10uA. Vcm spans the range from -1V and -4V.

    I could possibly benefit from a INA with lower Vos, however I couldn't find any really much better than INA118UP. I also considered the INA190, which is suggsted for currents in the uA range. However, INA190 has fixed gain, so it does not allow me to change scale.

    Could you please suggest me a better approach ?

    Kind Regards,

    Alberto

  • Alberto,

    The change in the Vos you see is NOT caused by a change in the input differential voltage, VM1, but rather due to a change in the input common-mode voltage, Vcm.  As you may see in the datasheet table below, the specified typical CMRR is 120dB (1uV/V), where CMRR = (delta_Vcm/delta_Vos), so for each 1V change in Vcm you may expect 1uV change in Vos. 

    This means for Vcm change from -1V to -5V you should expect Vos change of ~4uV and that is what you see below: 17uV-12.8uV=4.2uV

    Vos=-12.8uV for Vcm=-1V

    Vos=-17uV for Vcm=-5V

    Simulating the actual circuit for the Vcm=-1V and Vcm=-5V confirms that it is CMRR behind the change in Vos - see below:

    All in all, you do not need another part - all you need to do is to calibrate the system for different Vcm and account in your post-processing for the  changing input voltage offset vs Vcm as shown above.

  • Dear Marek,

    you are definitely right: Vosdoes change with Vcm. However, even if Vcm is stable, I can see a modulation of Vos. In my setup, I set the Vcm = -1V and I measure different offsets with increasing Vdiff. The effect is relevant when Vdiff is a few time larger than Vos. 

    Thank you so much for your time and support. If you do not have specific info on such an effect, I can close this case.

    E2E, as usual, is a great motivation to use TI hardware in my design.

    Best,

    Alberto

  • Hi Alberto,

    I am not sure how you measure the change in Vos with a change in Vdiff (VM1).  Simulating the variation in the input voltage, VM1, as expected I do not see any change in the input voltage offset, Vos = VM2-VM1 - see below.

  • Dear Marek,

    for very small Vdiff, of the order of few Vos, I see in the lab (not in simulation) a difference between the measured Vout and the Vout calculated with the INA gain equation. I interpreted such an effect as a modulation of Vos caused by Vdiff. Of course, that's just a 'guess' to justify the experimental evidence. For Vdiff larger than a few Vos, measured and predicted Vout values they do match within the errors. I am wondering why, for very small Vdiff, the INA seems to be not linear. Apparently, there must be some non linear effect, vanishing when Vdiff > Vos.

    Thank you again for your time.

    Best Regards,

    Alberto

  • Alberto,

    A difference between the measured Vout and the Vout calculated with the INA gain equation is most likely result of the gain error (limited by tolerance of Rg resistor) and not a change in the Vos.  Please take the reading of VM1 and VM2 as I had done above and find the offset by calculating the difference between the two numbers instead of measuring the output, which includes a gain error. Since any change in the input voltage, Vdiff, is a single-ended (left side of R2 does not change), the only modulation of Vos caused by Vdiff would be due to a change in Vcm by half of the change in Vdiff voltage - remember, Vcm = (VINP+VINN)/2. Thus, a change in the Vdiff of 50uV, would mean a change in the Vcm by 25uV resulting in change of delta Vos=~25uV/120dB = 25pV - well below the noise floor of INA118 so you could not possibly be able to measure it.  All in all, I believe any error you see at Vout is caused by a gain error and not a variation in Vos.

  • Hi Alberto,

    I would suggest the following test circuit:

    alberto_ina118.TSC

    The voltage dividers transform the input voltage Vin=0...10V into a current of 0...100nA through R1. If there's actually such a non-linearity or hysteresis, you should no longer see a straight line.

    I agree with Marek. I don't think that there's a modulation of Vos by Vdiff. I would expect just the opposite: The lower the input signal, the lower any non-linearity.

    But what makes me headache is the 47µ caps. If these are electrolytic caps, the measurement could be ruined by their leakage currents. Leakage currents of electrolytic caps heavily depend on the voltage drop across these caps and can show values of way more than 1µA. Leakage current decreases with time, but after turn-off and turn-on they start on a high level again.

    Ceramic caps with extremely high cap values can also show very strange DC performance. If they got damaged during the soldering (and they get very easily damaged by soldering...) they can show very strange leakage current effects.

    I would recommend to use metallized film caps made of PET or even PP.  

    Another issue that comes into my mind is thermoelectric potentials. And damage due to ESD.

    Kai

  • Dear Kai and Marek,

    I agree with you on blaming the caps for 'strange' behavior. They are all Multi-Layer-Ceramic (MLCC) by TDK and/or MURATA. Tested with our Keithley SMU, they showed leakage lower then 1nA. However, according to your suspects, I have removed them all. The issue is still there. We keep seeing a non-linearity in the low region of current range, let's say from 0 to 30nA.

    For me it is quite easy to test in the lab your scheme. Next week, I will let you know the results.

    Thank you all guys  for the help and support !

    Best,

    Alberto

  • Dear Kai and Marek,

    while preparing the setup suggested in your previous post, please have a look at this thread:

    The issue discussed in the post seems quite similar to the one I am experiencing: lack of linearity when Vdiff is in the range of a few Vos. In this case, the component is from Analog Device and the SPICE model seems to reproduce the distortion.

    As soon as I have experimental data, I will share them with you.

    Thank you again for all your support and understanding.

    Best,

    Alberto

  • Dear Kai,

    I have just simulated your setup, in order to plot the INA gain. Gain is just defined as INA_Vout/ INA_Vdiff.

    As you can see, the simulation clearly show a huge non linearity until che current flowing in Rsense (4.7k) does not reach 50nA. This is more or less what I experience in the lab. This is the TINA schematic:

    alberto_ina118_with_gain.TSC

    Have you any suggestion about how to avoid/moderate such a non-linearity in the very low range of Vdiff? I am wondering I should pre-amplify my signal in order to meet the INA only when it is large enough compared with Vos. Easy to say, hard to do, I guess. I would need an extra low offset opamp, with a Vos as low as 0.1uV or even less.

    Thank you so much for your time and support.

    Best,

    Alberto

  • Hi Alberto,

    I think it's only an input offset voltage or rather input offset current effect:

    alberto_ina118_with_gain_1.TSC

    Kai

  • Dear Kai,

    I agree with you. However, that means I can not use the INA to measure currents lower than 50nA... The non linearity is huge. Have you any suggestion about how to compensate it ?

    Best,

    Alberto

  • Hi Alberto,

    a simple offset voltage adjustment will help, either by hardware or by software. Here by hardware:

    alberto_ina118_with_gain_2.TSC

    Kai

  • Dear Kai and Marek,

    tuning Vref in order to compensate the Vos fixed the problem. I can now read the current with +/-2nA accuracy.

    Thank you so much for all your suggestions and support. E2E is a great opportunity for designers !

    Best Regards,

    Alberto