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.

OPA4388: TINA-TI simulations: schematic's not correct

Part Number: OPA4388
Other Parts Discussed in Thread: OPA388,

Hello Marek,

about your simulation for phase margin, in your schematics you connected the capacitor on the + input of OPA388 to ground while it is in parallel at the sensor.

I suppose it changes something, doesn't it?

Thanks for replies.....


PS: I have downloaded the PSPICE from OPA388 web page and run it in the CADENCE PSpice without get any of the behavior I have in the real circuit. Of course, the simulation I did could be not accurate....  

  • Maurizio,

    Are you referring to 1uF C11 shown below?

    If so, I included it because you showed it in your original schematic - see below.

    However, C11 is driven by the NTC circuit and not the amplifier's feedback and thus it does not materially effect stability - see below.


    But what it does do is to help filter out the thermopile thermal noise.  See below - Vout total noise with and without C11.

    After running AC stability analysis (that includes 1TH inductor and 1TF cap to make it look like open-loop), you must do post-processing;  AOL=Vout/Vbeta, invBeta=1/Vbeta and LoopGain=AOL*Vbeta = (Vout/Vbeta)*Vbeta = Vout

    Of course, in the real circuit you do NOT include 1TH and 1TF and to see phase margin you would have to use a gain-phase analyzer.   Without it, you may only indirectly determine stability of the circuit by looking at transient small-signal overshoot - see below.  By disturbing the system with a small-magnitude square current waveform, you may see the output has no overshoot and thus is very stable.

  • Yes, Marek, it's it!

    But in my schematic it is connected in parallel at the sensor and not between + input and GND.

    In my simulation, that changes the response curves.....

    and I have ever to learn something from your analysis, thanks. 


  • Yes, it would be better to connect the filtering cap, C31 (or C1 below), to ground BUT the output of U11 (U6 below) that drives the TH- in your NTC circuit is a reference DC voltage, which is effectively AC ground so there should hardly be any difference in its AC and filtering performance - see below.

    Btw, I see you have opened another E2E thread dealing with output drift that from the looks of it (since you claim replacing it with simple resistor seems to help) comes from NTC circuit  and not op amp - perhaps you should share the details of the implementation of NTC circuit for us to see if there is something we can identify that could cause the issue.

  • Hello Marek,

    in the simulation with the PSpice the results are different than yours because I looked for Vout/Vin. If I move Vin to - input and I use the same arrangement as yours, I have the same.


    but with Vin at the sensor output I have this 2 behaviors:

     with C1 in parallel to sensor

     with C1 to GND. It appears quite different, cause that I asked the question.....

    BTW, in the real circuit the offset with or without  that filtering capacitor (C1 in your sch) changes: from -11 mV with, to ~0 mV without, RTO. And now I'm thinking that my OPAMP could be damaged ... but damaged in a way it works fine except for OFFSET? Strange, absolutely strange....

    About your last comment, the NTC has its own circuit separated from the thermopile and I don't use the "classic" resistor + NTC but a constant current generator to better use the full scale of the ADC and to have a better resolution: think, I have an absolute accuracy better than 0.1 °C over the full range from -20 to 60 °C.  So, the NTC values have no relation with the thermopile output.

    I'm open to any suggestions, this thing makes me stuck because I cannot release the circuits for pre-production testing. now it appears like a witch-hunting...


  • Maurizio,

    The phase margin must be measured at the frequency where 1/beta (noise gain) intersects AOL curve.  However, you don't seem to show the this post-processing as shown below so no surprise you get different results.

    Also, I had assumed you use NTC driver only for cold-junction compensation but it is the thermopile that you rely on for measuring actual temperature. Thus, I'm not sure how to understand your last comment that " NTC values have no relation with the thermopile output."  Please show the actual implementation of NTC and thermopile circuits and/or constant current generator so we may understand how they are supposed to work - a picture is worth thousand words.

  • Thanks for the explanation, I was only common to measure the phase margin in DC/DC SMPS Innocent

    About NTC: the NTC is connected to another section of the OPA4388 and its output  goes directly to one of the ADC inputs: it's the SW which later calculates the temperature of the sensor body. There are no direct interaction between the NTC and the thermopile .... 

    The thermopile output is connected between V_BIAS (the CM voltage) and the + input of another section of the OPA4388; its output goes directly to another input of the ADC.

    The SW reads 3 channels: V_NTC, V_BIAS and V_OBJ, then calculate the sensor body temperature from the V_NTC and, with the sensor voltage (V_OBJ-V_BIAS) used the lookup table to guess the temperature.

    I know that many circuits, including the one suggested somewhere in the Amplifiers section of devices, use the NTC to change the CM voltage in a way to "compensate" the "cold junction temperature" but this analogue method is good only for small ranges of measuring temperature and it is not very accurate if the SW won't add some corrections. Or, that what I have understood. Then I have been using a different approach since the beginning, thinking at the easiest and cheapest way to perform the calibration: I use just 2 points calibration, one for NTC and OFFSET and the other for OBJ temperature.

    Look at the schematic:

    Thanks for any feedback, any....


  • I believe Raymond already assists you with this part of the application in a separate tread so I'll close this post.