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: TINA-TI simulation

Part Number: LMP7721
Other Parts Discussed in Thread: TINA-TI, TEST

Hi Team,

Our customer is simulated LMP7721 in TINA-TI for an electrometer application and would like to confirm if the simulation result is correct. According to our customer,

I am using the Spice simulator to build the design and have observed a wide variation in input currents. I want to know if this is correct.

The device is being used in a transimpedance configuration. The input current is from 0 -10pA. The feedback resistor is 10G Ohms.

The positive input to the device is set to 1.25V by a voltage reference - in the simulator I am just using a Voltage source.

With Vcc = 2.5V and VSS = 0V, the input current is 58fA. But by making slight changes to either the supply Voltage (3.3 or 5V) the changes in input current are huge. Also if I adjust the reference Voltage on the positive input I likewise get huge variations in input current which I don't understand.

Is this behaviour correct or is the real-life part better (more stable/predictable)...?

By increasing the supply Voltage to the LMP7721 from 2.0V to 5V the change in input current is significant.

Does the real part behave this way..?




  • Crazy high gain, might be a good place for a Tee network in the feedback, 

    And, normally the source R is shunt -making those changes the output is pegged to rail, so not working yet -put that 2.5V bias on the shunt source R also

    I went a little further with this, but the DC errors in the model have so much gain to the output it is always pegged to the rail. I tried a servo amp, not too successful yet if I try to keep the resistor summing into the inverting node high, if I make it 2Mohm the servo amp finds an output in range - for this model set. The probed voltage is at the servo amp output  so in range for the supplies, but would need to have DC error worst case to make sure on range. 

    And, everytime I changes something yes the sim input bias current changed - the LMP7721 is a transistor level model so that makes sense as it is picking up input CM changes. but of course those numbers are way over specified input Ib = perhaps a modeling mistake. 

    The servo of course AC couples the path, here is an AC sweep - really surprised it did not go flat until 100Hz. 

    ANd the file, 

    LMP7721 hi Zt with servo amp.TSC

  • Hi Michael - so in summary do you agree that the issue is with the model..?



  • It would seem so, as I drilled down through my files I in fact found a new model testing directory where I was apparently off looking at AC things on the 2019 model update, using that model in a very simple DC test shows an input bias current that is no where near Figure 1 in the datasheet. The DC probe is on the V+ input showing -2V for this +/-2.5V sim. The old model is similar in this sim. 

    Models are a multi faceted task and frankly quite difficult to get everything right  - I was at the time mainly interested in input C modeling to map into transimpedance designs correctly - you probably don't care, but the new model has way too high of input C in the model, easy to fix with external negative C which TINA supports - kind of useful actually, 

    More than you would ever want to know on this topic - this starts out with Aol validation vs the old model then moves on to input C extraction for the old and new models, 

    3187.Testing the old against the new LMP7721 TINA models.docx

  • Ken, 

    Yes, the macro- model does not correctly show typical IB at 25C.  Having said that, please notice that the LMP7722 typical IB of 3fA and max of 20fA applies ONLY at the input common-mode voltage, Vcm, at the mid-supply and moving it away from Vcm=2.5V on 5V single supply results in large increase in IB with Vcm>3V - see below.

  • And Ken, here I took the ESD diodes off the input stage nodes, Ib drops to zero. 

    So for the device, they must have used very low leakage ESD structures but the modeling used standard ESD cells - maybe just changing the Is in the diode model can fix this , this took it to zero,