OPA838: Spice model. Bias current dependent on absolute voltage.

OPA838 absolute voltage bias error.TSC

Morning Michael, 

There are several violations in your set up. 

1. OPA838 is not unity gain stable, do not test it at gain of +1 like this

2. Your input sine wave looks to be +/-3V exceeding the supply range.

Get the operating setup in range and try again, 

0882.OPA838 absolute voltage bias error.TSC

Hello Michael.

1. Ok. I have changed the gain to +6 according to datasheet.

2. The input is 0. I'm moving the whole circuit's reference to simulator 0. I have biased the input to mid supply

The model error is still there.

Michael Hallin.

To put a little more light on the issue: I have simulated with both 3.0 and 5.0V supply. The bias current is correct (-1.8µA) when bias voltage is half supply.

I had started to test this yesterday and came back to move it a bit forward today. 

These input DC term tests are more difficult than I had imagined, attached is one possible flow - I get that the input model is kind of working but appears to show a CM input R of 100kohm, which is too low for this part. 

Some ruminations on testing DC parameters for op amp macromodels.docx

and the file I was using, 

Input DC error testing OPA838.TSC

Great work.

Yes the input current varies with bias voltage, but it also varies for constant bias voltage when You put the triangular generator between simulator 0 and opamp supply.

This shows that something in the model, perhaps the 100K resistor is tied to simulator 0.

I cannot open your TSC file, but I have recreated your test setup and verified function with a OPA4354 opamp.

In the attached file I have added VG2. This moves the whole circuit, and should not influence the current measurement at all. On the OPA838 this is not the case at all, but interestingly the OPA4354 model also shows the effect, but much less.OPA838 absolute voltage bias error II.TSC

Keep after it, 

there are at least 3 levels to what you are looking at

1. Physical device with actual transistors I/O impedances and protection diodes

2. Data sheet info, some of that is ATE sourced and some is designer sims

3. customer macromodel - these are trying to emulate the device without transistors in this case - lots of interactions and possibly suspect implementations. 

The DC stuff is important, but somewhat secondary if the part basically biases up and works in AC sims. This is interesting to test models, but more important does the part offer the AC results it is being sold for. 

With 4 articles due end of Nov, will have to turn to those now - again, have fun.

I had meant to also mention there is a background project to update these models to what is call the GWL template - again a macromodel but much more methodical on these kinds of issues - perhaps that is nearing completion for the OPA838 model and it will give you better results. This current OPA838 model is what I would call spaghetti code, almost indecipherable (and not worth the time if a new one is imminent)