THS4140: Gain accuracy

Hi,

In our circuit, there is differential signal feeding the amplifier; this signal range between -1V and +1V, so after the amplifying stage with gain=10 we get output signal between  -10V and +10V.

My question here was regarding the expected amplified result in simulation, for +1V at amplifier input, because the result we get in the simulation is different than expected ==> should be +10V, 5V on each terminal and we get 9.97V, and the non-symmetric values for positive and negative signals.

I understand from your comments that you recommend to feed the amplifier input with DC for each positive and negative terminal, as shown in the below

Is this what you meant?

If yes, note the result has remained unchanged (still 9.97V and the unsymmetrical values).

Also, regarding your comment about feeding the differential source to the amplifier, it results in the value mentioned before, so the circuit seems to amplify the input signal, but not accurately ?

Thanks,

Arie

, so this raises another question - 

So you suggest that in the simulation circuit, I feed the amplifier input with diffe

Hello Arie,

There are a few factors that play into the error on the output. The input bias currents should only affect the common mode output of the device Vocm, and there is usually some type of offset on the Vocm pin that will contribute to the 33mV shift. You will also see that there is a slight difference between the two bias currents in the model which will contribute to Vdiff. Your concern of the gain, however, or the change of Vdiff is mostly caused by the differential offset voltage which should be somewhere between 1 to 7mV (from the datasheet) that will be gained by a noise gain of 11 V/V. Taking these factors into account, it is indeed possible to obtain this shift in gain in a real world application, especially with resistor tolerances coming into play.

Best,

Hasan Babiker

Arie.TSC

Hi,

Thanks for all the replies.

A few notes regarding Hasan's comments - due to same concern with offset impact, I checked in simulation with input of 0.5V instead of 1V at amplifier input, and the value received at the output was 4.98V,

so 20mV off instead of 30mV.

I believe that if offset would have been the main cause, I would expect 4.97V (assuming linear behavior)? Do you agree?

Another thing is the output load - please note that in our simulation we used a high value for the load; we can see that for a load < 1Kohm, there is close dependency between output value and load.

I assume (and I will also check) that if R5 + R6 in your circuit would be > 1K, the output value would be closer to the 9.97V (and respectively pos and neg terminals values) ?!

Additional note - well understood and agreed about real life resistor tolerances. In our actual board, we used 0.1% for both Rf, Rg and we measure (with high accuracy DVM) 4.018 Kohm and 401.8 ohm respectively, which is one of the reasons I had to pursue this simulation process, to clarify why the gain seen at the output is off.

Still, in the specific simulation, the resistor values do not affect the gain deviation, so if I understand correctly, I can trust the simulation results showing the deviation, as reflecting the amplifier parameters?

Thanks,

Arie 

Hi Hasan,

if it actually was the result of differential input offset voltage, I would expect an output offset voltage shift, like shown in this example. But the output signal amplitude would stay unchanged:

Arie1.TSC

This is in contradiction to the result of Arie's simulation, which does not show an output offset voltage shift but a change of gain.

Kai

Hi Arie,

I am from an era when we had no Spice models or couldn't afford the Spice simulation software. Later, we found out that the Spice models had so many restrictions that they really weren't helpful at that time.

From this point of view I would never expect that a Spice model is telling the whole truth. Running a Spice simulation is one mosaic piece of many others during the development. But the Spice model is not the Holy Grail...

Consider the Spice model as what it is: A nice tool. No more and no less. To be honest, I think, it's much more important to develop the special knowledge and skills to carry out serious measurements during the development phase instead of trusting a simulation too much.

It's highly welcomed that TI has begun to improve all their Spice models. But this is a long and stony way...

Kai