The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Accurate characterization of opamps
https://designers-guide.org/forum/YaBB.pl?num=1109151470

Message started by Vivek Sharma on Feb 23^rd, 2005, 1:37am

Title: Accurate characterization of opamps
Post by Vivek Sharma on Feb 23^rd, 2005, 1:37am

Hi,

I came across some new methods for characterizing differential circuits such as opamps. This is called the Middlebrook method and is described in detail in the book "The Designer's Guide to SPICE & SPECTRE" by Ken Kundert (Sec. 3.4, pg 77-84).

I have a little problem using one version of this method for characterizing my CMOS opamp.

In the setup, the noninverting input of the opamp is connected to AC ground (DC common-mode voltage), while the inverting output is connected to the output through a series connection of a voltage source (DC=0V, AC=1V), and a voltage-controlled voltage source. The vcvs is controlled by vout (output of the opamp), and has a gain of (1-beta), where beta is the feedback factor of the loop.

If you do the math, you can see that this would result in a closed-loop gain of 1/beta, while minimizing any errors due to common-mode gain and loading. Now for the problem:

When I use large values of beta (~0.9), everything is fine and I can characterize my opamp. However, at larger closed-loop gains and small values of beta, the trouble is that the DC operating point of the circuit is not set at the desired value since the VCVS treats DC and AC signals the same way. As a result, the output vout shoots off to the rails because the loop tries to enforce the virtual ground condition by making the 2 input voltages equal. This places the opamp in a low-gain region.

Is there a way to overcome this problem. I could try to use a set of VCVSs and a huge cap to filter out the DC and then provide this to the VCVS in the branch between the inverting input and the opamp output, but this seems like a bad idea.

Surely, someone can suggest a better way to use the Middlebrook method.

Thanks
Vivek