Hello,

I am interested to know if I have made an obvious error in my attempt to simulate feedback circuits.

Your comments would be very much appreciated, Thank you.

Summary:   I am implementing the g-parameter method of simulating series-shunt (voltage amplifier) feedback circuits as described in “The Designer’s Guide to Spice & Spectre” by Ken Kundert (starting on page 97.)  As a test vehicle, I am using a veriloga amplifier model of a single ended output op amp in a non-inverting gain configuration (The model has finite input capacitance and output resistance).  I did a quick check on this model by placing an ac source between the divider point on the feedback network and the inverting input and see reasonable results.  The individual g-parameter simulations look reasonable and the g-parameter that represents the closed-loop gain (g21) behaves reasonably.  However, the other feedback parameters that are obtained using the Cadence calculator as products and sums of products of the g-parameters give results that make no sense.  

Details:
Simulator:  Spectre
Test Vehicle:  A veriloga model of a single ended op amp with the following elements:
Tanh block representing input pair distortion
Slew rate block (1V/us)
DC gain block (80 dB)
First pole (100 Hz)
Second pole (10 MHz)
Output resistance (100 ohms)
Input capacitance (100fF)
These blocks are isolated by buffers so they do not interact.
This amplifier is set up in a simple non-inverting gain configuration with a gain of Av=101. The resistors are ideal models of 100k and 1k

Procedure:
I have worked through the presentation in “Analysis and Design of Analog Integrated Circuits” 4th ed by Gray and Meyer.  Chapter 8 Feedback section 8.5.3 covers “Practical Configurations and the Effect of Loading” for Series-Shunt feedback.  I have derived equations for closed-loop gain without neglecting reverse transmission through the amplifier or feed-forward through the feedback network to confirm I understood the equations in Designer’s Guide to Spice & Spectre for closed loop gain (3.119) and (3.120), basic amplifier gain (3.121) and feedback factor (3.122).  

I next used the h-parameter to g-parameter relationships to confirm I understood the expressions for feedback elements in terms of g-parameters:  basic amplifier gain (3.132), feedback factor (3.133), loop gain (3.134) and closed-loop gain (3.135).

Summary of parameters:
Del =g11*g22-g12*g21
h11 = g22/Del
h12 = -g12/Del
h21 = -g21/Del
h22 = g11/Del

The feedback elements in terms of g-parameters are given below.   (I have my output source set up in inverse polarity as compared with the text so I have a couple of sign reversals below):

A = a/(1+af) = g21             (Closed-loop gain)

T = -(g12*g21)/(g11*g22)             (loop-gain)

a = g21*(Del)/(g11*g22)                (basic amplifier gain)

f = -g12/Del         (feedback factor)

Next I set up the test bench as shown if figure 3.28 in The Designer’s Guide (I have my output current source polarity reversed compared with figure 3.28 but have taken this into account in my expressions.)  This was done with two test circuits, test bench A driving the input of the network with an ac voltage source of unity magnitude and test bench B driving the output of the network with an ac current source of unity magnitude.   (Please see attached schematic).

The simulation results for each of the g-parameters on their own look reasonable but plots of the combined expressions generated by the Cadence calculator do not make sense.  I would appreciate your comments if you can spot something I overlooked.  Thank you.

Regards,

DJS