0
\$\begingroup\$

Suppose we are given a fully differential (low voltage single supply) op amp with common mode feedback. It is common knowledge to break the loop in order to measure open loop ac performance for loop stability. In the first figure, there is a common setup to measure differential gain from input to output and measure the gain and phase response at the output (vop-von, differentially). The idea is to break the loop (using large inductor short at DC) for dc feedback only, but open loop to ac. The input on the other hand has a large capacitor for open to DC and short to AC. One can run a bode plot of the AC Gain and Phase, in order to look at Phase Margin of the differential gain.

Can someone show a working testbench setup for LTSPICE that will measure the CMFB (Common Mode Feedback) loop stability using AC analysis? A google search will return plenty of examples for stb analysis using a program like Spectre, but next to no examples for a more simple spice simulator like LTSPICE.

More specifically, given an OTA architecture (Folded Cascode Input and Dual Differential Pair common mode feedback) as in Fig 2., can you show a test setup (like Fig 1) with specific source stimulus and response measure location(s) to perform CMFB loop stability using AC analysis with LTSPICE only?

differential_gain_test_bench

Fig. 1 LTSPICE test bench to measure Differential AC gain and Stability.

folded_cascode_opamp_DDA_CMFB

Fig. 2 Fully Differential Folded Cascode Operational Amplifier with DDP CMFB

\$\endgroup\$

1 Answer 1

2
\$\begingroup\$

I learned to use Spectre (and other simulators) when there was no STB type source. So what I used (and still do) is a lowpass filter (RC) and a VCVS. Here I use that to keep the feedback loop working at DC but at AC the loop is open as the RC filter prevents an AC signal from getting through.

To make a Bodeplot of the AC transfer, I add a voltage source with AC = 1 in series with the VCVS.

This shows how I open the loop of an amplifier with feedback and add my DC-pass but AC-block circuit:

schematic

simulate this circuit – Schematic created using CircuitLab

R5 (10 Mohm) and C1 (1 Farad!) form the RC lowpass filter. The VCVS is simply a voltage buffer, it is needed because the output of the RC lowpass filter is a point with a very high impedance so a voltage buffer is needed.

When you plot the voltage at the point marked "loop output" you will get the open loop transfer.

For a common mode loop, you can use the same technique. If you find an example where a STB source is used in the common mode loop, you can simply replace the STB with my circuit.

\$\endgroup\$
5
  • \$\begingroup\$ Thanks, but I don't think it's as simple as substituting for every STB source. For one, a fully differential feedback would require 2 stbs and hence, 2 ac sources. There is some method spectre also uses to switch between FD and CMFB loops. \$\endgroup\$
    – pat
    Oct 11, 2021 at 17:25
  • \$\begingroup\$ There is some method spectre also uses to switch between FD and CMFB loops. Yes nice, I don't use it and I don't need it. Call me oldfashioned but these "handy" STB and that kind of nonsense makes engineers "lazy" and then often they doen't realize what they're actually doing and how to properly analyze feedback loops. \$\endgroup\$ Oct 11, 2021 at 17:39
  • \$\begingroup\$ My op, refers to a fully differential OTA with CMFB. Most of the papers show two stbs breaking the loop from differential to cmfb. e.g. eecis.udel.edu/~vsaxena/courses/ece614/Handouts/…. Could you change your answer to reflect that? Or show specifically, where and how to translate. \$\endgroup\$
    – pat
    Oct 11, 2021 at 17:40
  • \$\begingroup\$ You asked about the Common mode feedback loop, which is single ended. a fully differential feedback would require 2 stbs and hence, 2 ac sources No, I can analyze the open loop behavior of a differential system using only onw STB or the circuit I show above. Realize that you can make an ideal single to differential or differential to single converter using VCVSs. And a VCVS doesn't need to be grounded on one side, nor does an RC filter. \$\endgroup\$ Oct 11, 2021 at 17:41
  • \$\begingroup\$ Or show specifically, where and how to translate. It would be an excellent exercise for you to do that! How do you think I would analyze a diff. feedback loop using the method I show above which is single ended? Hint: the RC filter and the VCVS don't care if I connect one side to ground or to something else. \$\endgroup\$ Oct 11, 2021 at 17:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.