The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Analog Verification >> Analog Performance Verification >> Modular Bench for characterizing OTA
https://designers-guide.org/forum/YaBB.pl?num=1701422800

Message started by avlsi on Dec 1^st, 2023, 1:26am

Title: Modular Bench for characterizing OTA
Post by avlsi on Dec 1^st, 2023, 1:26am

Hi,

I came across this article from Ken.

https://designers-guide.org/analysis/diff.pdf

I have also seen this article on building reusable testbenches from

https://circuit-artists.com/testbench-templates-how-to-reuse-and-boost-simulation-efficiency/

I am curious to understand how to map differential amp characterisation to single ended ouptut OTA. I want the following parameters

AC
1. open loop gain
2. Unity Gain Bandwidth
3. Phase Margin
4. Gain Magin
5. Power supply rejection
6. Ground rejection
7. common mode gain

DC
1. Input common Mode range
2. Output swing
3. offset (mismatch based)

Noise
1. Noise

TRAN
1. settling time
2. slew rate
3. power consumption

Is there any existing reference to this kind of single bench in cadence?

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 4^th, 2023, 12:42am

I am using a balun at the input side however without DC negative feedback the output voltage is not at the right place.
Hence created a verilog A resistor that acts as a short during DC or TRAN however acts as open during AC or Noise analysis. I tried inductor with value of 1 GH however went ahead with this. I see that with this setup common mode gain, differential mode gain will come based on the balun input setup.

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 4^th, 2023, 4:04pm

Never do that. Use the stb analysis instead.

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 4^th, 2023, 5:27pm

Hi Ken,

My OTA is a single ended output with differential input. May I know how can the image you sent be transformed to test this?
And the resistors in feedback and input, what is the value, is it like 1G Ohms?

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 4^th, 2023, 5:40pm

This idea of using giant inductors, giant capacitors, analysis dependent switches, etc. is ancient designer wisdom. But it is wrong. And we have known it is wrong for a very long time. Even back in the 70's, Grey and Meyer described an accurate method for extracting the feedback parameters of a circuit. Their focus was not simulation, but hand analysis.

Unfortunately, this particular idea refuses to die, probably because the right approach was difficult for users to execute in a typical SPICE simulator. But that has not been true for a long time. With the addition of the stb analysis in Spectre, there is no excuse for using these old hacks that are simply not accurate.

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 4^th, 2023, 5:46pm

For single ended amplifiers the placement of the current probe is even easier. You need to find a wire that if cut would block all feedback. Splice the iprobe into that wire. So, you can often place the iprobe in series with the negative input of the amplifier, or in series with the output pin of the amplifer, or in series with the feedback resistor, etc. Presumably you can replace your giant inductor or analysis switch with the iprobe.

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 4^th, 2023, 8:24pm

Thanks Ken. How to run the common mode gain and differential mode gain with the same bench?
I am still hung over this Balun as it gives option to do common mode, differential mode, power supply gain, ground gain in the same bench.
May be I am missing a way to use this stb especially for common mode and differential in the same bench.

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 4^th, 2023, 11:55pm

If you want the loop gain of the common mode path, place the iprobe in the common mode path.

If you want to do both using the same testbench, just put both iprobes in to the circuit. Then specify the one in the dm path for the dm stb analysis, and the one in the cm path for the cm stb analysis.

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 5^th, 2023, 12:15am

Just to be clear, the two baluns in my figure above are solely for the purpose of measuring the loop gain with the stb analysis. For other analyses, such as closed loop bandwidth and such, you would also have baluns at the input and output of your circuit, assuming your circuit were differential at both the input and output.

Also, a single ended amplifier would not need the baluns for measuring loop gain. For your single-ended amplifier, which is differential at the input, you would use only one balun, and it would be placed on the input.

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 5^th, 2023, 12:43am

Thanks Ken, I tried to sketch the simulation ideas that were discussed in this thread.

Option-A - The one I or many are exposed to use with EDA tools to get Differential Mode Gain, Unity Gain frequency, Phase Margin using STB analysis
Option-B - BALUN and at input with d,c representing differential and common mode input respectively and p,n representing the 2 input terminals of OTA. the feedback switch is open for AC, short for DC and TRAN.
Option-C - My understanding is to use this approach to get both common mode and differential plots in the same bench. I am not sure whether the BALUN is neededin this.
Option-D - Removing BALUN from option-C gives us option-D.

Do share your thoughts on which approach is better suited for the differential to single ended.

For example, I would also want to measure transconductance at output of the OTA as it can be multi stage or some other variation (not just simple 5T OTA) and output open loop resistance.

Does any of the options warrant an improvement to measure the afore mentioned parameters ?

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 5^th, 2023, 10:56am

Conceptually all of these testbenches seem fundamentally wrong.

The things you are trying to measure are the properties of a circuit. You need to include that circuit in the testbench. If you don't include the circuit in the testbench then the results you get are not very meaningful. It appears as if you are trying to characterize the OTA itself, but an OTA rarely works alone. At a minimum you must accurately model the input and output loading and the entire feedback circuit. You leave your inputs disconnected. Are you driving them with voltage sources, current sources? Neither makes any sense.

Why don't you show us the circuit you want to characterize, and then we can help you build a testbench for it.

Comments on your given options:

Option A: conceptually correct if your circuit is a simple follower, but if that is the case I am not sure why you have a resistor in the feedback loop.
Option B: if that switch is used to measure loop gain then this option should not be used.
Option C & D: Not sure what kind of circuit you are trying to model here. I also see no need for the iprobe in series with the amplifier's positive input. It is not in the feedback loop, and so is not suitable for use in a stb analysis.

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 5^th, 2023, 5:28pm

Ken,

In our test bench, we use a voltage source to drive the input, which helps us characterize the OTA. The output is connected to both a metal cap and gate cap, and for simplicity, this setup is depicted as a capacitor load. The Device Under Test (DUT) here is a unity gain buffer.

Regarding option A, note that there's no resistor in the feedback loop. Instead, what we have is an 'iprobe' element, which is utilized for Stability Transfer Function (STB) analysis. This element is depicted as the text 'iprobe' within a box in the diagram.

As for options C and D, the 'iprobe' text refers to the same element used for STB analysis. In both these options, 'iprobe' is part of the feedback mechanism for STB analysis. However, the text might not be very clear in the PDF version.

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 6^th, 2023, 2:09am

My mistake on the resistor.

On each of your circuits you have two inputs shown. You cannot connect a voltage source to both. By not specifying the sources the result is confusing and ambiguous.

You cannot just draw a box around the OTA and say I am testing that. You are always testing the entire circuit, including the amplifier, the feedback circuit, the loading and the source impedance.

So you are not characterizing an OTA. You are characterizing an OTA in unity gain configuration with a specific load capacitance and zero input impedance.

Since your circuit is a single ended voltage follower, you don't need any baluns as your circuit is not differential.

If you want a universal testbench for a voltage follower, you would put an iprobe in the feedback loop, a voltage source at the input and a current source at the output. With the voltage source at the input you can measure closed loop gain, bandwidth and input impedance. With the iprobe in the feedback loop you can measure loop gain and gain and phase margin. With the current source at the output you can measure output admittance. And with the power supply sources you can measure supply rejection. Open loop bandwidth is a useless measurement and so should be skipped.

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 6^th, 2023, 5:31am

Thanks, Ken,

I've created a draft of the test bench (TB) in line with the method you proposed. Here's a breakdown of the test cases:

Test Case 1: Applying 1 V to _avdd_ac for VDD rejection measurement.
Test Case 2: Setting 1 V at _avss_ac to evaluate VSS rejection.
Test Case 3: Using 1 V for _vref_ac to measure closed-loop gain, AC analysis bandwidth, and output noise.
Test Case 4: Conducting STB analysis to determine loop gain, phase, and OTA UGBW.
Test Case 5: Sweeping _vref_dc to measure the input and output common mode ranges.
Test Case 6: Implementing a pulse on _vref_dc for settling time measurement.
Test Case 7: Exploring whether setting _iout_ac and plotting Vout versus Frequency suffices for effective output impedance calculation.

To streamline the process, some of these test cases can be combined. For instance, combining Test Cases 1, 4, 5, and 6 into a single test could be more efficient.

Title: Re: Modular Bench for characterizing OTA
Post by Ken Kundert on Dec 8^th, 2023, 10:14am

Presumably you don't really intend to put a dc source in series with an ac source. One source can generate both a dc value and an ac value. And it makes no sense to put two current sources in series.

You can simply further by replacing multiple AC analyses by one XF analysis. Just apply the XF analysis to the output and it will convert transfer function from each source to that output.

Title: Re: Modular Bench for characterizing OTA
Post by avlsi on Dec 8^th, 2023, 9:30pm

The image posted here is a representation of the analysis and in a real bench, its a single source of each type and both DC and AC are part of the same. XF is interesting option.

Can common mode gain be measured with this configuration, or is a different setup required?