On Sun, Dec 22, 2024 at 10:50 PM, Tom wrote:

Screen shot of Delta-Sigma Converter uploaded

... here:  https://groups.io/g/LTspice/album?id=299674

 

Specifically, here:

Photos > Delta-Sigma Converter > 2024-12-22_224815.jpg

 

Andy

 

 

On Tue, Dec 24, 2024 at 07:53 PM, Tom wrote:

It takes 3 sources to make a DI/DO amp with CM ref. I tested both E and G sources.

In general, G (current) sources are preferred over E (voltage) sources.  Current sources tend to result in fewer problems.

 

Oddly if the 4 feedback resistors are all equal you can swap the output connections and the circuit still works. That is something you can't do with a manufacture's device model. 

Have you thought about why?  What would make it work apparently OK with positive feedback rather than negative feedback?

 

One of the things not clearly mentioned in this thread, is the difference between a diff-in-diff-out operational amplifier, and a diff-in-diff-out controlled-gain amplifier.  The SPICE model in your earlier message has a 1E+8 voltage gain from input to each output, so it is an op-amp with that much gain, but of course it can be tweaked to any desired gain.  The one in the photo you uploaded appears to be a controlled low-gain amplifier, I think - although it does have local feedback through the two capacitors, so maybe it is supposed to be a high-gain op-amp.

 

Andy

 

Yes there are multiple types of "Fully-Differential Amplifiers". There is no set minimum requirements. For the case of feeding a differential ADC, a "true" Fully-Differential OpAmp is generally used.

 

The Delta-Sigma Converter uses the input amp as a differential-summing-integrator. The amp used by the original Author Holger was a XSPICE gain code model. I was having trouble getting my version created from the original netlist to work and changing the amp seemed to help. The "other" changes I made were probably the real reason I finally got the simulation to work.

 

Why swapping the outputs works for Fully-Differential amps made with E and G sources is unknown to me. The XSPICE code model also uses G sources and also works when outputs are swapped. The caveat being R1=R2=R3=R4.

On Wed, Dec 25, 2024 at 08:53 PM, Tom wrote:

Yes there are multiple types of "Fully-Differential Amplifiers". There is no set minimum requirements. For the case of feeding a differential ADC, a "true" Fully-Differential OpAmp is generally used.

I  would think that a non-Op-Amp would be needed in that case, perhaps a diff amp with a preset gain of 1 or 2.

 

What is meant by the word "true" in "'true' Fully-Differential OpAmp"?  Is there another kind that is not "true"?

 

Why swapping the outputs works for Fully-Differential amps made with E and G sources is unknown to me. The XSPICE code model also uses G sources and also works when outputs are swapped. The caveat being R1=R2=R3=R4.

Perhaps it is teetering on the brink of instability, and it only remains stable when the resistors are exactly matched?

 

Andy

 

The ideal fully differential amplifier is another description of a theoretical device called the nullor. A nullor has two ports: one port, has zero conductance, zero voltage across its terminals, and zero current into its terminals, named a nullator; the second port, has zero impedance into its terminals, and arbitrary voltage and current at its terminals, called a norator. The nullator and the norator must exist in pairs which are called nullors. A nullor is an ideal opamp which has floating differential output instead of single ended.

The interesting property of an ideal opamp with infinite gain is the polarity of the input doesn’t change the circuit function. This is an operational rather than a numerical property which isn’t shown in ordinary numerical simulation. This property is shared with nullors which have no port polarity preference.

The nullator-norator pair were named in a 1964 paper by Carlin on singular network elements as part of unofficial competition among circuit theorists to find the most fundamental network elements. A european theorist Tellegen, had a fit and criticized carlin’s paper. 
Textbook examples: Bruton, rc active circuits; mitra, analysis and synthesis of linear active circuits 

 

The interesting property of an ideal opamp with infinite gain is the polarity of the input doesn’t change the circuit function.

It seems that the polarity should matter.  I get it that the differential input voltage tends towards zero, to the degree that the voltage gain from input to output approaches infinite.  But the polarity of an incremental input voltage (call it dv), should determine the polarity of the differential output voltage, shouldn't it?  Surely it can't really be arbitrary, can it?  Maybe I'm not understanding what you're saying.

 

Andy

 

Andy, I’m away from my library now so I can’t consult dusty sources, but the idea of polarity indifference for infinite gain amplifiers is illustrated in the technique using “stamps” or templates to build or modify admittance matrices for circuit elements. The simplest case adds a passive one port—typical r,L,and c elements by constructing a stamp of four entries in a square array to add to the existing matrix. The stamp overlays the matrix to add the two positive diagonal terms and the two off diagonal negative terms at the appropriate locations corresponding to the element nodes. For active two port devices, the stamp values are no longer symmetric and the four matrix locations correspond to the two controlling nodes and the two controlled or output nodes. In the case of infinite gain, the stamp doesn’t add infinite entries but rather prompts the combining or deleting of appropriate rows and/or columns.  This is where the polarity indifference shows up.  Notice this is just a mechanical mnemonic device, the actual theoretical circuit analysis is the actual basis.

As an aside, I wrote a C program that analyzes linear networks symbolically. It accepts r,L,c,g,m,ideal opamps, and controlled sources to produce two port parameters— z,y,h,g,t,and s symbolically.

As an example the ideal opamp(the nullor model) still drives the summing junction to virtual ground regardless of the polarity of its inputs. Using a controlled source in place of the nullor produces the expected sign changes in the results as a function of the input polarity.

Again, this is still empirical rather than rigorous.

Jeff

 

On Thu, Dec 26, 2024 at 08:06 AM, Andy I wrote:

I  would think that a non-Op-Amp would be needed in that case, perhaps a diff amp with a preset gain of 1 or 2.

 

The term "OpAmp" is misused by me. TI and AD use Amplifier. They make multiple types designed to feed differential ADCs and almost all rely on external resistors to sett gain.  Gain is dependent on usage and ADC chosen.

 

 

What is meant by the word "true" in "'true' Fully-Differential OpAmp"?  Is there another kind that is not "true"?

 

 

Another term misused by me but it's to point out the requirements a simple E/G sourced amp may not have compared to a "real life" one.