Why Virtual Short Concept in an Op-Amp is only applicable to negative feedback

Edaboard member LvW has posted a LTspice example many years ago. The positive feedback solution converged in case of an ideal OP with infinite bandwidth but failed with a real OP. Not sure if I can locate the old thread. The theoretical explanation is probably in the same direction. Input differential voltage is never zero for a real OP.

I don't recall that post, but I can simulate it with variable gain. With ideal Vio = 0 , Iin=0 and Rout=0, Rf=0 the error is purely a function of Vout not saturasted and Vdiff = 1/Ao , gain.

http://tinyurl.com/ytw6qqdd Now anyone can see the correlation. But typically 1e6 gain for BJT types but there is a wide range of choices and CMOS types tend to be much higher.

RIP RAP (Robert A. Pease. Inventor of the best Op Amp before uA 741 was even thought of and the bandgap reference diode and an icon of Silicon Valley in the 70's

WHAT'S ALL THIS STUFF ABOUT INPUT OFFSET, ANYHOW?.

• https://archive.org/details/Bob_Pease_Lab_Notes_Part_1/page/n19/mode/2up?q=Bob+Pease
• https://archive.org/details/Bob_Pease_Lab_Notes_Part_3/mode/2up?q=Bob+Pease
• https://ia803107.us.archive.org/26/items/Bob_Pease_Lab_Notes_Part_5/Bob_Pease_Lab_Notes_Part_5.pdf
• https://ia902808.us.archive.org/22/items/bobpeaselabnotespart8/Bob Pease Lab Notes Part 8.pdf
• https://ia803102.us.archive.org/10/items/Bob_Pease_Lab_Notes_Part_9/Bob_Pease_Lab_Notes_Part_9.pdf
• He was famous for all his monthly journal articles called "What's All This Stuff About ___ Anyhow?"https://archive.org/details/Bob_Pease_Lab_Notes_Part_1/page/n19/mode/2up?q=Bob+Pease
• World Class Analog Design https://www.jotrin.com/userfiles/downloadfile/202003121610523330.pdf

and a thumbnail of my old boss who once looked like RAP long before this 2015 reunion of Unisys staff in Winnipg at the Peripherals factory for all of Unisys.. I was TE Mg with 10 ATE's raised floor office with 6x 1100 mainframes, 2 Liebert ACU's , 2 Liebert pwr conditioners and 50 1" dual port SCSI cables to the test floor. with hundreds of minicomputers, Apple ]['s and custom testers.

for 9 sigfigs (rediculous but for the detail oriented folks)

I'm not completely sure if the original question is adressing the same problem, here's the mentioned old thread, instructive in any case:

Keep in mind the virtual null ONLY exists in linear operation, because when the output is saturated, the LOOP GAIN is ZERO.

Remember Vin+-Vin-= Vo/Ao.

+ve feedback relaxation oscillator work this way, without a virtual null, yet have enough -ve feedback to toggle the output .with RC delays..