Except that, if the likes of Bob Pease thought simulation to be a mere foolishness, should anyone dispute it?

Bob used to take a computer up onto the building of the parking garage next to building F and throw them off once a year. He used to say:

"I don't hate computers, computers hate me. I LOATHE them...."

Bob was smart enough to know when a computer was lying and that was too often.

Our IC designers used lots of sims with various degrees of usefullness. But, they all knew what aspects could NEVER be simulated like stability and compensation. The managers lived under the delusion that you could do a paper design of an IC and go to fab no worries. That's pure BS. NEVER happen, the only time parts worked first pass were when the designers had so much experience they knew what worked form EXPERIENCE. Those designers are rare.

In our group we had a power regulator product family (LP396X) that famously came out and did not work, it oscillated terribly. It cost the company so much money and bad PR, it went all the way up to the CEO. After that, the standing orders basically were:

"If it oscillates, EVERYBODY GETS FIRED."

The designers knew we had to really be careful and thoroughly evaluate the part before release. They hired a new hot shot designer who brought out a part that was "flexcap" stable, meaning it was stable with input/output caps that had ANY value of ESR (equivalent series resistance).

So the part was about to be released (always a hurry) and I was taking data and saw something funny. By manipulating the source impedance to my lab power supply, I got the thing to break into oscillations. I found the range of input capacitor ESR values over which it was unstable.

So I go back to the designers office:

ME: "You gave me data showing the part was stable with any input cap. Did you test this?"

DESIGNER: "Of course. We ran simluations with ESR values all the way down to zero Ohms."

IDIOT! The advertising copy I had done had to be cancelled (It was already at the magazines). If I hadn't caught that we would have released another oscillator. Sims are creating a whole crop of similar idiots.

BTW: here's the part that Mr Simulator came an inch away from releasing had I not caught the problem:

http://www.national.com/ds/LP/LP38501-ADJ.pdf

Here is what I added to the D/S to keep from losing all the lawsuits that would have followed:

All linear regulators can be affected by the source impedance
of the voltage which is connected to the input. If the source
impedance is too high, the reactive component of the source
may affect the control loop’s phase margin. To ensure proper
loop operation, the ESR of the capacitor used for CIN
must not exceed 0.5 Ohms.

 

Switching regulators put what is basically negative resistance load on the source: As the input voltage to the switcher goes up, the current goes down. That caused me grief once (thankfully, it was a home project). IIRC, I was trying to run a switcher off the output of another switcher. You probably have a more sophisticated term for this effect. I am not an expert on switchers.
I can't disagree with most of your comments about simulators. I am a 70 year old retired circuit design engineer (mostly analog). Naturally, I spent many years designing, analyzing, and breadboarding my circuits before they went to production.
For me, a simulator is a time saver. It reduces the number of breadboards I have to build. It also allows me to play "what if" with much more ease than a breadboard will allow. It does not eliminate the need to breadboard circuits, though. I too am appalled by the number of hobbyists who go straight from a sim, or someone else's schematic, direct to PCB.
The last 10 years of my career, I worked as a logic/circuit design engineer, designing some of the support logic in DRAM chips.
Agilent EEsof EDA was an essential part of the design process, and was very good at predicting circuit behavior. Of course, it depended on having accurate models, which was the job of the modelling group.

 

Informed skepticism seems appropriate as one approaches any modelling scenario.

RF engineers, meteorologists, environmental & climate scientists, stock brokers, economists, architects, civil engineers, geologists, market researchers, spin doctors .... all need to question what their computer modelling tells them.

We used to say "A bad workman blames his tools".

 

I played with it too...

hello I was just wondering what affect does the the last two Lt1819 before the summer at the end are they essential to the design

 

hello I was just wondering what affect does the the last two Lt1819 before the summer at the end are they essential to the design

You mean the voltage followers made up of U4 and U5.
The short answer is, U4 is required, but theoretically, U5 isn't. U5 is needed for high speed signals, in order to keep the path delays balanced.

 

You mean the voltage followers made up of U4 and U5.
The short answer is, U4 is required, but theoretically, U5 isn't. U5 is needed for high speed signals, in order to keep the path delays balanced.

I tried taking off U4 and U5 from the circuit ,U5 didn't get affected this much but U4 it was horrible to look at what affect U4 got on the signal, I built this circuit on a breadboard it was kind of acceptable but it was not as good as you i can see it from your circuit .

 

<snip> I built this circuit on a breadboard <snip>

This is why you are having problems.

See reply #2; part of which reads:

<snip> it may be layout related as you would need a very carefully thought-out and placed layout to keep out oscillations and distortion.<snip>

Kamran Kazem