Interesting Simulation Problem

...snip...

Nope. You missed it. As usual >:-} ...Jim Thompson

Just curious. How often has simulation revealed something, you didn't already know, or suspect?

Reason for my question, when I first started Engineering, my mentor looked over my shoulder at my several pages of equations and laughed uproariously. He said, you'll get lost in all those equations. Instead use these simple models, then when you know what you have, verify with those lengthy equations. He was right. Thus today, rarely, if ever, does the simulation do anything but verify.

In his case? Every project. He designs bugs in and simulates them out. Most programmers work that way, too.

Good point. Jim says he has a 20,000 transistor analog CMOS circuit. It has to be low standby power, so he's trying to figure out how to use Spice to find floating nodes.

Why didn't he design it to not have floating nodes? Does he not understand his own design?

Or someone else designed it, in which case he still doesn't understand it.

Analysis, either mathematical or using a simulator, is a poor replacement for design and understanding.

Not very often. Occasionally it reveals a latch-up/lock-up I didn't anticipate. And virtually _all_ CMOS OpAmps require simulation to get the loopgain compensation correct... the Level=3 school boy equations (the only ones hand-tractable) don't even do a good job at gain prediction.

Yep. That's what I've attempted to convey... design with your head, then verify via simulation.

The neon bulb modeling exercise was just to prove to myself that I could create a model for a very abrupt slope change while maintaining finite derivatives. Otherwise I have no use for a neon bulb... haven't used one in a circuit for probably 50 years. ...Jim Thompson

I'm always amused by your insistence to jump in with both feet and display your profound ignorance >:-}

Here's how the floating node situation comes about...

You design a circuit for "form, fit and function" (as they say :).

THEN: You go back and add switches (this is CMOS analog FYI) to put the circuit in sleep mode... extremely low leakage :-}

Which is why you never post circuits with component values ?:-) ...Jim Thompson

As opposed to designing it hazard-free in the first place. You'd make a great programmer, if you ever learned to program.

I have posted lots of circuits with values, including whole schematic sheets of actual products, and PCB photos. I've also posted lots of conceptual circuit doodles without values.

You mostly post circuits without circuits! Namely PDF waveform results without any circuit at all. Or just claims without content. And you almost never post netlists that other people can run and verify.

I share and discuss, you keep secrets and boast. Different personalities. No wonder we don't get along.

Pretty good weaseling, a B+ at least.

I'm a big fan of belt-and-suspenders, though, especially if it doesn't take any extra effort. Using gmin and split supplies, plus some simple script to parse the .op or .raw file, could be quite a good use of time.

Nobody's going to place and wire 20000 transistors by hand, so there has to be some combination of manual and automatic generation. Even using the greatest care, you could have some bug in your design software, misnamed nodes or something, and an automated way of finding that with zero effort would be pretty useful.

Cheers

Phil Hobbs

Place, yes, wire, no... wiring is usually automated (that's part of what LVS is all about... the netlist describes the connection), then the layout engineer rubber-bands and re-routes to suit.

Besides the whole 20,000 aren't done at once. I/C layout is by convention (and ease) done cell-by-cell. These days it is also common to have at least three metal (wiring) layers, or more.

Misnamed nodes isn't an issue (unless forced, node naming is automatic). Identifying those that can go wandering off is.

Hopefully I'll find a way to do this without splitting the supplies (necessary if you use the gmin approach). Splitting the supplies make me nervous because I keep running into modeling types that want to call the substrate node 0 (zero), when it should be...

?SUB|[@SUB]|~SUB|0|

which assigns substrate connection to node 0, _only_ if the substrate node isn't specifically declared.

Searching thru my symbol library to make sure I got that script correct I noticed that, in PSpice, I have 846 personalized symbols :-[ ...Jim Thompson

You don't have room to talk! Did you figure out that bloggs screw up yet?

Jamie

It does work as advertised - his claim is not at all believable.

Putting red LEDs in pairs with a basic resistor ballast hung off 6v will easily match or beat his suggested method. The buck converter on that step up ratio would be lucky to get 80% efficiency and is a lot more complex to implement.

PWM regulating of the current would be another choice. The losses could then be as low as Vsat.Imax plus dissipation in a smaller series resistor to ensure chains share current more or less equally.

Most LEDs can survive a higher pulsed current provided that their power dissipation is not exceeded. More of your supply voltage gets converted into light this way since Vf rises with increasing current.

The screw-up seems to have been in the original circuit which seemed to hav e a resistor in series with each LED to regulate the current through it, so that roughly half the power being dissipated was dissipated in the series resistors, creating heat but no light.

What Fred presumably had in mind was to put all the LEDs in series and cont rol the current through the string with a single boost converter.

The voltage drop across the single current sensing element is then a small fraction of the voltage drop across all the LEDs in series, and - in a prop erly designed system - rather less than the voltage drop across any one of the original series resistors. Effectively, you use pulse width modulation to to get the desired voltage output to drive all the LEDs and this can be pretty efficient, if you don't scrimp on the inductor and the switch.

It seemed to be well beyond what the OP was interested in doing, or capable of understanding, but if you think that Fred screwed up, you probably didn 't understand what he had in mind.

His proposal would have roughly doubled the battery life for the same light output, which is pretty dramatic, purely by regulating the LED current mor e cleverly.

And you still don't get it do you?

I hope you were never involved in any life saving applications when you were in the job force..

Jamie

I am glad you finally fest up to something.

And don't bother to reiterate on the original thread, it obviously slid under your radar.

Jamie

Jamie's approach to spelling is original, to put it kindly,

I think you are being disrespectful of monkeys here. They could at least be relied on to type random character strings. Jamie's output is rather more structured and repetitive. Not well structured, but he does tend to make th e same mistakes rather frequently.

Useless.