PSpice Ratiometric Simulations

This might get you started. Got the basic idea from elsewhere on the web , so no credit to me :0) It does a worst case analysis 1% on all resistors , but I have NOT done the 0.1% ratio calculation. I'm sure someone can improve , but it will get you started. Use LTSpice.

Version 4 SHEET 1 880 680 WIRE 176 -32 16 -32 WIRE 320 -32 176 -32 WIRE 432 -32 320 -32 WIRE 16 16 16 -32 WIRE 176 16 176 -32 WIRE 320 16 320 -32 WIRE 16 160 16 96 WIRE 176 160 176 96 WIRE 320 160 320 96 WIRE 176 240 16 240 WIRE 320 240 176 240 WIRE 176 256 176 240 FLAG 176 256 0 FLAG 432 -32 Vout SYMBOL voltage 16 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 1 SYMATTR Value2 "" SYMATTR SpiceLine "" SYMBOL voltage 176 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value 2 SYMATTR Value2 "" SYMATTR SpiceLine "" SYMBOL voltage 320 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value 3 SYMATTR Value2 "" SYMATTR SpiceLine "" SYMBOL res 0 0 R0 WINDOW 3 -70 117 Left 2 SYMATTR InstName R1 SYMATTR Value {wc_a({R1},tola)} SYMBOL res 160 0 R0 WINDOW 3 -32 119 Left 2 SYMATTR InstName R2 SYMATTR Value {wc_a({R2},tola)} SYMBOL res 304 0 R0 WINDOW 3 68 68 Left 2 SYMATTR InstName R3 SYMATTR Value {wc_a({R3},tola)} TEXT -264 -24 Left 2 !.param tola=0.01 TEXT -264 296 Left 2 !.function wc_a(nom,tola) if(run==1,nom,if(flat(1)>0,nom*(1+tola),nom*(1-tola))) TEXT -264 -64 Left 2 !.step param run 1 40 1 TEXT -264 328 Left 2 !.tran 1 TEXT -264 24 Left 2 !.param R1 = 1k TEXT -264 64 Left 2 !.param R2 = 2k TEXT -264 104 Left 2 !.param R3 = 2k

Parameterize the _value_ of R2 and R3. Then Monte Carlo on the parameter and the ratio? ...Jim Thompson

You could make R2 and R3 each from two resistors in series, and apply different Monte Carlo rules to them, specifically lock R2A and R3A to the same value and flog them 1%, but let the other two be independent and vary 0.1%.

Something could be done with a Y-configuration representing R2 and R3, but takes more thinking.

Jim Thompson a écrit :

The real question being: does one need spice for such a simple answer that should take no more than 2 or 3 lines?

You didn't include the "2 or 3 lines" ?:-) ...Jim Thompson

Come on... I carefully wrote 1 2/3 lines to hint you in the right direction, leaving you 1 1/3 line to complete the answer :-) Now you have between 20 and 30 characters left. That's getting to be tough.

I was yanking your chain. It's a simple Algebraic expression. I don't (generally) do Monte Carlo... I do worst case. If I pass worst case I know I'll pass _any_ Monte Carlo. ...Jim Thompson

You can use the known ratio tolerance, if you change the circuit part with R2, R3, V2, and V3 to a de Thevenin equivalent circuit, applying the resistance tolerance to the internal resistance and the ratio tolerance to the equivalent EMF of the de Thevenin voltage source.

Worst case in this instance means only testing eight points, so one does wonder why they can't start there...

Some managers get a hard-on over Monte Carlo... you know the type, they live in Californica, vote leftist, and buy Cadence tools >:-}

In my ASIC world it's easier to do worst case... maybe 15-120 simulation runs depending on the component type-count... and it's easier to evaluate the results by eye. ...Jim Thompson

I could see doing Monte Carlo either if it runs faster than worst-case, or if some of your "goodness" to parameter relationships weren't monotonic (meaning that the worst case isn't on a vertex of your parameter space), or if you (a) have a good grasp on the actual probabilities within your ranges, and (b) aren't passing worst-case but think you could get sufficient yield in the real world.

I don't know if I believe either (a) or (b), I'm not sure that there are any cases of non-monotonic relationships like I'm hypothesizing, and I'm not sure than you really get a good statistical sampling in Monte Carlo unless you do more testing than you would need for worst case.

Maybe it's just that "Monte Carlo" sounds like a better place to be than "worst case".

The main non-monotonic worst cases are usualy in filters, where the resonant point is included in the range of values...

Charlie

I was once in the casino at Baden-Baden, Germany. Talk about feeling out-of-place... people were betting more than I made in 5 years :-(

But nice restaurants... like "Der Kleiner Prinz" ;-) ...Jim Thompson

er

en

rst

es

on

I read an article about by who travels around to casinos, the only place he really felt poor was one of the asian casinos where new rich chinese were shoveling piles of money on the tables like there was no tomorrow

-Lasse

It's been a while since I was at either, but the Macau casinos looked pretty grubby compared to Monte Carlo. A lot of money changing hands at both though.

Best regards, Spehro Pefhany

Stay in the casino awhile and the two rapidly become one. :(

Cheers

Phil Hobbs

Free drinks!