Is there something wrong with PMOS model in the LTSpice XVII?

As you say, "When I tried a some [sic erat scriptum] specific fet it worked."

You've answered your own question.

LTspice is notorious for having LTspice-specific models tweaked with special effects to make LTspice look super speedy.

I'd do two things, avoid LTspice-specific models and be sure to set solver to "Alternate".

Make that _three_ things... treat LTspice XVII with caution until it is completely debugged.

Did you try your circuit on LTspice IV?

Complete schematic at...

Message-ID:

...Jim Thompson

On closer examination of the circuit... WHERE in the world did you get that circuit ?:-)

I suspect _ideal_ diodes 'D' may actually be the problem... though it looks to be a crap circuit in general. ...Jim Thompson

That circuit appears to have a stable state that toasts the mosfet.

change L to something much larger and the circuit will oscillate.

reasons are obvious if you check out the stock values on spice defaults

Jamie

When PMOS Source is at +20volts and Gate and Source are at zero, the fet should conduct. Now the circuit takes 4mA. This same happens with LTSpice IV. I have no problems with NMOS, what I can remember.

There is no problems with the simulation when I change fet with "Pick new MOSfet".

How can I see what PMOS values are?

The symbol should have a model attribute... and the model should be in one of the libraries.

BEWARE: LTspice has taken to encrypting their models, so you may not be able to tell what the parameters are.

Oooops! Loading up your schematic... that PMOS is a so-called break-out device... you need to fill in the parameters, otherwise it defaults to who-knows-what... which is why it works when you "Pick new MOSfet". ...Jim Thompson

Forget to add: Place "hand" cursor over MOSfet, right-click to set/see parameters. ...Jim Thompson

Thank you. Just that there was wery little to choose from. VDS, Rds on and Qgate. Not for instance Vthres.

I played some more with circuit and noticed there definitely is a stable non oscillating state. But I couuld make work with realistic part values.

Is there a better circuit without ICs? For 2-3 1W leds and around

Of course not... LTspice relies on hokey LTspice-specific models to demonstrate/exaggerate how fast it is >:-}

...Jim Thompson

How much current is needed by a "1W" LED? ...Jim Thompson

Well, I aimed for 300mA. Max is around 1A.

Try the following. I arbitrarily picked a PMOS, and didn't spend any time optimizing, but it's something for you to experiment with.

Version 4 SHEET 1 1208 800 WIRE 112 -272 -112 -272 WIRE 256 -272 112 -272 WIRE 576 -272 256 -272 WIRE 256 -192 256 -272 WIRE 112 -176 112 -272 WIRE 576 0 576 -272 WIRE 112 16 112 -96 WIRE 368 16 112 16 WIRE 528 16 368 16 WIRE -112 112 -112 -272 WIRE 368 112 368 16 WIRE 576 224 576 96 WIRE 576 224 496 224 WIRE 608 224 576 224 WIRE 768 224 688 224 WIRE 864 224 768 224 WIRE 864 240 864 224 WIRE 496 288 496 224 WIRE 112 304 112 16 WIRE 768 320 768 224 WIRE 256 352 256 -112 WIRE 256 352 176 352 WIRE 864 352 864 304 WIRE 256 432 256 352 WIRE 368 480 368 192 WIRE 368 480 320 480 WIRE 576 480 368 480 WIRE 864 480 864 416 WIRE 864 480 656 480 WIRE 864 528 864 480 WIRE -112 656 -112 192 WIRE 112 656 112 400 WIRE 112 656 -112 656 WIRE 256 656 256 528 WIRE 256 656 112 656 WIRE 432 656 256 656 WIRE 496 656 496 352 WIRE 496 656 432 656 WIRE 768 656 768 384 WIRE 768 656 496 656 WIRE 864 656 864 608 WIRE 864 656 768 656 WIRE 432 688 432 656 FLAG 432 688 0 SYMBOL voltage -112 96 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 20 SYMBOL npn 320 432 M0 SYMATTR InstName Q1 SYMATTR Value 2N3904 SYMBOL ind 704 208 R90 WINDOW 0 5 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName L1

SYMBOL res 848 512 R0 SYMATTR InstName R1 SYMATTR Value 2.7 SYMBOL res 672 464 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R2 SYMATTR Value 100 SYMBOL schottky 512 352 R180 WINDOW 0 24 64 Left 2 WINDOW 3 24 0 Left 2 SYMATTR InstName D1 SYMATTR Value BAT54 SYMATTR Description Diode SYMATTR Type diode SYMBOL diode 848 240 R0 SYMATTR InstName D2 SYMBOL diode 848 352 R0 SYMATTR InstName D3 SYMBOL res 240 -208 R0 SYMATTR InstName R3 SYMATTR Value 10k SYMBOL res 96 -192 R0 SYMATTR InstName R4 SYMATTR Value 1k SYMBOL cap 752 320 R0 SYMATTR InstName C1

SYMBOL pmos 528 96 M180 SYMATTR InstName M1 SYMATTR Value RSR020P05 SYMBOL npn 176 304 M0 SYMATTR InstName Q2 SYMATTR Value 2N3904 SYMBOL res 352 96 R0 SYMATTR InstName R5 SYMATTR Value 22k TEXT -128 744 Left 2 !.tran 0 300m 0 TEXT -128 784 Left 2 !;op 0 300m 0

...Jim Thompson

Pmos requires the gate to be negative w/r to source for conduction between source and (negative) drain.

RL

A typo. Source is at +20V. Gate and DRAIN are at zero volts. The fet should conduct.

I tried the circuit Jim Thompson sent. It worked but I noticed , when I tried to make switching frequency higher, that it will stop oscillating with certain values. Either L1 or C1 must be large. I quess these kind of supplies need a Scmitt-trigger or flip flop to switch the fet either fully conducting or fully off.

LM

It does have Schmitt action, observe base of Q1 and gate of M1.

Not quite sure why is needs large values. I suspect that the time constants get forced by the required high value of current sense resistor, so a different sensing method is in order... maybe use a comparator, even though you said no I/C's ;-) ...Jim Thompson

I'll try it. What is perhaps happening is that I'll move from discretes to ICs, whether I wanted it or not

Leif M

Thanks again. It worked(of course). I'll test some more to get some idea of when it works and when not. Leif

It's easy enough to get going,with non-default model parts, and larger L values. A PNP bipolar switch will react much more predictably with the switched-current drive. Even so, I doubt that very high frequency operation was anticipated.

RL

Thumbing thru a book, "Feedback in Analog Circuits", by long time buddy Agustin Ochoa, Xtreme Spectrum (*) & Zarlink chip projects, I found an example analyzing a switching regulator.

Ochoa suggests that it is best analyzed in a similar fashion to PLL analysis, except use duty-cycle as the loop variable.

Which led me to realize that hysterical (:-) loop control is self defeating... with regions of non-oscillation and yet no ripple reduction.

Conclusion: Clocked works best.

(*) 7,010,056 (Spread Spectrum) System and Method for Generating Ultra Wideband Pulses ...Jim Thompson