free energy!

I've got my own issue with bipolar junction transistor models. What get pub lished are Gummel-Poon models, which don't model inverted transistors all t hat accurately, and - in particular - can't model the "squegging" mode in b ipolar Baxandall class-D oscillators with big feed inductors.

The VBIC model ought to be able to manage this, but it's parameters are tre ated as "commercial in confidence" by transistor manufacturers, so I've nev er been able to get my hands on one.

The paper only mentions "squegging" in a footnote at the bottom of page 752 , but the behaviour blocks the obvious route to minimising the harmonic con tent of the output waveform. MOS-FET-switched versions of the circuit don't squeg, but mostly end up being a bit more complicated.

is a actually an LTSpice .asc file, which will work fine after you have cha nge the extension from .htm to .asc.

That's NOT a Sallen-Key filter, it's a simple audio amplifier, posted years ago (2002) when someone here asked about dead band distortion.

Why did you have trouble recognizing such a "bizarre circuit"?

Clueless? Or just so anxious to be nasty you missed the whole point?

Posted just to illustrate that PSpice DOES compute power correctly.

However:

• D:\Projects\Expments\Audio\DeadBandReal.sch

• Schematics Version 9.2.3

• Fri Jun 14 15:25:38 2002

.OPTIONS SOLVER=0 .PARAM DEADBAND=1 SLEW=10 RBYP=40

• From [PSPICE NETLIST] section of C:\Cadence\PSD_14.2\tools\PSpice\PSpice.ini: .lib "C:\PSpice\SymbolLib\mylib.lib" .lib "C:\PSpice\SymbolLib\nom.lib"

.INC "DeadBandReal.net"

.PROBE V(*) I(*) W(*) D(*) NOISE(*)

.END

• Schematics Netlist *

R_R1 OUT 0 4 V_VSRC2 N_1 N_2

• ROUT=0.1 VP=15 VN=-15 PM=45

Concatenate .CIR and .NET above, and you can run this on LTspice, just do file, open, choose the .CIR file, then RUN.

Wowee! That does not prove that the simulator you use has no flaws.

I designed $ millions of chips BC (before CAD), some as long ago as 56 years ago are still selling.

Check out the LTspice List. I suggest that all of you who use LTspice subscribe, so that you're aware of its pitfalls.

e.g. It almost categorically can't run Microchip models. PSpice can. ...Jim Thompson

That's because you're not recognized as a professional in the industry

...Jim Thompson

published are Gummel-Poon models, which don't model inverted transistors al l that accurately, and - in particular - can't model the "squegging" mode i n bipolar Baxandall class-D oscillators with big feed inductors.

treated as "commercial in confidence" by transistor manufacturers, so I've never been able to get my hands on one.

Why would I be? I've worked on the electron-beam microfabricators that wrot e the masks for the assorted layers in the process of creating integrated c ircuits, and on electron-beam testers that looked at a single transistors o n those integrated circuits when they had been built and were working, but that was just electronic engineering.

There'd have needed to be a need to know on my part to facilitate something that they wanted done before I'd a "commercial in confidence" disclosure.

Well, U1 sure looks like a S-K lowpass to me, and at 33k*1uF is a bit slow for audio--the corner frequency is 3 Hz. Also it apparently has no output, and two of the resistors on each base don't look to be doing much.

Apart from that, it's entirely sane and rational. ;)

Cheers

Phil Hobbs

Duh! You and Larkin like to play team-tag-assholes against me... but you blew it, AGAIN... you looked inside the BEHAVIORAL MODEL for U1 (BU7262, OP-AMP-CONFIG, from the Device Models & Subcircuits Page of my website), an OP-AMP... the 3Hz is the open-loop corner of the Op-Amp.

There's no S-K Filter ANYWHERRE in that circuit.

The circuit was just posted to show PROPER computation of power dissipation. ASC file coming... I will look forward to your "expert" analysis >:-}

Apart from that, you're a perfectly sane and rational AMATEUR... who couldn't even properly analyze how the model works >:-}

...Jim Thompson

You posted the picture I'm talking about--look at it yourself. I sure didn't go spelunking round your website. In the picture you posted, U1 is clearly a 3-Hz Sallen-Key. That photo ain't no audio amp. I saved a copy of it in the Wayback Machine for future reference:

or .

If that isn't what you thought you posted, that would partly explain your confusion.

And regarding the above bluster and mud slinging: Where I come from, we know a trick worth two of that one.

Cheers

Phil Hobbs

The reference file showing PSpice power computation was...

as is the .CIR and .NET above; how it got snipped into referring to the LTspice ASC file I don't know, but the LTspice ASC file _does_ show negative power... apparently because LTspice keep the +/- terminals the same for voltage direction, even if the resister is flipped.

So, irrespective of S-K or audio, LTspice _can_ report incorrect power. ...Jim Thompson

Really, Jim, you're snarling in all directions and not making sense any more. That's sad.

Maybe R6 is installed "backwards", with the current direction wrong relative to the voltage nodes. Resistors do have a current direction in LT Spice, and I sometimes flip them over to see the current the way I want. We know that resistors don't generate power, so it's easy to fix.

Confusions abound:

(1) It wasn't my "bizarre circuit". Someone here posted it as a question back early in the year. I downloaded the file, but was too ill at the time (Septic Shock) to analyze it (and have since forgotten what the original question was ;-) I picked it as an example arbitrarily because it came up first in a search for *.asc

(2) Suspecting, from the equation that appears when you apply the ALT key over a component, I found that the flaw is committing to the +/- terminal convention of the component.

(3) Thus I flipped R6 end-to-end and got negative power.

That _is_ a flaw in LTspice.

I posted this (previously)...

to show how PSpice does it.

I'm not peddling PSpice... you can't even buy PSpice with the features I have anymore. Cadence/OrCAD has completely neutered the product to be only PCB-specific... they want you to have to buy Cadence/Virtuoso/Spectre for the big $$$$.

Maybe I'll have my heirs auction off my version ;-) ...Jim Thompson

You're making the world far more difficult to deal with than it really needs to be....

By circuit analysis convention all currents are defined INTO the nodes. That's where Mikey blew it.

Let the plebes respond first as to how to calculate power dissipated.

HINT: Kirchhoff's Laws apply ;-) ...Jim Thompson

In this case, I flipped R2 to reverse the current direction. LT Spice got both equations right.

I can't make LT Spice produce any, even transient, negative power into a resistor. Maybe you can.

I think my negative averaged power was probably a windowing issue into a capacitor. Capacitors can absorb or deliver energy.

It seems to be related to when you change something in a circuit, whether you do a "Save" or not. "Save" is good, not saved produces the anomaly.... seems to blow LTspice's mind as to current and voltage sign direction. Strange.

Do you have Solver=Alternate set? I have that set by default, cures most mismatches between LTspice and industry standard simulators. ...Jim Thompson

HINT: There's life after the trapezoidal rule. ;)

Cheers

Phil Hobbs

And the horse may talk ;-) ...Jim Thompson

How'd you get it to say "POWER in R6"?

The upside down graph is easy enough, flip the resistor after setting up the plot and the current is reversed but the voltage is measured on original nodes.

OK, that makes negative power. Flip the resistor *after* you run the simulation but before you do the alt/click thermometer thing, so the post-sim calculation and plot uses old RAW data but a new equation.

The simple fix: don't do that.

Of course it does.

Hey, look at this one!

Observe the displayed equation when pressing the ALT key (power, NOT integrated). ...Jim Thompson