cool high voltage generator

Your exceeding you device MAX recomended voltage again, John.

Cheers

Now I'm afraid to open a tube of these parts.

And if I connect the proper power supplies, it pulls 100 mA, just sitting there closed-loop.

I had a simulated CMOS 3-inverter oscillator which drew IIRC 1E38 amps at 5V. I didn't build it for fear it would destroy the world's power supply infrastructure.

I just hope the design never falls into the wrong hands.

Cheers

Le Mon, 16 Sep 2013 19:31:26 -0700, John Larkin a écrit:

With that fast an amplifier you're not supposed to run it for such a long time...

It would be far too easy for some naiive student to build a gigavolt or even a teravolt flyback converter.

I find myself worrying about power dissipation in Spice models, using, say, 1K resistors when 1 ohm is more logical. A million volt source and a million ohm resistor is a good source, but I'm reluctant to waste all those virtual watts.

I have used 1e15H inductors as DC-only shorts. Like for biasing an AC coupled amp, where I don't want to see noise say, or low frequency deviations from the resistors.

(Now I just need to find some in Farnell and I'm all set).

I did some superconducting simulations with something like `1E-10 ohm resistors.

Won't work in San Francisco because most of that voltage will get taxed away.

It's the shipping costs that will getcha.

For realistic simulations use behavioral parts that behave like real-world parts...

...Jim Thompson

What's weird is the direction. The V+ pin *sources* current. If you open the V- pin, it sinks current.

There must be some independent (ie, unpowered) current sources inside the chip, sneaking current through some junctions or something.

Yup, 23 mA comes out of the V+ pin, into a big load resistor to ground, but it stops at 23 kilovolts. Violates conservation of energy big time. That should be illegal.

The 100 mA supply current, with +-5 supplies, is also dead wrong. I sometimes use the supply pins as signal sources, into cascodes or optocouplers, and I like my supply models to work.

For that the models are often too primitive, too much behavioral content in them. Even with LT switchers I often see issues that are definitely there in real life and the sim won't even show a whiff of it.

My personal opinion is that LT goes overboard with behavioral modeling, and clearly does not spend the time to A-B compare the behavioral model against a simulation of the true, device-level, netlist.

Probably a task assigned to junior engineers... you know, those right-out-of-college types with PhD's >:-}

One thing I notice LT is doing is gradually encrypting all their libraries.

Even the "passive" components are being encrypted... bead, cap, inductor and resistor are already done.

Likewise I note that any chip with a .SUB extension is now encrypted.

I'm trying to decide if this is good or bad. I note further that using alternate solver seems to give more accurate results (much to the annoyance of those here on S.E.D who know it all, who don't like the spike size this shows up :-)

My suspicion, since the instructions in the manual say...

"There is no .option to specify which solver is used, the choice must be made before the netlist is parsed because the two solvers use different parsers."

This smells like maybe the alternate solver uses a real netlist, otherwise why "different parsers"... and the alternate solver is visibly slower.

Save all your old libraries, if you still have them >:-} ...Jim Thompson

Good luck with that! I know I got into trouble big time once trying to model current use by an opamp. Most model, if they model anything at all, just put a resistor to ground from the sources to model the 'normal' supply current. There are also a lot of them that have current sources, or internal '0' nodes that can create interesting results. The most common one in PSpice was when the user didn't put down a '0' ground, but pulled a different one out of the libraries. If there is no ground in the schematic, you get a nice obscure 'floating nodes' error message. If there is a '0' in the middle of the opamp model, then everything in the design gets scaled to that node as ground!

Modeling the actual current consumption so that you can play those cascode tricks requires you to write your own model, and you have to really understand the performance of your opamp to get that to work correctly. A non-trivial exercise!

Charlie

Back in the ol' Microsim days, there was actually a small lab with scopes, curve tracers, signal generators and other gear that was used to verify models. Was used extensively during the original digital part modeling days. Probably too much trouble today...

Charlie

I am impressed with the LT manual, in that it expressly recommends NOT to use the global '0' inside of subcircuits.

I don't find it difficult >:-} ...Jim Thompson

I wouldn't think so. Are companies now too cheap to assign one engineer and one technician full-time to the task?

(Of course, I have to lament, some of those MicroSim models weren't so nifty-swifty >:-} ...Jim Thompson

Superconductors are supposed to have zero resistance, but on the high-field NMR magnets, the fields occasionally jump down (by a few PPM per jump) as the current paths shift.

I float opamps hundreds of volts off ground. And none of the opamps that we buy have ground pins!

Or, just give up simulation and design the old-fashioned way.