SPICE gets it wrong

Nevermind I found the link

There isn't unlimited current for sure; even for large signals the 3904 definitely doesn't have infinite output admittance.

Randomly-picked gate stoppers and ferrite beads and stuff are finally kind of a kludge; hey! I thought the whole point of a driver circuit for a BJT was to slam gobs of current in and out of the base fast-like!

What's the frequency response of the C3/R8/L3/R6 network in isolation? Band-reject?

B+D has 10 of 'em still in stock, $1.97 each.

Yeah, in practice it's something like 100mA, which is still quite reasonable, and much more than the steady-state base drive...

Ooh, Q5 and Q3 are actually 2N4401 (and, perhaps unsurprising, all the other "MMBT"s are just lies pulled from my library -- they're actually "2N"s), which means more current for the same hFE, give or take. That makes a difference.

With D4 removed, everything else more or less as shown, I noticed the on-time was significantly reduced: because almost half the shunt current (via R13) came from base drive alone. That was partly, but not wholly, made up for by the whopping 8us turn-off time for the TIP31! :-)

So that means about half an amp through the poor Q3. Which doesn't add up; R3 is only delivering about 2mA, so that implies a fairly high hFE, 250 -- at current, no less. (That would be reasonable for a ZTX651, but a 2N4401 maybe not as much.)

Incidentally, PBSS303NX also works fantastically in place of Q6, and 2N4401 is also capable of being abused hard enough to do the job. 2N3904 still manages to switch, but it looks kind of ugly... :)

It does, at least in so much as C3 allows "some". (It's also rate limited by C1 + R3, hence the long turn-on delay.)

Yes, it's parallel resonant (with lossy L and C giving a low Q), strung between two low-impedance nodes. If r_e is on the order of 1 ohm, then the LF cutoff frequency is 20kHz. Notch center at 3.8MHz, then reaching a shelf at 33MHz where the 22 ohm takes over.

In practice, it seems the LF cutoff is much higher than 20kHz, seeing as Q3 is more dynamic than that (it will swing up as high as C1+R7 lets it go, then D4 pulls in). L3 may also be saturating in part, as it's on a ferrite toroid.

Hmm, I estimate it's 70% inductance at 400mA, so maybe not much. It's 5 turns on a T37-43 (Fair-Rite 2643375102), FWIW.

In any case, it inserts a lot of "squishiness" between Q3 and Q6, and that also allows C1+R7 to be more effective (on some transistors, they aren't necessary at all).

Tim

In the right application, TIP31 is quite a nice switch (it makes a good driver for a low-dropout regulator, for instance). Power transistors usually don't

In the right application, TIP31 is quite a nice switch (it makes a good pass transistor for a low-dropout regulator, for instance). Power transistors usually don't saturate as nicely. That's no design error, it's not easy to guard against second breakdown without gumming up the ON operation of a transistor.

(obligatory lowpass negative regulator...)

The TIP31C spec was SO loose because it was a pioneering plastic part ( the 'C' is a voltage breakdown bin, it's that old...) which was intended to be cheap with high yield. Various manufacturers who co-opted the part number (to compete on the cheap part) weren't making the same part, just meeting the spec.

It's hardly surprising that the SPICE models are junk. The manufacturer tries hard not to spend extra test time on the low-end product. At a guess, these parts will not be similar batch-to-batch, nor manufacturer-to-manufacturer. TIP31 and 1N914 are similar, that way.

I rarely use bipolar transistors any more.

They're cheap.

A 2N7002 or FDV301 costs us 2 cents.

MTP10N10 is 100 volts, 10 amps, TO220, 37 cents.

The threshold voltage of 7002s is horrible. Anything with a decent threshold voltage is more expensive than that and NPNs are less than a couple of cents anyway. A dual or a dual pre-biased saves a bunch in placement costs, too, though we can't use them as I/O.

On an infinite heatsink?

The FDV301 is a low-threshold part. Also 2 cents.

I have done some fun things with 2N7002s. This is 50 volts into 50 ohms in 1.8 ns, transformer isolated:

In this application, I couldn't find a modern mosfet that was any better than the ancient 2N7002.

According to the data sheet, it can dissipate 75 watts. It can switch a kilowatt. Fast.

It doesn't need a lot of base current either.

You shouldn't have titled your subject as "SPICE gets it wrong".

It should have been "Spice got it exactly right for the inadequate model that was presented". ...Jim Thompson

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

Thinking outside the box... producing elegant solutions.

"It is not in doing what you like, but in liking what you do that is the secret of happiness." -James Barrie

And, it won't work in the low-voltage LDO negative regulator

nor lots of other places. Bipolars' Vbe is usefully low and predictable.

On a sunny day (Tue, 09 May 2017 16:32:11 -0700) it happened John Larkin wrote in :

Maybe for high end stuff like you are making, but for many down to earth things they are very useful, bipolars are basically current amplifiers. MOSFETS voltage to current converters. Both have their field of application where they are best. And there are some nice RF ones too.

On a sunny day (Tue, 09 May 2017 19:40:10 -0700) it happened John Larkin wrote in :

I exected a circuit diagram ;-(

Anybody can draw that scope picture in gimp.

:-)

On a sunny day (Tue, 09 May 2017 22:05:42 -0700) it happened Jim Thompson wrote in :

True 'math got it right' in Epicycles too:

Sorry, that's too valuable to give away. I will say: drive the gates hard and ignore the data sheets.

You can buy it and shoot your own scope pix!

We have a 100 volt version, using different fets.

I'd like to go higher and faster, with GaN fets and fancier transformers, but I don't think there's a market.

If there is some gate supply voltage available, a mosfet LDO can drop millivolts.

There are commercial p-fet based integrated LDOs too.

On a sunny day (Wed, 10 May 2017 07:23:04 -0700) it happened John Larkin wrote in :

Let me guess, you used an ethernet transformer...?

I see the 2N7002 goes for 8 Euro cent. But I am sure your design adds a significant amount of added value :-)

There is a little oscillation at the start of the pulse.

Maybe some science field, was just reading CERN started a new linear accellerator: https://home.cern/about/updates/2017/05/brand-new-linear-accelerator-cern