Faster BJT in levelshift

Nah, don't use that, JT is right it's a stinker.

Something like this maybe?

assuming you want to use cheap discretes, and not buy a nice driver chip.

I can't tell if Q1 is a badly drawn BJT or an even badlier drawn MOS of some sort. Been taking drafting and scanning lessons from Jan? ;-)

Tim

How about 2N7002 instead of BC846 ?

Does the PNP follower B-E junction get breakdown?

piglet

The problem is Miller capacitance. A cascoded 2n7002 would fix that, a one-component fix.

Cheers, James Arthur

Hey, that's what we call feature creep over here :-)

But yes, if the +12V jumps around too fast this isn't going to work unless you use more parts and that would blow the budget out the window. That probably only leaves the strategy of negotiating a good deal on one of the faster BJTs.

But, 'fast' and '50V' in a BJT are mutually exclusive. That's why the 2n7002 cascode is helpful.

He doesn't really need the PNP--he can use the pre-driver to do that.

Vcc Vcc -+- -+- | | [R1] | | |/ Q3 Q1 +---+----| 2n7002 | | |>. ||--' | | ||. | ===

Cheers, James Arthur

Not really. I brought a couple of examples but they cost more than a BC846. So they would only work if the price can be negotiated down accordingly.

However, I suspect that such strategies add way too much cost in this case. When you add parts you can't negotiate down to the old totals.

AoE must totally confuse you.

There would still be a passive pullup on the upper nfet.

Why not use a pfet in the output, and get rid of the floating 12 volt supply?

If the cascode has too much bandwidth then, though, the arrangement looks kinda scary, sort of like a Colpitts oscillator:

That's enough for small drivers, especially since the cascode dramatically reduces that node's capacitive loading.

That works as long as you handle the shoot-through problem, it just takes more parts.

A bootstrap will generate the floating +12V supply--that's just a cap and a diode.

Cheers, James

I only added the 2n7002. 3 cents by the reel over at Digikey.

If you can get one of the others for cheap and with a Miller capacitance close to zero, sure, go for it.

Meanwhile, the 2n7002's an option with no uncertainty about negotiations, availability, etc.

Cheers, James Arthur

I assume the +12 supply rides on V+. The budget is already blown.

Why two n-fets in the output stack?

The passive pullup on the big nfet kills the speed of the rise, and the pulldown to the source kills the speed of the fall. Faster transistors won't help either.

Or fewer parts.

If DC response isn't needed. The problem is underspecified.

Sort of, except those are amplifiers, with linear bias, etc., and you're working with digital style on/off signals.

Well, didn't youstabee. 2SC3502 was available until recently, and I thought I had seen another like it still available but don't remember the number. There's a fair number of ZTX parts with those kinds of ratings (but not wattage).

Maybe not 3 cents though :)

Tim

Congratulations, you discovered the reason for gate resistors and hard saturation. :^)

In other words, this is only relevant during commutation -- which is a very important consideration. With modern SuperJunction FETs, you might see a bright and sharp tone burst at 300MHz!

(Curiously, there are a few devices that have remaining transconductance while in "saturation" -- JFETs are one, toobs are another -- so they need to be stable even in the saturated state. Most MOSFETs and all BJTs poop out under these conditions and don't have any gain to oscillate with.)

(Hmm, all simple BJTs, not Darlingtons..?)

Tim

The gate part of the symbol reveals it's a MOSFET.

The slowdown issue is driving capacitance. John's circuit can drive the capacitances arbitrarily fast.