Faster BJT in levelshift

Except that a BJT doesn't work. See my post above, or piglet's below.

D2 is already an emitter follower in the .asc. You don't really need it if the drive is stiff enough, so I ASCII'd D2 as a diode, to save Klaus a penny.

R1 is already a current source. (Think about it.)

Cheers, James Arthur

Hmmm?

Saturated BJT works just fine. Emitter to GND, resistor divider from input to base to GND. Yes, it's inverting now, deal with it. :)

Partially, but not fully -- you get some of the advantage of the "Q capacitor" used in 60s audio amps (the pull-up is bootstrapped to the output), which gains you the output swing, _but not the gate swing_, so it's no better than the same circuit referenced to ground (whigh might be driving the low side).

Like this thing, which is a CCS-loaded stage, driving a follower, driving

2N7002/BSS84, driving another follower, driving the final gate: Faster than standard gate driver ICs (though at some expense in shoot-through current, due to the lack of dead time in the middle "CMOS" stage -- a matter of convenience).

The front end looks like this,

except IRF510 --> BSS84+2N7002. "In" is a logic level signal, with whatever level shifting / current limiting is needed to drive the base in the usual way.

A CCS is ~2-3 times more efficient than a resistor, because it applies the same dV/dt the whole way, while drawing the same quiescent current. It's more parts, but it can be a very reasonable tradeoff depending on one's priorities.

The overall alternative, I guess, would be a pull-up to a fixed (B+ + 12V) rail, which would be less convenient than the bootstrap method, and perform significantly worse anyway. (A CCS, would work as well as otherwise, but dissipate the full B+ + 12 volts, instead of just the 12.) I don't consider that an alternative at all, as the bootstrap is obvious enough, and good on its own. :)

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: https://www.seventransistorlabs.com/

Ye-e-e-esh >:-} ...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 http://www.analog-innovations.com |    1962     | 
              
     It's what you learn, after you know it all, that counts.

That's nice, thanks

I tried a simelar version, but with BC846 and BC856 earlier today, in PSpice, but there was an issue when the voltage at the output is switching

This one seems not to have the problem, I will check if it was a model difference between LTSpice and PSpice

I had 50 ohms in series with V2 to emulate the output impedance of the microcontroller pin driving the FET source, but I do not think that is the reason

Cheers

Klaus

But you just ruined the whole point of the cascode--eliminating Miller feedback--which was being proposed as a fix for Klaus' original problem.

Holy crap Batman! 14 semiconductors, nine passives and a hand-wound choke? Twenty-four parts? Are you trying to give poor Klaus a stroke?

Klaus is worried about pennies, so adding a bunch of parts to make the thing faster than it needs to be is all wasted money, to Klaus.

As drawn, the improved Tringlotron has a risetime under 250nS, easily meeting Klaus' spec.

If Klaus wants to go faster, changing the cascode FET to a BSS145 cuts that in half. (I think that's more due to the sloppy capacitances in LTSpice's FET models than the actual parts, though.)

Cheers, James Arthur

Yes, that's why I gone that route. PFET is simple, but performance of 2xNFET is a lot better for same price

Cheers

Klaus

Nice, will try that out tomorrow :-)

Cheers

Klaus

"it's a bitrex circuit! Engage suck mode! Randomize all values!"

But Miller isn't what you're up against, it's Coss. Which is a whopping

30pF (give or take Vds) for a 2N7002. Equivalent to an additional ~30nC gate charge at the other end.

Coss can't be magically cured by cascoding -- that capacitance still has to be charged and discharged, and the current for that charge has to come from the pull-up resistor.

Cascode keeps Crss from screwing things up, or you can pick a device that has little enough Crss to behave. An MMBTH10 is about 1pF, and that's "Coss" and "Crss" together.

In my experience, 2N3904 is even good enough. :)

And you can still cascode the BJT if you insist, without sacrificing saturation voltage; but, of course, at the expensive of a couple more parts.

It was an example. My apologies for assuming you would understand the relevant part of it as an example; I will restrict such things in the future:

Because that's much easier to follow..(?)

I wasn't aware that you were the moderator of this public newsgroup and that we cannot have a discussion about variations on level shifting circuits.

I shall bow to your seniority and abstain from this conversation in the future.

I've seen a half dozen various 2N7002 models, and none of them is very good...

I don't recall that any made gross errors in Coss, but the gate circuit varies a lot.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: https://www.seventransistorlabs.com/

I'm pretty sure, but--because I couldn't find it specified--not 100%, that Miller was Klaus' original limitation.

That's what's in the LTSpice model, but OnSemi lists Coss for their

2n7002K as 4pF (typ), 8pF (max.) 8pF makes a time constant with R1 (560 ohms) of about 4nS, worst case.

Given Q=CV, you're assuming 1kV. :-)

The node only needs charging to 30V, so if we assume 10pF total for Coss and some stray, that's 300pC when fully charged.

Yes--I've got a reel of MMBTH10's in my junk box. But the MMBTH10 is only a 25V transistor, and Klaus needs 50V.

Of course we can have a discussion. You and I were just talking about different things. I was trying to address Klaus' goals, and you were trying to turn that into a much-faster, more complicated driver that goes well beyond what Klaus needs.

Cheers, James Arthur

Don't bother, Spice automates that for you.

I did change the transient analysis duration, to speed things up. At some settings, I see the giant current spikes. At others, it spends forever setting up the initial conditions.

I think people will be breadboarding for a while longer.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Is that a "maybe" or a "yes but I don't want to say Win made an error"?

It was just a "yes", without note of the poster. ...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

| 1962 | It's what you learn, after you know it all, that counts.