Bipolar transistors for SMPS?

Considered cascodes? Cheap hv fet on top of a sot-23 low threshold fet, or something like that.

John

I haven't used these for anything near 100 volts, but have made some very nice, high efficiency switchers with lower voltage versions.

They have a Toff longer than 100ns, but most of that is storage time, not transition time. Some other nice choices:

120VCeo 75VCeo:

--
John Popelish

The specified storage times apply to the specified conditions. Pull charge out of the base faster and it goes quicker.

--
John Popelish

The technique was well established in the days when BJT's were commonly used for switchers. There is a limit to how fast you can pull the charge out the base, set by the resistivity of the base region (which tends to be high) and the distance from the base metalization to the middle of the base volume. Look for a transistor with low rbe and an interdigitated (sp?) base structure.

--
--Larry Brasfield
email: donotspam_larry_brasfield@hotmail.com
Above views may belong only to me.

Hello All,

What is your favorite (cheap) bipolar transistor for switcher applications? I mean something that can do 100V or so, tolerate an amp or more and turn off in much less than 100nsec.

Reason I ask is that lots of low voltage gear cannot drive the gate of a FET high enough to guarantee a resonable conduction. There are some very-low Vgs versions but usually they can't do 100V or so. Same for bipolar, lots of gold-doped RF gems but they cannot stomach higher voltages either.

Regards, Joerg

Hi John,

Thanks. Seems they make some good stuff in the UK. But these transistors have pretty long storage times. That's ok in slow converters but when you want to keep the magnetics small it can become tight.

Regards, Joerg

Hello John,

Sometimes that ain't so easy. Often all you have is a 3.3V supply, or down to 2.5V in two-cell operation. No negative swings that could help in wiping the charge, except for some capacitor tricks. Then you have to drive as hard as the min hfe spec says, plus some margin.

The only option would really be the good old baker clamp while giving up some efficiency for not reaching the famously low Vcesat.

Regards, Joerg

Hi Tim,

Yes, indeed they do. AN22 is really good. The old trick of a cap across the base resistor is in there as well but there we are limited to whatever capacitance the driving chip can sustain. Thanks to the fact that most people use FETs this is usually spec'd. AN22 had another nifty scheme in there that I hadn't seen before, kind of a kickback inductor-transistor scheme. That was really interesting. They've got some clever engineers over there in the UK. Maybe I should have another Tadcaster ale tonight.

Regards, Joerg

In article , Joerg wrote: [...]

A few suggestions:

Disable the Baker-clamp for the first 75% of the on time. If you are making a simple squarewave converter you can do that.

Add an anti-cross-conduction circuit to live with the storage time.

Make a small DC-DC converter to make enough voltage to drive MOSFET gates.

--
--
kensmith@rahul.net   forging knowledge

Or a cascode circuit.

RL

Hi Ken,

That is a good suggestion, it could eke out a few more percent in efficiency especially when running on fairly low voltages. It would add another little transistor though but that one can be a 'penny device'.

75% would be a bit much at todays frequencies. Most of my switchers run at several hundred kHz or even above a MHz. Then when the transistor storage time is >500nsec I've got a wee problem.

It's usually single transistor circuitry, no cross conduction. The concern is that the situation becomes odd when the storage time represents a good part of the desired duty cycle, or worst case could exceed it. Light load is another matter. That would require to enter into a pulse skipping mode and then the whole switcher begins to sound like a wood splitter hitting a knot.

That's what Winfield Hill had suggested in another thread about really low voltage single cell applications. It does make things complex though because most of the time the driving device itself isn't rated for use above 3.3V. That would require a level translator in addition to the helper-SMPS.

Regards, Joerg

"Joerg" wrote

Lets get this straight:

Vce max = 100 volts base/gate drive of 2V Ic > 1A Toff ~ 35nS ~$0.02

And I imagine (why not):

hfe > 200min Vce sat < .2V @ 1A

This sounds like two devices. Or is this a design for a 2-260V input power supply?

At what Vce do you need the 1 amp Ic? At what Ic do you need the 100 volt Vce?

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
Remove spaces etc. to reply: n o lindan at net com dot com
psst.. want to buy an f-stop timer? nolindan.com/da/fstop/

Hi John,

Sorry, should have mentioned that it's not just the available devices that are limited in voltage but usually there isn't much there in supply voltages besides 3V or so.

Otherwise a cascode or something with active level translation would work. Without a decent VCC all there is left would be a step-up transformer or a dedicated lil' helper switcher to create a higher rail but that usually blows the cost budget out of the water.

Regards, Joerg

Hi Nicholas,

That would be nice to have and 2 cents sounds even better.

It's not an actual design, just a rough patch I keep running into at times. Others do, too, for example folks who want to control big stuff with new micro controllers that were migrated to a new process and can't be operated at 5V anymore.

In my case it is sometimes the generation of variable voltages, say,

3-50V or 3-80V kind of like a function generator. All from low battery voltages, ideally just a couple of cells. Well, ideally one cell but that's mostly not in the cards because logic chips become expensive in that range.

Near or at saturation. In continuous current mode it could be a few volts but that can be avoided.

Close to zilch, when turned off.

Regards, Joerg

Well then don't run them so fast.

Have you considered making it a constant on time instead of a constant frequency converter. This way, the timing of the Baker clamp is constant and the switching chips power decreases as the load decreases.

A constant on-time converter avoids this problem. You can make the circuit fade from constant on to constant frequency. A lot depends on the exact chip you are using. If the oscillator connections are brought out, you can play with them. On some where they aren't, the shut-down pin is really a part of the oscillator circuit and clever yanking on it can slow the running frequency.

Unless, you run the main switcher chip on the helper supply.

--
--
kensmith@rahul.net   forging knowledge

In that case, you need a fast, high-voltage, hi-beta bipolar transistor, in my professional opinion.

Hey, how about a darlington? The big guy doesn't saturate so turns off fast. First switcher I ever did was a 24-to-5 buck, with a 2N2905 driving a 2N3055 at 20 KHz, pseudo-PNP connection. The 3055 didn't saturate so turned off quick. Error amp was an uncompensated 709, I think.

John

Dreamer! Don't you realize what happens to device dimensions when voltage goes up?

I haven't been following this thread... really busy doing real work ;-)

What was the original problem?

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Hi Ken,

Sure, but then cost and size of the magnetics increase.

That would indeed be a great way to use a gated Baker clamp. On light loads it'll still behave like a Harley, kathumpah...kathumpah... but the ripple can be muffled with a larger output cap.

Yes, that can be done. Most of my designs are discrete anyway, no PWM chips.

Well, yes if you use an off-the-shelf PWM chip. But that is often too costly and then you are stuck with either logic chips or a micro controller which won't be happy above 3.5V or so.

Regards, Joerg

Hello John,

Yes, that's what I am looking for. But it has to be a 'Walmart' version, not some boutique part that costs through the nose ;-)

Ah, the good old 709. I wasn't much of an opamp fan in those days, mainly because they cost too much and you could design blazingly fast amps for less money around transistors such as the AF126. But I had a tremendous respect for the 709 while everybody else was going for the

741. That would have been like buying a car with an automatic transmission...

Darlingtons are great but in converters where Vin is 3V it is tough to stomach the efficiency penalty, since they can't go below a diode drop for Vce on the big one.

Regards, Joerg