PNPs with high BV_EBO

Peanut gallery -- what about MOSFETs? PMOS or depletion. Usually rated 20V G-S.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website:

There's always germaniums :)

NT

Your circuit arrangement is not clear. According to TI, class G is simply complimentary emitter followers with switched supply voltages. For example, see

PowerWise?< Class G versus Class AB Headphone Amplifiers

Figure 3. Basic Class G Design Using Split-Supply Class AB Output Stage

Figure 4. Music Output Example

In this application, the transistors can never be reverse biased more than half a volt or so.

Your application is completely different. You are not amplifying music, you are supplying power to a resistive load. Why do you need a PNP, and why can it be reverse biased by 12V?

Please explain!!

The HP 547A current tracer probe used a pnp with high Veb in the series pass regulator but I have no idea what part number.

piglet

Seems the HP-545 logic probe used same hi BV-eb pnp part as the 547. Manual lists as 1853-0389 made by HP themselves, good luck buying those!

piglet

piglet

There are a couple of problems. First, I have no negative supply, so it's hard to get the PMOS to turn on as a follower. Second, it's a class B driver, so handling the uncertainty in V_GS is somewhat difficult.

It's not that big a deal to add a diode, but it's not exactly elegant and it steals drive current from the other end of the totem pole. (It has to be a shunt diode because I can't stand the drop of a series diode.)

Cheers

Phil Hobbs

I'd post the circuit except that it's a customer job. The topology is an asymmetrical bridge. One side is a current conveyor with a gain of about 500, and the other is a normal emitter follower totem pole arranged to keep the outputs symmetrical about VCC/2. It has to be pretty small, so there's no space for thermal pours.

The TEC is a very low-Z load, but its potentially large thermocouple voltage makes it a four-quadrant device. That means that a fixed supply voltage for the driver requires a lot of headroom, which leads to a lot of dissipation most of the time. So fixed supplies are out.

A Class D amplifier uses one or sometimes two PWM outputs with LC filters. In small audio amps, the speaker's inductance is often enough to work with, so you can get away with no filter at all. The hash this generates is a real problem with TECs, because there's a lot of capacitance between the TEC elements and the cold plate, so the hash gets into the diode laser bias and causes spurs.

In Class G, you use an agile switching power supply and a linear output stage. That way you can keep the dissipation low while preserving the naturally good filtering provided by the output stage, which works like a capacitance multiplier.

Class G isn't as efficient, but it's a good fit for this job. (To be pedantic it's probably Class H really, but who's counting?)

Cheers

Phil Hobbs

Thanks. Unfortunately the 2SA1615 is obsolete, and anyway it was too big for this part of the circuit. (It's the driver stage of a Sziklai.) A PNP version of the 2SD2704K would be perfect, but there ain't one, it seems.

Cheers

Phil Hobbs

It that only a transient condition?

I remember, from way too long ago, some PNPs that had high Vbe breakdown. 2N2906 maybe? The data sheets say 5 volts, but they are probably just playing it safe.

That means that a fixed supply

Seems like an h-bridge switcher with good filtering would work. Or a single buck switcher with an outrageous filter and then a reversing switch downstream, SSRs maybe. Does having controlled gain in the linear stage help?

You could do a switcher-based control loop and literally add your c-multiplier filters, which would only lose one or two jd's.

The spurious spec is outrageously tight, though. It's the quietness of the driver that's the selling point. (Very low BOM cost doesn't hurt either.) The downside is that there are a fair number of parts.

Yup. I'm using some super-low V_CEsat transistors (ZXTN25020), and have a higher voltage supply available, so I don't even lose the whole junction drop.

One consideration is that this is actually a customization of an existing design, so the NRE budget isn't lush and there's licensing revenue as an upside.

Cheers

Phil Hobbs

Is RMS noise the issue, or spectral lines? Spread-spectrum (ie, cheating) would help the latter.

The thermocouple voltage fights you, so it's a steady-state condition when you're running at large delta-T. It can be a good couple of volts depending on the number of junctions in the TEC.

Probably, but you know how I hate midnight phone calls. ;)

I just increased the drive resistance (which I can do since the Sziklai effectively has a beta of about 50000) and stuck in a BAV99W from base to emitter.

Cheers

Phil Hobbs

Both. This driver is about 30 dB quieter than any commercial one I know of. The spread-spectrum thing would probably help, but I only get one crack at this. New SMPS chips always require a bit of assing around with, because they're all individuals.

I should try doing a SS switcher sometime when things calm down a bit.

Cheers

Phil Hobbs

I like TPS54302 lately. It's well behaved.

The old germanium standby for PNP chopper/oscillator use (which similarly benefits from high BE breakdowns) was 2N404. Fast-forward half a century, and its descendant is CMPT404A, 'PNP Silicon chopper transistor'

VEBO is 25V; Mouser has a few thousand in stock.

Search engines don't do much good on the more obscure qualities one sometimes wants... I found this one by playing a hunch that the old part number would be echoed in its replacement. That was a sound marketing ploy: the part CAN be found that way.

I recently did a 250W (total combined capacity) PSU with under 40dBuV noise (300kHz RBW, 20-300MHz). Just stock parts, solid ground planes, and some extra filtering on the input and outputs.

Nothing very special I suppose, but the customer isn't seeing any trash in their ADCs/DACs/FPGAs, so we're happy.

To do much better I think would require sacrificing efficiency and/or adding shields.

Your application would probably be reasonable in

Interesting, thanks. That's by far the closest so far.

Cheers

Phil Hobbs

This used to be more popular in chopper applications.

Consulting the bipolar database and restricting to smd parts, I see:

- - 200T MAT SOT23 TO236 (SC59) LLG 2Q BC856 - -

- BT404 Si pnp 225mW 25V 24V 12V 150mA 150>C

- 20 30/400 CZH SOT23 LLG 2M - - - CPH6520 Si pnpdual 350mW 30V 25V 15V 150mA 150>C

- KT218A9 Si pnp 200mW 80V 80V 30V 50mA 125>C

- KT218E9 Si pnp 200mW 20V - 20V 50mA 125>C

- MMBT404 Si pnp 350mW 25V 24V 12V 150mA 150>C

- 20 30/400 MOT SOT23 TO236 LLC 2M - -

RL

that table is well scrambled :/

NT