Safe operating area

This does a miliamp:

A proper flyback supply with UC3842 would do a fine job, or if low noise is required, it could be followed with a FDPF2N50 and error amp in a circuit similar to this one.

Tim

Is this for a series regulator? That only needs to drop from the filter capacitor to the output node, so a 300V regulator with 350V input only needs a 50V transistor operating point. It's likely that your normal operation won't get near the SOA limits anyhow.

If they don't say then chances are you can construct your own from the max collector current, the max collector-emitter voltage, and the max power dissipation (paying close attention to heatsinking, which has a strong effect on power dissipation).

I think most Spice-based simulators can take a .TRAN analysis (voltage and current versus time) and change axis variable to get a real time V-I locus.

If you're into such play toying, as I am, you can also plot a template overlay, from the device data sheet, onto the simulation plot:-) ...Jim Thompson

OP asked for 0-300VDC, that's different?

Although one could pre-regulate something like Tim's inverter over a range, say 100 to 320V and post regulate with a linear stage?

That would reduce running V across the output transistor.

Grant.

I just looked at three hi V transistors from online catalogue and two had SOA curves. A lot of the transistors weren't packaged well enough for heatsinking. I don't think there's a big problem finding suitable transistor?

Nice circuit, maybe I could spend a bit more time winding own transformers for this type of thing? More fun than plugging components values into an app. note circuit?

In another post I suggested pre-regulating the inverter, sorta like LCD backlight inverters do to regulate the CCFL current.

Seems most CCFL controllers use a common circuit one finds on the Internet -> regulate V+ into centre tap of a 2 transistor PP inverter driving hi V xfmr, with a tank cap across collectors. Sense output current (voltage in this case) to control the inverter V+.

One might follow the inverter's rectified DC with linear regulator for quieter output?

Grant.

Except under short circuit conditions. You generally want current limiting in these circuits, so you generally want to stay out of avalanche, too. The SMPS is intrinsically slower than the postreg, so it'll be a few milliseconds before it realizes, oh hey, you don't need all this voltage.

If your load can tolerate peaky currents, it would be possible to go with avalanche anyway, being careful that the transistor can tolerate that overload.

Tim

Use a voltage multiplier diode string instead... derive the AC from a high freq source so you can use small caps in the string..

P.S. you'll need high speed switching diodes of course..

Lots of HV regulators work that way. The coarse adjust on the old Fluke (?)410 was a switch on the transformer taps, because the regulator tubes would overheat or generate X-rays if you ran their voltage high. Semiconductors also overheat, but second breakdown (thermal inhomogeneity across a broad base) is mainly a problem in high-current transistors, maybe a 'few tens of milliamps' means a small enough emitter area that there's no significant SOA limit.

The input-V selection doesn't need to be regulated, just adjusted.

IIRC, the 2N3439 in my example was good up to a couple mA at the maximum voltage used, and that that was much less than rated dissipation, but since I only needed 1mA, that's fine.

Tim

Just looked up the 2N3439; at DC, it's good for 200 mA at 50V (with infinite heatsink). The 2 mA point on the SOA curve is nearly full rated Vce.

It was a 1A transistor, 450V rated. That SOA limit is FIERCE.

I guess this is why most HV regulators use stacked transistors or scale the input voltages to the pass element.

Hi Tim, whit3rd and all others,

When I look into the datasheets of NXP's high voltage transistors (300...400V types like BF620, BF720, BF820, BSP19, BST39...), I see that none of them contains data about the SOA. Do you think that I can just calculate the max. DC collector current from the power dissipation (considering heatsinking, as you mentioned) divided by Vce? Is it really that simple? What about second breakdown? Is this effect not an issue any more in a modern transistor? Is this the reason why the datasheets do not contain these SOA diagrams any more? I am not absolutely sure about this, but I vaguely remember that "in former times" almost every transistor datasheet contained such an SOA curve. Are modern transistor so much different from the older ones?

Regards, Christian