Forums

High voltage pnp's (250 volts)

Started by Hul Tytus April 14, 2020
On Wednesday, April 15, 2020 at 11:33:48 PM UTC+2, Winfield Hill wrote:
> Klaus Kragelund wrote... > > > > Gatedrivers can be had for below 0.2 USD at volume > > > > But the OP is going for a cheaper solution. The > > level shift cap solution is about 10 times cheaper > > He's going for a $0.02 solution? A p-channel part > will more than wipe out that savings, compared to > n-channel. Maybe not compared to bipolar, but I'd > check what LCSC and the Pacific rim has to offer. >
Oh, maybe I am guessing. But using a BJT seems like he is cost driven It's true that a HV PNP will cost more than the NPN. The drive method dictates a PNP. If he shifts to NPN, he needs the gatedriver, or use an old style transformer based driver, but then the price of the high side switch (NPN) is very low, contrasted by high cost for the transformer On the other hand, a 0.5A DP350 PNP is almost as cheap, looking it up https://www.diodes.com/assets/Datasheets/ds30624.pdf That brings the levelshift idea and PNP down to very low cost Cheers Klaus
On Wednesday, April 15, 2020 at 11:50:47 AM UTC-4, Hul Tytus wrote:
> James - I should have used ATTINY13 (Atmel's AVR series microcontroller in an > 8 pin package) rather than AT13. The output, with 3 (cmos) pin's wired together, > is, by spec, +/- 75ma. To alter your schematic a bit, the base of the fzt957 > is connected to the emiter through a 10k resister and a diode. It's connected > to the ATTINY13's output with 1 3300pf cap directly and another 3300pf cap in > series with a 580 ohm resistor. > Along the lines you suggest, checking with a more powerful driver might > provide some insight. I'm hoping though to find a more effective pnp device, > probably one designed as a switcher.
With an oscilloscope you can easily tell where the problem is. If the leading edge is slow, the PNP isn't getting turned on hard enough. If the PNP isn't fully on, d.c. drive. If the falling edge is a ramp, you're not turning it off hard enough. It's not the transistor that's giving you trouble. It's the drive. Cheers, James Arthur
   The FZT957's poor performance turns out to be caused by the device itself. 
Looking through it's datasheet again, I saw the, previously unseen, typical 
rise & fall time spec: 109ns and 2500ns...

Hul


Hul Tytus <ht@panix.com> wrote:
> Thanks - I'll take a look.
> Hul
> dagmargoodboat@yahoo.com wrote: > > On Wednesday, April 15, 2020 at 11:59:06 AM UTC-4, Hul Tytus wrote: > > > > > > dagmargoodboat@yahoo.com wrote: > > > > On Wednesday, April 15, 2020 at 9:26:11 AM UTC-4, piglet wrote: > > > > > On 15/04/2020 2:16 pm, dagmargoodboat@yahoo.com wrote: > > > > > > On Tuesday, April 14, 2020 at 5:29:20 PM UTC-4, Hul Tytus wrote: > > > > > >> I've been ... driving the transistor from 3 pal'd AT13 > > > > > >> outputs through 3300 pf parrallel with 3300pf in series w/580 ohms along with a diode & 10k resistor to maintain bias. > > > > > > > > > > > > I'm having trouble understanding exactly what that looks like, > > > > > > but it sounds way too wimpy to drive a FZT957. Turning it on and > > > > > > off quickly will take a short hefty burst, hundreds of mA at least. > > > > > > > > > > > > Here's my best guess at your drive (I don't know what an AT13 is). > > > > > > > > > > > > (view in Courier) > > > > > > C2 +250V > > > > > > 3.3nF -+- > > > > > > C1 .--||--. | > > > > > > 3.3n | | D1 |<' Q1 > > > > > > AT13 |--||--+--R1--+-->|--+----| FZT957A > > > > > > x3 | 580 | | |\ > > > > > > '--R2--' | > > > > > > 10K | > > > > > > > > > > > > > > > > > > > > > > > > > > > > I took AT13 to mean an Atmel AT-Tiny 13? > > > > > > > That makes sense. So maybe the 2nd 3.3nF is a level-shift? I've > > > > added it (C1) to the sketch. Looks wayyy too wimpy. > > > > > > > > As my earlier reply to Win shows I assumed he was driving the PNP base > > > > > from 5V levels around ground and the load was at negative HV below > > > > > ground but your interpretion also makes sense! I wonder if a simple > > > > > capacitive bootstrap N-ch FET would do? > > > > > > > > > > piglet > > > > > > > There are super nice P-FETs that are easy to drive, without a BJT's > > > > pesky saturation or storage time.
> > > > > > James - about your "super nice p channel fets". Know of any 250volt versions > > > that can be driven by 5 volts ("logic level")? That would be an attractive > > > direction. > > > > > > Hul
> > "Logic level" cuts down the candidates. At Digikey, there's only one (!).
> > https://www.infineon.com/dgdl/Infineon-BSP317P-DS-v02_04-en.pdf?fileId=db3a30433b47825b013b604df1d959f2
> > We don't know enough about your duty cycle and such to know if the BSP317 > > (or two or three in parallel, if you'd rather) will fit your needs, > > though.
> > Cheers, > > James Arthur
Hul Tytus wrote...
> > The FZT957's poor performance turns out to be caused by the > device itself. Looking through it's datasheet again, I saw > the, previously unseen, typical rise & fall time spec: > 109ns and 2500ns...
Hul, an amp at 250V implies quite a bit of power. I trust you're not creating an inefficient design, saving 50 cents in parts cost during manufacture, but costing the user $3 in extra electricity used over its operating lifetime? -- Thanks, - Win
On Thursday, April 16, 2020 at 9:37:19 AM UTC-4, Hul Tytus wrote:
> The FZT957's poor performance turns out to be caused by the device itself. > Looking through it's datasheet again, I saw the, previously unseen, typical > rise & fall time spec: 109ns and 2500ns... > > Hul
But those times are a result of the manufacturer's test setup, not the way we use the devices in real life. Those test fixtures often drive the device with 0..10V through a weenie resistor, with no speed-up cap. In this case, the t.off is spec'd at i.b=-50mA, that's why it's so slow. You should be pulling hundreds of mA out of the base to turn the FZT957 off *hard*. You're going to find, looking around, that the FZT957 is not some poorly-designed geometry that's 1/10th as good as it should be :-). It's a low-saturation device optimized for switching power supplies. But you'll find that making efficient 250V x 500ns x 1A pulses with a bipolar transistor -- or even a FET -- takes a bit of doing. Cheers, James Arthur
On Thursday, April 16, 2020 at 4:12:19 PM UTC+2, Winfield Hill wrote:
> Hul Tytus wrote... > > > > The FZT957's poor performance turns out to be caused by the > > device itself. Looking through it's datasheet again, I saw > > the, previously unseen, typical rise & fall time spec: > > 109ns and 2500ns... > > Hul, an amp at 250V implies quite a bit of power. I trust > you're not creating an inefficient design, saving 50 cents > in parts cost during manufacture, but costing the user $3 > in extra electricity used over its operating lifetime? >
Well, that is an issue I have heard before for sure You need to be competitive, so always doing designs with lowest possible loss is not an option Cheers Klaus
Klaus Kragelund wrote...
> > Winfield Hill wrote: >> >> Hul, an amp at 250V implies quite a bit of power. I trust >> you're not creating an inefficient design, saving 50 cents >> in parts cost during manufacture, but costing the user $3 >> in extra electricity used over its operating lifetime? > > Well, that is an issue I have heard before for sure. > > You need to be competitive, so always doing designs with > lowest possible loss is not an option.
It's one thing to be competitive, it's another to be eeking out every last cent for profit. Klaus, you have a history of designing for low manufacturing cost, but also for high efficiency, higher than most of us manage when we're trying. -- Thanks, - Win
Phil - placed an order today for the mje15035. Thanks for the suggestion.

Hul

Phil Allison <pallison49@gmail.com> wrote:
> Hul Tytus wrote:
> =====================
> > I'm hoping > > there are some pnp devices in the relatively low 1 amp and 2 amp range that are more useful than the fzt957 units. Any suggestions?
> ** The MJE15035 is gonna be hard to beat.
> https://www.onsemi.com/pub/Collateral/MJE15034-D.PDF
> BTW: your Q is all over the place, like mad woman's breakfast.
> .... Phil
Hul Tytus wrote...
> > Phil - placed an order today for the mje15035. Thanks for the suggestion. > > Phil Allison: >> >> ** The MJE15035 is gonna be hard to beat. >> https://www.onsemi.com/pub/Collateral/MJE15034-D.PDF
I have always thought of the MJE15034 and MJE15035 transistors as linear amplifier parts, rather than switching transistors. It'll be interesting to see how they work out for you. Let us know. -- Thanks, - Win
Winfield Hill Nit Picks Me Again:

===================================
> > > > Phil - placed an order today for the mje15035. Thanks for the suggestion. > > > > Phil Allison: > >> > >> ** The MJE15035 is gonna be hard to beat. >> > >> https://www.onsemi.com/pub/Collateral/MJE15034-D.PDF > > > I have always thought of the MJE15034 and MJE15035 > transistors as linear amplifier parts, rather than > switching transistors. >
** Same goes for most of the parts suggested here.
> It'll be interesting to > see how they work out for you.
** Which is totally irrelevant to the doubt raised. With near constant Hfe of 200 up to 1A, typical Ft of 80Mhz plus Vce sat of 0.2V at 1A falling to and 50mV at 100mA - you never know, it might just work out OK ? Want another fight over mS v ms ? Can you spell the word " pedant " aloud please ? ..... Phil