Anyone had experience with these power transistors?

[bcarso]

Some time ago On Semi announced some power bipolar parts with in-package isolated diodes. As transistors they are identical to the popular "Motoshiba Toshirola" 2SA3281 and 2SA1302 devices. The TO3P package has been modified to incorporate two more leads for the diode.

The part numbers are NJL1302 and NJL3281.

A fellow consultant sniffed at these because he said Why didn't they make some driver parts as well? which I thought was a trifle uncharitable at the time. I suppose you could use them as the driver parts as well, for that matter, although it's a lot of silicon for any but really high power apps.

Anyway, I see they are still being made, and I have an inquiry in to their tech support about some questions (like what is the degree of electrical isolation between the diode and transistor, and max differential voltage), but in the meantime I wondered if anyone in here has had experience with these puppies.

Used properly they should allow a significant improvement in amplifier performance, permitting one of the bugaboos of class AB design to be handled well, namely the time lag between the device die temp and the biasing transistor and the resulting distortion and/or destruction.

Although I'm getting increasingly comfortable with the fidelity of really good switchmode, I think there's still a place for linear.

 

[chrissugar]

I don't have any experience with them, but reading the recent blog at www.tubecad.com I thought it would be great to do some tests because the concept is very good for amp design.
http://www.tubecad.com/2006/10/blog0084.htm

chrissugar

 

[bcarso]

Thanks Chris. Indulging in an off-topic digression (hey I started the thread!), I like his discussion of the friend who tied the AX7 grid to B+, following an egregious schematic error:

"After I had rewired the amplifier, my friend told me that the resistor had to attach to the B+ and that the schematic said so. He was right; the schematic held one nasty typo. It actually took quite a bit of persuading to get him to go along with the correct wiring; such is the power of the printed page. In the absence of understanding, only pleas based on authority can persuade. On the other hand, if he had understood how the amplifier topology worked, he wouldn’t have needed my help. "

"In the absence of understanding, only pleas based on authority can persuade." How very very true. I may have this laminated and hung on the wall.

It reminds me of the situation with an engineer I was shepherding a bit through a design, one of her first efforts to do something beyond the databook app notes and their ilk. The part I suggested she use was the dual OTA from pre-Caen-fire Philips, their P/N NE5517 (essentially equivalent to the LM13700).

I would assure her that the part would function down to quite low voltages, and would walk her through the schematic for the part, discuss base-emitter voltage drops, saturation, voltage swings, etc. etc., and she would nod her head as if understanding (she was Chinese, with a seemingly good command of English, but I've found there is sometimes a tendency to nod knowingly without real comprehension).

Then I would say So---it will work in this circuit with a 5V supply if we use it in this fashion, right?. And she would get a disturbed look on her face and say But the datasheet says +/- 15V. :roll:

Finally I just told her to do it, and I would accept responsibility if it didn't work. I also got a verbal assurance from someone at Philips who knew how transistors worked, but it was clear that, for her, the power of the printed spec was prevailing.

 

[bcarso]

Well so far no response from tech support. I would think that a possible 50k piece annual usage would get someone's attention. But maybe they didn't know the answers to my questions. It wouldn't be the first time.

 

[Svart]

It's likely tech support doesn't know. I've contacted various manufacturers on a number of parts, especially this elusive schottky bridge rectifier that solved many problems for me and a motor controller design... Which seemed to only exist on paper as absolutely no one had them in stock and the manufacturer couldn't even tell me if they were being made anymore.

It was like the twilight zone!

Distributor: "oh yeah we have those, I see 20 right here in the computer" Me: " can you please check the stock manually for me?"
Dist.: "sure hold on."

wait....
elevator music...
more waiting...
more Muzak...
more waiting..

Dist.: "hello sir? It seems that we don't have any.."
Me: "are you sure? what about the stock listed in the computer?"
Dist.: "well I went to where the parts should be and it's empty"
Me: "the box was empty?"
Dist.: "well, no, the whole box and everything is gone."


Are these parts integrated on die or are they just stuffed into the same housing? It would be nice to have them on die like most MOSFETs.

I still try to find a way to use IGBTs where needed though. Would something like that work for you or do you need the base's attributes?

 

[Guest]

I use a fast recovery diode right on top of the driver transistor (a differential pair transisrot + diode), this device would fit better. However, no need for such currents in drivers, so smaller dies would fit perfectly.

 

[bcarso]

Still no word from ON... I'll just buy some and see wtf is up I'll bet there's just an isolation diffusion tub, so probably a hard limit to the delta V of diode and transistor.

IGBT's would be a problem since the threshold voltage is all over the place compared to bipolars, and the tempco of drain current is different and dependent on operating point.

Sanken also has some interesting parts I was put onto by Duke Aguiar yesterday. They are integrated Darlingtons with temp comp diodes. But they pretty much have to be used in complementary pairs, as they have found it easier to fab the schottky diodes in one part and the regular PN in the other. Adjustment resistors externally are mandatory, and only the common-collector connection (complementary e-follower) is practical at first blush, unless you want to go to a whole lot of trouble.

The part numbers are SAP09, SAP10, and SAP16, with a P or N suffix for the polarity. The diodes are tied on one side to the base internally. They also have 0.22 ohm R's in the emitters, but you can get direct access if you don't want to use them. They warn you that the resistors are likely less robust than the transistors.

 

[bcarso]

[quote author="Wavebourn"]I use a fast recovery diode right on top of the driver transistor (a differential pair transisrot + diode), this device would fit better. However, no need for such currents in drivers, so smaller dies would fit perfectly.[/quote]

Thanks. I was revisiting Self while looking for a Linsley-Hood book I've misplaced and noticed his remarks about sticking the bias Q on top of TO3 packages, claiming a much shorter thermal tau compared to adjacent on the heatsink (wonder if he tried boring a hole and sticking it underneath?).

 

[Svart]

Still, putting a diode on top of a TO3 is barely tracking the true temp of the die, right? you are really only tracking the average temp of the whole case and heatsink. This opens up a whole new set of problems and attributes like heatsoaking, which can both hurt and help. If the die needs to be cool to be stable and the heatsink has inadequate cooling, then the mass of the metal will keep the whole mess hot for a long time. However, this can also be used to keep a more constant average temperature too.

Then again, I use MOSFETs for most of my switching and rarely use BJTs for power so what do I know..

I suppose I would thermally bond the diode to the bottom of the T03 case, likely as close to the housing as possible, at least keeping the diode on the same metal as the die.

 

[Guest]

[quote author="bcarso"]
The part numbers are SAP09, SAP10, and SAP16, with a P or N suffix for the polarity. The diodes are tied on one side to the base internally. They also have 0.22 ohm R's in the emitters, but you can get direct access if you don't want to use them. They warn you that the resistors are likely less robust than the transistors.[/quote]

If they blow up like very fast fuses it is Ok, otherwise I don't need such resistors :green:

 

[bcarso]

Finally got a response---the responsible party is out of town until next week.

 

[Rossi]

I feel your pain. There's a guy in a microphone company who's actually responsible for PR and questions from press guys like myself. He only responds when he's away. Well actually the mail robot does it for him. Fortunately I know some other guys in the company who are not responsible for my questions but answer them gladly. The engineers behind the products are always eager to talk about their creations. :thumb: Drop the middle man!

 

[bcarso]

Well, now they've replied to a friend who asked the same questions, prompted by me, later than my request, and they have still not gotten back to me yet. I wrote them a slightly testy one, speculating that the friend's ieee.com email addy might have gotten priority, and that I was "underwhelmed."

 

[bcarso]

Finally got a repsonse just now. The thermal tau is "about" 100ms (fast compared to other means of temp sensing) and the withstand voltage between diode and transistor is >600V. They did not disclose whether the diode is on the same chip or not, and I requested I-V curves. (EDIT: the I/V curves are actually lurking in the datasheets---just not captioned very clearly. The diode looks like a 1N4936, about, i.e., a nominal 1A fast-recovery part).

Interesting parts. I think I will do something with them when I get a good excuse.