High power transistor

[Cliff Matthews]

this is not "how far can a push a transistor" its how much more head room can i put between what i need and the transistors maximum output. So for example, if i need 15 Amps continuous, its handy to know i've got a maximum of 30 Amps.

The mj2955 does the job well, but i'm looking for a transistor that has less voltage sag, is rated to dissipate more power (not to max it out) but for greater head room and more efficiency.

when i tried a high power darlington, when it wasn't oscillating, its efficiency was good, it ran cooler, no voltage sag under load etc. But it did oscillate periodically. That was a 50 Amp collector darlington, so all i'm after is better head room for the same current use. The two i've found are the 2N5684 pnp and the MJ4502 pnp transistor. all i'm looking or, is are they viable options for better head room than the MJ2955.

These are the issues I want to know as well. That upper graph shows the Hfe sag way down (after 5 amps especially), but it regains when the heat goes back up. Wrapping my head around all the parameters is difficult at my age.

[Cliff Matthews]

I hate to state the obvious with 8 transistors in parallel you have 8x the reserve... No single transistor can do that, and as I said, thermally, there is NO single transistor that can dissipate as much as a few in a parallel...

Best case to junction numbers I've seen was maybe 0.5W/K, with few TO220 transistor in parallel you can do better than that...

I think the OP wants simplicity in the circuit, without all the ballast resistors etc..

[davelectronic]

I'm fine with ballast resistors, i'm after efficiency. This is going to be a taste it and see thing i think. Thanks for everyone's help and input. 

[David Hess]

I'm fine with ballast resistors, i'm after efficiency. This is going to be a taste it and see thing i think. Thanks for everyone's help and input. 

If it is a linear regulator, then its efficiency is completely defined by its input and output specifications.  The power difference has to be dissipated somewhere and the temperature rise of a given amount of air flow is completely defined by the power.

[2N3055]

I hate to state the obvious with 8 transistors in parallel you have 8x the reserve... No single transistor can do that, and as I said, thermally, there is NO single transistor that can dissipate as much as a few in a parallel...

Best case to junction numbers I've seen was maybe 0.5W/K, with few TO220 transistor in parallel you can do better than that...

I think the OP wants simplicity in the circuit, without all the ballast resistors etc..

Well I'm sorry to be party pooper but at 30-40 A there is no simple and reliable in the same box.. Actually paralleling is as simple as it gets..
For instance, since you run individual transistors at lower currents, you will have less of a HFE nonlinearity too.. 
Of course, it's engineering, everything has pro et contra...  That is why I suggested designs that already work well...
That is simpler than inventing something from the scratch..

Best way to do this is with a preregulator, or if there is no need for super quiet source, to go switcher route..
And that is even less simple..

[davelectronic]

Earlier successful attempts with TIP2955.'s and 7812 regulator. The 4 x on a cpu heatsink got up to 72°C with 50% duty cycle use at 10 Amps, just over. I will try these two other transistors, first as a single, then in parallel. If they don't oscillate i'm hopeful for better efficiency.

[davelectronic]

I don't  want a single transistor to take the place of x4. I'm looking for like for like improvement and efficiency. So replace 1x mj2955 for mjxx and 4x mj2955's for 4x mjxx where xx is the unknown at this stage. See maybe what i'm getting at !

[davelectronic]

I hate to state the obvious with 8 transistors in parallel you have 8x the reserve... No single transistor can do that, and as I said, thermally, there is NO single transistor that can dissipate as much as a few in a parallel...

Best case to junction numbers I've seen was maybe 0.5W/K, with few TO220 transistor in parallel you can do better than that...

I think the OP wants simplicity in the circuit, without all the ballast resistors etc..

Well I'm sorry to be party pooper but at 30-40 A there is no simple and reliable in the same box.. Actually paralleling is as simple as it gets..
For instance, since you run individual transistors at lower currents, you will have less of a HFE nonlinearity too.. 
Of course, it's engineering, everything has pro et contra...  That is why I suggested designs that already work well...
That is simpler than inventing something from the scratch..

Best way to do this is with a preregulator, or if there is no need for super quiet source, to go switcher route..
And that is even less simple..

To step outside the box and try unknown scenarios.. And switching would present issues, in some circuits. Only because i want it for powering RF equipment.

[2N3055]

In that case,  experiment away and have lots of fun doing it... 

At 30-40A eye protection is recommended... 

[davelectronic]

In that case,  experiment away and have lots of fun doing it... 

At 30-40A eye protection is recommended... 

Cheers buddy

[Seekonk]

Safety in numbers.   Even if one component is 40A, the SOA of that device is just about the same as a lower current device in a linear use.

I had a military project and the presentation by the Air Force had a bullet....Radio Shack Compatibility.   I always liked that.  There was nothing in it from Radio Shack.  It just meant that it used all off the shelf stuff.  Don't like using transistors that are exotic.

[David Hess]

Take a look at the configuration of the larger Astron linear supplies.

The 37 continuous amp RS-50A uses 8 (!) x 2N3771 (30A 150W) transistors in parallel divided onto two different heat sinks with 4 transistors each.  The 57 continuous amp RS-70A does the same thing but includes forced air cooling.

And that is *with* foldback current limiting and a transformer secondary optimized to supply the optimum unregulated DC voltage.  At 37 amps, the input filter capacitance of 70k microfarads allows about 4 volts of input ripple so the unregulated peak voltage is about 22 volts DC; I can measure it if someone is really interested but I think this is a close enough estimate.  My guess is that power dissipation is then about 250 watts lost for 500 watts out or 66% efficient.

[David Hess]

In that case,  experiment away and have lots of fun doing it... 

At 30-40A eye protection is recommended... 

One advantage of the metal can TO-3 package is that the destruction is contained.  Plastic packages tend to explode producing shrapnel which sticks in things like the ceiling, my face, etc.

[davelectronic]

I'm quietly looking for useful old new stock to use. Another reason i look for dated parts. Herein the uk the lm317k is expensive,  i found a few for cheap, linear technology manufacturer, £13 for x10 bargain i'd say. And that's it really,good hardy cheap usable components, mostly semiconductors. I got bitten by cheap Chinese fakes, so just use a handful of trusted sellers I'm ok with.

[David Hess]

Also do not forget the LM350 which is a 3 amp LM317.  The LM338 is the 5 amp version.  All are or were available in TO-220 *and* TO-3 packages.  You might run across either cheap.

At some point I picked up a tray of TO-3 packaged and a tube of TO-220 packaged LM337s which are the negative 1.5 amp version of the LM317.  They are more finicky about stability but can use an external NPN instead of PNP pass transistor.

[Cliff Matthews]

** Not High-jacking thread, but in #20, I was asking for light reading the top MJ4502 graph. Is this a percentage of rated typical hFE?

[2N3055]

** Not High-jacking thread, but in #20, I was asking for light reading the top MJ4502 graph. Is this a percentage of rated typical hFE?

Yes that is correct..
Typical rated hFE is specified at certain operating point, this is spread over whole range..
And it also shows the spread over temperature...

[davelectronic]

Not wanting to start another thread...
As its the same type of topic.

Does anyone know how i can stop a single mj11015 darlington power transistor from oscillating intermittently. This is in a 7812 circuit, and the above trasistor is the pass element.  The hfe is high at 1k. Any thoughts on this ? Because if i can cure the oscillations, it would be a rock solid transistor with no voltage sag under load.

[Cliff Matthews]

You mention oscillations often, so is it measured on a scope? At what percentage of load at what freq?
Is this not just a case of the 7812 cutting out because the pass device is not carrying enough of the load?

[CJay]

It's essentially the simple truck dropper circuit, crude and not particularly reliable but it does work and can be useful, you gain thermal protection if the pass transistors are thermally coupled to the 7812 for instance.

Not sure what you mean by 'efficiency' though, no matter which transistor(s) you use they will always have to 'waste' the same amount of power as heat so there's no efficiency gain by using a different transistor.

Definitely investigate the LM723, there are a ton of very useable designs out there, PW Marchwood is a good one for radio work and is not prone to bursting into oscillation (as being discussed on another thread here, the 78xx series can have all sorts of issues)

(And I think we've discussed this on another forum, TM1?)

[David Hess]

Not wanting to start another thread...
As its the same type of topic.

Does anyone know how i can stop a single mj11015 darlington power transistor from oscillating intermittently. This is in a 7812 circuit, and the above trasistor is the pass element.  The hfe is high at 1k. Any thoughts on this ? Because if i can cure the oscillations, it would be a rock solid transistor with no voltage sag under load.

Make sure you are bypassing enough current around the Darlington for the 7812 with an external base-emitter shunt resistor.  Add a base series resistor.

I might implement the circuit as shown in the National Semiconductor application note but with 2 diodes in series to compensate the 2*Vbe of the Darlington.

[davelectronic]

The circuit with the mj11015 was put together in simulation on another forum by other members,this was after i built it to fault find. Several members did this in simulation and it showed no issues. And one member was building it, to establish the problem, he hit on the same problem. Eventually he emailed a friend that worked for linear technology, but i heard no more after that. So thought I'd throw the question out here. The only thing i could work out was maybe it was that transistors massive gain of 1K that might have been the issue. It was always when loaded the voltage jumped about a bit, but only a volt or so up then back down to normal. It didn't do it continously. Unfortunately i don't have the use of a scope yet. I know your going to tell me it might not be oscillations, but the general feeling in the other forum and by multiple members thought it was. Thanks for the replys and help from everyone in this thread.

PS. Just a high to a CB good buddy CJay (-;

[dmills]

And as for efficiency, it is a LINEAR REG, to a first order (Vin - Vout) * Iout, maybe plus 10% if your devices have really low beta.

For me, I would never try to find a higher power device in that class, parallel devices with small current sharing resistors is a much better way to go (Thermally, and in terms of effective SOA), but then I probably would not be trying a linear reg at that current level anyway.

Regards, Dan.

[David Hess]

The circuit with the mj11015 was put together in simulation on another forum by other members,this was after i built it to fault find. Several members did this in simulation and it showed no issues.

There is enough variation in transistors and regulators that a simulation would likely miss the marginal cases.

I only used a Darlington for this once and since it was just a quick hack and worked, I did not bother to do stability analysis.  I was also very conservative about the external base-emitter shunt resistor and external emitter resistor to control open loop gain.

[davelectronic]

I'm going to exaust all options with viable transistors, then move on to the LM723 as a circuit option for a linear psu. Finally end up building a permanent units for HF radio use.