LM317HV With Pass-Transistor.

[neo]

I put this into beginners because this is a semi stupid question.
Originally mention in this thread, https://www.eevblog.com/forum/beginners/variable-regulator-pass-transistor/
Was this circuit,


What i want to do is replicate it, I've attempted it but i fried the 2N2907, i figured out why, it was a mechanical fault, and so i intend to try again.
I want to use the LM317HV because i have a massive, the thing takes up a couple square feet, 7 amp transformer which when rectified to DC gives me 55 volts. This eliminates the LM338K (which i have) leaving me with only this as an option.

Toying with just the transistors in LTSPICE, that is to say omitting the LM317 because i cannot find a way to simulate it, the circuit works acceptably using the 1 ohm resistor in series with the LM317 and .15 ohm resistor to the transistor.
That gives me up to 4.5 amps out with safe margin, if i connect another 2n3055 in parallel to the first one but with the collector going through another .15 ohm resistor, this would give me 5 amps out while keeping the LM317HV well within it's bounds.

My questions are;
Will this work reasonably well for what it is?

Have i overlooked something? Particularly something that might lead to more magic smoke?

My aim here is not for perfection, though that would be nice, but instead to get 45V out. That leaving overhead for the resistive drop, the LM317HV drop and any other losses in the circuit. Can this circuit with my proposed modifications do that?

What about at lower voltages? Will the voltage drop inside the LM317HV, and it's wattage limits, be a problem? If so i have a variac i could use on the non-adjustable power supply feeding this circuit.

I have a handful of 2n3055 on hand,
hFE=110 Vf=534mV
hFE=110 Vf=534mV
hFE=50   Vf=608mV
hFE=63   Vf=495mV
hFE=25   Vf=500mV
hFE=54   Vf=515mV
hFE=297 Vf=525mV
hFE=449 Vf=505mV
I'm thinking the top two, because of how closely matched my transistor tester says they are. But they also have much lower gain than the bottom two 

Thank you to anyone who has read this far, please and thank you for any help.

It might be an important note that this isn't about practicality as much as it is using what i got and learning from it.

[Paul Price]

Your proposed circuit is poor one to model to accomplish your goal:
(1) The 2N2907 transistor will always be destroyed by excessive  base to emitter  current if there is only a short overload.
(2) If you want to parallel the pass transistor, you must also demand more drive current from the 2N2907 already on a suicide mission.
(3) The pass transistors would better  be PNP power types or else  you will won't be able to balance their current share with emitter swamping resistors.
(4) Once the 2N2907 shorts out, the 2N3055 would likely quickly fair with a short also and likely fry any circuit you are powering with this supply as well.
(5) Excessive voltage drop across your circuit shown will cause it to quickly overheat and shut down if you choose to select a low voltage output and attempt to operate it at high currents. Massive heatsink required.

Read about emitter swamping resistors for current matching with  pass transistors.
Study  the examples given in the LM317 datasheet that show how to correctly  boost current output with PNP and NPN pass transistors.. Simpler circuits, well designed, often work better than more complex part count solutions.
Study the LM340T datasheet showing circuits using a PNP pass transistor for power boost.

[Ian.M]

LTspice will happily simulate conditions that would let the magic smoke out of a real circuit, and one of the essential things to do before building a simmed circuit is check worst case dissipation.

Your idea will certainly let the magic smoke out if the output is ever shorted.  Take a look at the 2N3055 S.O.A graph (the one in the OnSemi datasheet is probably clearest).   Examine the DC line and note the  40V, approx 2.8A point.   Above 40V the current capability falls off rapidly due to secondary breakdown, and by 55V, a 2N3055 can only handle slightly over 1A, and only if you can keep its mounting surface at 25 deg C.    A 2N3055 really isn't a good choice for voltage drops much over 40V.

The Sziklai pair emitter current is limited by R4.   Worst case: the LM317 can pass 3.4A (but not for long - its thermal limiting will reduce the current rapidly), and there will be about 4.4V across R1+D1.   Less 0.7V Vbe drop in Q2, this gives about 3.7V across R4, and a current of about 8A, the bulk of which is passed by the 2N3055.   The 2N2907 only has to handle about 8% of it, less than its 600mA Ic_max.   However it doesn't have a high enough power dissipation rating to handle that for long so a beefier driver transistor is needed.   Whatever you use is going to need to handle the 2N3055 base current at the current limit, so needs a DC SOA rating suitable for 10% of the 2N3055 emitter current at the full input voltage.

Also,at 5A out, you risk smoking the transformer - A secondary feeding a bridge rectifier + reservoir capacitor is only good for a continuous DC current of 62% of its RMS AC rating, so your 7A transformer is only good for 4.3A out.  In practice a 15% overload isn't going to cook a big transformer like that quickly, so you'd probably get away with it for half an hour or so.  However if you left it running at full load current overnight e.g. charging a big Lead Acid battery, odds are you'd burn it out.

To simulate a LM317, you need a model for it.  I've attached the one I use.   Unpack the .asy and .lib to the same folder as the .asc sim you want to use it in.

N.B. there is a flaw in that boosted LM317 circuit even at low input voltages - if you make a bad choice of PNP transistor and diode, its possible for it to increase the minimum load current to regulate above the 10mA required by the LM317, provided by the divider resistors feeding Adj.  If so, add a resistor of between 50R and 100R across the 3A diode to delay the transistor turn-on.

[David Hess]

Current through the base-emitter junction of the 2N2907 drive transistor needs to be limited.  The 2N2907 is small for driving a 2N3055; in the past, an Ic=4A Ib=2A 2N3054 would have been used to drive a 2N3055 but of course you need a PNP part.

I like the idea of using a separate PNP drive transistor for each 2N3055 but maybe it is better to acquire some power PNP transistors to make things simpler.  In the past, silicon power PNP transistors were much more expensive than silicon power NPN transistors so the more complex circuit made more sense.

Note that a minimum controlled output capacitance is important for good transient response.  Do not use a giant low ESR output capacitor.  I would not use a small low ESR capacitor either; they belong at the load if necessary.

To protect whatever the circuit is powering, the transformer, and against fire, I might add an SCR crowbar circuit triggered by excessive output voltage to blow a fuse at the input.  This would activate in the event of a pass transistor short.

SPICE is not going to be very useful for modeling the thermal effects in this circuit.  If the 317 is properly thermally coupled to the pass transistors, then its built in thermal and safe operating area protection can protect them as well.  Note that properly in this case as described below is not the same thing as maximally coupled.

Base-emitter shunts should be added to all transistors.  Did you get the idea for this circuit from one of my posts which mentioned it?  My example includes base-emitter shunts as shown below from the National Semiconductor 1980 Voltage Regulator Handbook.

[neo]

I got the idea from one of Ian M.s posts, but like most my ideas it is brilliant on paper but completely useless in practice.

[Ian.M]

I have a handful of 2n3055 on hand,
hFE=110 Vf=534mV
hFE=110 Vf=534mV
hFE=50   Vf=608mV
hFE=63   Vf=495mV
hFE=25   Vf=500mV
hFE=54   Vf=515mV
hFE=297 Vf=525mV
hFE=449 Vf=505mV
I'm thinking the top two, because of how closely matched my transistor tester says they are. But they also have much lower gain than the bottom two 

I'm also very suspicious of some of those transistors' h_FE readings.  The last two have far far too high measured h_FE  and are probably defective - either fakes or leaky.  The one with h_FE of 25 is probably also no good as well, assuming the same <1A test current.

The first two,  h_FE 110 are probably the real deal - their gain is right where it should be for a current for Ic of a few hundred mA according to fig 3 in the OnSemi 2N3055 deatasheet.   The remaining two are also probably OK.

It would help to know what test current your h_FE tester uses for power devices.

However none of them are going to be much use with 55V unreg in as mentioned earlier, unless you use a far more complex multi-stage regulator circuit.

Probably the best option to do something useful with your transformer would be to get a DPS5005 DC input PSU module.

[neo]

I have a handful of 2n3055 on hand,
hFE=110 Vf=534mV
hFE=110 Vf=534mV
hFE=50   Vf=608mV
hFE=63   Vf=495mV
hFE=25   Vf=500mV
hFE=54   Vf=515mV
hFE=297 Vf=525mV
hFE=449 Vf=505mV
I'm thinking the top two, because of how closely matched my transistor tester says they are. But they also have much lower gain than the bottom two 

I'm also very suspicious of some of those transistors' h_FE readings.  The last two have far far too high measured h_FE  and are probably defective - either fakes or leaky.  The one with h_FE of 25 is probably also no good as well, assuming the same <1A test current.

The first two,  h_FE 110 are probably the real deal - their gain is right where it should be for a current for Ic of a few hundred mA according to fig 3 in the OnSemi 2N3055 deatasheet.   The remaining two are also probably OK.

It would help to know what test current your h_FE tester uses for power devices.

However none of them are going to be much use with 55V unreg in as mentioned earlier, unless you use a far more complex multi-stage regulator circuit.

Probably the best option to do something useful with your transformer would be to get a DPS5005 DC input PSU module.

It is just a cheap Chinese tester. The last i bought new from arrow, but i may have broken it, the rest have been kicking around since Methuselah was a kid.

[Zero999]

Why not use the ordinary LM317 and a cascode? I haven't done a detailed simulation, but it should work.

[neo]

Tip31C and 32C, any chance those are not very specific?

[David Hess]

Tip31C and 32C, any chance those are not very specific?

There is nothing special about them.  They have the voltage and current specifications to make good drivers for a 2N3055.  I might use the 2N5192/2N5195 or BD441/BD442.

[neo]

Tip31C and 32C, any chance those are not very specific?

There is nothing special about them.  They have the voltage and current specifications to make good drivers for a 2N3055.  I might use the 2N5192/2N5195 or BD441/BD442.

Thanks, i think i might have some 2n5192/2n5195s

[Zero999]

Tip31C and 32C, any chance those are not very specific?

There is nothing special about them.  They have the voltage and current specifications to make good drivers for a 2N3055.  I might use the 2N5192/2N5195 or BD441/BD442.

Thanks, i think i might have some 2n5192/2n5195s

Those look reasonable to me.