30V 10A power supply keeps blowing 2N3055s

[anacierdem]

Hi,

I'm working on building a current limited bench power supply at http://electronics-diy.com/30v-10a-variable-bench-power-supply.php. 

Schematic;



One thing that I've found so far is the power rating of R7 should be high as it is dissipating power at high output levels.

The major problem I'm having is that I keep frying at least one of the output transistors when testing with a 2.2Ohm load at 20V output.
This should draw about 9 amps when not current limited. I'm using 4 3055s and believe that it must be within spec even if I use a single transistor - with current limiting set at ~3 amps.
Check the SOA at http://www.onsemi.com/pub/Collateral/2N3055-D.PDF. Current unregulated voltage is ~32VDC.

Here is a capture on the scope;



The first channel is the voltage drop across the ~0.24 ohm shunt resistor and second channel is the output voltage after applying the load.
The limiting seems to work nice to me - voltage drops to about 10V after 300us.

BTW after working and blowing some trannies, I put a random inductor to limit instantenous current, here it is for reference;




The scope result is with this inductor in place, and from one of the initial tries before frying a transistor. The inductor does not resolve the problem
but clears irregularites on the scope as expected. I cannot repeatedly apply the load and prevent frying them. Shorting the output immediately
kills transistors ~70% of the time, which I want to prevent.

What do you suggest? How can I protect the output transistors and what is the possible cause of this behaviour?

[danadak]

.1 ohm ballasting R's seem too low a value to ensure current
sharing.


https://electronics.stackexchange.com/questions/77045/transistors-in-parallel


Consider doing spice sim with range of expected Hfe/Beta of 3055's.


Regards, Dana.

[anacierdem]

Thank you for the tip. But the thing I dont understand, this does not seem out of the SOA even for a single transistor at the lowest current setting? Am I misssing something?

[Andy Watson]

I would include a resistor, 200-500 , between base and emitter of the 2N3055s to mitigate against leakage induced thermal-runaway.

[oldway]

This schematic is wrong....If ajusted to upper side of R11, there is no current limit at all....that's a true transistor killer ....

I am waiting you correct the errors and explain what you do....Whe are in the beginner section and you have to learn.....
I will help you...
First at all, current sense (pin 3 of 723) must always be connected to output.
That is already wrong on the schematic.
Second, current limit (pin 2 of 723) must be connected to cursor of R11 to ajust the current limit.

723 is limiting current when voltage between pin 3 and pin 2 is becaming higher than more or less 0.6V (base/emitter conduction of internal transistor of 723).

For measuring current, whe have (R5 and R6) two 0.47R in parallel = 0.235R 
If current reach 2.553A, 723 will start limiting current.....that's wrong, we need 10A current limit.

The resistance must be 0.06R to limit current at 10A...this could be achieved using 3 resistors in parallel...can you calculate this ?

For lower current limit down to 0, whe add to this voltage a positive voltage of 0.7V coming from the 7V reference output (pin 6 of 723).

I thing you must now be able to correct this schematic....

NB: I apology, I can't be laughing of errors of a beginner, I must help him...that's the reason why I have edited my post....

[Jay_Diddy_B]

This schematic is wrong....If ajusted to upper side of R11, there is no current limit at all....that's a true transistor killer 

1+

I would also suggest following actions:

1) Test the circuit with a single 2N3055 transistor at 1/4 of the output current. This will remove the need for sharing and be cheaper if the unit fails.

2) I would remove the top from one of the failed transistors, and post a picture here.



We can tell by the size of the silicon inside if you have a real 2N3055 or a fake.

Regards,

Jay_Diddy_B

[anacierdem]

This schematic is wrong....If ajusted to upper side of R11, there is no current limit at all....that's a true transistor killer ....

I am waiting you correct the errors and explain what you do....Whe are in the beginner section and you have to learn.....
I will help you...
First at all, current sense (pin 3 of 723) must always be connected to output.
That is already wrong on the schematic.
Second, current limit (pin 2 of 723) must be connected to cursor of R11 to ajust the current limit.

723 is limiting current when voltage between pin 3 and pin 2 is becaming higher than more or less 0.6V (base/emitter conduction of internal transistor of 723).

For measuring current, whe have (R5 and R6) two 0.47R in parallel = 0.235R 
If current reach 2.553A, 723 will start limiting current.....that's wrong, we need 10A current limit.

The resistance must be 0.06R to limit current at 10A...this could be achieved using 3 resistors in parallel...can you calculate this ?

For lower current limit down to 0, whe add to this voltage a positive voltage of 0.7V coming from the 7V reference output (pin 6 of 723).

I thing you must now be able to correct this schematic....

NB: I apology, I can't be laughing of errors of a beginner, I must help him...that's the reason why I have edited my post....

OK, let me explain further;

I'm indeed testing the circuit with R11 adjusted to the lower end (full voltage drop on the ~0.24ohm resistor).

I was calculating drop accross 723s limiter transistor drop as ~0.7V thus the current limit I'm expecting is 3A, which I measure as 2.8 A at lower voltage outputs (at about 12V output current limiting works OK).

I'm currenntly not testing the full 10A range but 3A. I'll then move on to higher currents.

I didnt understand "For lower current limit down to 0, whe add to this voltage a positive voltage of 0.7V coming from the 7V reference output (pin 6 of 723)." part of your post, could you clarify further?


I'll draw a corrected schematic later on.

[anacierdem]

Here is the die;

[wraper]

Here is the die;

This is a fake crap.

[anacierdem]

Here is the die;

This is a fake crap.

That explains my problem IMO.

How to identify a genuine one?

[Ian.M]

You *WILL* exceed the S.O.A. if you short the output to ground with the current limit anywhere near max.  A modern genuine 2N3055 is right on the edge of failure at about 3A @ Vce=38V dropping to about 1.5A @Vce=48V, all with the mounting flange at 25 deg C.  The derating curve goes to zero at 200 deg C, a range of 175 degrees, so for every 10 degree rise you need to derate the max Ic at that Vce by 5.72% (call it 6%).  You have stated the Vce is about 32V at Vout of about 10V, which means your unreg DC bus is at about 42V.   IMHO, its right on the bleeding edge before you have *ANY* heatsink temperature rise, and you'll probably need a minimum of 6x 2N3055 to have enough safety margin.


Also, its possible to use the emitter resistors that *MUST* be there for even current sharing as the current sense resistor.  Simply use 10R resistors to join all their top ends to a common point to feed the LM723 current limit pin.

IMHO that die is very suspect.  It looks to have an active area of about the same width as the top of the pins, and genuine modern 2N3055 dies are typically about 50% wider, giving double the area and are marginally specced  compared to N.O.S. 1960's ones which have *MUCH* greater die area.  Also the white silicone is suspect

[anacierdem]

Indeed I have incortectly stated Vce I meant unreg DC is 32 volts. Monitoring temperature to keep it around 30C.

[Jay_Diddy_B]

Here is the die;

This is a fake crap.

I agree that this fake.

A good transistor for this power dissipation will include a copper heat spreader between the silicon die and the case. This is a genuine audio power transistor MJ15025:



There is no easy way to tell which transistors are fake. The only reliable way is to buy them from a trusted source.

For this application you might be better using 2N3771, but again you need to make sure that they are genuine.

Jay_Diddy_B

[anacierdem]

Thank you for all the feedback! I'll not buy from this provider anymore parts.

I think now I have a general understanding of 723 and I have some of them lying around.
Will design a similar but better circuit myself and try to build that with proper transistors.

The reason why I have chosen this design was that it was looking simple, but Murphy is
everywhere As it is a hobby project for me, it will take some time. I will post if I manage to make it work!

Do you have any final suggestions for building a linear power supply? Or should I work on SMPS, which
always seemed much more complex to me. And there will be switching noise right?

[anacierdem]

You *WILL* exceed the S.O.A. if you short the output to ground with the current limit anywhere near max.  A modern genuine 2N3055 is right on the edge of failure at about 3A @ Vce=38V dropping to about 1.5A @Vce=48V, all with the mounting flange at 25 deg C.

When the output voltage is 25V with 32V supply, maximum possible CE voltage is 7V right? this is the point where the transistor fails with 2.2 ohm load on output (~9A CE current when unlimited). This state will only last for about 300-400us according to the scope. But there is also the possibility that they blow after current limiting kicks in which will bring output voltage to ~6.6V (theoretically) at 3A. This makes CE voltage 25 Volts, which is below the SOA limit of 4A. Is my rationale right?

Even we go outside the SOA, 4 of 3055s should allow for some margin. I'm convinced that the fake transistor is the culprit.

[Zero999]

This schematic is wrong....If ajusted to upper side of R11, there is no current limit at all....that's a true transistor killer 

1+

I would also suggest following actions:

1) Test the circuit with a single 2N3055 transistor at 1/4 of the output current. This will remove the need for sharing and be cheaper if the unit fails.

2) I would remove the top from one of the failed transistors, and post a picture here.



We can tell by the size of the silicon inside if you have a real 2N3055 or a fake.

Regards,

Jay_Diddy_B

Are those real?

I thought the die in a 2N3055 is much larger.

[Kleinstein]

There are different version of the 2N3055. The very large die could be a very old one, sometimes called 2N3055H - it's also slower and more robust.

Some manufacturers now seem to use rather small dies and to a certain degree can compensate for this by making the die also thinner and this way get slightly better thermal contact. Still chances are the very small one are fakes, at least I remember a slightly larger die, maybe twice the area.

[Ian.M]

Why not set up one of the suspect transistors on a large fan cooled heatsink in a constant current circuit from your unregulated rail?  Use an emitter resistor and a low voltage adjustable supply on the base, a fuse and an ammeter in the collector circuit, and a voltmeter between collector and emitter. Adjust the base voltage starting from 0V to set it up to be within the SOA (about 80% of SOA Ic limit for the measured Vce) and monitor it till the heatsink temperature stabilises, continually re-adjusting it to maintain 80% of SOA limit after derating for temperature (plot a graph in advance so you can correct quickly), then leave it to soak test.   A genuine one will stand up to that indefinitely. A fake wont - odds are it will blow before you even get it set to 80%. and at that point you should raise a dispute that the supplied transistors don't meet specifications so aren't of merchantable quality.

Once you've got good ones, note that a 2N3055 will blow in tens of microseconds if a transient takes Ic above 15A, even with a low voltage across it.  Make sure your current limit circuit is *ALWAYS* fast and effective.

With max 32V Vce, a 2N3055 is theoretically good to about 3.5A @ 25 deg C (from DC SOA).   Lets assume a maximum heatsink temperature of 80 deg.  Call it 85 deg C at the transistor case.  That's a 60 deg rise, so Ic needs to be derated by 36%, giving about 2.24A.

That's right on the bleeding edge for four pass transistors if you try to set the current limit to 10A with the output shorted.  Five transistors would be safer and six should be very robust.

[Zero999]

There are different version of the 2N3055. The very large die could be a very old one, sometimes called 2N3055H - it's also slower and more robust.

Some manufacturers now seem to use rather small dies and to a certain degree can compensate for this by making the die also thinner and this way get slightly better thermal contact. Still chances are the very small one are fakes, at least I remember a slightly larger die, maybe twice the area.

Yes, even though modern electronics are smaller, the laws of physics still apply which limit how small the die of a power transistor can be.

I've recently destroyed some 2N3055s, in developing a voltage clamping circuit to absorb the over voltage generated by a DC motor drive, when it performs dynamic breaking. The problem was the circuit was working perfectly, then, unknown to me, someone changed the drive settings and increased the motor torque, causing harsher breaking. The problem was solved by using a higher power transistor.  I might crack some of them open.

[Twoflower]

The smaller die was also mentioned to me. Even as the 2N3055 is very old and widely used is there any modern replacement which can be used in circuits like this? Maybe even with a more modern housing, as this might be the mounting easier and probably even the thermal parameters might improve.

[wraper]

The smaller die was also mentioned to me. Even as the 2N3055 is very old and widely used is there any modern replacement which can be used in circuits like this? Maybe even with a more modern housing, as this might be the mounting easier and probably even the thermal parameters might improve.

There are a lot of better BJTs which are more cost effective and I'm wondering why 2N3055 is still used. Not to mention that modern 2N3055s have nothing common with old ones. Much smaller die and overload robustness. Also I doubt that modern 2N3055s are unique designs. More likely manufacturers just take another currently produced transistor with suitable characteristics, place die into TO-3 case and call that 2N3055.

[oldway]

Where did you buy the 2N3055 you use ?
Do not forget to draw a corrected schematic.

[Zero999]

The smaller die was also mentioned to me. Even as the 2N3055 is very old and widely used is there any modern replacement which can be used in circuits like this? Maybe even with a more modern housing, as this might be the mounting easier and probably even the thermal parameters might improve.

I replaced the 2N3055 with the MJ14002G in my voltage clamping circuit and it worked perfectly, but it's much more expensive, possibly out of the original poster's price range. Other possibilities are the 2N3771G or MJ802G.

One thing to bear in mind is that, it's not just the current, voltage and power ratings which are important but the safe operating area. Some high current, high voltage, transistors are designed to be used as a switch, rather than a linear power amplifier and are completely unsuitable for this application.

[bson]

The smaller die was also mentioned to me. Even as the 2N3055 is very old and widely used is there any modern replacement which can be used in circuits like this? Maybe even with a more modern housing, as this might be the mounting easier and probably even the thermal parameters might improve.

2SC5242.  Comes in O and R grades according to h_fe.  The TO-3P form factor is easy to work with, just a single mount hole.

[Twoflower]

How about the 2SC5200RTU? From a fast scanning through the parameters it looks like that would work as a replacement or did I oversee a important parameter? The SOA looks identical to the old guy. And it looks like it's not that expensive like the MJ14002G.