Mastech HY5002 power supply died. Worth fixing?

[Artlav]

For the second time a power supply dies on me.


First time it was HY3005 - 30V, 5A linear power supply.
That one worked for a few years, before dying with 50V across it.

Then i got a HY5002 - 50V, 2A. More volts, less current.
Totally expected it to last a few years or more, since it should have less heat and stress...
And then it died less than a month later, giving out 80V!
The silver lining is that this time it was honestly reporting that it gives out 80V, instead of showing ---.

So,
is this thing worth repairing?
That's about the only sort of power supply that i can buy here, others with comparable specs being almost an order of magnitude more expensive.
Anyone have experience with these or making them more reliable?

I'll appreciate any info on this thing, schematics, common failures, etc - these seems to be common all over the world, so someone must have fixed one before.
Haven't tried diagnosing it yet since it's past midnight here.
At the time of a failure it was set to 35V, and there was a brief short circuit.

One more thing - the first, 30V 5A one, had a loose bolt rattling inside when i bought it.
Ominous sign about quality...

[grumpydoc]

Probably (one of) the pass transistors has a CE short, might be an easy fix

[poorchava]

Easy fix.  Manufacturer has overstressed the output transistor(s) to use cheaper ones.

What pass transistor is used in this one anyway?

[Artlav]

TLDR version: Pass transistor dead. Exact replacement unsourceable. Which transistor parameters matter in a linear supply?

Long version:
Came across a schematic of a similar unit over in this thread: https://www.eevblog.com/forum/projects/psu-protection/
However, so far it was not needed.

Opened the thing up.
Looks similar to HY3005, only with a square transformer instead of a toroid:


A surprisingly large input capacitor.
Can this be the beast that kills it if short circuited?
It would dump all the stored energy (23.5J of it) through the entire power circuitry of the supply.
Off the top of my head, that's a pulse of over a 1000A for a few 100s of microseconds across a 15A rated transistor. 


The transistor is MJ15015, which appears to be a high voltage 2N3055A.
Made by Transun... which googles up as a travel company.
Interesting diversification strategy. 


Anyway, the thing is dead - all it's pins are shorted together.


Which raises the question - what to replace it with?
For one, i haven't found MJ15015 on sale anywhere that does not take a month to ship.
For two, the part appears to be already highly overrated for the task, which does not fit well with the "overstressed the output transistor(s) to use cheaper ones" theory (unless this one is way overspeced).

Any advice?
Closest thing i can get is MJ15003 ( http://www.digikey.com/product-search/en?vendor=0&keywords=MJ15003 ), which is 20A, have twice the gain and only 1V saturation voltage (instead of 5V).
Do any of these matter, especially Vcesat?

[poorchava]

You should be looking at a replacement which has bigger safe operating area than this one (you already know, that equal one won't work). Obviously the voltage should be same or higher. Current is less important, because you've got a 15A transistor in a 2A power supply. It all comes down to safe operating area and power dissipation. When replacing transistors, you also need to make sure that current gain is enough or you may fry the base drive transistor (if it's not fried already, which could be the case with badly shorted pass transistor).

If I were you, I'd just get 2 or 3 high power transistors (whatever... 2N3055, TIP33C, BD249C, TIP35C, TIP142, 2N5886), make sure that base drive is adequate, add some load balancing emitter resistors and off you go

[MagicSmoker]

I concur with poorchava that this an SOA (Safe Operating Area) violation. The Wikipedia entry on SOA is a good start for understanding this, but ON Semiconductor (formerly Motorola) has a really good app note (AN1628) that explains it in more detail.

Attached is the SOA curve from the datasheet of the MJ15003 transistor you found at DigiKey. The datasheet says this transistor is rated for 250W max collector dissipation, 20A max collector current, and 140V max collector-emitter voltage (with the base open-circuited). You would think it is grossly overrated for use in a 50V/2A linear regulator application where the maximum input voltage is 80V; after all, the worst-case collector dissipation would be 160W. Setting aside for the moment whether that little heatsink on the back of the Mastech PSU can dissipate 160W for very long (it can't), the SOA curve shows that with 80V drop across the MJ15003 transistor the maximum allowed collector current is only ~1.6A. That results in a total power dissipation of 128W, or about half what is supposedly permissible.

SOA is highly dependent on the fabrication process used for the transistor so two different transistors with identical "DC" ratings could have very different SOA curves. From my cursory perusal of other NPN transistors in a TO-204 (formerly TO-3) case available at DigiKey, I would go with the MJ21196 instead of the MJ15003.

[edit - corrected collector current rating to 20A]

[dannyf]

The transistor is MJ15015, which appears to be a high voltage 2N3055A.

Pretty much any power transistor would work.

[macboy]

I have attached a couple schematics of similar units. Not exact, but they all use a similar common design. You can also try the Mastech forums or e-mail their support and ask nicely for the schematics. In my experience, a helpful human who speaks english is likely to respond, which is a welcome surprise these days.

[dannyf]

I have attached a couple schematics of similar units.

I used the same design (LED1/LED2) in my CC/CV design - mine is mcu driven and the opamps are powered with floating supplies and theirs analog + opamps with fixed power supply.

[macboy]

I concur with poorchava that this an SOA (Safe Operating Area) violation. The Wikipedia entry on SOA is a good start for understanding this, but ON Semiconductor (formerly Motorola) has a really good app note (AN1628) that explains it in more detail.
....

This would be true, except that these power supplies use a variable input side voltage provided by switched taps on the transformer. Typically, one of four input voltages is selected, based on the actual output voltage. The schematics I provided above show this.

[Teneyes]

While you have the PS open, you may wish to add more pass transistors. There is room on the heat sink.  I modified mine from HY3003 to a HY3005. I also changed Voltage adjustment to a 10 turn pot. See schematic above.  Also check output cap. C80, 470uf  hard to find near front banana posts.  Good luck.

[Bored@Work]

For one, i haven't found MJ15015 on sale anywhere that does not take a month to ship.

The transistor is not too difficult to get

http://au.mouser.com/Search/Refine.aspx?Keyword=MJ15015
http://www.digikey.com/product-search/en?WT.z_header=search_go&lang=en&site=us&keywords=MJ15015&x=0&y=0&formaction=on

Both companies that have small stock and can ship immediately. There are likely other sellers out there with stock.

[Artlav]

I replaced the transistor with one MJ15003, and so far it works nicely.
Not other damage found.
MJ21196 might have been better, but it's not easily available as well.

I still have the 3-transistor board from the HY3005, so i think i'll try populating it. The main board have the unpopulated layout for three, so upgrading it to higher current version (or just higher-reliability low current one) should be easy.
I'll post about it if/when i try that.

Thanks all.

The transistor is not too difficult to get

One transistor: $2.5
Transcontinental shipping: $120

Nope.

[Bored@Work]

One transistor: $2.5
Transcontinental shipping: $120

Nope.

You know, it always amazes me when people spend more time finding excuses than finding parts. It took me five minutes to find a reseller in Russia selling the transistor. And I don't even speak the language.

[mariush]

From 2 uk pounds: http://www.ebay.co.uk/sch/i.html?_odkw=MJ21196&_sop=15&gbr=1&_fcid=168&_osacat=0&_from=R40&_clu=2&_trksid=p2045573.m570.l1313.TR0.TRC0.H0.XMJ15015&_nkw=MJ15015&_sacat=0

Some are shipped from UK or other Europe countries (for example the third result from PCArena-Bristol at about 8 pounds incl. shipping), so they should arrive to you in about 10 days at most.

[megajocke]

I replaced the transistor with one MJ15003, and so far it works nicely.

a
That should be an excellent replacement. Assuming it's actually what it says on the tin. Counterfeiting of power semiconductors is pretty common -- see for example the following page: http://sound.westhost.com/counterfeit.htm

If you still have that original broken transistor it might be interesting to decap it and compare the innards to the pictures on one of the subpages of the above page showing what genuine Onsemi/Motorola parts look like inside. With a dremel you could take the top off, or if it's of really low quality, squeezing it in a vise might be enough to pop the top off! A MJ15015 die should be somewhere inbetween the size of a 2N3055 and a MJ15003. And if you find white goop inside all bets are off...

I found the following page for "Transun", by the way:
http://htecl5094.en.china.cn/

That doesn't really inspire any confidence. The link to their main website is dead too...

A real, up-to-spec MJ15015 might have worked fine in that power supply. Judging by the number of relays, the number of transformer wires and the schematics of similar units it looks like the design has 8 different selectable transformer voltages so I would expect the voltage drop in normal operation to be less than 10 V over the transistor.

It is more difficult though to determine whether or not the transistor can take the energy stored in the smoothing cap when the output is shorted suddenly as it needs to discharge the capacitor from 80 V to 10 V at a current of 2 A max (the current limit) which would take 160 ms.

A somewhat pessimistic approximation would be to consider a rectangular pulse of the same peak power and energy dissipated at the peak voltage. This is 160W @ 80V with an energy of 14.4 J which means an equivalent time of 90 ms. This is within the 25 degree C rating (160 W @ 80 V for 100 ms), but if the transistor is already hot it might be outside the rating -- but this is a pessimistic approximation so then again it might be okay after all!

But if you consider your MJ15003 replacement (assuming it's genuine!) the above pulse (which is a pessimistic approximation) is within the derated 100 degree case temperature rating (about 160 W @ 80 V for 100 ms), so there should be no need to worry! A 100 degree case temperature is probably pessimistic too which means you'll have even more margin in reality because the transistor won't be dissipating more than 20-30 W or so in worst case steady-state.

One would do a similar check for the previous stage transistor, assuming worst-case minimum beta for the pass transistor, to be sure that part is also up to the job.

[Artlav]

You know, it always amazes me when people spend more time finding excuses than finding parts. It took me five minutes to find a reseller in Russia selling the transistor. And I don't even speak the language.

Yep, i know i'm lazy.
Anything more complex than "give money, take parts" at local shop needs a serious reason.

Assuming it's actually what it says on the tin.

Hm. Interesting.
That got me curious about the original Transun transistor, one that burned up.

In the ensuing interrogation, the thing responded to squeezing, prying, hammering and chiseling with true chineese stoicity, so i had to escalate to a dremel.
Here is what i found:

The case looks to be good quality, but what about that gunked die?
Removing the gunk reveals this:

Looks more like the fake than the real images, and quite a bit tiny.
What do you think?

The new ones come from reputable reseller, so they are likely to be genuine.
Don't really want to ruin a good part on groundless suspicion, so i'll have to wait for it to fail first.

[SeanB]

About the right size for a 15A die, if it was fake it would be a lot smaller, about 1/3 of each dimension ( silicon is expensive) and would not have had the sealer on it.  Just typical poor assembly, but looks like a big enough dice. Probably died from SOA being exceeded, so adding 2 with an emitter resistor to share current would likely last longer. If doing that use a TO247 packaged part with an isolated tab, that generally can be easily adapted to fit a TO3 mounting with a little work.

[dannyf]

What do you think?

Looks too small.

The new ones come from reputable reseller, so they are likely to be genuine.

TO3s bought off channels (ebay, surplus sellers, etc.) are likely of questionable quality. I would try a real place (like digikey).

In this case, pretty much any device will work.

[Alex Eisenhut]

Looks too small.

To me too. 2N3055 is a power transistor, to me it should look like this inside



Of course, how can you know? I don't know. But old 2N3055s I have all have very thick metal, like 3mm.

My theory is that if you are using thick metal, you're not going to cheap out on the innards.

www.pmccomponents.com/data/TR/2N3055.PDF

Maybe if you weigh the part on a good scale, the heavier the better, I'd guess...
PMC 2N3055 16.4g
ST MJ2955 12g