Cheapo power brick repair and schematic

[Th3_uN1Qu3]

I had this here power brick from a monitor i fixed last year. Primary cap had failed, sending tons of 100Hz ripple and damaging a cap in the monitor as well. Repaired the monitor, replaced the primary cap, drilled a bunch of holes into it so it doesn't fail again (it was pretty tight in there), got paid. A few days later i get called that it failed again.

I decided to swap this brick for a laptop power brick with its feedback loop adjusted to output 14v instead of 18v. This power brick was 12v 3A, but the laptop brick wouldn't be stable at anything lower. The caps in the inverter were 25v so no problem there. Monitor worked fine and still does.

Upon taking the brick apart everything tested good, but the fuse was blown. I used a piece of thin wire instead of the fuse (as it was a small, uncommon variety of fuse) and it worked. Not sure why the fuse blew. So i adjusted it to 14v instead of 12v and used it for charging UPS batteries. Worked fine until today when i accidentally shorted it. It went BANG.

Turns out that wire i used for the fuse was a bit thicker than i thought... Damage is as follows: MOSFET, source resistor, snubber diode and a PCB trace. The circuit seems to be the standard self-oscillating 2-transistor flyback affair with TL431 and opto, just like 5vsb circuitry of ATX PSU. It has no connection to earth ground. Input filtering is two small chokes and an X cap. Switching transistor was a 6N60, snubber diode a UF1xxx, the rest of the code is burned out. I have no idea what the source resistor was as it has a non standard color code. Beautiful.

Primary cap used to be a 68u 400v Su'scon, this got replaced by a 47u 400v Samxon as i couldn't find 68u and 100u wouldn't fit. Secondary caps are 2x 1000u 25v Su'scon, it has a pi filter too. The only protection i could find was a 15v 1W zener on the output, which i had to remove as it went short when i bumped the voltage up to 14v. Nothing to save the switching transistor it seems.

This is probably not worth fixing (i have bigger and better bricks), but it's simple and low parts count so i'll try to lift the schematic and maybe unwind the transformer too to learn a bit more about flybacks. Expect updates in a few days.

[tom66]

It's probably a design similar to this. Admittedly, it's a very clever design, using only two transistors and a few passive components. But, it has major shortfalls.

It's good for up to 20W maybe, but 15W is probably a sensible limit. Not sure how they managed to make it power a monitor, unless it was a small 15" or 17".

Unlike your PSU this is actually short circuit protected: when a short circuit occurs, it starves the bias winding of power, causing the circuit to switch to low frequency oscillation (10-20 Hz.)

However, it has poor efficiency (~65-70%), drives the opto near limits (40Vce) and is unstable in some circumstances.

The primary cap is actually fairly critical, as the supply becomes hideously unstable with a capacitor that's too small. 100 Hz ripple, but also lots of other frequencies.

[Th3_uN1Qu3]

Originally Posted by tom66

Not sure how they managed to make it power a monitor, unless it was a small 15" or 17".

The monitor was a 17". And the brick isn't as weak as you think - it could put out 6A for short periods of time. The voltage dropped when i did that though. And it did run hot.

[b700029]

On the topic of cheap PSU toplogies, there's a schematic of a single-transistor-with-feedback supply floating around on the Internet.

[tom66]

Originally Posted by b700029

On the topic of cheap PSU toplogies, there's a schematic of a single-transistor-with-feedback supply floating around on the Internet.

That is impressive - I suppose they use the optocoupler transistor as part of the oscillator. I'm guessing it has absolutely no short circuit or overcurrent protection.

[Th3_uN1Qu3]

Just noticed the little bridge rectifier is blown too. This will definitely be stripped down for parts, not before i lift the schematic tho.

[tom66]

Heh. Looks like the components are designed as antifuses: the bridge rectifier shorts on overload, throwing the breaker .

[Th3_uN1Qu3]

Nah, can't blame the bridge when the fuse wire was twice thicker than it should have been. My fault there. The original fuse had blown without a reason like i mentioned in the first post.

I've made some progress with the schematic, but i have to pull the SMD caps one by one and measure them, there's also an unmarked SMD zener. After i finish all that i can pull and test the transformer too and then unwind it. That would be tomorrow since while busy measuring i suddenly started to feel sick, i got the chills. Must be from the long busy days and lack of proper sleep in the last couple weeks. I'm better now and i'll be going to sleep.

It is indeed a classic 2-transistor circuit with opto feedback. The driver transistor is a 2N5551 SMD aka MMBT5551. It appears like they just threw the parts on there because the order of the part numbers on the silkscreen makes little sense. The only interesting part until now is the snubber, which is an RCD circuit wired from drain to +HV with an additional capacitor across the resistor.

[tom66]

Originally Posted by Th3_uN1Qu3

Nah, can't blame the bridge when the fuse wire was twice thicker than it should have been. My fault there. The original fuse had blown without a reason like i mentioned in the first post.

There is a reason fuses aren't replaced with simple pieces of wire. There is quite a lot of engineering behind a fuse, including safety and failure time. Even if your wire could survive, say, 500mA, can you guarantee it will fail quickly at 1A? I have run 1A cable at 3A for extended periods of time with no failure. Fusing current <100ms for some cables can be 10-20x the continuous current rating.

Originally Posted by Th3_uN1Qu3

I've made some progress with the schematic, but i have to pull the SMD caps one by one and measure them, there's also an unmarked SMD zener. After i finish all that i can pull and test the transformer too and then unwind it. That would be tomorrow since while busy measuring i suddenly started to feel sick, i got the chills. Must be from the long busy days and lack of proper sleep in the last couple weeks. I'm better now and i'll be going to sleep.

These two-transistor circuits are very critical of component values. They will often oscillate with any set of values, but efficiency will be very poor and ripple current will toast caps quick. Optimising the values can be done with lots of calculations, or my favourite method is to tweak the values until it works in a circuit simulator.

A good 2-transistor circuit can be efficient, I've been able to optimise one to 85% before. It's very tricky because it's so critical of component values. The MOSFET often dissipates only a few hundred milliwatts, so you don't even need a heatsink on it.

Originally Posted by Th3_uN1Qu3

It is indeed a classic 2-transistor circuit with opto feedback. The driver transistor is a 2N5551 SMD aka MMBT5551. It appears like they just threw the parts on there because the order of the part numbers on the silkscreen makes little sense. The only interesting part until now is the snubber, which is an RCD circuit wired from drain to +HV with an additional capacitor across the resistor.

Hmm. I wonder why they used a 160V transistor? The peak voltage the transistor usually sees is the gate voltage, which is set by the zener, and is typically around 18V. The simulations I posted seem to work with 2N3904s and 2N2222s, and those are cheap 30-40V parts.

Interesting snubber design. For output power <20W, the snubber is mostly unnecessary. Peak voltage at drain is around 400V even without snubber, so using a 650V device is sufficient. Snubber is necessary at AC power on and power loss though as voltage stress is very high then.

[Th3_uN1Qu3]

Done with the primary side. Tested all SMD caps and the zener (18v). Now for the secondary which won't be too much work, and the transformer.

Edit: All done. I also want to correct a few inaccuracies, as follows: The output caps are 1500uf before and 1000uF after pi filter, both 16v not 25v like i said. They are Su'scon SD. Interestingly the feedback is taken after the pi filter... i'll have to check the compensation component values, this is interesting.

Attached LTSpice simulation file. The differences between sim and real circuit are: I guessed the value of the source resistor R7. D3 isn't 4148 but something a tad beefier, but i haven't got a clue what it is exactly. The zener is supposed to be 18v instead of 15v, and the mosfet is 6N60 not 8N60, so efficiency is a bit higher in simulation. If the component numbering order seems weird, that's exactly how it was on the silkscreen. All other values are accurate including the ESR of the output caps.

It doesn't look that bad if you ask me.

[Th3_uN1Qu3]

As for the input filter (not pictured), it's pretty interesting. It has one 0.33u X cap from neutral to live with two 1Meg resistors in series to discharge it. Two small inductors in series with each input leg of the bridge rectifier. They are not wound on a common core but on two separate small yellow/white toroids. They measure only 30uH each so i don't know how effective they are really. At least they tried.

Two 2n2 Y caps from neutral and live to ground (not just one) and here's where it gets interesting - the Y caps aren't referenced to earth ground (which like i said isn't connected) but to secondary ground instead. Basically this is a PSU with the earth ground omitted. Just like an ATX PSU plugged in an unearthed outlet, it will bite you if you touch secondary ground and earth ground at the same time. There's a 3rd 2n2 Y cap joining the secondary ground with the primary ground (negative of primary cap). I've seen this before and i understand it's good for EMI filtering, but to me it just seems like a very, very, very bad idea.

[tom66]

RMS ripple through the cap before the pi filter is 3.3A, but it's rated for 1.5A. Wouldn't be surprised to see that bulge... that's just plain bad design.

Then again, I had a similar problem with a two-transistor circuit. I got ripple down to around 2.7Arms, so I had to spec a 3300u 25V cap.

[budm]

"There's a 3rd 2n2 Y cap joining the secondary ground with the primary ground (negative of primary cap). I've seen this before and i understand it's good for EMI filtering, but to me it just seems like a very, very, very bad idea." That is very common practice to get the EMI down, and the cap is always X/Y type for safety, the value is low enough for the leakage current.

[Th3_uN1Qu3]

Now that you reminded me: Both secondary caps tested bad even though they didn't bulge. The 1500uF one had a whopping 550 mOhms ESR (instead of 55 it should have had), and the 1000uF 110 mOhms instead of 58.

[ben7]

Originally Posted by Th3_uN1Qu3

Now that you reminded me: Both secondary caps tested bad even though they didn't bulge. The 1500uF one had a whopping 550 mOhms ESR (instead of 55 it should have had), and the 1000uF 110 mOhms instead of 58.

Hehe, and did you use your new ESR meter your making?
Sometimes the caps dry out, or in the first place they were poorly made

-Ben

[budm]

That is very commonly used snubber circuits.

[Th3_uN1Qu3]

I typed resistor instead of diode there, i was talking about the additional 220pF cap across the diode.

And yes i did use the ESR meter i'm making - if i had one already i wouldn't be making one.