MCH-K605DN

[BradC]

G'day All,

I ordered an MCH-K3010D from fleabay and a K605DN turned up in the box. I pulled it out of the box to hear a bit of a rattle, so I thought I'd pop the lid.
Out fell a 750 Ohm 5W resistor. There are 2 others on the board, both with broken legs.

I've notified the seller, but due to the Chinese holiday they aren't back until the 10th, and I don't want to go poking it with a soldering iron until I find out if they want it back.

Given I have nothing else to do with it right now, and I'm sure as hell not going to plug it in, I thought I'd take some pics. Barring the dicey resistors and the obvious mods to the primary rectified rails, it actually looks pretty well put together. I figure it'd cost me as much to send it back as it cost, so let's see what they come back with.

[BradC]

Given they got everything else wrong, I suppose I should have checked to make sure that bodge on the back of the board wasn't a voltage doubler for 120V *before* I powered it up.
Thankfully the only carnage is the primary caps and the fuse.

[BradC]

Righto then. As you might imagine, this power supply has a very similar schematic to the K305D.

The primary capacitors are 470uF/250V. The thing is they are 18mm diameter and 40mm height with a ~7.5mm lead spacing. The closest I can find in the diameter and lead spacing is a 50mm high cap, and that doesn't have a hope of fitting. 45mm is the tightest I'd get in there at a squeeze.

For the life of me I can't find a cap > 330uF with those specs outside of seriously dodgy looking E-bay listings :

https://www.ebay.com.au/itm/2PC-250V-470uF-Radial-Electrolytic-Capacitors-For-PCB-LCD-Mount-105C-18x40mm-/264427329200

I pulled the existing caps and they both measure ~400uF & < 0.1Ohm ESR. Of course one has a pretty seriously bulged top and bung now. My gut says just replace them with 330uF quality caps and be done with it.

Right now it has been re-assembled and is running a 175W load test (all my poor little SkyRC BD200 can do with a max input voltage of 35V) and it seems to be coping ok. Nothing excessively hot and it hasn't blown up.

If I'm replacing the primary caps I might as well replace the other 5 leaded electros while I have it open. I can get replacements for all those.

Given I'm planning on reducing the designed primary capacitance to 70% of the schematic, does that sound like a sane thing to do?

[BradC]

Thought I'd have a look at its behaviour under current limiting. Ewwww.

All readings are x10. I was using a cheap probe and it doesn't tickle the selector ring on the scope.

Using a 12V 21W lamp as a load. Current limited to bring the voltage down to ~6.5V :


So, Max volt ~ 8.,8 and Min ~4.0. Average is ~6.5 (which is what it read on the display)

Ok then, so what happens as we start to get out of current limiting?  Bursts of limity noise goodness :



Same but AC coupled with a bit more amplitude :


So, this'll make a decent battery charger. I'm not sure I'd let it loose near anything valuable, particularly when approaching the current limits.
Steady state output noise is ~50mv.

[BradC]

Right, so it was crap as a battery charger also.

Trying to charge a set of 14.4V SLAs with a current limit of 3A it never got near 3A, rather bouncing around 1.8-2.2A with a revolting waveform again. When not in CC mode it was fine (so as the voltage came up to set point it behaved nicely).

Some investigation shows the current limit is implemented by hitting the compensation/soft start pins pretty much in an "on/off" fashion causing the poor PWM chip to bounce around all over the place. There's also soft start, over-current and over-temperature inputs into those pins, so I had to figure out how to change the current limit feedback without affecting any of them.
 
So, a bit of poking and a quick mod to have the current comparator act on the other side of the PWM and it's now behaving nicely. There is a bit of lag when you hit the current limit slowly (ie, as I wind the pot down it takes a second or so before the limit kicks in (ie, lamp is drawing 1.8A and as I slowly turn the pot down it kicks in at 1.5A). Once the limit kicks in it's fully linear and that behaviour is dependent on the distance between current and setpoint (ie put a big load on and it kicks in almost immediately, same for winding the knob down fast).

I might have a play with that and see if I can improve it, but right now I have a linear current control that behaves as it should and is now satisfactory for charging batteries.



As as result I could also remove the extra 1500uF cap on the output as this seems to make the supply more stable. Need to do some more tests, but it passed all the step load tests I could throw at it.

[BradC]

Further testing.

On heavy loads I was still getting an unstable oscillation while under CC mode at initial connection.

Unlike the attached schematic where B2 is used to keep tabs on the primary current, this particular PSU omits all that and splits RL1 into RLA & RLB. One of these is used as feedback to the CC circuitry surrounding U1-A and the other disappears into the fan control circuit somewhere around U2. Large current spikes on the output then cause the over-temp control input via D3 & R40 to turn on Q7. This puts the PWM protection into an oscillation loop and we see the waveforms shown previously.

The problem is with 1500/3000uF charged up to the setpoint voltage, it doesn't matter what happens in the current feedback loop, it will *always* trigger this over-current protection and send the supply into oscillation with a low impedance load as it is measuring the dump from the caps into the load. You wouldn't see that on the primary side measurement. This would appear to be a "cost down optimisation" in the later model boards.

Removal of R40 stopped this happening, and this has stabilised the CC loop nicely regardless of load or step changes. I dropped C3 (in the feedback path of U1-A) from 0.1uF to 0.01uF. This sped up the current limiting considerably. Another order of magnitude lower causes the supply to oscillate.

The only issue I have now is removing R40 removes the "last resort" thermal shutdown, so I need to get into that area around U2 and figure out what is going on to fix the overcurrent protection and allow me to restore that function.

The CC limit now takes ~15-20ms to activate and pull the current down, so it's not ideal. But fine for battery charging.

During the experiments I broke the R50 feedback resistor which caused the output caps to charge to somewhere north of 100V. Turns out ~3000uF at > 100V pretty much instantly vaporised the ends of the H7 lamp filaments I was using as a load. Lucky I caught it before the caps went off (they are rated at 63V and got really warm quite quickly).

[BradC]

So thinking you are smart by removing the cycle-by cycle catastrophic current limiting works a treat until you short the output to set the CC value, at which time it shorts both primary switchers and immolates the fuse.

Cue RS order for a new set of MOSFETS and a new resistor to replace the one I so smugly removed earlier.

[tautech]

Is the K305D schematic more or less correct for these higher voltage models ?

I've had a K305D for years and beaten it up a bit and the only issues have been display drivers dying.
It's done lots of FLA and SLA charging.

[BradC]

Is the K305D schematic more or less correct for these higher voltage models ?

Almost identical. I spent some time tracing out the over-current protection. They've taken U2-B out of the thermal protection circuit and set it up as a ~10A current limit. So there is no thermal feedback into the PWM loop, and the gross current protection works on the output.

Any current peak exceeding 10A halts the PWM. With 3000uF on the output that happens when you connect a low impedance load (Like an SLA). That causes the unit to oscillate on the current protection.

My next experiment is to slow down the over current protection by a couple of cycles to see if I can get to to stabilise. None of this would be a problem on the original design with the over current protection on the high side.

I'm still waiting on a k3010d (which I originally ordered) to arrive, so I'll have a poke at that when it turns up. If I can repair the 605 and make it work properly then it's a bonus as it didn't cost me anything.

[BradC]

Replaced the primary switchers. Probably over-capitalized but also replaced the 2 x 1500uF output caps with a single 1200uF close to the transformer and left out the caps post current sensing. The primary caps (C39 & C44) were replaced with 330uF 250V units and the housekeeping supply (C28) with a new 400V 10uF. There was also an additional 560uF on the secondary that got replaced (somewhere akin to C32 on the schematic). Oddly enough, the replacements are all significantly larger than the units they replaced.

Dropped the 3.9k resistor in the modification to 890 Ohms as when really winding the voltage up the current rose to 2.5A no matter what the limit was set at, but if I removed the resistor entirely the supply turns into an oscillator. Also removed Q14 & Q15 (they were causing amplification of the oscillation under CC mode) and changed C26 from 3300pF to 4700pF. It's a *lot* happier on the whole and it did pass the "short the output at maximum voltage" test (and any other abuse I could dream up). It now behaves when charging batteries also.

There is still some fairly significant noise while under CC and a heavy load, but it's a *lot* better than it was. Dropping the bulk capacitance on the secondary side improves this behaviour, but then the overall output noise increases. 560uF was really too small, and 1500uF was a bit much. 1200uF seems to be the sweet spot with this particular unit, which is lucky as that was what I ordered. A real 105C 1500uF 63V cap would not physically fit in the box.

I'm actually pretty impressed by this little unit. I think I got a dud. It had been dodgied up to 120V, the load resistors look like factory floor sweepings and bits were falling off out of the box.

I'm not enamoured with the behaviour of the current limit circuit. Out of the box it was taking some ~40ms to respond and then created the waveforms posted at the start of the thread. After the modification it's a lot better behaved with regards to output ripple, but it's still only twice as fast as it was.

For a 60V 5A PSU at ~$140 it's probably pretty reasonable. For a PSU that cost me $22 in parts and a bit of a learning experience it's probably pretty good value. I'll be interested to look at the MCH-k3010d if it ever arrives.

[tautech]

Thanks Brad for writing up your findings and mods.

I've been pretty happy with the K305D's and never had any comebacks from the few dozen I've sold.

I'll be interested to look at the MCH-k3010d if it ever arrives.

I hope one of these are OK for you but one of my customers tried one and reckoned it was a POS compared to K305D so temped as I've been I didn't bring any of the 10A models in. 
Anyways, I look forward to your findings.