Mercury 3A 13.8V Regulated Switchmode Power Supply

[tom66]

Another eBay find for £6...

Got it here, plugged it in. It powered up, but the LED on the front flickered a lot and the fan speed was unstable. (Yes, a 41.4W supply with a fan. This tells me everything I need to know about how efficient it is.)

Put it under 10 ohm load - output just sagged with ~13.8V pulses.

After 10 minutes of warming up, it started running fine. Ripple was crap though at 300mVp-p (with 10 ohm/1.38A load), so it probably has some bad caps on the output, or just too small ones.

Theory:- Startup/bias/Vcc capacitor is bad.

However, I cannot open the case due to a damn proprietary screw that the OEM has used. Arrghhh! Why can they not use a normal screw, like people actually have drivers for? And surely, a normal screw must cost less...

Plan is to add an adjustment pot on this, fed into the feedback circuit, so the output can be varied from ~5V to maybe ~45V or above. I have some 1000u 50V Panasonic FM and some 680u 50V Nichicon HE available for this purpose. I might just keep it as a 13.8V supply, but it's pretty noisy so not much use for anything that requires a lot of precision. I don't do CB.

Pics soon, when I figure out how to get it open!

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

I had to make my own screwdriver to get out the screws. It worked . I took a cheap multitool and cut a small dedent in the flathead part, the screws came out easily.

PCB was glued into case... clearly not designed to be repaired... cut the glue and got the board out.

Overall, the build is pretty poor. At least it's got an EMI filter, and a bridge rectifier. But the components look like they have been fired from a poorly aimed catapult. Some of them have hit the PCB at odd angles. Most aren't flush to the board. The soldering is also poor; most of the PCB is hand soldered. I could do better hand soldering blindfolded. (I say most - the board is clearly wave soldered, but all of the power components have obviously been hand soldered with lots of cold joints, possibly because the wave soldering wasn't good enough. Which makes you wonder how bad the wave soldering could have been...)

Good things:
- 100uF input capacitor. This is more than enough. It's a ChangX but that's okay for a primary. Given the L&C PSU I took apart only had 110uF and it was a "350W".
- Decent output caps (1000uF x 2)
- Decent size transformer
- EMI filter
- Bridge rectifier
- Inrush limiter (NTC)
- Very good heatsinks, especially for the output power
- Compensated TL431 feedback, unlike the last power supply...

Bad things:
- Obviously designed by someone without a clue on how to lay out a PCB; for example, the filter caps aren't next to each other, one is near the transformer and the other is on the other side of the board.
- About 20 different versions of this power supply on the same PCB.
- No pi filter.
- As above, poor soldering, components aren't flush to the board, etc.
- Cheap secondary caps. (What else do you expect...!)

I found the startup cap (47uF 25V "UNITED")... measured good capacitance but ESR was 1 ohm... which is quite high although maybe not unreasonable for a 47uF.

Replaced it with a 100uF 50V Chemicon KMG. PSU now starts reliably, there's a ~1 second delay from power up due to the high capacitance.

Ripple measuring 300mVp-p at just 1A load is too much. So I removed the original "Gather" caps (which tested okay on the ESR meter) and replaced them with some FMs (1000uF 50V.) I also added a coil which should be around 2mH (~18 turns, 30mm toroidal core.) Result: ripple is now 100mVp-p. Much better. I could probably reduce it further, but the largest cap I have above 35V is 1000uF and I don't have any decent premade coils that aren't SMD.

Next step: Switch and adjustable output.

[Th3_uN1Qu3]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

That coil can't be 2mH. Maybe 20uH.

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

That coil can't be 2mH. Maybe 20uH.

I thought 18 turns was a bit low for 2mH; but the online calculator said so! I used it in a buck converter which struggled to run with 100µH, but was okay with 200µH, so I'm guessing it's at least 100µH.

[Th3_uN1Qu3]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

2mH is 2000 uH. Not 200uH. Yes it could be 100u at light load, powdered iron cores have a "soft saturation" curve which means the inductance decreases with increasing current, without completely saturating the core tho, so it doesn't become a straight wire, but the inductance decreases by a certain amount. The calculator by Micrometals can tell you those details.

[PeteS in CA]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

Hmmm ... that looks like it could be a flyback design, in which the "output inductor" is the transformer primary. If that is the case, adding that extra inductor could have affected the control loop considerably. What did your ripple look like? If it was 100Hz, your input 'lytic could be inadequate or bad. If it was spikes repeating at the switch frequency, those might be fixed by adding a ferrite rod inductor instead of that big toroid.

[PeteS in CA]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

If that core is a Micrometals T124-26, that inductor is ~18.8uH (~18,800nH).

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

Hmmm ... that looks like it could be a flyback design, in which the "output inductor" is the transformer primary. If that is the case, adding that extra inductor could have affected the control loop considerably. What did your ripple look like? If it was 100Hz, your input 'lytic could be inadequate or bad. If it was spikes repeating at the switch frequency, those might be fixed by adding a ferrite rod inductor instead of that big toroid.

I made sure to put the inductor after the control loop feedback. The PCB is designed to take an inductor in that place, but they've placed a wire jumper instead.

Ripple was at 33kHz, the switching frequency presumably. I'll take a pic of the scope shot later. I'm going to test it on the variac soon to see how well it will cope with low line voltages. (Just out of curiosity.)

Just did a test with a 3.3 ohm load. That's 4.8A. PSU was able to power it, voltage dropped to around 13.7V (0.1V drop from no load.) Ripple was only 150mVp-p... Pretty impressive. The PCB says 2A/3.5A/5A though, so the only difference in versions might be the MOSFET and/or diode - or perhaps none at all!

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

If that core is a Micrometals T124-26, that inductor is ~18.8uH (~18,800nH).

It might be. I have just realised where I put my 100µH Bourns inductors. Maybe I should use one of those in place? Saturation current is around 8A from memory.

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

Uh oh, pulled the supply out and found it wouldn't start again. Same ticking on and off.

Tested my cap... 1.2 ohm ESR?! How did that happen... I didn't test it beforehand, maybe it's normal for KMG? (It's a new cap, I put it in a PSU briefly but pulled it later.)

Anyway, the smallest I have in ~25-50V is 470uF (Chemicon KY.) Supply works from that now, but takes 5 seconds to start. Not really a problem for me.

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

Okay, so I looked at the PCB again and I noticed my cap for the output of the L-C wasn't soldered in properly... hmm... I soldered it back in and then the PSU was unstable. I obviously didn't look carefully at it, as *it does* take feedback from after the coil. Broke the trace and moved it to before the coil; now it's stable and the ripple is very low, apart from a bit of ringing on the switching waveform. Any way to remove this ringing? If I could get rid of this ringing, the noise is acceptable for a lab bench supply - 10mVp-p tops with a 10 ohm load.

[tom66]

Re: Mercury 3A 13.8V Regulated Switchmode Power Supply

Okay, so the mod is complete, but unfortunately, the maximum output is around 35V, which is exactly the same as my current bench supply. I was hoping for closer to 45-50V. The minimum is about 12.5V, below which the Vcc for the controller (UC3842) drops out, and the output turns on and off at several Hz.

At least I should be able to float the bench supply (optional output earth) and put them in series for epic justice. And having a second supply doesn't hurt. And it's probably capable of 5A at lower voltages. Need to determine the maximum power; possibly check the sense resistor on the primary. (Primary current is limited, so the output power will be limited by this.)

There is a certain point - around 24-27V - where the supply is unstable with no load (1Vp-p), but I'll try later with a load and I'll see if that changes.

I will attempt to configure it for a minimum output. The potentiometer is configured to set the maximum output, in addition to linearising the output voltage over the full span of the potentiometer, using three resistors.