ADT MPT-301 – an early CWT ISO unit?

[momaka]

So, I got this PSU from a computer I picked up off of Craigslist last summer. I can't find a date code of when it was built. There is nothing on the warranty sticker or the caps to help me find out. However, I can only assume it is as old as the Intel D845WN Pentium 4 motherboard it was powering, which is from 2001. The PC looks like an original build and didn't appear to have been modified in its life, so I don't think my guess is wrong.

Anyways. Most of you probably know how I do this now with the pictures and descriptions. If you want to skip some reading, then just open the pictures and skip to the Summary section. Here is a picture of the back of the PSU:


And here is the label of the PSU to see who made it:
http://www.badcaps.net/forum/attachm...1&d=1443215464
Hmm, the model number says MPT-301? Macron?! But where is Macron's UL number and the Macron-specific "AMD Intel / ATX12V TEST PASS" sticker?
I tell you what – this PSU definitely is not a Macron. Just look at that QC sticker (on the side of the PSU) and the fan grille again in the first picture. Seems familiar?

If anyone said Channel Well, you got it right. That specific QC sticker is the big giveaway (with the multiple stamps). They are just trying to trick us again like they did with the TurboLink PSU I have (which had a Deer UL number). And now I will give you some proof of that – strip the PSU naked!
http://www.badcaps.net/forum/attachm...1&d=1443215464
At first, it looks like a Deer or L&C. In particular, you just don't see the TPS3510 supervisory IC like you do with other ISO units. However, the output caps being all Fuhjyyu and the 5VSB design strongly suggest CWT ISO series. Sure there are no "ISO": marking next to the fuse, but that is most likely because this design is a very early version of the ISO series, or perhaps a close cousin.

Either way, it is definitely not the best CWT build quality. Some things to note in the above picture: lack of input filtering, thin 300V, 20 AWG wire on the primary side, bulged caps, and lack of 3.3V torroid (suggesting cheap linear regulation.) Another (not visible) downside: the fan – Rulian Science RDM8025SA, rated for 12 V and 0.13 A… and almost seized .

But hey, at least there are some PI coils on the secondary side. In fact, almost each rail has one… well, almost (will cover that "surprise" in a bit.) As for the soldering, it is overall decent and the gaps between high voltage traces on the primary side are good. Here is a picture:
http://www.badcaps.net/forum/attachm...1&d=1443215464

Oh, and most notable thing in the entire PSU is the 12V rail filter cap: a Fuhjyyu TNR 16V, 1650 uF cap! .

Seriously, what's up with that capacity? They could have easily advertised it as both a 1500 uF or 1800 uF cap, and it would still have been within 20% tolerance either way. But I guess that is Fuhjyyu quality for you. And now I have this rare sample in my collection box .

Finally, there is the surprise: that big, white, wirewound resistor. At first, I thought it was a load resistor of some sort. But it is rated for 0.035 Ohms resistance and 10 Watts, so no way it can be a load resistor (even on the 3.3V rail, such resistor would pull 311 Watts!) Instead, it is actually used in place of the PI coil for the 3.3V rail. Why? - I have no idea. Perhaps to stabilize regulation of the 3.3 V rail since they used only a single 1x Fuhjyyu TNR 16V, 220 uF cap for filtering? Regardless, I just replaced it with a proper PI coil (since the PCB is marked for one anyways) and changed the 220 uF filtering cap with a 1500 uF one (which I am planning on changing it again, likely for a 2200 uF cap). 3.3V regulation is achieved in a linear fashion with a MOSFET, so the output should be pretty clean either way.

Summary

Primary Side:
http://www.badcaps.net/forum/attachm...1&d=1443215464
- RS405 bridge rectifier (rated for 4 Amps), no input filtering at all
- two Fuhjyyu LP 200V, 680 uF, 22 x 40 mm (dia. x h) bulk caps
- two 2SC2625 NPN BJT transistors (TO-3P package) for the main PS
- two Fuhjyyu TNR 50V, 2.2 uF, 5 x 11 mm electrolytic caps for the BJT drive circuit
- 2SC5353 NPN BJT for the 5VSB main switch, used in a 2-transistor design
- critical 5VSB cap: Fuhjyyu TNR 50V, 1 uF, 5 x 11 mm, electrolytic
- main transformer appears to be slightly wider than size 35, but also slightly shorter.

Secondary Side:
http://www.badcaps.net/forum/attachm...1&d=1443215464

3.3 V rail:
- Linearly regulated from 5 V rail with 40NF03 MOSFET (100V, 35A @ 100C)
- 1x Fuhjyyu TNR 16V, 220 uF, 6.3x11 mm on the output.
- 15-Ohm load resistor
- 0.035-Ohm, 10W resistor instead of a PI coil on the output

5 V rail:
- D83-004 (30A. 40V) schottky rectifier
- 2x Funjyyu TMR 10V, 1000 uF, 10x20 mm caps with a PI coil (8 mm core, 6 turns) between them
- 150-Ohm load resistor

12 V rail:
- STPR1620CT (16A, 200V) fast recovery rectifier
- 1x Fuhjyyu TNR 16V, 1650 uF , 10x23 mm cap with a PI coil (6 mm core, 20 turns) before it.
- 470-Ohm load resistor

-12 V and -5 V rails:
- 4x FR153 1.5 Amp ultrafast diodes (two per -5 V and -12 V)
- 1x Fuhjyyu TNR 16V, 220 uF cap (for each rail) with a PI coil (3-4 mm core, 34 turns) before it.
- the -5 V rail has a 150-Ohm load resistor and -12 V rail has 470-Ohm load resistor.

5 VSB rail:
- FR-302 (3A @ 75C) fast recovery rectifier
- 2x Funjyyu TNR 16V, 470 uF, 8x14 mm caps with a PI coil (4 mm core, 16 turns) between them
- 10-Ohm, 3-Watt load resistor!

5VSB section:
Like with the other two CWT units, I decided to take some time again to draw the 5VSB circuit. See below:


All I have to say about this circuit is that it is very similar to the other two CWT PSUs I posted. It is a 2-transistor design with a 1 uF, 50 V critical cap.

Some 5 VSB load tests:
- With no load, the PSU 5VSB circuit was drawing 5.3 Watts @ 0.60 PF. Output voltage stable at 5.13 V.
- With a 1.1 A load (incandescent light bulb), the 5VSB was drawing 20 Watts @ 0.49 PF and output voltage dropped to 4.85 V.

The worst part about this circuit is the 10-Ohm minimum load resistor on the 5VSB rail – it draws 2.5 Watts all of the time. It may seem like nothing, but considering the size of that resistor, it can really make it hot. Just look at what the resistor solder points did on the isolating plastic sheet under the PCB:


Overall…
This is not that terrible of a PSU, especially considering the fact that it was still working with bad caps up to the point I went to pick it up. So what I did is I (temporarily) recapped the 3.3 V, 5 V, and 12 V rail with some Chemicon KZJ 16 V, 1500 uF, 10 mm dia. caps (and yes, I am aware that they are crappy caps, but I had just pulled them from a dead Xbox 360 motherboard the previous day and they tested okay on the ESR meter.) I am also aware those are some very low ESR caps not quite suitable for PSUs. But I figured this would be a good experiment. Surprisingly, the PSU did not oscillate at all and powered that Intel D845WN motherboard fine again. The other modifications I did were to swap the 10-Ohm 5 VSB dummy load resistor with 75 Ohms and put a proper PI coil (again, from an Xbox 360 motherboard) in place of the 0.035-Ohm wire-wound resistor on the 3.3 V rail. Here is a picture of the temporary "repair":
http://www.badcaps.net/forum/attachm...1&d=1443215464

I did not run the PSU for too long though, since I have not fixed the seized fan yet. But when I do, I will test it for a bit longer. For right now, I just know that it is working "fine". Given that the primary transistors are 2SC2625 BJTs, I think the PSU is worth recapping and fixing.

[Wester547]

Originally Posted by momaka

So, I got this PSU from a computer I picked up off of Craigslist last summer. I can't find a date code of when it was built. There is nothing on the warranty sticker or the caps to help me find out.

The flyback and feedback transformer have what appears to be a "0201" datecode on their labels, and the main transformer "0143". So, presumably, sometime around January 2002 or thereabouts.

Either way, it is definitely not the best CWT build quality. Some things to note in the above picture: lack of input filtering, thin 300V, 20 AWG wire on the primary side

The wire gauge is important, but as is the actual material the wires are made of (copper or aluminum).

Seriously, what's up with that capacity? They could have easily advertised it as both a 1500 uF or 1800 uF cap, and it would still have been within 20% tolerance either way. But I guess that is Fuhjyyu quality for you.

Those 1650uF 16V Fuhjyyus and Jamicons were found in some older FSP units as well. And Fuhjyyu quality... not in love with Fuhjyyu anymore? Kidding, kidding... the fact that the Rulian Science fan had nigh seized probably helped the Fuhjyyus along to their doom.

Regardless, I just replaced it with a proper PI coil

The coil itself is actually made of ferrite; "PI" just references the arrangement just like an "L" filter (one capacitor and one coil in series). I'm sure you know this, and I don't mean to impress upon anyone as a terminology nazi, I just want to make that lucid. You can call them whatever you like, of course, I'm not trying to dictate anyone's choice of words.

- main transformer appears to be slightly wider than size 35, but also slightly shorter.

Maybe it's some variant of an EI-35 or EI-33 transformer?

So what I did is I (temporarily) recapped the 3.3 V, 5 V, and 12 V rail with some Chemicon KZJ 16 V, 1500 uF, 10 mm dia. caps (and yes, I am aware that they are crappy caps, but I had just pulled them from a dead Xbox 360 motherboard the previous day and they tested okay on the ESR meter.)

The worst of the KZGs/KZJs/KZVs/TMVs/TMJs by far are the 3300uF 6.3V 10mm diameter 25mm high KZGs. Not to sound like a broken record, but those are the ones most infamous for bulging in storage without ever having a voltage applied to their plates or without their electrolyte having ever conducted any current. The other seems to be a tad more stable, especially if you keep them cool, but that doesn't mean they too aren't ultimately unstable, junk capacitors.

Very nice and very thorough teardown. I'd be curious to know what the D845WN had in terms of capacitor brands lest it needed to be recapped as well. IIRC, that board uses SDRAM, so it should really benefit from the linear regulated +3.3V output (being directly powered by that rail).

[Per Hansson]

Originally Posted by momaka

Oh, and most notable thing in the entire PSU is the 12V rail filter cap: a Fuhjyyu TNR 16V, 1650 uF cap! .

Seriously, what's up with that capacity? They could have easily advertised it as both a 1500 uF or 1800 uF cap, and it would still have been within 20% tolerance either way. But I guess that is Fuhjyyu quality for you. And now I have this rare sample in my collection box .

Finally, there is the surprise: that big, white, wirewound resistor.

I laughed out load when I read this, thanks man, that's Fuhjyyu for ya!
And that resistor, man I was thinking "hmm, looks like someone dropped a resistor in the PSU"

[goodpsusearch]

The resistor acts like a fuse for overpower protection for 3.3V? :P

[momaka]

Originally Posted by Wester547

The flyback and feedback transformer have what appears to be a "0201" datecode on their labels, and the main transformer "0143". So, presumably, sometime around January 2002 or thereabouts.

I never really thought to look there. You have golden eyes!

Originally Posted by Wester547

The wire gauge is important, but as is the actual material the wires are made of (copper or aluminum).

True. But when I see 20 AWG, I always frown by default.

Originally Posted by Wester547

And Fuhjyyu quality... not in love with Fuhjyyu anymore? Kidding, kidding... the fact that the Rulian Science fan had nigh seized probably helped the Fuhjyyus along to their doom.

Haha, of course I still love Fuhjyyu
I think they died both because of the fan and the fact that this is a CWT PSU (Fuhjyyu and CWT don't appear to mix well together).

Originally Posted by Wester547

The coil itself is actually made of ferrite; "PI" just references the arrangement just like an "L" filter (one capacitor and one coil in series). I'm sure you know this, and I don't mean to impress upon anyone as a terminology nazi, I just want to make that lucid. You can call them whatever you like, of course, I'm not trying to dictate anyone's choice of words.

Actually, it is good that you correct me, since I never really knew the proper terminology for them. I just saw this terminology used here way way back and that's how I remembered it.
That said, not all of these inductors have a ferrite core. Some simply have air cores.

Originally Posted by Wester547

I'd be curious to know what the D845WN had in terms of capacitor brands lest it needed to be recapped as well.

It did need to be recapped. And speaking of how terrible the 6.3V 3300 uF KZG are... well, that's what the D845WN had - three out of four on the CPU V_core were bad. The rest of the capacitors were all small blue Teapo (SEK?) and Chemicon LXZ for the CPU VRM high side.

Originally Posted by Per Hansson

I laughed out load when I read this, thanks man, that's Fuhjyyu for ya!
And that resistor, man I was thinking "hmm, looks like someone dropped a resistor in the PSU"

I'm glad you enjoyed my post
Yeah, that resistor really was just shoved in there.

Originally Posted by goodpsusearch

The resistor acts like a fuse for overpower protection for 3.3V? :P

Not really.
A short-circuit on the 3.3 V rail could still pull 3.3/0.035 = 94 Amps through that .
Probably there to stabilize regulation because the 3.3 V rail only had a single 220 uF cap as filtering. Either that, or they just ran out of parts at the factory and stuck whatever they could find . I just have no other explanation.

[Wester547]

Originally Posted by momaka

I never really thought to look there. You have golden eyes!

Thanks. Usually, when I open up a PSU, the date codes are the first thing I look for, on all the components.

Actually, it is good that you correct me, since I never really knew the proper terminology for them. I just saw this terminology used here way way back and that's how I remembered it. That said, not all of these inductors have a ferrite core. Some simply have air cores.

You're right, I've seen some Huntkey PSUs omit the core completely and just leave the coils, namely on the +3.3V rail. I don't think that's a good thing. Maybe better terminology for the coils would be "coils with ferrite cores" or "coils with air cores." For those who don't know, a PI filter is an output filter with capacitor(s) and a coil in between them, in series with them.

It did need to be recapped. And speaking of how terrible the 6.3V 3300 uF KZG are... well, that's what the D845WN had - three out of four on the CPU V_core were bad.

Yup, the 3300uF 6.3V KZGs are the most notorious preggers. You will be hard pressed to find a board with those still intact (IE a board where at least one hasn't bulged or silently failed). They mingled ultra low ESR electrolytics with solid state capacitors in the VRM low circuit, on these boards. Sometimes the Fujitsus were known to fail because of poor seals but otherwise they were pretty bullet proof.

The rest of the capacitors were all small blue Teapo (SEK?) and Chemicon LXZ for the CPU VRM high side.

Teapo SK, 85*C general purpose capacitors (SEK is 105*C). Sometimes you will find 2200uF 16V 13x25 Teapo SC capacitors in the VRM input as well on these boards. I rarely find them bulged, but because Teapo fail without showing it so often, that doesn't really say much.

[Per Hansson]

Originally Posted by Wester547

You're right, I've seen some Huntkey PSUs omit the core completely and just leave the coils, namely on the +3.3V rail. I don't think that's a good thing. Maybe better terminology for the coils would be "coils with ferrite cores" or "coils with air cores." For those who don't know, a PI filter is an output filter with capacitor(s) and a coil in between them, in series with them.

Well now if we are gonna nitpick we can't really call the ferrite-less coil a "coil with air core"
Because otherwise I have an "imaginary air filled goldbar" that I want to sell you

[Wester547]

Originally Posted by Per Hansson

Well now if we are gonna nitpick we can't really call the ferrite-less coil a "coil with air core"
Because otherwise I have an "imaginary air filled goldbar" that I want to sell you

Yes, that's true.

I guess what I'm trying to say is, to my knowledge, the coils with the ferrite cores are better than the coils with the air cores. But even the coils with the air cores are better than nothing of course.

[momaka]

Originally Posted by Wester547

You're right, I've seen some Huntkey PSUs omit the core completely and just leave the coils, namely on the +3.3V rail. I don't think that's a good thing.

I think it also has to do with the design, though. Maybe the higher inductance of a ferrite core would cause oscillation problems. I've seen this on other power supply brands as well, and again on the 3.3V rail, so maybe there is a reason. In particular, I always see it on mag-amp regulated 3.3 V rails.

Originally Posted by Wester547

They mingled ultra low ESR electrolytics with solid state capacitors in the VRM low circuit, on these boards.

Indeed. There are 5x Nichicon NA 560 uF caps on that board as well - probably what kept it from crashing. And yes, RAM is fed straight from 3.3 V rail, since I don't see any other regulators. Only small Teapo SK bypass caps here and there.

Originally Posted by Wester547

Teapo SK, 85*C general purpose capacitors (SEK is 105*C).

You are right again. I knew SEK probably wasn't right, since those are the brown-sleeved Teapo's (that Bestec loves to use). Should have gotten off my lazy ass and looked at the board . It is literally sitting atop my parts shelf about 6 feet behind me.

Originally Posted by Wester547

I guess what I'm trying to say is, to my knowledge, the coils with the ferrite cores are better than the coils with the air cores.

Well, it depends. You can have an air core with the same or higher inductance of a ferrite core. The coil size will just have to be much larger. But it will be more linear (though, I doubt that matters here).

This is a nifty tool I've used before to calculate some air-core inductors:
http://www.m0ukd.com/calculators/air...or-calculator/

[Wester547]

Originally Posted by momaka

I think it also has to do with the design, though. Maybe the higher inductance of a ferrite core would cause oscillation problems. I've seen this on other power supply brands as well, and again on the 3.3V rail, so maybe there is a reason. In particular, I always see it on mag-amp regulated 3.3 V rails.

Well, Huntkey do include a ferrite core in many variations of the same design, that's why I took it as cheapness on their part.

Indeed. There are 5x Nichicon NA 560 uF caps on that board as well - probably what kept it from crashing. And yes, RAM is fed straight from 3.3 V rail, since I don't see any other regulators. Only small Teapo SK bypass caps here and there.

The KZGs are more likely to go open circuit than the bad HMs. The bad HMs would likely fail in a partial short circuit manner, which means the NAs wouldn't have made the same difference as stability goes. And there are freestanding FETs in the VRM output on this model - not good for heat output.

You are right again. I knew SEK probably wasn't right, since those are the brown-sleeved Teapo's (that Bestec loves to use). Should have gotten off my lazy ass and looked at the board . It is literally sitting atop my parts shelf about 6 feet behind me.

Bestec either use the SEK or SJ (equal to KZE like SM) series in their PSUs, and sometimes SK on the input. Whilst both SEK and SJ are prone to failure, neither are as bad as CapXon KM and KF (equal to LXY I think).

Well, it depends. You can have an air core with the same or higher inductance of a ferrite core. The coil size will just have to be much larger. But it will be more linear (though, I doubt that matters here).

Yes, a higher uH value is better. I meant to say that if all specifications were otherwise equal that the ferrite core would be superior.

This is a nifty tool I've used before to calculate some air-core inductors:
http://www.m0ukd.com/calculators/air...or-calculator/

Yup, that's a neat tool, like the resistor color code calculator. More useful than the lameXtreme PSU calculator for sure.

[momaka]

Originally Posted by Wester547

More useful than the lameXtreme PSU calculator for sure.

Anything is better than those online PSU calculators. Doesn't matter if you select onboard video and an Intel Atom CPU... they always recommend some astronomically high power ratings for the power supply... like 400 Watts minimum, lol.

[Stefan Payne]

Originally Posted by Wester547

Bestec either use the SEK or SJ (equal to KZE like SM) series in their PSUs, and sometimes SK on the input. Whilst both SEK and SJ are prone to failure, neither are as bad as CapXon KM and KF (equal to LXY I think).

Well, I think Teapo SZ may be worse than those...

Just pulled one of those out of my Straight Power E6 - barely used!
I think it wasn't used until I got it, so it's about 7 Years old with probably less than 100h of use...
And that cap failed...

[Wester547]

Originally Posted by Stefan Payne

Well, I think Teapo SZ may be worse than those...

Just pulled one of those out of my Straight Power E6 - barely used!
I think it wasn't used until I got it, so it's about 7 Years old with probably less than 100h of use...
And that cap failed...

Yes, Teapo SZ is Teapo's worst series (worse than KZG). Not that Teapo SM is noticeably better. It came to my notice that Teapo with relatively recent datecodes (2005+, especially the smaller ones) are much more likely to bulge than the older ones (not that those wouldn't silently dry up).

[momaka]

Originally Posted by Stefan Payne

Just pulled one of those out of my Straight Power E6 - barely used!
I think it wasn't used until I got it, so it's about 7 Years old with probably less than 100h of use...
And that cap failed...

Yes. Sometimes, caps sitting unused can be harder for them than if they were in use. But in this case, Teapo is just cheapo stuff , and that's why they failed quicker in storage.

[Wester547]

Originally Posted by momaka

3.3 V rail:
- Linearly regulated from 5 V rail with 40NF03 MOSFET (100V, 35A @ 100C)
- 1x Fuhjyyu TNR 16V, 220 uF, 6.3x11 mm on the output.
- 15-Ohm load resistor
- 0.035-Ohm, 10W resistor instead of a PI coil on the output

Are you sure that it's a 35A, 100V part? "40NF03" usually suggests a 40A (at 25*C, probably around 30A @ 100*C) part whose drain-source voltage is rated for 30V. If that's a STP40NF03L, it's rated for 28A @ 100*C, and for only 30V, not 100V.

I do have a question about linear regulators, now that we're on the topic. I understand that a linear regulator like LM78xx or 79xx will manage to keep the output voltage constant regardless of the input voltage, so long as the input voltage is higher than the output voltage. Is this also true if you use a MOSFET (such as an N-Channel FET) as a linear regulator instead (as in this case)?

[momaka]

Originally Posted by Wester547

Are you sure that it's a 35A, 100V part? "40NF03" usually suggests a 40A (at 25*C, probably around 30A @ 100*C) part whose drain-source voltage is rated for 30V. If that's a STP40NF03L, it's rated for 28A @ 100*C, and for only 30V, not 100V.

You are absolutely right!
I don't even know where I got these numbers from. Worse, I wrote down "40NF013" in my notebook. Makes me wonder if I am even conscious some days.
Thankfully, your note sparked some further investigation on my part. I opened the PSU and checked: the MOSFET is indeed an ST P40NF03, rated for 40A and 30V.

Thanks. I'm glad there is someone that actually reads my posts .

Originally Posted by Wester547

I do have a question about linear regulators, now that we're on the topic. I understand that a linear regulator like LM78xx or 79xx will manage to keep the output voltage constant regardless of the input voltage, so long as the input voltage is higher than the output voltage. Is this also true if you use a MOSFET (such as an N-Channel FET) as a linear regulator instead (as in this case)?

Yes.

With the MOSFET, however, the output can nearly reach the input voltage, if desired (that is, if the MOSFET is fully turned ON), unlike the 78xx/79xx series regulators, which usually need to have the input voltage at least 2V higher than the output.

[momaka]

Right… after having this PSU sit on the shelf for more than 5 years now, I decided it really must have had enough - time to fix it up and get it going. It had accumulated more dust (both inside and out) from sitting than some of my PSUs that are actually in use. Imagine that!

If you’re thinking, gee, why even bother with this sorry-looking thing?… to which, I can’t really give any logical explanation either. At this point, I already have enough other recapped and well-working ATX PSUs. I figured, maybe restore this PSU and put it back in the original P4 system that it was with. After all, it’s a low-power Pentium 4 PC - hardly anything challenging to power up. Also, while fixing that crappy KDMPower MI-X8775CD PSU, this one started looking like a much better PSU all of a sudden.

The first part I started with, was to change all the small caps and add the missing input EMI/RFI filter parts. On that note, this PCB was not easy to work - there was very little space for the EMI/RFI choke and X caps. And it’s even harder if you’re trying to do it with all-scavenged junk parts from other stuff like I was. However, I managed to find just the right combination of parts to implement a slightly-better-than-nothing basic EMI/RFI filter.

Namely, I had a spare RF choke that I pulled from a dead Nichicon PlayStation 3 (fat) power supply, along with a 0.22 uF and 0.015 uF Matsushita X2-class caps from a Panasonic CRT TV board I saved ages ago. The small 0.015 uF cap probably wasn’t even worth installing. However, it fit perfectly under the tiny space left on the PCB under/ in front of the bridge rectifier (after removing one of the primary caps and bending/adjusting the bridge rectifier position slightly.) Once that was placed, I proceeded to solder the RF choke, leaving the 0.015 uF cap sandwiched in there and impossible to get to anymore. Next came the 0.22 uF X2 cap - it was bigger than the board silkscreen (seriously, could CWT have left less space there ) so I had to solder extension leads on it and position it mid-air over the fuse slightly. At least the 2.2 nF Y2 caps fit properly (these are brand new, by the way.)

While poking in that area and with one of the primary electrolytic caps removed, it occurred to me to test its capacitance. Again, these are Fuhjyyu LP 200V, 680 uF in 22 x 40 mm size. However, I was shocked (though not literally ) that my meter only measured it at 445 uF. That’s not anywhere near 680 uF nor 560 uF! Heck, it’s a bit low even for 470 uF… but passable, I suppose. The other cap (which I didn’t remove and measured in circuit) read 412 uF. I’ll give it the benefit of the doubt that perhaps the other components in the circuit were playing with my meter and hence the even lower capacitance. But in any case, why do these large Fuhjyyu caps have such small capacity? For their size, they should read normal. I doubt there’s a small cap hiding inside that big casing, because the cap itself actually felt to have normal weight. I guess no point in trying to figure that mystery out - it’s a Fuhjyyu after all! On the positive side, that capacity should be just about enough for the 300 Watt rating given by the PSU label… if that’s to be trusted in any way.

Now let’s move on to the next part: the actual recap. First, I did the 5VSB circuit.

The small 1 uF “critical” cap on the primary was changed with Nichicon PJ. Meanwhile, the output was recapped with a United Chemicon KZE 10V, 1000 uF 8 mm dia. cap (for the 1st filter spot), followed by a Rubycon ZLH 6.3V, 820 uF, 8 mm dia. cap (for the 2nd filer spot.) I also replaced the 75-Ohm load resistor I had put with 2x 100 Ohm, ¼ Watt resistors in parallel (for a total of 50 Ohms as a dummy load on the 5VSB.) With this, the 5VSB drew about 3.3 Watts from the wall idling with no load, which is reasonable for a 2-transistor self-oscillating flyback design.

Now remember those 5VSB tests I talked about in post #1 above, where the 5VSB dipped down to 4.85 V with a 1.1 Amp load? Guess what - the new caps took care of that. The 5VSB was 5.13V with no load and 5.12V with 1.1 Amps of load - i.e. the voltage was largely unaffected by my load. One might think the higher capacity of my new caps must be helping… and it probably is. However, I checked the old 16V, 470 uF Fuhjyyu TNR caps that were on the 5VSB originally, and one of them read 8.5 Ohms of ESR, while the other was just slightly over half Ohm ESR! No wonder the 5VSB acted up. That silly hot 10-Ohm resistor originally in there probably cooked the 5VSB output caps to death. In fact, I actually suspected this before I even recapped the PSU with the KZJ caps. So that’s one reason I never put it in use - and now I’m glad that I didn’t, because the 5VSB circuit was just a failure waiting to happen. So for those of you who rely on visual inspection to decide if caps are good or not: this is exactly why you can’t use visual inspection only!

Next, let’s see the rest of the recap.



Given that the 3.3V rail is regulated in a linear fashion and that it uses the 5V rail for its supply, it only made sense to have good proper filtering on the 5V rail. So while the original caps on the 5V were just a pair of 1000 uF Fuhjyyu TMRs, I used 2x 10v 2200 uF Rubycon YXJ. These are the most entry-level low-ESR caps you can probably get – even less so than Panasonic FC, Nichicon PW, and United Chemicon LXZ. They are closer to Nichicon PS / PM, and United Chemicon KYA… i.e. really just a notch above general purpose - which is actually perfectly fine for this PSU, given it’s an old half-bridge design with low switching frequency.

Meanwhile, the 12V rail got a single 16V, 2200 uF UCC KYB cap in 10 mm dia. I figured, if the original 1650 uF Fuhjyyu was supposedly enough for the 10 Amp rating of the 12V rail on the label, then 2200 uF should surely be enough to raise that to 12-13 Amps. The 12V rectifier is up to snuff already, being rated for 16 Amps (and this is an H-bridge PSU.)

The -12V rail also got a similar treatment as the 5V rail: an entry-level low ESR Nichicon PJ rated for 16V, 470 uF in place of the original 16V 220 uF Fuhjyyu cap. But I was a bit stumped what to use for the -5V rail. I could use another 16V 470 uF Nichi PJ… but that seemed a waste for such unused, low-power rail (in addition to being low on those Nichi PJ caps, which I use mostly just for the -12V rail on older PSUs.) So I looked through my cap boxes, and for a moment even considered a general purpose Nichicon VZ. But the thought of using a GP cap, even in this application, didn’t seem right. Out of curiosity, though, I measured the ESR on the two 220 uF caps from the -12V and -5V rails. And guess what? They both showed about 0.4 Ohms ESR. My VZ, on the other hand, showed about 0.25 Ohms - not great either, but still 2x better ESR and with higher capacity, too. So I ended up using the 16V 470 uF VZ on the -5V rail.

Lastly, the 3.3V rail: here, I also pulled another lazy turn by leaving the UCC KZJ cap in as a test. After all, this is a linearly-regulated rail, so it shouldn’t stress the cap a whole lot or care too much about its ESR. However, just as a “safety net”, I also added a 10V 220 uF UCC KY cap in parallel with it (right in the spot where the 3.3V rail dummy load resistor used to sit before.) Thus, even if the KZJ went completely open, that UCC KY should still keep the 3.3V regulator happy - after all, 220 uF is what the 3.3V came with from the factory.

more in next post...

[momaka]

And now, about the original Rulian Science 80 mm fan… it was starting to stick. It still worked, but spun slow. No doubt I can repair it (and I did!) But what bugged me about it is that it had the cheaper round plastic housing, whereas the PSU case had 4 additional cutouts in the exhaust grille for a “normal” fan with rounded case edges (not sure how to describe it, so just see the picture below.) As such, the original fan would suck air in through these 4 cutouts rather than blow air out through them, which is just silly. So I decided to swap the fan with another.

This “new” “DF0802512sem” fan (whatever brand that is) came from a junk Linkworld PSU I replaced about a decade ago for someone. It was a little dusty, but running well otherwise. It had the normal case with rounded corners on the output side, so it could use the 4 extra exhaust vent stamps on the PSU’s case.

While at it, I also had to decide what to do about the fan speed. This PSU came with no fan controller. Instead, its fan was wired to run at full 12V all the time (and probably why that Rulian Science fan was starting to seize up.) I played around with some diodes and resistors on my bench at various voltages… and determined that 6V is about where the fan is relatively quiet but still moves a decent amount of air for a lower PSU load. To get 6V without designing a speed controller, I connected the fan leads between the +5V and -5V rails with two parallel 100-Ohm resistors in series with the fan. In retrospect, I might have to come back and revise this, due to the -5V rail dipping a bit too much and making the fan run only on 5.5V. But for the time being, it seems to push enough air for lower loads on the PSU.

Speaking of the case, this was the last part. I highly dislike restrictive air intake holes – something often seen on many cheapo PSU’s. So I had to change that.

This is where it really pays to have a good corded drill. It made quick work of drilling all those holes with ease. What took the majority of the time… deburring! In particular, those holes close to the bottom of the case just can’t be accessed easily, so I had to manually file them down. Nonetheless, I’m very pleased with the results and the PSU should be able to cool down a lot better now.

A few more mods I did but didn’t show: before finally installing the PSU PCB back in the case: swap the crappy thin 20 AWG, 300V wires from the plug to the PCB with slightly better ones (though not by a whole lot): 20 AWG, 600V, 105C. And while at it, I also swapped the voltage selector switch wiring from a junk GenMax PSU. Likewise, here I went from 20 AWG, 300V to 20 AWG, 600V. The wires appear to be real copper too.

All in all, that more or less concludes the mods and work I did on this PSU. I won’t tell you how much time (days) I spent on it. But let’s just say I did enjoy it. Up next: a few simple load tests. Stay tuned.

[momaka]

Well, I can’t really say precisely how the PSU outputs did in terms of ripple, because I don’t have an o-scope to measure it. I doubt it’s bad, though. What isn’t so good, however, is the voltage regulation on the 12V rail. Namely, the 12V rail is very weak. How did this PSU even power a 12V-based system in the first place?!

For starters, after recapping, modding, and putting everything back together, I connected the PSU’s input AC in series with a 470-Watt heating element to limit power, should the PSU have problems anywhere. As stated, I was able to test the 5VSB up to 1.1 Amps of load, and the voltage hardly changed at all from its nominal 5.13V. The main PS told a different story, though. Using a single 12V20W halogen bulb on the 12V rail and a 12V6W incandescent bulb on the 5V rail, I turned on the PSU and checked outputs. The 3.3V was sitting comfortably at 3.356V. The 5V rail was also more or less happy at around 5.15V. The 12V, however, was around 11.3V - out of spec on voltage! :\ Ok, maybe it doesn’t behave at lower power with a heavier 12V load? So I switched these load bulbs around (with the 12V20W now drawing only about 5.5 Watts from the 5V rail and the 12V6W bulb drawing 6 Watts from the 12V.) This improved things a little, with the 12V rail going up to 11.6-11.7V, while the 3.3V and 5V remained mostly unchanged.

To make sure the series 470-Watt series heating element wasn’t playing with the results, I removed it and performed the same test mentioned above. The results were exactly the same. So I did one more test: put both 12V20W and 12V6W bulbs on the 5V rail, with the 12V left unloaded. I powered On the PSU and… finally the 12V looked more normal at around 12.1V. But all this told me is the PSU doesn’t like heavy 12V loads and prefers 5V-heavy loads… which is exactly the opposite of what I imagined would be the case, given that the 3.3V rail is generated from rectified 5V rail, thus suggesting that there would always be a decent load on the 3.3V/5V rail output.

This called for a real-world test: connect the PSU to a motherboard and see what happens. And I just had the perfect motherboard for this: a Biostar NF325-A7 socket 754 motherboard that just recently quit on me (no POST, but CPU eats power.) The motherboard does not appear to use the 5V rail for anything, but does use the 3.3V rail for the RAM and chipset, both of which don’t draw much current. On the other hand, the CPU (Athlon64 3400+ Clawhammer) heats up and draws almost full-load power (90 Watts from the wall, using another PSU with APFC for more accurate power measurement… so with the inefficiencies factored in, that gives about 70-75 Watts on the 12V rail), despite the board not POSTing. In other words, this was a perfect 12V-heavy and light 3.3V/5V load. I connected the ADP MPT-301 PSU to this motherboard and hit the power button. But the PSU was like, Nope! This ain’t for me. The CPU and PSU fans spun up for a about a second and then the PSU just shut off. Hitting the power button again and again did the same thing each time. I check the maximum voltages the rails would reach, using my multimeter’s “Max” function (however accurate that may be, since the PSU doesn’t really stay On for that long.) The 12V rail reached - at least according to the meter - 7.5V. Meanwhile, the 5V rail didn’t make it past 2.5V. It was clear the PSU just wasn’t liking this 12V-heavy and 5V-light load.

So I tried something else. I have this 3000 Watt, 240V heating element from a scrapped dryer. It has many turns of Nichrome wire (I plan on making a basic PSU load tester with this some day… whenever that will come.) So I can easily put jumper leads on any part of the Nichrome wire to use as a load… which is what I did here – I connected the 5V rail to a set of coils for approximately 2x 5.555 Amp loads (11 Amps total). Of course, with the resistance of my jumper leads, the actual resistance was higher, so I am guesstimating (was too lazy to hook a meter to try to measure the current) I loaded the 5V rail with somewhere around 9 Amps only (i.e. 45 Watts). With that and the 12V20W halogen bulb on the 12V rail, I powered up the PSU again. And guess what? – It ran perfect. The 5V rail was at around 5.03V, while the 12V rail was comfortably sitting at 12.08V. The 3.3V was stable at 3.347V too. The -5V was still a tad too low from the fan’s measly load, giving only about -4.6V. Likewise, the -12V was also low at -11.2V… but that’s how these two negative rails were pretty much in all of my above tests. 5VSB was solid at 5.13V, regardless of the power state of the main PS.

With this last test above, my Kill-a-Watt meter was showing a power draw of about 94.6 Watts from the wall. I should note here that I also had the 12V6W bulb on the 5VSB, which makes the PSU draw about 6.6W from the wall. Therefore, the main PS was likely pulling only about 88 Watts from the wall. And given that I had only about 45 + 20 + 2-3 Watts total on the output of the main PS… that comes out to 67-68 Watts. Therefore, this suggests the efficiency was somewhere around [(67/88)*100] = 76% - not bad for an old clunker like this!

Of course, I’m still disappointed about the weak 12V rail performance. I guess this will be yet just another recapped PSU to add to my 5V-heavy collection – and it seems like I have more of these than I do units with good 12V rail. Given the current crazy prices for “retro” PC hardware, though, maybe this isn’t such a bad thing, as I know some people are looking for old 5V-heavy PSUs like this. This may not be the strongest one out there, but it will likely be fine for just about any system, save perhaps for something like a high-end dual Athlon MP system with a Radeon 9700/9800 video card (which also draws a lot of power from the 3.3V and 5V rails.) That being said, I’ll probably come back to it again and do more tests to see how much power I can get from the 12V rail with various loads on the 5V rail. I suspect it won’t be much. So more than likely, I will need to make a revision label for the label on the PSU to indicate the outputs more accurately. After this, I may also remove the 4-pin 12V CPU connector – after all, if the PSU can’t power a 12V-based PC properly, what’s the point in keeping it there… which is a shame, as this one uses proper 18-AWG wire, unlike it’s newer brother, the Turbolink LC-A420ATX. I might even swap the rectifiers around too. But that will be left for another day.

Anyways, I think that's more than enough words already about this PSU. It probably wasn't worth wasting any time on, but some of you already know me by now - I waste time on a lot of useless crap like this.
Hopefully there's some useful modding info and ideas in all of the above mumbo-jumbo, though.

[momaka]

Alright, just for a quick update on this one again, if anyone is even following this thread (if not, at least this might help me to remember what I did. )

So basically, I was testing a few old motherboards tonight. One of them, an Intel D845WN, was giving me trouble – looked like it wouldn’t work in the case, especially with the CPU heatsink installed and tightened, due to previously coming with a CPU HS bracket that severely warped the PCB. So I removed it from its case and laid it on my bench with the CPU heatsink only resting on the CPU (no retention brackets installed.) Anyhow… I was a little mad after a while trying to get that board to POST and I didn’t have another PSU handy on my bench, so I thought why not try this ADP MPT-301 again and see if anything smokes, considering its weak 12V rail.

Nothing smoked, however. Instead, the PSU ended up working fine – no overheating or unstable operation (once I got the RAM issues sorted, that is.) Then again, it might be worth noting that this Intel D845WN motherboard is what was coupled with the ADT MPT-301 in an older custom PC tower. So if this PSU had powered that motherboard for all its life prior to the recap above, it surely shouldn’t have any problems after (otherwise, that would mean I done goofed on something.)

In terms of hardware, I had a 2.8 GHz Pentium 4 Northwood (400 MHz FSB version) CPU running in the motherboard when I did this test (a slight upgrade from the 1.6 GHz P4 Willamate that originally came with the build.) So with that, a Radeon 7000 video card (one of my test AGP cards), and a single 3.5” HDD, the 12V rail did dip as low as 11.60V under full CPU load… but that’s still in spec. Otherwise, when idle, the 12V rail was sitting at around 11.84V. The 5V rail, on the other hand, hardly moved at all: 5.15V when the CPU was at 100% load and taxing the 12V rail, and 5.12V when the CPU was idle and 12V rail not used as much. So the 5V rail really is stable as a rock on this PSU.

Power-wise, my Kill-a-Watt was indicating about 60 Watts when the PC was idle and 110 Watts as the worst case when loading the CPU the most I could. I probably ran the whole system for a good 1 hour at medium to high load, trying different benchmarks and stress tests, as I was also curious what temperatures I would get with the CPU heatsink I was using. In any case, given that a PSU like this has typically around 75% efficiency, I think it’s fairly safe to guesstimate the output DC side power consumption as around 45 Watts idle and 83 Watts with max load… which is more or less inline with the maximum CPU TDP (69 Watts typical, 80 Watts max.)

Thus, the ADP MPT-301 is actually capable of running something basic like a Pentium 4 system from its day. Sure the 12V rail was dipping a bit on the low side… but it was still in spec, at least voltage-wise. Given that the 12V rail is rated for 10 Amps max and that I was loading it down with maybe 6-7 Amps total, I think the results are reasonable here.

The last thing I did after all the long CPU benchmarks and stress-tests: turn off the PC, unplug, and quickly take apart the PSU to check temperatures inside. After all, don’t forget that I installed a mod in it to slow down the fan to run on ~6V. When I opened the PSU, surprisingly nothing was too hot. The primary heatsink was very warm, but not hot by any means – most certainly less than 50°C. The secondary heatsink was by far the hottest component, though I could still put my hand on it and not get a burning sensation, suggesting the temperature was around the low to mid 50’s Censius… and that’s with a room ambient temperature of 72-73°F / 22.5°C. This suggests the secondary heatsink may reach low to mid 60’s °C in an average PC case, where the temperature is usually closer to 30-35°C. Nonetheless, I think these results are pretty normal for an older and less-efficient PSU like this. Therefore, I think I’ll scratch the ideas I had at the end of post #19 and instead call this PSU pretty much done / ready to use. I just might change the label a bit at some point to more accurately reflect the maximum power this PSU can provide. While the primary may be capable of 300 Watts with the BJTs in there, I think the rectifiers on the secondary and the voltage regulation will probably limit this PSU to closer to 200 Watts usable… and that’s if both the 12V and 5V rails are heavily loaded, which doesn’t really happen often in a real-world PC (be it modern or older 5V-heavy design.).

Thus, I’ll go ahead and call this PSU complete.