LC Power LC420H-12-US (office series) rebuild

[Pentium4]

I'm sure some of you are familiar with this brand who normally is in Europe. I want to use this in a very low powered system. The PSU in this system now is way overpowered and I want to use it for something else.

This system at its peak uses a mere 75W so this thing will be fine I believe. However, I do think it needs some modifications.



Overhead shot:


Look at those input wires! Nice. They're all 18 gauge with lots of insulation. The input is solid and is only missing MOVs. Speaking of, does someone know of a MOV part that I can have on hand from Digi-Key or Mouser in general to have on hand for PSUs that don't come with MOVs?

The switchers are 13007 TO-220 unfortunately with a small heatsink but I won't be pushing this past maaaybe 120W peak.

The transformer is of good size, no issues there.

On the secondary, we've got:
45A FET on the 3.3V (40N03GP) This must be derived from the 5V?
x2 20A schottky on the 5V (S20C45C)
20A ultra fast on 12V (U20C20C)

Here's my main concern with this, there is only one 1000uF 16V (!) cap on the 12V rail! And they jumpered where a PI coil should be. I was going to put in a 2200uF 16V Chemi-Con KYB. The thing is, the ESR is half of the original. Would it be better to put in a 1000uF 16V Rubycon YXJ and then add in a PI coil from the 12V rail of a scrap Solytech PSU?

[CapLeaker]

Re: LC Power LC420H-12-US (office series) rebuild

I guess I’ll leave that up to you what you wanna do with or what and what not to mod.

I’d like to say I am a lazy f*ck but in realty I am just waaay to busy running around all day fixing crap. I’ve got tons of power supplies running in various shapes, voltages, tons of computers, other critical equipment, Blabla…

What computer PSU’s go: I run them unil they are dead for whatever reason. I have a spare for each PSU, as this stuff runs 24/7/365. I swap out the broken PSU, repair / refresh the old PSU and it goes back to the shelf. I am not making myself the head if it has a pi filter, Blabla… because the computer is not (special? Yes it is, ver special!) a power hungry high end monster. These are low end computers, that run certain software 24/7/365 as a requirement in order to keep this outfit going and there is plenty of RF around in that spot.
It’s just a low end computer! It will be fine!

[Pentium4]

Re: LC Power LC420H-12-US (office series) rebuild

I understand where you're coming from. However, I want to improve it if I can. I have some time for this and I do care about my equipment as well as enjoy working on and improving hardware. Especially since I don't have access to lots of free equipment like I used to.

This is for one of my personal computers but I'd still rather just work on it once and forget it, like so many of my projects in the past.

[CapLeaker]

Re: LC Power LC420H-12-US (office series) rebuild

I guess you just answered yourself the question you asked. Filter it is!

[PeteS in CA]

Re: LC Power LC420H-12-US (office series) rebuild

Depending on the cooling air, I'd guesstimate that PSU would be OK for 150W-200W.

On the output filter, the 'lytic handles the switch frequency ripple, and the ferrite rod inductors filter out spike noise from switch device and rectifier turn-ons and turn-offs. IMO, if you want to have some confidence that spike noise won't mess with your computer's operation, I'd go with the YXJ cap and ferrite rod inductor (I assume the KYB is larger and would prevent installing the inductor as well).

[Pentium4]

Re: LC Power LC420H-12-US (office series) rebuild

Hey Pete, good to see you!

The KYB cap is actually 10x30 so it would fit in the same footprint. But it has an ESR of 0.027Ω versus the Chun Shing 1000uF 16V 10x20 cap has an ESR of 0.06Ω which is quite a drop. I would like to keep the controller happy so it doesn't create more ripple. So I think the Rubycon YXJ 10x16 @ 0.062Ω is probably the best bet with the PI filter.

The PI filter is coming from another half bridge unit on the same rail, do you think this is the most likely outcome for it being compatible?

[PeteS in CA]

Re: LC Power LC420H-12-US (office series) rebuild

I'm pretty sure the inductance of ferrite rod inductors is pretty low, so I doubt a slightly different part would affect stability.

I might not post a lot, but I look in on most days.

[momaka]

Re: LC Power LC420H-12-US (office series) rebuild

The input is solid and is only missing MOVs. Speaking of, does someone know of a MOV part that I can have on hand from Digi-Key or Mouser in general to have on hand for PSUs that don't come with MOVs?

I don't have any, but I remembered I had an Apevia 680W PSU that I had taken a picture of the MOV on its input (not across each cap), and it had a part number TVR10471. So maybe look up something similar to that? This would be for an MOV across L-N AC input, though. If for installing across 200V-rated caps, then look up a part number from an MOV from another PSU.

For the TVR10471, I found the following specs from datasheets:
Max Continuous Voltage AC (RMS): 300
Max Continuous Voltage DC: 385V
Varistor voltage @ 1mA DC: 470V (423-517)
Max surge current 8/20 us: 4 kA
...
which led me to the following parts on Digikey:
MOV-14D471KTR
B72214S0301K101
ERZ-E08A471
V10E300P

I'm no expert on MOVs, though, so not sure if those are 100% perfect fit for use across L-N use.
As for MOVs across each cap, I don't have any part numbers I can look at right now to find any comparable parts.

The switchers are 13007 TO-220 unfortunately with a small heatsink but I won't be pushing this past maaaybe 120W peak.

Should still be fine for up to 150-200W load if they are genuine 13007.

Here's my main concern with this, there is only one 1000uF 16V (!) cap on the 12V rail! And they jumpered where a PI coil should be. I was going to put in a 2200uF 16V Chemi-Con KYB. The thing is, the ESR is half of the original. Would it be better to put in a 1000uF 16V Rubycon YXJ and then add in a PI coil from the 12V rail of a scrap Solytech PSU?

Why not add the missing PI coil and put in the 2200 uF KYB?

Well, I guess the question is if you intend to use this PSU with a 5V or 12V -based PC. If 5V -based (i.e. Pentium II/3 or Athlon/Athlon XP), probably fine to go with the same 1000 uF capacitance, since most of the load will be on the 5V rail and not the 12V rail... so the YXJ cap should do. But if the PC will be 12V -based, then a higher capacity cap on the 12V rail would filter more ripple current there... so in that case, the KYB might be better, provided the PSU is OK with it (should be, in most cases.) In any case, adding the missing PI coil shouldn't hurt anything.

I'm pretty sure the inductance of ferrite rod inductors is pretty low, so I doubt a slightly different part would affect stability.

Indeed.
They are mostly meant to provide additional filtering for high frequency noise from the switching.

Those rod inductors are somewhere in the range of 1 uH or less, whereas the output toroid is typically 10+ uH on each rail (IIRC from one of Unique's post, actually somewhere in the 20-something uH range.) Unfortunately, my cheap meter can't measure less than 0.02 mH (20 uH), so I can't really measure the windings of output toroids properly.

Thus, adding 1 uH to 10-20 uH is only a 5-10% change in inductance, which probably isn't enough to upset the feedback loop.

I might not post a lot, but I look in on most days.

Sadly, I've been doing this as well.

My current excuse is I haven't unpacked a proper clicky keyboard yet... and I really dislike typing on laptop keyboards made after 2006-2007.
In reality, I'm just struggling to do as much in a day as I could before in terms of electronics repair / analysis.

[Pentium4]

Re: LC Power LC420H-12-US (office series) rebuild

I don't have any, but I remembered I had an Apevia 680W PSU that I had taken a picture of the MOV on its input (not across each cap), and it had a part number TVR10471. So maybe look up something similar to that? This would be for an MOV across L-N AC input, though. If for installing across 200V-rated caps, then look up a part number from an MOV from another PSU.

I couldn't find any in any of mine that were visible. Surprisingly, I usually just see the two slots in between the two big caps. I ended up buying some of these:


I picked these because I thought that you want the Max DC voltage rating to be 200V so that it protects the main caps? Do you think these are still going to be suitable for non-PFC PSUs where it has two slots between the big caps?

Should still be fine for up to 150-200W load if they are genuine 13007.

They are. However, the fan control thermistor is just sitting on the board in the middle of the output cables That won't do them any favors.

Why not add the missing PI coil and put in the 2200 uF KYB?

My initial thoughts were that the ESR was too much lower than the original. I think I remember you saying for the sake of the feedback loop staying happy, ESR is more important, but for output ripple to be lower, you want capacitance to be higher? Cause I have a 3300uF 16V Rubycon PX that has the same ESR as the 1000uF 16V CS WE. Would that work too? I know that's not ideal but I'm curious. I feel so rusty

Well, I guess the question is if you intend to use this PSU with a 5V or 12V -based PC. If 5V -based (i.e. Pentium II/3 or Athlon/Athlon XP), probably fine to go with the same 1000 uF capacitance, since most of the load will be on the 5V rail and not the 12V rail... so the YXJ cap should do. But if the PC will be 12V -based, then a higher capacity cap on the 12V rail would filter more ripple current there... so in that case, the KYB might be better, provided the PSU is OK with it (should be, in most cases.) In any case, adding the missing PI coil shouldn't hurt anything.

It's going to be in a low powered 12V system. It's honestly so low powered that even the 1000uF YXJ would probably be fine, but I'm still curious what would be best cap wise. I will definitely add the PI coil

It's a Celeron G6900, one NVME drive, one 3.5" SATA HDD, one DVD drive, 2 RAM sticks. No video card. The CPU uses a mere 14W at 100% CPU usage and 2W idle. Crazy how efficient it is! The PSU uses 8W running with a paper clip, I'd honestly be surprised if it even crosses 50W total.

[momaka]

Re: LC Power LC420H-12-US (office series) rebuild

I couldn't find any in any of mine that were visible. Surprisingly, I usually just see the two slots in between the two big caps. I ended up buying some of these:
https://www.badcaps.net/forum/attach...1&d=1692316721

I picked these because I thought that you want the Max DC voltage rating to be 200V so that it protects the main caps? Do you think these are still going to be suitable for non-PFC PSUs where it has two slots between the big caps?

Probably should work fine, though I don't have much knowledge when it comes to picking MOVs.

Also, electrolytic caps can take a certain surge voltage that is higher than their rated voltage. For 200V caps, I think most manufacturers allow 20-50V higher (so about 220-250V max DC) for short periods of time (a few seconds, IIRC.)

So technically, the MOV could be rated a little higher than 200V.

That said, I just looked through my pictures collection more thoroughly today so I could see which units have MOVs between the primary caps. Many did, but most are heatshrinked (usually done to contain smoke and shrapnel damage when the MOV explodes from a big surge... though it's not a requirement.) However, I found one PSU that didn't have a heatshrinked MOV... and ironically, it was the ePower EP-350XP you gave me a while back. Perhaps it was also dumb luck, but I found a picture of that PSU where I could read the part number on one of the MOVs without having to take apart the PSU. The MOV was a TVR10271.

From the TVR10271 datasheet:
Max Continuous Voltage AC (RMS): 175V
Max Continuous Voltage DC: 225V
Varistor voltage @ 1mA DC: 270V
Max surge current 8/20 us: 1 time 2.5 kA; 2 times 1.3 kA

So it looks like you're pretty close with your selection. I doubt the slightly lower voltage rating will be an issue.

They are. However, the fan control thermistor is just sitting on the board in the middle of the output cables That won't do them any favors.

Hey, who knows, maybe that is the hottest area in the PSU. (I doubt it, though.)
This is one of those cases, where if you have a thermal camera, you can see which is the hottest part of the PSU and place the thermistor there. Usually it's either the secondary heatsink & close to the rectifier that passes the most current or it's the output inductor. The BJTs on the primary should run relatively cool up to about 40-50% PSU load. For 13007 BJTs, that would be about 100-125 Watts. At least that's been my observation with H-bridge PSUs: the primary heatsink is always much cooler than the secondary up to 150W load.

My initial thoughts were that the ESR was too much lower than the original. I think I remember you saying for the sake of the feedback loop staying happy, ESR is more important, but for output ripple to be lower, you want capacitance to be higher? Cause I have a 3300uF 16V Rubycon PX that has the same ESR as the 1000uF 16V CS WE. Would that work too? I know that's not ideal but I'm curious. I feel so rusty

Other way around : feedback loop cares more about the capacity... or more specifically, that at least a certain capacity is present. Anything higher is good and only makes the feedback loop more stable.

ESR matters more in terms of how the output filter is designed for the PSU (i.e. inductance of output toroid and capacitors' capacitance, along with taking into account the PSU switching frequency.) The output filter is an L-C -type filter, but there is also a small series resistance in there, coming mostly from the inductor's wire DC resistance and the capacitor's ESR. So that essentially makes it an RLC "tuned" circuit. Decreasing its series resistance causes this circuit to produce more and more of its own resonant frequency. This is called "ringing", because it's analogue to how a bell or a string on a musical instrument vibrates when it's struck. Depending on how the feedback loop is designed, this ringing noise/ripple from the filter may be completely ignored or completely mess up the regulation of the PSU because it can be at a frequency different from the PSU switching frequency, and thus not expected by the feedback look. Most PSUs don't have a problem if the ESR of the output caps is decreased several times relative to what was originally there. But some do. So that's why it's a bit of an "experiment" when caps with to low of an ESR are used... and hence why polymer caps are not recommended for recapping PSUs, unless the PSU output filter was specifically designed to work with ultra-low ESR caps or polymers.

If the oad on the 12V rail is expected to never go over 5-8 Amps, the 1000 uF YXJ cap should probably do OK. Ripple might get quite high at the 8-10 Amp mark (or possibly even over the ATX spec.) Upping the capacity to 1200-1500 uF should make things run more quiet in terms of R+N. And 2200 uF would be OK for up to 12-14 Amps load.

In regards to the 3300 uF Rubycon PX cap - that actually might work quite nicely in there, provided it fits. The only caveat is that it's a 2000h -rated cap, so shouldn't be used if it's close to a high heat source... but I don't think that's the case here with this PSU.

It's going to be in a low powered 12V system. It's honestly so low powered that even the 1000uF YXJ would probably be fine, but I'm still curious what would be best cap wise. I will definitely add the PI coil

It's a Celeron G6900, one NVME drive, one 3.5" SATA HDD, one DVD drive, 2 RAM sticks. No video card. The CPU uses a mere 14W at 100% CPU usage and 2W idle. Crazy how efficient it is! The PSU uses 8W running with a paper clip, I'd honestly be surprised if it even crosses 50W total.

Nice!... and LOL! - The PSU wastes almost as much power as the CPU itself under nominal loads. But even if you used a 80Plus Bronze or Silver PSU, it would still probably waste as much power, unless it was a very low-power -rated PSU (like under 150 watts.) Most PSUs, regardless of 80Plus ratings, still tend to get rather poor efficiency under 10% load.

I set up an SFF desktop PC for my parents, and it has a Pentium G3260. Though being a 55W TDP CPU, I rarely see the system use more than 30 Watts with regular use. Idle, I think it went down to 15-20W and peaked around 40W doing very CPU & GPU intensive stuff (some older games.) Runs nice and cool. Great for when it's hot in the summer. In the winter, though, I usually switch to my power-hogging. It's not wasted power if it goes towards heating your house and if that's what you want.