Infinity IN08-450 450W Power Supply Review

[c_hegge]

You can send them my way any time. I collect them throughout the year. The roundup is all of the cheap and nasty PSU's I've collected throughout the year, so feel free to contribute a cheapie any time. PM me if you're interested.

[Wester547]

That's the key, switching speed. You can build a higher powered PSU with half-bridge, but the pennies you save there cost elsewhere. You'll need decent heatsinking, more bulk storage & larger main transformer. For example a Delta/Newton 350W dual P2/3 era PSU (~2000), employed a 10A bridge, 2x 1000uf primaries & ERL39 transformer all for only 15A@12V & 35A@5V. Now, the VRD specs were quite different then, but when circa 2010 half-bridge designs have 4x1.5A diodes, 2x560uf primary caps & ERL35s all on ATX spec 2.01 PCBs yet tout >20A@12V - call me unimpressed. That's without even looking at the secondary side.

Well, I did know that MOSFETs have quite an advantage over bipolar transistors, though I've recently read that the difference in switching speed (between a MOSFET and its bipolar transistor 'equivalent') is gigantic. If that's the case, it explains why those older half bridge Huntkeys exploded when loaded to their peak rating after only 10 minutes at 40-50*C room temperature (in those Hardware Secrets reviews) - they all used FJA13009 (TO-3P), FJP13009 (TO-220), FJP13007 (TO-220), KSE13007 (TO-220), KSE13009 (TO-220), KSE13009L (TO-3P), or 2SC3320 (TO-3P) bipolar switching transistors. They were built well otherwise, employing either 120mm or 140mm fans, thick heatsinks, lots of input and output filtering, 39-44 size transformers, a largely ovespec'd secondary, huge input capacitors (820uF-1000uF) and large rectifying bridges (15A without a heatsink), and yet the switchers still went bang at their labeled rating... probably because half bridge only uses bipolar transitors on the primary.

Not only that, though, but this review of a very low end Huntkey suggests that even with their thermally controlled fans, the fans do not rev up fast enough to compensate for the increased load and heat (I used to have a Huntkey very similar to it. It didn't give me problems, but after I upgraded to another PSU, I found a bloated KSC "GL" capacitor on the +5V rail, 2200uF/10V - guess I should have expected that since it was literally touching a resistor next to it, and since the +12V rail's PI filter coil was literally touching it too...). Since cooling is a huge factor in how well PSUs perform, it's no surprise those older PSUs could only do 100W below their label continuously (though I have to admit that I've never been that impressed by Huntkey's choice of secondary capacitor filtering, either... and maybe undersized, dipped red coupling capacitors had to do with it since they were half-bridge?).

That older Dell/Delta/Newton isn't a single switch forward, is it? Those older Dell Newtons were very overbuilt and extremely well designed (the Dell Hipros and Lite-ons of that time, likewise), off loading an ample amount of stress from the PSU and allowing even crappy capacitor brands to hold up well in them. I still have a Dell/Newton Power 250W (NPS-250GB) whose Ltecs have not failed yet after 18,000 hours of use and 5,000 power cycles (since 2001). It has very thick heatsinks, a very powerful fan, a single switch forward design, a great transformer design, an overspec'd primary and secondary side, and a very nice voltage filtering circuit, so while I know bad brands can go at any moment I wouldn't be surprised if it lasted until 30,000 hours of use, assuming the sleeve bearing (Sunon) fan doesn't croak before that (it's still blowing out as much air as it always did). And I also think Newton Power underrate their PSUs brutally when you see how well they're built inside.

[c_hegge]

From the P2-3 era, it could quite possibly be half bridge, although it's been a while since I've dealt with a PSU that old. I occasionally pull 250W PSUs from older Dell Dimension 3000 and 4700 PCs (usually Newton). They are single transistor forward, and are often still holding up fine even after 7+ years.

[steenss]

That older Dell/Delta/Newton isn't a single switch forward, is it?

Nope. Solid as a rock, though. Don't have them anymore. Non-standard ATX wiring. These also had large can size low/non-aqueous Ruby/UCC caps. This is the case:
Ahem, circa 1998...

I occasionally pull 250W PSUs from older Dell Dimension 3000 and 4700 PCs (usually Newton). They are single transistor forward, and are often still holding up fine even after 7+ years.

I've got an IBM PPC PSU somewhere. What a nice piece of work. Dual PCBs, double forward, indy secondary, Rubys/UCC large cans everywhere. Can't remember the OEM. PM sent.

[Th3_uN1Qu3]

This may sound silly, but... I take it there isn't really such a half bridge design that uses MOSFETs on the primary, is there?

Yes there is. Look up the L6599D. It's used in some 230W power bricks from Delta (rebadged DAP015D) but it's the same thing. Resonant half bridge with MOSFETs. Efficiency way above 80+. Just because it's not common, it doesn't mean it doesn't exist or can't be done. It's just that driving a double forward is easier.

Even the old 13009s in TO-3P can put out some decent power if you know how to use them. I've pulled 650W from one pair, with only a regular 80mm fan for cooling. And they did that for hours, at 35C ambient temp. They didn't blow - i just turned it off because it got unbearably hot in the room from the load resistor. I did that 2 years ago.

I also know of someone who did 4kW with a full bridge of 'em, but can't verify the story.

[c_hegge]

Yes there is. Look up the L6599D. It's used in some 230W power bricks from Delta (rebadged DAP015D) but it's the same thing. Resonant half bridge with MOSFETs. Efficiency way above 80+. Just because it's not common, it doesn't mean it doesn't exist or can't be done. It's just that driving a double forward is easier.

Thanks. I haven't seen that one before.

[Wester547]

This may sound silly, but... I take it there isn't really such a half bridge design that uses MOSFETs on the primary, is there?

Yes there is. Look up the L6599D. It's used in some 230W power bricks from Delta (rebadged DAP015D) but it's the same thing. Resonant half bridge with MOSFETs. Efficiency way above 80+. Just because it's not common, it doesn't mean it doesn't exist or can't be done. It's just that driving a double forward is easier.

Well, it was posted on the Badcaps.net forum that considerable modifications would need to be made to the primary circuit of at least a run-in-the-mil half bridge PSU in order to use FETs instead of BJTs. I don't think most half bridge PSUs would take FETs without the linked changes, and as you said, it's easier to use FETs in forward topology than half bridge. I didn't think it couldn't be done, though, which is why I asked. I agree that efficiency is important, but so long as it's above 70%, other things matter to me lots more, like the amount of ripple in any given design. TO-247/TO-3P parts obviously have the advantage of covering more surface area when it comes to heatsinks (their ability to conduct heat away) and power dissipation. Speaking of resonant half bridge and those ST FETs, though, I have read that there are in fact BJT/MOSFET hybrids...

Even the old 13009s in TO-3P can put out some decent power if you know how to use them. I've pulled 650W from one pair, with only a regular 80mm fan for cooling. And they did that for hours, at 35C ambient temp. They didn't blow - i just turned it off because it got unbearably hot in the room from the load resistor. I did that 2 years ago.

I remember reading that. Not sure if you'd be able to do the same thing with an 115V input as that would be less efficient but yes, cooling does make a big difference, so does hugely overspec'ing the secondary side, as you have said before. In fact, that probably makes the biggest difference and that's probably how the commonly known OEMs design PSUs such that they are not stressed and last a while even with underpar capacitors - hugely overengineer the secondary side. That is probably where the efficiency really is affected, though I think the primary side matters lots too, like the "On" resistance specification of FETs and the input capacitors, etc.

I think the biggest kicker though is switching speed, with regard to switching on times, switching off times, fall times, rise times, etc.... BJTs are measured in millionths of a second but MOSFETs are measured in billionths of a second so I think the difference between them at least in that respect is very large (as c_hegge said, switching on and off times are when switching losses are at their highest). Of course, a number of things in the overall design probably do limit the difference between them, like power dissipation, the overall capability of the PSU, the superior turns ratio of half bridge, a full wave rectified output (which I think considerably increases what the filtering on the secondary can do... but then you can get away with overspec'ing less in overall wattage potential with FETs on the primary, at least I have noticed that often in PSU reviews) etc. I have also read that you can fit more current into a smaller die with BJTs (probably because they handle voltage drops better than MOSFETs in terms of current increase), and that BJTs, with a high enough drive current and voltage, have a RDS(on) resistance that is half that of the highest quality FET. In the application of SMPS, though, I think MOSFETs are by far supreme. An interesting note about them though is that their switching speed decreases with temperature rises.

I also know of someone who did 4kW with a full bridge of 'em, but can't verify the story.

I can believe it with full bridge.