Power supply build quality pictorial. part 2

[Behemot]

Re: Power supply build quality pictorial. part 2

You got your math wrong by one decimal place. 100kHz is 10us, 40kHz is 25us.

Oh , yeah, right, wrong unit conversion

[Th3_uN1Qu3]

Re: Power supply build quality pictorial. part 2

Nope. He reckons that the limit comes from half bridges having a 50% duty cycle, and that only one diode is used in the math, like in a forward design. If we used both diodes and his formula (with the same 50% duty cycle), we get double the rectifier's rating.

Mains is 50% duty cycle too... Oh wait, mains is a sinewave so that's nicely symmetrical by definition, but that's another topic.

The current output of a half bridge for a given voltage (or any power supply for that matter) has nothing to do with the duty cycle. It has to do with the input voltage and the turns ratio of the transformer.

[c_hegge]

Re: Power supply build quality pictorial. part 2

^
And that's exactly how I've always understood it - The duty cycle of the switchers doesn't affect what the secondary rectifiers can handle.

[Th3_uN1Qu3]

Re: Power supply build quality pictorial. part 2

Or to put it in another way - if the input voltage is high, the diodes can run out of steam even at a low duty cycle of the primary section.

[momaka]

Re: Power supply build quality pictorial. part 2

The way I calculate the maximum safe current for the secondary of a single forward PSU design is this:
Take the full rating of one side of the schottky diode (i.e. half the capacity of the whole schottky diode) and add to that half of the full rating of the other side of the schottky. The total is the continuous output of the PSU for the rail on which that schottky diode is on.

So let's say we have a SBL3040PT (a 30 A rated schottky) on the 5V rail. Each side of the SBL3040PT can do 15 A. So for one side we have 15 A and for the other we have 15/2 = 7.5 A. That gives 15 + 7.5 = 22.5 A max continuous on the 5V rail. Round down to the nearest Ampere, and you get a number that pretty much almost always agrees with the ratings that you'd find on a Delta, HiPro, and LiteON PSUs given that schottky diode and single transistor forward topology. Just so we are clear - this is a formula I came up with and may not be entirely accurate. But since it gives consistent results with honestly rated PSUs from big manufacturers such as Delta, LiteOn, and HiPro, IMO it is a safe formula to use when calculating continuous output power. And again - this is for single transistor forward topology only (and maybe double forward too?).

[Behemot]

Re: Power supply build quality pictorial. part 2

I think in reality it is that they are over-rating it some. I still believe that arround the current per device value there is the minimum it should be to comply with PSU rating.

[Wester547]

Re: Power supply build quality pictorial. part 2

I believe no matter the schottky rectifier that the switching transistors, the transformer, and the components that limit the maximum amount of wattage a PSU can deliver limits the amount of power on all rails (you'd probably have to calculate it evenly enough so as to be able to achieve the labeled rating). If you add up all the values of the individual rails you'll probably find yourself at least 50-100W over the (maximum continuous power output) rating of the label. Of course, for some very overbuilt PSUs, this may be accurate - for others, not so much.

[Behemot]

Re: Power supply build quality pictorial. part 2

Yeah but we all know Delta, Hipro and LiteOn are among the very overbuilt ones most of the times. Just OEM. Look at all the better PSU's at Hardware Secrets, almost every one of them can be overloaded for another at least 100 W. The thing is, these are usually overpowered in terms of several hundreds %, semiconductor wise. It is because of efficinecy here too.

[Th3_uN1Qu3]

Re: Power supply build quality pictorial. part 2

I believe no matter the schottky rectifier that the switching transistors, the transformer, and the components that limit the maximum amount of wattage a PSU can deliver limits the amount of power on all rails(you'd probably have to calculate it evenly enough so as to be able to achieve the labeled rating).

It is because of efficinecy here too.

Exactly. In the cheapies, a pair of 13009 switching transistors blow at 250 to 300W output. I was able to pull 650W output from a pair, with a single 80mm fan for cooling, and they held up just fine. How? Overspec'd secondary rectifiers, which yielded 84% efficiency at max output, and 90% at the 300W level.

Now compare that with the 70% typical efficiency of cheapo PSUs, which drops even lower as the supply approaches its limit (and inevitably, blows up). It's a tough job getting rid of all that heat, and those puny heatsinks typically found in such low-end (mis)designs don't help either.

[tom66]

Re: Power supply build quality pictorial. part 2

But in the worst case, 70% efficiency only means the switching transistors have to handle ~430W at 300W load - so how did you manage 650W? Is it perhaps because the poor quality diodes don't switch fast enough and this causes greater transients on the transformer primary?

[Th3_uN1Qu3]

Re: Power supply build quality pictorial. part 2

But in the worst case, 70% efficiency only means the switching transistors have to handle ~430W at 300W load - so how did you manage 650W? Is it perhaps because the poor quality diodes don't switch fast enough and this causes greater transients on the transformer primary?

In the worst case, the efficiency drops towards 60%, and that's when they usually 'splode. If it would keep 70% efficiency at 300W, then the transistors would survive.

I used a rather beefy heatsink, but it didn't come from anywhere else than a PSU, some of them have larger heatsinks. They still blow up due to the low efficiency and lack of proper protection circuitry (or rather, the misconfiguration of such - it does exist but it's set too high so it's useless), but the larger heatsinks can take 100W or so with a typical 80mm fan running at 2400 rpm.

I had thermal protection set to kick in at 90C on the primary heatsink, and it didn't, so temps were below that. This test was made last summer, with 32C+ room temp, with the supply and load i think it got close to 35C. After 2 hours of continuous running i shut it off because i couldn't stand the heat in the room anymore. The supply is still sitting in a corner somewhere, i had issues with the compensation, and as it turned out later i had a mistake in my TL494 PCB, which made the compensation not work at all. I'll get back to it someday, it's meant to do +/-60v, for powering an audio amp.

650W at 84% efficiency gives 124W dissipation, close enough to your example don'tcha think?

I am currently investigating quasi-resonant operation, if i can push the efficiency towards 90% i could aim for the 900W level with the same pair of 13009s. Given the relatively low currents required for audio amplifiers as opposed to a computer PSU (higher impedances means higher voltages and lower currents, so we got several amps at 100-some volts instead of dozens of amps at 12v), it should be doable.

[tom66]

Re: Power supply build quality pictorial. part 2

Sounds interesting. One point I would make is I have a quasi resonant server supply. Actually two of them, 12V at 98.0A each, won them on eBay for £6. The heatsink for the transistors is tiny(!) compared to the PFC stage and output stage; but it does get very warm under no load (I'd estimate 10W dissipation idle, though I haven't measured it.) I wonder if the topology has poor efficiency at low loads.

[Th3_uN1Qu3]

Re: Power supply build quality pictorial. part 2

I wonder if the topology has poor efficiency at low loads.

It does. Quasi resonant means it only hits resonance at full load - where the efficiency is needed the most. It is also the simplest to make, but more on that when i actually have one up and running.

[c_hegge]

Re: Power supply build quality pictorial. part 2

In the worst case, the efficiency drops towards 60%, and that's when they usually 'splode. If it would keep 70% efficiency at 300W, then the transistors would survive.

And that's what I find usually happens. If you watch the power draw on an overloaded PSU, you will not that it slowly increases as it warms up. It will usually increase by a couple of watts every 10 seconds or so for the first minute, then it will start increasing much faster. Once the efficiency drops to around 65%, the power draw suddenly skyrockets (maybe 10-20W/second), and then:


My guess is that at 65% efficiency at 350W, the plastic washers melt and they come off the heat sink.

[Wester547]

Re: Power supply build quality pictorial. part 2

If the amps a diode package can deliver are lower in a single and double transistor forward topology, would the voltage it can deliver be lower too? For an example - if a TO-247/40A/45V schottky from OnSemi could theoretically do ~29A in a single transistor forward topology, would the voltage it could deliver be lower than 45V as well?

[Th3_uN1Qu3]

Re: Power supply build quality pictorial. part 2

My guess is that at 65% efficiency at 350W, the plastic washers melt and they come off the heat sink.

TO-247 devices don't need the plastic washers, and they still blow up in the same way. There's just too much heat to dissipate, because of the poor efficiency of the secondary rectifiers.

If the amps a diode package can deliver are lower in a single and double transistor forward topology, would the voltage it can deliver be lower too? For an example - if a TO-247/40A/45V schottky from OnSemi could theoretically do ~29A in a single transistor forward topology, would the voltage it could deliver be lower than 45V as well?

The voltage rating on the diode isn't the maximum output voltage, it's the maximum reverse blocking voltage. When the diode is not conducting, it can hold off a certain amount of voltage. If you increase the voltage past this level, the diode shorts out.

This reverse blocking voltage is the absolute maximum peak voltage that can come from the transformer, without the diode blowing up. Power supplies are designed so that at maximum line voltage and minimum load, the transformer output voltage remains below the reverse voltage rating of the diode, with a 20-25% margin.

I've been warned before about putting 45v rated diodes on 12v rails to get higher current output and higher efficiency, but my own measurements show that the transformer voltage of a typical ATX PSU on the 12v rail is 24v usually, 30v max, so a 45v diode is still safe, but as always, YMMV.

To answer your question: This reverse voltage rating applies in the same way regardless of topology.

[tom66]

Re: Power supply build quality pictorial. part 2

In general you can exceed the Vbreakdown of a diode without damage (avalanche breakdown), but you will cause extreme power dissipation if it is continued for more than a few ms. Which is why they blow up. Isn't the Dell 3008 monitor well known to have issues with 170V diodes going short? On just a 24V supply?

[Wester547]

Re: Power supply build quality pictorial. part 2

I've been warned before about putting 45v rated diodes on 12v rails to get higher current output and higher efficiency, but my own measurements show that the transformer voltage of a typical ATX PSU on the 12v rail is 24v usually, 30v max, so a 45v diode is still safe, but as always, YMMV.

To answer your question: This reverse voltage rating applies in the same way regardless of topology.

I've seen many manufactures put 100V-200V+ (sometimes more) diodes (or schottky rectifiers) on the +12V rail... is that dangerous?

[c_hegge]

Re: Power supply build quality pictorial. part 2

TO-247 devices don't need the plastic washers, and they still blow up in the same way. There's just too much heat to dissipate, because of the poor efficiency of the secondary rectifiers.

True, but I find that TO-247s usually take a bit more to make them blow up. I haven't killed TO-247 switchers below 400W before, but I've killed lots of TO-220 13009s at 320W or so. Having said that, though, it's possible that the slight increase in surface area contacting the heatsink could be helping too.

[PeteS in CA]

Re: Power supply build quality pictorial. part 2

True, but I find that TO-247s usually take a bit more to make them blow up. I haven't killed TO-247 switchers below 400W before, but I've killed lots of TO-220 13009s at 320W or so. Having said that, though, it's possible that the slight increase in surface area contacting the heatsink could be helping too.

There's a big difference between TO-220 and TO-247/TO-3P packages in their ability to conduct heat from the die to the metal tab of the package.

STMicro MJE13009 datasheet: Thermal Resistance, Junction-to-Case = 1.14 *C/watt

https://Datasheet for Fairchild Semi...47 600V MOSFET: Thermal Resistance, Junction-to-Case = .58 *C/watt

TO-247/TO-3P packages can also fit a larger, lower R(DS On) die, which translates to lower power dissipation when the device is fully on (but higher dissipation during switching time, due to the higher gate capacitance). If a TO-247/TO-3P MOSFET is switched quickly at a reasonable switch frequency, it will dissipate less power than will a TO-220 MOSFET operated at the same Drain current. Or a TO-247 can be at the same die temperature as a TO-220, but operating at a significantly higher Drain current.