Diode ratings in push pull converter

Indeed you have been swindled. Unfortunately you are not the first nor surely the last person who will be swindled by "L&C Technology" power supplies. That company clearly has no moral standards, they simply slap stickers on their products with whatever fanciful ratings they want with no regard to the actual powersupply construction.

As I write I have a L&C technology so called "300W" powersupply sitting in my lap. I might add the powersupply failed after maybe 3000 hours of operation (at much less than 300W output). The powersupply has three substantially discolored regions on the PCB indicating prolonged extreme heat. This powersupply also uses the same basic output rectifiers as yours does (two 10A center tapped TO-220 schottky diodes and two 3A FR302 devices). It has two 330uF 200V input capacitors.

To be within specification of just the output rectifiers both of our powersupplies should be rated for an absolute maximum of no more than 10A

Now adays I don't normally buy computer powersupplies unless I get a chance to at least peek through the holes in the side with a flashlight first. Signs of reasonably decent construction I specifically look for are large input capacitors (two 470uF or more for 250W, 680uF or more for 300W+), and a large main power transformer that uses a round center post. Transformers that use a round center post are inherently superior to those that use rectangular center posts. For a given cross sectional area the circle has the smallest perimeter, which directly translates into better copper wire utilization. These types of cores also normally have a much taller than wide window area which also improves proximity effect losses for high frequency operation.

When Franc made his post, I questioned his determination of the FR302 diodes, but now you confirm it in another L&C Technology PS. Two 3A half-wave rectifier diodes could be claimed to deliver 6A continuously (although without a proper temperature derating). The L&C Technology 17A rating doesn't speak well for company's honesty.

The diodes would appear to drop nearly 2V at 17A, at 25C, less when hot. Let's call it 1.5V; that's about 13W dissipation (50% duty). It's doubtful they can do this for more than say 30ms, starting cool from a current *much* lower than 6A. Perhaps this is long enough to do something useful, spin up a disk drive, etc., but the discrepancy between the 6A datasheet and 17A L&C spec'd numbers boggles the mind.

Another thing that boggles the mind is using el-cheapo $39 power supplies with thousand-dollar processors, display monitors, etc.

Well okay, but a low voltage schottky diode operated at reasonably decent current density has a typical forward voltage more in the range of 0.4-0.5V. In your typical computer power supply which uses a half bridge topology the secondaries are center tapped. So rectification for each rail is normally achieved using a single TO-220 or TO-247 packaged dual diode in the typical full wave rectifier (for center tapped secondaries) configuration. So only one diode is effectively in series with the output at any given moment in time. The 12V rail usually uses a TO-220 packaged 100-200V ultrafast recovery diode with a forward voltage approaching 1V.

This does raise the interesting question of whether the power rating stamped on the PSU is the input power or the output power.

If it's the input power then 200W power supplies can easily be upgraded with a resistor :-)

Regards, Alan

O.T.O.H.:

That power supply probably costs less than USD 5,00 per piece in lots of

PeeCee equipment is notoriously s***ty - it's amazing how often it happens to work in the general case while anyway near the "edge" - Boom! (literally)!! Probably the very same principles are used here as in the olde days with power transistors for switchmode: Everybody *knows* that the salesperson lies through his teeth, after all his lips are moving, so we just de-rate by 60% to compensate.

I have a generic "400W" ATX PSU which I suspect is a fake. This is its circuit diagram:

"LC-B250ATX ch. Y-B200-ATX ver. 2.9 JNC Computer Co."

The label on the cover reads as follows:

L&C Technology Inc Model: LC-B400ATX

The +3.3V and +5V rails are each rectified by a S10C40C dual diode, and the +12V rail is rectified by two FR302 fast recovery diodes.

According to their datasheets, the FR302 is rated at 3A, and the S10C40C is rated for 10A.

Printed on the PCB is the following legend:

F5A/250V (for 180W -- 235W) F7A/250V (for 300W -- 350W) F6.3A/250V (for 250W)

The fuse is 5A. :-(

The PCB is labelled "Y-B200 ATX Ver 2.8" which also suggests a 200W rating.

Can anyone offer an opinion as to the real rating of this PSU? How do the diode ratings in the push pull converter affect the PSU spec? Can one assume that the total wattage is close enough to ...

Pt = (3.3 x 10) + (5 x 10) + (12 x 6) + (5VSB x 2) + (-5/-12 wattage) = 175W approx

- Franc Zabkar

they probably parallel both diodes. in theory this gives twice the current rating, if they share equally (which they probably do as they are on the same die so at the same temperature)

the diode drop is a large chunk of the 3.3V output power - say 1V at

read the diode drop off the datasheet (taking account the individual diode current where 2 are paralleled) and add it to each supply voltage in your calc, and you'll be a lot closer. It doesnt account for xfmr and choke Cu losses, switching losses or any primary FET conduction losses, but you'll be fairly close.

in a push-pull or forward or half- or full-bridge converter the average diode current is the output current.

I have no idea where you got your current values from....

without specifying the input voltage, it is not much use looking at the fuse rating. Even then its not too helpful (unless you take into account the pulsating nature of a rectifier/diode filter, not that hard if you like maths)

but if its a 5A fuse it probably wont like more than 3Arms long term - the fuse will eventually fail at any more than that.

Cheers Terry

I have seen both approaches in pc smps high-current outputs - often a pair of dual diodes on opposite sides of the same heatsink. the datasheet for the particular schottky mentioned specs 0.6V at 10A, ie

Cheers Terry

Jolly good points.

I have a mate who has several hundred videogame machines. He has used pc smps for a long time, as they are cheap and readily available. About 6-7 years ago he stopped buying really cheap ones, and started buying (comparatively) expensive ones from a reputable manufacturer. His smps failure rate plummeted to almost zero (his techs were replacing one every week or two), and the logic board failure rate likewise became negligible.

I think the answer to "how can you make a good smps for so little money" is simply that you cant. Cheaper is invariably nastier.

Cheers Terry

On Wed, 16 Mar 2005 19:35:27 +1300, Terry Given put finger to keyboard and composed:

No, the anodes of each diode attach to opposite ends of a center tapped winding. The rated average forward current is 5A per diode, giving a total rating of 10A for the package. As each diode only conducts during alternate half cycles of the oscillator, the rating of that particular rail would be 10A.

I'm not concerned with the power output from the *winding*, except insofar as it reflects the efficiency of the PSU. Instead I'm trying to determine the power available to the external load, ie I'm comparing this figure to the one on the PSU's rating sticker.

That's what I've always thought, but after seeing several "200W" PSU's with 150W parts I was left to wonder whether there was something I didn't understand.

I'm not sure which current values you mean, but one set was printed on the ratings sticker, the others came from device datasheets.

Here is a diagram of a push pull converter:

A diode rated at X amps should be able to carry 2X amps during each half cycle of the oscillator. The other diode carries 2X amps during the alternate half cycle.

The fuse ratings were printed on the PCB. I included this information to confirm that this particular configuration of the PCB was intended for a 180W-230W application. The suitability of the fuse was of no particular concern to me.

- Franc Zabkar

On 16 Mar 2005 04:57:39 -0800, Winfield Hill put finger to keyboard and composed:

I was careful not to say that this was actually an L&C PSU. The sticker says L&C, but who knows who really made it? I posted several links to what may be the real (?) L&C unit. These look entirely different. Even the low wattage L&C enclosures look different to mine.

Who knows what a quality PSU really costs to make, especially in the PRC, and by prison labour? One wonders if a genuine 400W generic PSU is really significantly inferior to a branded product. Remember, mine a 180W PSU with a 400W label, so it never really had a chance.

- Franc Zabkar

On 16 Mar 2005 02:50:25 -0800, Winfield Hill put finger to keyboard and composed:

Yes, I did. In fact I'd traced it before I found the online circuit. I'd even determined the pinouts of the mysterious combo PWM controller/supervisor/3V3 regulator IC. Any differences in my implementation of the circuit are in unpopulated sections (missing filter caps, linked inductors), and in the wiring of the fan. My fan is fed directly from the +12V rail, whereas the circuit diagram shows a thermistor controlled version (this is available on my PCB, but not used).

Meticulously. :-)

Yeah, right. :-)

The PSU label has a separate C-tick certification number. I've emailed the relevant authority.

- Franc Zabkar

On Wed, 16 Mar 2005 04:45:18 -0700, "Fritz Schlunder" put finger to keyboard and composed:

As stated previously, this PSU does not look like any of the L&C PSUs in the URLs I posted elsewhere in this thread. Maybe it's a fake L&C unit put together by a third party. After all, a fake version of a generic brand would not be subject to the same scrutiny as a fake Enermax, for example.

I believe the various diodes are good for their rated current up to 75 degC.

- Franc Zabkar

Paralleling a pair of diodes is not uncommon. If I was sneaky I would have said "possibly" not probably :)

as you have discovered, the rail "rating" doesnt have bugger all to do with the diode rating. And the rated average current is given at some (unspecified) setup having a thermal resistance which keeps the diode at a suitable temperature. Depending on how you cool the device, it is possible to run at much higher currents than the data sheet suggests (Ifsm being the single-shot limit, usually constrained by adiabatic heating, or perhaps bond wire fusing). Conversely if you did a dreadful job of cooling the diode, reliable operation (now theres the snag) would mandate a lower-than-rated current. My experience with cheap smps is that they beat the snot out of the diodes, which often fail.

for a reputable manufacturer, its what the label says. For a cheap manufacturer, its often a lot less - depends on how long you want it to live for, and under what conditions. Cheap smps ratings are a bit like audio amp power ratings in that regard - wildly optimistic if not complete fiction.

when a design is "optimistically spec'd" (thats being kind) like this, its the losses that determine the medium- and long-term power available to the load, because they are the bits that will die. But you cant really calculate that, you pretty much have to measure it. I have done that on several occassions when evaluating OEM parts - the results are sometimes pleasing, often horrifying.

cap ripple current ratings affect long-term operation too, and are usually woefully inadequate. These are wonderful for techs, they are a never-ending source of income.

the mentality of the lying bastards that wrote the label. But as Fritz pointed out, dynamic and static loads are often quite different, and marketers usually use the former in place of the latter.

maybe. depends on overall thermal performance. The S10C40C datasheet gives 0.48V@5A = 2.4W and 0.6V@10A = 6W, so the peak die temperature rise is 2.5 times larger with a 2x increase in current. again, thermal considerations dictate whether or not this is a terribly bad thing, but it will probably decrease the life of the part, due to the effect of temperature cycling on the die-Cu interface.

if you know the ac input voltage and dc bus capacitance, you can calculate the 100/120Hz ripple and diode conduction angle, then calculate the rms current it will draw - assuming a diode/cap rectifier. I doubt a cheesy smps like this will have a UPF universal input.

Cheers Terry

belabouring the point: 75C case temperature. Add to that the drop across the heatsink, then internal ambient, which will likely be a fair bit hotter than external ambient.

Cheers Terry

The schematic indicates that a dual diode body is to be laid out for the 12V rail at D14 - not FR503s. FR503s might be used in the negative rectifier, as these are inicated as being discrete.

A lot of power is easily deliverable by the higher voltage rail, with a TO220 dual rectifier employed. This is also where the extra power is usually needed.

An SB1040 is also not necessarily a Mospec S10C40 - this could be an autocad designation simply identifying a footprint or bin number.

I suspect that the correspondence between the schematic and the unit you're disassembling is not as perfect as you believe.

A single schematic can easily represent a number of models that differ in power rating - just as the same board art can be populated with components producing different total output ratings - you'd realy need to read the BOM in question or identify individual parts present in a correctly labelled unit.

RL

On Sat, 19 Mar 2005 21:45:45 GMT, legg put finger to keyboard and composed:

The artwork can accept either a TO-220 package or the bigger TO-3P. It appears the PCB can be configured for 180W operation or 350W, depending on the choice of caps, inductors, transformers, and power semis. Another difference is that my PSU has 4 rectifier diodes at the AC input whereas the circuit diagram shows a bridge. The artwork is designed to accommodate either. If anything, the circuit diagram of the LC-B250ATX PSU shows a fully populated PCB - my "LC-B400ATX" has missing parts.

That should be FR302, and they *are* used in the +12V rectifier. Do you know of any PC PSU that uses 3A diodes on an 0.8A rail? :-)

An SB1040 is a single diode in a TO-220 package. An SB1040CT is a dual diode.

I traced most of the circuit before I discovered it on the Net. The circuit references appear to be identical. Even the transformers bear similar part numbers, ie EI-33ASG, EE-19N, EE-16(TOP). The former is shown on the schematic as an EL-33ASG - I don't know if this is significant, or if it is a typo.

Agreed. I've identified the parts and they clearly cannot come within a bull's roar of the spec claimed on the label.

- Franc Zabkar