Still a little confused how switchers/MOSFET's work

[Pentium4]

So when you look at a datasheet for a switcher or MOSFET, how can you determine how many watts they are good for? I know around the forum I've heard from you guys that 13007's are good up to 250W and 13009's are good around 350W, how do you determine this? (Other than testing them and experience) Do you just do V x A but putting into account the turn off time and operating range? Thanks

[eccerr0r]

Re: Still a little confused how switchers/MOSFET's work

When a transistor specifies watts, this is how much power it can dump out as heat before it fries. It has only a small correlation with the overall PSU wattage - basically this waste heat is one part of the inefficiency equation. The more inefficient the design the lower the watts the whole PSU can pass to the load.

The waste heat the transistor generates is Watts=current^2 * resistance of the channel. The lower the resistance the less waste heat (but usually reducing resistance also reduces switching frequency (due to the limits of "real" semiconductors) and you get more dynamic/switching losses that way.)

So it's not just the transistor, it's also coupled with the design.

[Pentium4]

Re: Still a little confused how switchers/MOSFET's work

Thanks for the useful info I'm guessing the main reason they die is inadequate heatsinks or bad airflow?

[eccerr0r]

Re: Still a little confused how switchers/MOSFET's work

Yes, a lot of the time it's bad heatsinks or exceeding limits of the semiconductors. A lot of the time you can go over the limits for short periods of time but if due to unforseen circumstances they last *longer* than expected, this will kill the part. Popular reasons? Bad capacitors, underspecced resistors that fry, protection diodes that die, ESD...

BTW, the power loss calculation is for MOSFETs. For bipolar junction transistors the "wattage" is also dissipation watts, but their heating is mostly relative to their saturation voltages times the currents through their respective terminal - but a bit more complicated as base currents are a significant contributor to dissipation (then again mosfet dynamic current is also significant too... but it's not the transistor dissipating it, usually).

[Wester547]

Re: Still a little confused how switchers/MOSFET's work

Another thing to consider is that MOSFETs "switch" much faster than bipolar transistors. They are much more efficient, and that is important because it is during switch "on" and "off" time that the switching losses are at their highest. Even a single MOSFET can switch much faster than its bipolar transistor equivalent (or a pair of them therein, though in half bridge, if I'm not mistaken, only one bipolar transistor is used at a time). You can have a PSU with half-bridge topology that can output great wattage, but that will require great heatsinks, huge bulk storage (input capacitors), a large bridge rectifier, a hugely overspec'd secondary (so as to help the primary), and very good cooling, just to get 70% efficiency or higher at demanding loads (though probably 80% efficiency at lower loads, but it is below 70% efficiency that transitors usually skyrocket in power dissipation and go BANG). Since the forward converter uses one or two MOSFETs (single or two switch forward), that already allows you to save pennies in many other places and still gain more efficiency (though forward topology isn't necessarily more efficient than half bridge in every aspect).

This is also why the move to synchronous rectification and buck DC-DC converters (MOSFETs) improves upon efficiency lots and allows for smaller PSUs to have less cooling, smaller heatinks, and yet do so well (80+ efficiency) at the same time. Want an example of what I mean? Compare Hardware Secret's review of Huntkey's Greenstar "450W" to their review of the Antec VP350 (Delta 350W). Notice how the Huntkey has 1000uF Teapo primaries, a 15A rectifying bridge, TO-3P 13009s, an 140mm fan, and large components otherwise, for the most part, and yet only can do 360W with ~74% efficiency at 40*C to 50*C room temperature before running out of steam. It exploded at 450W and hard at that. The VP350, on the other hand, is a single forward design that uses a single STW12NK90Z (TO-247) MOSFET, 470uF CapXon primaries, an 120mm fan, an 8A rectifying bridge that doesn't get as much airflow as it could, and decent components otherwise, and all that for 350W at 76% efficiency, under the same conditions, (it did burn at 475W, though), which is arguably better than how the Huntkey performed with clearly bigger components.

That's what I mean. Granted, since eccerr0r did note the advantage of greater power dissipation, it's true that the TO-3P 13009s have a power dissipation of 130W while the STW12NK90Z can take up to 230W without burning, which is quite a difference, but another thing to consider is how much power dissipation is improved and how much the discrepancy is minimized when you add good heatsinks and copious cooling. MOSFETs can also work at a much higher switching frequency than bipolar transistors can, though bipolar transistors also work better at lower frequencies, but then when rectification at a lower frequency takes place, bigger components must once again be used.

It's no wonder why those cheapo, half bridge PSUs can't do anything with their ripple and noise in spec - the use of bipolar transistors, despite the massive ineptitude of the rest of the PSU, does not help that. To answer the thread a little more clearly, though, I have found that while the wattage capability of a PSU depends heavily on the sum ability of all the components, that the switching transitors seem to take the most heat (if they cannot switch fast enough they explode), along with the input capacitors (the storage of energy, and cooling of course is a major factor). The transformer and rectifying bridge matter too but I have found those to matter less in determining what a PSU can ultimately do (the bridge rectifier "rectifies" AC to DC - that's about what you can draw from the mains, and the efficiency in that regard... saturation of the main transformer can damage the switchers and in turn, an overload of the switchers can damage input capacitors). Having an overspec'd secondary can nicely augment what the PSU is capable of, though, by increasing efficiency, taking off stress, and keeping the switchers happy.

[Pentium4]

Re: Still a little confused how switchers/MOSFET's work

I'm not sure why I didn't see this response before. Thank you very much wester for the awesome post. So much good information and good comparisons. I will have to read it a few times to take it all in but thank you for breaking it down for me and spending the time to type it out much appreciated!

[Wester547]

Re: Still a little confused how switchers/MOSFET's work

Well, to be honest, how much power a MOSFET or bipolar transistor can dissipate does probably curtail the difference between them (if a MOSFET has exceeded its safe power dissipation, no matter how fast it switches, it goes bang), but not so much that you can't get away with overbuilding less and achieving more wattage in a design that uses MOSFETs as the main switcher instead of BJTs.... BJTs do also seem to improve performance wise at frequences below 70KHz which is what I think the primary side actually sees (I believe Th3_Un1qu3 went into that) in a SMPS. MOSFETs can operate at higher switching frequencies. Last but not least, in the application of SMPS, MOSFETs obviously are supreme, but in other applications, like (if I'm not mistaken) audio, there might be a difference in favor of BJTs. Also, if you use BJTs in the application of a saturated switch, they actually might not do better, but in the case of a linear switch, not so much... a MOSFET can be fast in both. BJTs also have an advantage in their 'On' resistance ability if the drive current is high enough, but then their driving voltage is lower than that of a MOSFET. A MOSFET also may not be able to block voltages as well.

However, the biggest kicker, I think, is that MOSFETs have switching on and off times that are much, much faster (macroseconds vs. nanoseconds) than BJTs, which I believe is a larger difference between them than any advantage BJTs have. To be clear, I'm not really an expert on any of this, I have just observed that single and double forward designs seem to get away with overspec'ing less and doing more than half bridge PSUs. MOSFETs are definitely more efficient, though, which is why synchronous rectification requires a lot less filtering and much less cooling by way of its efficiency. However, a linear regulated DC-DC converter is very inefficient (it squanders a lot of power, and the higher the output voltage you use linear regulated DC-DC conversion with, the more wasted power) even with the use of MOSFETs (+5VSB is linear regulated and uses a flyback transformer, and doesn't even receive airflow - that is why those filtering capacitors are often the first to go) and is not to be confused with buck or synchronous (both of which are very efficient, though much more complex).