How a Switchmode Power Supply Works

[shovenose]

Did the article get revised with this new information?

[LongRunner]

Not yet

[shovenose]

Oh. Well hopefully c_hegge can do it or I can. I'm not super knowledgable about the actual contents of the article but I can copy and paste no problem.

[c_hegge]

I will do it, I just can't right now. My PC is currently packed away while HardwareInsights gets a new workshop/test lab/photo room. Once that gets done, I'll post a few pics of it all here, and then I'll be revising some of the articles. I should be ready to go within the next week or two if all goes according to plan.

[c_hegge]

OK. It's been revised

[LongRunner]

You're getting closer, but not quite there:

While there are two types of PSUs out there – linear and switching/switch-mode power supplies (or SMPSs).

The "while" no longer makes sense.

As I mentioned before (perhaps you missed it), there's no rule that Y-capacitors have to be ceramic (or that X-capacitors have to be plastic film). See for example Vishay MKP338 6 Y2. In many aspects, the film types are superior, but ceramics cost less and generally work well enough for small sizes.

The sub-classes usually used are X2 (pulse tested to 2.5kV) across-the-line and Y2 (pulse tested to 5kV) from line-to-earth, though some units have X1 (pulse tested to 4kV) capacitors across-the-line (occasionally even a mixture of X1 and X2, with no explained purpose); Y1 capacitors (pulse tested to 8kV) are used across the isolation barrier in "double insulated" switching supplies (albeit rather foolishly, as explained here).

For that matter, the non-safety-rated 400VDC PET film capacitors are unsafe at 230VAC even if surges are ignored (the 400VDC version of Vishay MKT373 for example is specified for 220VAC absolute maximum). Of course, I wouldn't feel safe in using capacitors with a surge rating of 640V at best on any mains.

Also, you forgot to mention that there needs to be a bleeder resistor (typical value of 330k—1M), ideally controlled by a CAPZero IC, to discharge the X capacitors after the unit is unplugged and prevent the user from getting a shock from the plug/inlet pins.

(although the primary side rectifiers are typically rated for a much higher voltage)

To be specific, the bare minimum is 400PRV, but almost everyone prefers to play it safe and use 600PRV (or more), even most of the junk manufacturers (I suppose the cost difference is negligible, though).

There's a bit of a typo:

With the switch closed, one side if the incoming AC voltage is connected in between the caps. This charges one for the caps during the positive half-cycle and the other in the negative half cycle.

Should be "of", "of", and "on" respectively

The value of [the bleed/divider] resistors is not absolutely critical. A lower value allows more charge to flow between he capacitors, but reduces the efficiency slightly. Values of between 100K and 220K ohms are typically used.

A bit of clarification: The absolute value isn't critical, but it's essential that both are equal. (Of course, there is the odd junk PSU that just doesn't care about following this rule; I believe one was mentioned on Badcaps.net.) And regarding the illustration you made, black text on a deep blue area is very hard to read, so I suggest that you change it to white.

[Low power factor] occurs when devices don’t draw current completely in sync with the AC cycles, resulting in only part of the AC cycle being properly used.

That is the case for reactive (primarily inductive) loads (such as a traditional magnetic-ballasted fluorescent lamp, or a lightly loaded induction motor or mains transformer), but switching supplies aren't reactive loads (and can't be, because the rectifier prevents any current from going back into the grid) — instead, they draw a highly distorted current waveform (I did link to an explanation — perhaps I didn't draw enough attention to it). (The secondary rectifier and capacitor-input filter in a linear supply will do the same thing, although the transformer's winding resistance will result in a somewhat less awful input power factor.) (In principle, the X capacitors draw a small leading current, but that's of no consequence except possibly in standby mode.)

The PPFC inductor works simply by smoothing out the current waveform. While not so effective, it's at least extremely reliable (whereas APFC adds more parts to fail), although that's a rather small consolation.

Ideally, the fuse should blow, but in many cheap units, the fuse is rated much higher than it needs to be. This can lead to high voltage and high amounts of current continuing to be fed into the shorted transistor, which causes it to explode.

I'm curious about this one. Based on my calculations, the correct fuse rating (for the full input voltage range) would be around 2~2.5A per 100W out for a low-efficiency unit without PFC (so 4 or 5A for 200W, 6.3 or 8A for 300W, etc.).

I don't think it helps that there's no fuse between the primary capacitors and the switchers, for that matter. A failure mode specific to BJTs called "second breakdown" can lead to their demise under load even if the cooling is adequate. Power MOSFETs that overheat, on the other hand, can go into straight thermal runaway as a consequence of their positive temperature coefficient of on resistance.

If the transformer core is overloaded, it goes into what is known as saturation, which often damages the switchers, causing them to fail as if the switchers themselves were being overloaded.

That's a common misconception; the truth is here. (Well, I would guess that the feedback in a SMPS would cause the flux density to be nearly constant. But not actually being a SMPS engineer, that'll have to do.) In any case, at switching frequencies the flux density can't even approach saturation without excessive eddy current losses.

Also, it looks like you missed this tidbit from one of my previous posts:

…most standby supplies do have both capacitors.

Anyway, let's soldier (or solder? ) on…

[c_hegge]

The "while" no longer makes sense.

Fixed

As I mentioned before (perhaps you missed it), there's no rule that Y-capacitors have to be ceramic (or that X-capacitors have to be plastic film). See for example Vishay MKP338 6 Y2. In many aspects, the film types are superior, but ceramics cost less and generally work well enough for small sizes.

That's true, but I'm talking about typical situations here. 99.99% of PC PSUs use plastic film X and ceramic Y Caps

The sub-classes usually used are X2 (pulse tested to 2.5kV) across-the-line and Y2 (pulse tested to 5kV) from line-to-earth, though some units have X1 (pulse tested to 4kV) capacitors across-the-line (occasionally even a mixture of X1 and X2, with no explained purpose); Y1 capacitors (pulse tested to 8kV) are used across the isolation barrier in "double insulated" switching supplies (albeit rather foolishly, as explained here).

Added.

Also, you forgot to mention that there needs to be a bleeder resistor (typical value of 330k—1M), ideally controlled by a CAPZero IC, to discharge the X capacitors after the unit is unplugged and prevent the user from getting a shock from the plug/inlet pins.

Added.

To be specific, the bare minimum is 400PRV, but almost everyone prefers to play it safe and use 600PRV (or more), even most of the junk manufacturers (I suppose the cost difference is negligible, though).

Added

There's a bit of a typo:

With the switch closed, one side if the incoming AC voltage is connected in between the caps. This charges one for the caps during the positive half-cycle and the other in the negative half cycle.

Should be "of", "of", and "on" respectively

Fixed

A bit of clarification: The absolute value isn't critical, but it's essential that both are equal. (Of course, there is the odd junk PSU that just doesn't care about following this rule; I believe one was mentioned on Badcaps.net.) And regarding the illustration you made, black text on a deep blue area is very hard to read, so I suggest that you change it to white.

Fixed the first bit. Disagreed on the second bit. The blue is not that dark. I can easily read it standing a few meters back from my desk.

That is the case for reactive (primarily inductive) loads (such as a traditional magnetic-ballasted fluorescent lamp, or a lightly loaded induction motor or mains transformer), but switching supplies aren't reactive loads (and can't be, because the rectifier prevents any current from going back into the grid) — instead, they draw a highly distorted current waveform (I did link to an explanation — perhaps I didn't draw enough attention to it).

I know. That was more of a general statement. I have added the bit about how they only draw current at the peak of the AC Cycles (non-linear load)

The PPFC inductor works simply by smoothing out the current waveform. While not so effective, it's at least extremely reliable (whereas APFC adds more parts to fail), although that's a rather small consolation.

I'm curious about this one. Based on my calculations, the correct fuse rating (for the full input voltage range) would be around 2~2.5A per 100W out for a low-efficiency unit without PFC (so 4 or 5A for 200W, 6.3 or 8A for 300W, etc.).

Maybe if the efficiency was only 50%. If the NTC thermistor is well chosen, you shouldn't have to go that high. By that rule, we would need a 16A fuse for 800W and a 20A fuse for 1KW. The mains is only capable of 15A in the US (and 10A here). Such a fuse would be pointless as the breaker in the household would trip before the fuse blows.

That's a common misconception; the truth is here. (Well, I would guess that the feedback in a SMPS would cause the flux density to be nearly constant. But not actually being a SMPS engineer, that'll have to do.) In any case, at switching frequencies the flux density can't even approach saturation without excessive eddy current losses.

Fixed

Also, it looks like you missed this tidbit from one of my previous posts:

…most standby supplies do have both capacitors.

Anyway, let's soldier (or solder? ) on…

Fixed.

[LongRunner]

The blue is not that dark. I can easily read it standing a few meters back from my desk.

It is on my monitor (a Dell U2713HMt).

Maybe if the efficiency was only 50%. If the NTC thermistor is well chosen, you shouldn't have to go that high.

The fuse is affected by the power factor, so with a typical power factor for no PFC, the fuse rating needs to be around 50~70% (or about 2 steps on the standard rating sequence) higher than for a unity power factor (all else being equal).

By that rule, we would need a 16A fuse for 800W and a 20A fuse for 1KW. The mains is only capable of 15A in the US (and 10A here). Such a fuse would be pointless as the breaker in the household would trip before the fuse blows.

I don't believe any sensible manufacturer makes a unit that large without active PFC. With active PFC and 80% efficiency, the correct fuse rating would be about 2~2.5A per 160W of output. I believe this is the formula:

PO ∕ (VImin × E × PF)

Where PO is the rated output power, VImin is the minimum input voltage, E is the efficiency, and PF is the power factor (both at full load, of course).

Also:

Y1 capacitors should never be used, as they can actually feed interference back into the earth on the mains.

It's their use across double insulation that causes the problems. Using them with an earth connection (as is required for Y2 caps) is alright. The point is, it's not the capacitors themselves at fault — it's how they're used.

[c_hegge]

It is on my monitor (a Dell U2713HMt).

OK, I'll fix that. My Samsung BX2240 lacks IPS (which yours has). My Lenovo thinkpad 1838 tablet, however, has it (isn't ironic that my tablet has a better display than my desktop PC in some respects? ) and the blue actually looks lighter on it.

The fuse is affected by the power factor, so with a typical power factor for no PFC, the fuse rating needs to be around 50~70% (or about 2 steps on the standard rating sequence) higher than for a unity power factor (all else being equal).

Yes, you are correct. One of those dull moments I guess.

It's their use across double insulation that causes the problems. Using them with an earth connection (as is required for Y2 caps) is alright. The point is, it's not the capacitors themselves at fault — it's how they're used.

I know. I have changed that to "should never be used in this application".

[LongRunner]

I know. I have changed that to "should never be used in this application".

The changed statement still doesn't make sense, because it still makes no mention of double insulation (in contrast to earthing).

Using Y2 capacitors with no earth connection (as would be the effect of running a desktop PC from an ancient unearthed mains outlet) would result in the long-forbidden "safety class 0", but would otherwise have much the same effect as the use of Y1 class in that manner.

(which should ideally have a small resistor in parallel with them to discharge them when power is disconnected)

It's not an "ideal" — it's a requirement (more so with larger X capacitors). Without the resistor, they will only discharge when the voltage on the primary capacitor(s) falls below the voltage on the X capacitors, and that's a bit too slow.