Powering AC Switch Mode Power Supplies with DC

[nbritton]

Does anyone have any insight on whether you can power AC adaptors that have auto-ranging and active PFC directly with high voltage DC? I know you can power traditional full wave bridge rectified SMPS directly with DC, but I'm wondering if you can do the same with modern auto-ranging active PFC SMPS. In particular, I'm wondering if I can power Apple's MacBook AC adaptors directly from a high voltage DC battery bank.

http://www.righto.com/2015/11/macbook-charger-teardown-surprising.html

The Apple SMPS is designed for 100 to 240 volts AC. So given that VDC = VAC * sqrt(2), would this power supply be compatible with a 141 to 339 volt DC battery bank? Also since only half the rectifier diodes would be in use if powered by DC, would it be smart to use a higher voltage so that less current goes through the active diodes? If it were powered by AC, each diode in the bridge would have a duty cycle of 50%. So if I doubled the DC voltage that would drop the current to 50%.

[David Hess]

Does anyone have any insight on whether you can power AC adaptors that have auto-ranging and active PFC directly with high voltage DC?

Isn't that redundant?  The active PFC either has enough input range to support a universal input or it does not.  I have never seen one with any sort of extra auto-ranging circuit like a doubler.

PFC should not care if the input is DC but I cannot say that there are not some finicky implementations.  Operating at a higher DC voltage will lower the current through the input rectifiers (and increase the current through the boost rectifier do to decreased duty cycle) but I would hope they are derated enough for this not to be a problem.

The only risk I know of is safety.  The input protection fuse and any power switch is not going to be rated to break high voltage DC.

[Ian.M]

Just in case there is a dropper resistor based startup circuit, you should *NOT* use a DC input voltage exceeding the max rated RMS AC input voltage.

[David Hess]

I would hope that the startup resistor is derated enough for that not to be a problem.

[Ian.M]

V²/R

It will certainly be rated to withstand 240V RMS +10%, but if you apply 340V DC it will be dissipating double what the designer expected.  Even if the resistor (or resistor chain) has enough margin on its rating, the excess dissipation is likely to shorten its life and may cause problems for other nearby components.

[David Hess]

I checked some designs and the startup resistor was derated to about 66% (90K and 1 watt) so 340 volts DC would indeed be too high for safety.

[station240]

I did an experiment where I powered a small PC power supply with PFC directly from a 150V solar array.
I removed the main input filtering, and bridge rectifier before I tested it
It did work but I found I needed a larger 400V capacitor, and to be honest higher wattage of solar cells.

141 to 339 volt DC seems like a very large range for a battery bank.

[T3sl4co1l]

Many PFC controllers assume pulsating input, and get confused when fed DC.

The definition of PFC still applies for DC, so in principle, nothing should be different -- but a lot of implementations make lazy assumptions like this (or take advantage of whatever simplifications that can be made in an AC circuit).

So, YMMV.

Tim

[David Hess]

But what simplification could they make which would prevent operation with DC?  Is there an application note discussing this or a reference design which has this problem?

I have heard before that some PFC power supplies have problems with modified sine inverters but this strikes me as a really poor design feature considering how common they are.

[T3sl4co1l]

A possible feature might be DC restore, zero crossing detect, lookup tables (hey, these things often have thousands of transistors -- an entire MCU worth of stuff inside) and so on.

(Yes, all this just to avoid a pair of BJTs in a multiplier configuration.  Ain't ICs wonderful?)

And since their design spec is limited to mostly-sinusoidal mains of known world frequencies (and occasionally up to 400Hz, but ones qualified for that might be the good kind anyway), they don't /need/ to follow the general definition.  So, you get things that don't do things...

Tim

[Someone]

But what simplification could they make which would prevent operation with DC?  Is there an application note discussing this or a reference design which has this problem?

You are right to ask for references, I dont know of any and to be accurate  we could look at the chips used for the specific apple power supplies and see they all use a multiplier based approach which will work fine with a DC supply. T3sl4co1l seems to like saying lots of things and feeling important while not listening to the actual requests made in posts.

[T3sl4co1l]

You are right to ask for references, I dont know of any and to be accurate  we could look at the chips used for the specific apple power supplies and see they all use a multiplier based approach which will work fine with a DC supply. T3sl4co1l seems to like saying lots of things and feeling important while not listening to the actual requests made in posts.

If you can't use your imagination, I'm sorry for that...

I'm afraid I don't have any specific examples that I've studied and shown they do/not work with DC.  If you'd like, you can send me a broad selection of examples and I can check if they work or not; I'll need somewhere to send the bill, though.

A quick search returns few ICs; this one says so in the description, surprisingly enough: http://www.mouser.com/ds/2/149/fairchild%20semiconductor_fan9612-320438.pdf

Yes, bypassing the PFC altogether will most likely work; the usual output is around 400VDC.  If anything is powered from the PFC controller's AUX (LV DC) supply, or enables/power-good signals are used with another controller, you may need to fill in for those as well.

Tim

[David Hess]

If you can't use your imagination, I'm sorry for that...

I'm afraid I don't have any specific examples that I've studied and shown they do/not work with DC.  If you'd like, you can send me a broad selection of examples and I can check if they work or not; I'll need somewhere to send the bill, though.

I do not have any examples either despite looking.  All I have is anecdotal reports occurring under unusual conditions (like UPS operation do to power disturbances) which by themselves could destroy any power supply.

A quick search returns few ICs; this one says so in the description, surprisingly enough: http://www.mouser.com/ds/2/149/fairchild%20semiconductor_fan9612-320438.pdf

I must have missed it in the datasheet or misunderstood the context of your statement.  What part of the description says that the Fairchild FAN9612 does not work with DC?

Features:
...
Works with DC and 50 Hz to 400 Hz AC Inputs

The measured peak value is then used in the following
half-line cycle while a new measurement is executed to
be used in the next half line cycle. This operation is
synchronized to the zero crossing of the line waveform.
Since the input voltage measurement is held steady
during the line half periods, this technique does not feed
any AC ripple into the control loop. If line zero crossing
detection is missing, the FAN9612 measures the input
voltage in every 32 ms; it can operate from a DC input
as well. ...

I am dubious of this scheme using peak detection instead of RMS measurement (Is Fairchild avoiding a patent?) however it looks like they went out of their way to make sure it works with DC and modified SIN inputs.  I suspect their peak detection method fails with high crest factor waveforms but that would be a very unusual operating condition.

[T3sl4co1l]

I must have missed it in the datasheet or misunderstood the context of your statement.  What part of the description says that the Fairchild FAN9612 does not work with DC?

Just to clarify, I was looking for examples that do or do not (definite rather than undefined) work on DC.  Most are undefined; that was one which actually states it (positive definite).

I am dubious of this scheme using peak detection instead of RMS measurement (Is Fairchild avoiding a patent?) however it looks like they went out of their way to make sure it works with DC and modified SIN inputs.  I suspect their peak detection method fails with high crest factor waveforms but that would be a very unusual operating condition.

AC, DC, funky waveforms, who knows... you can't win.

It would be pretty reasonable, at least, to expect that very quickly changing waveforms will have poor response (since the current loop can only react so quickly), but given that limitation -- a general approach should work with any waveform, frequency* and crest factor.

*Limited further by the voltage regulator loop, so you should avoid frequencies from almost-DC to 43Hz or whatever (else you get large output ripple, poor PFC, and poor regulation, because, what do you expect?).

So, these are the basic limitations we should expect from a general PFC.  Anything that's further limited probably has a reason (be it patent evasion, design shortcuts, or good old fashioned laziness).

Tim