Rectifying and switch-mode power supply - input ripple?

[joeyjoejoe]

I'm taking 24VAC 60Hz, rectifying it, and using a switcher to get 3.3VDC.

Right now, after the full-bridge rectifier, I have a single 470uF polarized cap. I'm designing it around a 300mA nominal load on the 3.3 rail.

At 350mA, the DC input voltage sags by about 1.6V between AC cycles (so it hits around 29V at the low point. Some voltage drop from diodes)

The output is rock solid. About 30mV noise peak to peak with some very relaxed probing setup.

There's plenty of overhead that the switcher can follow the voltage. I'll order some more capacitors and try it out at 330uF, 220uF and maybe 100uF. Since these are 50V capacitors (35V is a little close for comfort), the size drops quite a bit and selection improves. Two 100uF's in parallel might be a nice sweet spot.

My question is, how low can I go? Should I care about this droop on the input side at all by reducing these input capacitors a bit?

[james_s]

You're not going to completely eliminate the droop without using a ridiculously enormous capacitor and even then there will be some small amount. As long as the input voltage is always high enough for the regulator to function under worst case conditions (low input voltage, capacitor at the low end of the tolerance range, max load) I wouldn't worry about it.

[T3sl4co1l]

A good rule of thumb is to keep ripple under 20% or so for electrolytics at mains frequency or 2nd harmonic.  A small cap like that (say 100 or even 47uF) might not even get very warm, but it won't have a long life regardless.

220 or 470uF sounds fine for that, yes.

And yeah, the switcher doesn't care, and except for hold-up near zero crossing, you could just about get away with nothing.  If you wanted a higher power factor and longer life, maybe use ceramic or film capacitors -- the ripple will probably make audible buzz from ceramics, but won't otherwise do anything beyond normal aging, and film are fine.  The big downside is the high cost of ceramics that size (10s of uF at 50V will be a half dozen or so 1210 chips), or the bulk -- and cost -- of films that size.

Aluminum polymers are, in many ways, the dense, polarized equivalent of film caps at low voltages; but they're also not great at high ripple, I think.  Better than electrolytic, certainly.  A 50V 47uF part probably would last quite a while here.


Better question: why not rectify mains instead, i.e. use an offline power supply in the first place?  3.3 and 5V outputs are cheap and plentiful.  Maybe not with great noise performance, but also not hard to filter.

Tim

[james_s]

Maybe he's already got 24VAC available and doesn't need additional isolation. I did just that years ago with a HVAC system that had a 24VAC control transformer. I built a buck regulator to power something, it was a lot simpler than a complete off-line SMPS.

[T3sl4co1l]

Indeed, have worked a few projects like that -- also an HVAC oriented client, they use 24VAC frequently in their controls, whether for compatibility, interfacing, or just plain old tradition.  Still, worth asking what drives the requirement, if anything.

Tim

[joeyjoejoe]

Indeed, it is to re-use existing 24VAC available for HVAC equipment.

I did a quick search and SMD electrolytics are available at fairly large values to my surprise. Would two 220uF's versus a single 470uF offer any benefits other then the reduced resistance? Will higher quality (eg XXX hrs @105c versus XXX @85c) capacitors have any effects on how well they hold up to this ripple rule of thumb?

[T3sl4co1l]

I would just put in 220 or 470 and be done with it.  50V and 105C will last longer than 35V or 85C, and longer still from long lifetimes versus short (2000hr?).

There's also a few 85C 15khr types, which is basically like saying 2000hr 105C anyway, but the 85C rating presumably means it degrades faster at elevated temperature than a properly rated part would, but the low-temp extrapolated/expected lifetime is just as good.  So that would be fine too, unless your ambient is going to be obnoxiously high (but you're probably not sticking this thing on a motor casing or condenser coil, right?).

Tim

[joeyjoejoe]

Thanks for the helpful info. I'll solder in a 220, see how it goes, and be done with it. Pretty happy with this IC and layout I have, can now copy/paste this for any relevant circuit I need to stepdown.

[MrAl]

I'm taking 24VAC 60Hz, rectifying it, and using a switcher to get 3.3VDC.

Right now, after the full-bridge rectifier, I have a single 470uF polarized cap. I'm designing it around a 300mA nominal load on the 3.3 rail.

At 350mA, the DC input voltage sags by about 1.6V between AC cycles (so it hits around 29V at the low point. Some voltage drop from diodes)

The output is rock solid. About 30mV noise peak to peak with some very relaxed probing setup.

There's plenty of overhead that the switcher can follow the voltage. I'll order some more capacitors and try it out at 330uF, 220uF and maybe 100uF. Since these are 50V capacitors (35V is a little close for comfort), the size drops quite a bit and selection improves. Two 100uF's in parallel might be a nice sweet spot.

My question is, how low can I go? Should I care about this droop on the input side at all by reducing these input capacitors a bit?

The lowest you can go depends on the lowest voltage the regulator needs to be able to produce the required output.  This means the droop can only go that low or else you cant get the constant 3.3v output.