Transformerless power supply.

[buzzp]

I have the need to convert 120VAC to 24VAC. A 1VA transformer is to big for the package. I need to design a supply to source about .75 VA. I really don't want to use a resistive supply as this generates alot of heat (5-7W) and is very inefficient. I am thinking of a capacitive circuit, cap in series with the AC to limit the current using the reactance of the cap. I believe they use these types of circuits in appliance control.
Below is one design link using half wave rectifier. This is microchips app note AN954 and is good for about
15mA.

http://ww1.microchip.com/downloads/en/AppNotes/00954A.pdf

(forgot how to insert pics)

Below is yet another link using full wave rectifier from microchips document TB008.

http://www.e-sonic.com/whatsnew/Microchip/power/TB008.pdf

I have always known about a capacitive supply but never really dove into it much until now. I believe my requirements are feasible using such a design, just not sure if I am going to gain much board space using this method over a transformer depending on the requirements of the front end cap. I can come up with the capacitance, voltage, etc on my own just not sure if UL has any requirements for this cap to be class X capacitors. I always thought these were more for suppression but I thought of using them but the footprint is large compared to standard eletrolytics. I am thinking class X would be overkill, especially if I use transient protection such as an MOV and/or other suppression. Microchip does mention the second circuit requires R1 and R2 in order to achieve UL approval. R1 is for current limiting and R2 is for bleeding off the charge. I don’t believe the fuse is necessarily required as it only protects for the line and neutral from being miswired.
I will not be using either one of these circuits directly. However, I will likely be using full wave rectifier with the components on the line side to limit the traces with 120Vac on them (spacing). I need about 55mA at 12VDC but I am thinking I will change by biggest load (relay coil @12VDC is about 37mA) to operate on a different voltage (maybe as much as 115) to reduce the current requirements.
Anyway, I am just looking for some advise on the front end cap, as far as if I should use class X caps (I think not necessary) as well as any other advise from others who have used these types of circuits. Yes I kow the isolation is gone going this route without the use of a transformer or optics. Building a switching supply probably won't save me much space either and I would rather go the route of a resistive supply than building a switcher for this design. Thanks in advance.

 

[benta]

buzzp, 0.75 VA at 24 volts is around 30 mA. At 230 V 50 Hz, a 470 nF X2 cap will give you that. These are available with 15 mm RM from Epcos. At 120 V 60 Hz you'll need 680 or 820 nF, those are available in that size as well.

Benta.

 

[buzzp]

Benta,
For your quick calculations are you just using the reactance of the cap to come up with the current? So for 470n, the reactance at 50Hz is ~6800ohms, and 230 peak (rectified) is 325V. 325/6800=48mA (maybe you used 230 here instead for 33mA). If this is not how you came up with those numbers then can you tell me how you did real quick?

In any case, I have a sample X2 cap package to play with. I am going to test a cap circuit and play with the ROHM devices AND revisit my log book where I decided it would not be real feasible to use cap supply. Thanks Benta.

 

[buzzp]

I ignored the zener voltage above. It still dont make the numbers jive.
I think you just did not use peak voltages in your estimates. (230-24)/6800=30mA

 

[benta]

buzzp, the peak voltage is uninteresting.
For a quick "dirty" calculation, go with rms voltage and current, that's enough. You can compensate for output voltage (24 V) if you like, but that's just 10% anyway.
The point is, how can you keep the storage capacitor happy and full. And here you just need the rms current.
However, my approximation ONLY holds true for a circuit with:

A series cap followed by:
A full-bridge rectifier followed by:
A Zener diode and smoothing cap.

Again, don't forget inrush current series resistor and series cap discharge resistor.

Benta.

 

[buzzp]

I guess I can just use RMS throughout except on the cap voltage ratings, etc.
I'll be playing in the lab this afternoon with some circuits. I did plan on using the full wave bridge though. I am not sure why they did not use them in the circuits I pasted the links too.
Anyway, thanks for the input.

 

[buzzp]

Alright, I did some testing with what I have right now (47V zener) and can get what I need with a minimum of 800-900nf.

I don't want to use X rated caps in this design but I have not ruled it out yet. There will be front end transient protection in the form of an MOV and short circuit protection with a fast blow fuse. My question is should I use metallized polypropylene or metallized polyester film caps(assuming not using X2 caps)? I think polypropylene is the better selection but the choices are slim in finding the capacitance I need and the voltage. They are out there but cost more and are in a bigger footprint. I still think the higher dielectric strength of the polypropylene caps mandates there use. I have not spent a bunch of time pricing X2 caps but aren't these made of polypropylene too, typically? If I use metallized polypropylene film caps, I think it would be wise to check into X2 cap pricing as I have a feeling they are very comparable.
Any thoughts on using metallized polyester film here instead?

 

[Renovator1]

buzzp - I suspect you are beginning to realize why transformerless supplies are not often used... When you run the numbers to select the series capacitor (and consider the potential regulatory issues), a small mains transformer - even a SMPS! - doesn't look so bad anymore. That said, here are my *opinions*:

Don't use MOVs for transient suppression as they tend to catch fire when they fail and they aren't even much cheaper than TVS diodes, aka TranZorb, Transorb, etc. - TVS diodes for AC line use are comprised of two high power zener diodes connected cathode to cathode.

Consider using a PTC thermistor as both an inrush current limiter and auto-resetting fuse. The resistance of a PTC is low when cold, true, but at the power level you need it will still be high enough to be very effective.

You can put the TVS before of after the PTC thermistor; if you go with a MOV, though, put it *after* the PTC so when it inevitably fails the PTC will at least prevent the whole thing from catching fire.

X caps are usually (always?) constructed out of metalized paper, polypropylene or polyester because the metalization will vaporize in the the event of a pinhole fault (from a high voltage spike, a weak point in the dielectric, etc.) Pick whichever dielectric bests fits the budget, alloted space and performance requirements. Some believe that paper X caps are the best, but there isn't that much difference to me for this application.

Use the RMS voltage to determine the amount of series reactance (i.e. - capacitance) needed. "Regular" capacitors are rated for DC volts and their equivalent AC voltage rating can sometimes be shockingly (um...) low. X (or suppression) caps are always rated for a minimum of 250VAC which takes some of the worry out of the equation.

 

[buzzp]

Your right about the tranzorb. I will be using that instead of an MOV, no matter what I use for power. I mispoke that. I already priced them and they are cheaper than most MOV's.

I do think X caps are all metallized film, paper or plastic because of the self healing properties. I have not used them but in reading it sounds like they are only made in those three varieties as far as X2 caps.

In looking for the proper voltage rating cap (I won't use any DC rated cap without an AC rating, its just me) and looking for the capacitance I need, assuming using two in series, is somewhere around 1.5-2uf each. The metallized polyester film caps are smaller and two of these are still smaller than one X rated cap of around 1u. The size is probably small enough to fit where I need it to fit but they are polyester. Using the appropriate protection, the polyesters should be fine, I would say and it sounds like you agree.

As far as the PTC, I will be looking into this next as someone suggested. I had not considered it but I think it would be a good safety feature unless I go with a fuse (the design would be such the whole unit would have to be removed to change the fuse, removing it from all power).

If I could SMT, I would be much happier! Thanks for the reply and advice.

 

[buzzp]

I don't think using a capacitive supply is going to work unless I use only ONE metallized polyester film cap, non X rated. I have been thinking about the idea of using two. I can see the advantages of using two in series but wanted some feedback as to some field experiences using only ONE in a non-industrial environment, assuming front end transient protection like a tranzorb. Did they fail often? Also, did they have an AC rating at all? I guess, to me, if they have an AC rating and the appropriate transient protection is used to spare the cap the high voltage, then they should be reliable. Not as reliable as two in series but this is not a critical application by any means. Being size sensitive, using two caps would require the use of perpendicular PCB's, then I might as well go with a transformer.
Just looking for success and horror stories of using only one metallized polyester film, non X rated, cap, with an AC rms rating >150V. I will need about a 1u to get 40mA at 24VDC @120VAC in (needs to operate down to 105VAC), requirements changed a little for the better.
Thanks

 

[OperaHouse]

Another direction....a friend of mine uses a LED driver chip. This is basically constant current and any excess has to go into a zener. The upside is the device he makes can operate from 24V-250V AC or DC and it is surface mount. Haven't found the lit on the chip which has fallen into the abyss.

 

[Renovator1]

buzzp - I use metalized polyester and -propylene caps all the time in snubbers - a very demanding application - and have yet to see one catastrophically fail *unless* it was hit by lightning. Also, it just occurred to me that you don't need to use an X or Y rated cap because the cap will not be connected directly across the line (a resistor will be in series with it). Also, I personally prefer metal oxide resistors whenever it is expected that some abuse will have to be endured.

 

[OperaHouse]

"have yet to see one catastrophically fail" Got to agree, but that does bring a chuckle. My old boss wanted to get some suppressors CE marked that didn't have X caps in them. It didn't matter that I told him they had to be X rated caps or we would have to get our supplier certified. He spent 40K on the testing and they stopped 30% into testing. Every cap failed open. Our products were almost totally powered by caps. We used this little TECATE 250 VAC cap that was about half the size of any comparable cap. I even used 3 in series for 600V applications. Can't remember one coming back.

 

[buzzp]

I can not believe, wait yes I can, that someone would spend 40k out of pure stubborness. Surely the test house must have told him as well as you.
I agree, CE marking would require X caps. I don't think, in this case, the X caps would lead to a 'better' design, from the point of two caps in series. With my design, I am proposing using only one cap with transient protection and a fuse (or PTC possibly, still need to check out).
I suspect, if things go well, I will get a chance to use SMT. I also suspect CE marking will be needed, if this happens. Its just too bad I don't have this option now. Of course, I don't have multilayer board options either so cramming all these thru hole parts in the limited space, on two layers, will not be very desirable for trace routing/hardening for immunity issues. I can only do my best with the parameters.
I have learned alot trying to put one of these cap supplies together and do apprectiate all the feedback. I believe I have close to a final design done (except all of PCB and front end protection) as far as the schematic.