[yawg3d]

Hi everyone,
 
the following design is meant to be used as a PSU for various CMoy-like projects. In its current form, it should take 8V AC from a wall wart and output +-8V DC. The reason for this approach is that I do not want to mess with mains wiring myself (otherwise I would have gone with a double-secondary transformer).
 
 

 
 
The design is based on datasheet material and some textbook ideas.
The values for C2 and C3 are taken from datasheets and should be okay (C2 will be tantalum). The value for the filter caps C1 are based on a quick-n-dirty Spice simulation of the rectifier/filter stage which says that as long as the load draws less than about 250mA, the input wave form will stay above 8V at all times. Diodes are meant to be 1N4001s.
 
I would be very happy if some experienced folks could have a quick look to check whether (a) the design looks sane and (b) I can get away with 16V caps or whether I should go for 25V to be safe.
 
Thanks !

 

[Avro_Arrow]

Yes, it will work.
Good practice says use caps rated at 2x the expected voltage.
If you expect 8 volts, a 16 volt cap should be good enough.

 

[00940]

It could work but barely. 8Vac in this configuration would ideally give you +/-10.5Vdc. The regulators need at least 2v of headroom to operate so you're a bit short. As soon as you'll draw some current, you won't have enough headroom (quick sim show ripple going down to 9V at 250ma).
 
What could save you is using a wallwart which is more powerful than you need. Lightly loaded it could produce a bit more than 8Vac.

 

[yawg3d]

Thanks for your input! New plan, possibly better: Using a 12V AC supply (so amplitude 16.8V), 12V regulators and 35V 2200uF caps. Simulation gives a load of about 1.5W (although purely resistive), and the ripple never drops below 15V, giving at least 3V of headroom to the regulators. That should be okay, no?
 

 
Alternatively, using a 9V AC supply for the 8V regulators in my schematic, with 25V caps (I don't actually need +-12V, +-8V would be more than enough...) :
 
 

 

[00940]

Either should be ok. I'd take a 1A version in either case.
 
Btw, for 9V input, 16v caps are fine and for 12V input, 25V caps are fine.

 

[yawg3d]

That would be great, as I can save a lot of space with 16V caps (or go for more uF at the same size). I'll shop around a bit and see what I can find .

 

[yawg3d]

Do I need some form of inrush current limiting for those caps? The simulation shows 7A of charging current during the first charging phase, which is way too much for my AC Adapter and the diodes of course. Are there any real-world effects which do not show up in the situation and help prevent this? If not, what should I do?

 

[00940]

Neither the transformer nor the caps are perfect, this will limit the inrush current a lot. Don't worry. And even lowly 1n4001 have a 30A peak rating on surge (just don't use 1n4148 for PS, they have higher drop than 1n4001).

 

[agdr]

Quote:

Do I need some form of inrush current limiting for those caps? The simulation shows 7A of charging current during the first charging phase, which is way too much for my AC Adapter and the diodes of course. Are there any real-world effects which do not show up in the situation and help prevent this? If not, what should I do?

 
Add 1.53 ohms of series resistance in your 9Vac spice voltage source model (or just add a 1.53R resistor in series with your ideal voltage source).  That is the secondary series resistance I just measured on a CUI EPA090100-P5-SZ 9Vac 1A wall wart that I have
 
http://www.digikey.com/product-search/en?x=0&y=0&lang=en&site=us&KeyWords=epa090100-p5-sz
 
Like 00940 says you will find that will dramatically lower the charging current in the input filter caps and through the transformer secondary (and diodes).  It can still be significant however, and you need to make sure the ripple current rating of those capacitors from their data sheet is adequate.  Remember to de-rate the capacitor's ripple current for frequency by the little table the cap data sheets give.  At 50/60 Hz the rated ripple current (often given at 100kHz in the data sheet tables) is usually only 0.6 or so of the listed value.  You may (probably) need to use two of the 2200 caps in parallel to split the ripple/charging current spike between them to stay within bounds, if the de-rated ripple current of one 2200uF is not enough. I would leave at least 20% overhead (extra) on that de-rated ripple number on each cap.
 
Also consider using a LT1963A (positive) and LT3015 (negative) adjustable regulator adjusted for +/-8V, respectively, instead of the LM fixed voltage regulators. Both of those LT low dropout voltage regulators only need about 0.5V across them at 250mA which would recover that 1.5V you are short on each rail.  Both LT regulators also have significantly better noise ratings than the older LM parts.  The LT regulators don't need those protection diodes across them either. The downside is more expensive and a few more parts on each (resistors and one small capacitor) to set the voltage.
 
And consider going with Schottky rectifier diodes, like the STPSL260
 
http://www.mouser.com/ProductDetail/STMicroelectronics/STPS2L60/?qs=sGAEpiMZZMtQ8nqTKtFS%2fBdOQ7hRVdjrGbiVpKcsU7g%3d
 
that would only have about 0.2Vdc of voltage drop at 250mA vs. the 0.7Vdc or so for the standard silicon rectifiers.  That would recover another 0.5V or so that you are loosing on each rail. The Schottky's don't have quite as large of a current surge rating so that is why I've picked a 2A unit here. If you want to add a snubber to reduce any chance of secondary ringing caused by the rapid switching of the Schottky's, add a 0.01uF cap across the 9Vac line near the diodes, followed by a 0.068uF in series with a 110R resistor, also across the 9Vac line.  The 0.01uF cap lowers the ringing frequency while the 110R resistor dissipates the energy.  For the details see Hagerman's famous paper
 
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CCwQFjAA&url=http%3A%2F%2Fwww.hagtech.com%2Fpdf%2Fsnubber.pdf&ei=NwDBUYqSHY7S8wSZ8YCwBg&usg=AFQjCNGf7gd2pGvl6aR7O-3if76UrugW4w&bvm=bv.47883778,d.eWU
 
Good luck and have fun!

 

[splaz]

I'm not an expert, but I'll still throw some thoughts out there.
 
^ I was going to suggest as above, even if just LM317/337, only a few more components for adjustable regs then you have a lot more flexibility.
 
Are you getting boards made, etching, just using perf/protoboard ?
 
If you are having them made it may be an idea to get even more flexible and allow for different rectifier configs, say dual secondary/centre tap transformer instead of voltage doubler just by tweaking your layout to allow for multiple configs.

 

[yawg3d]

So much information - great stuff !
 
@agdr: I need to double check those ripple specs again. I wasn't really aware of that until now (and surprisingly, most head amp diy PSU designs out there do not seem to use parallel caps there). The LT regulators are a bit expensive. I may well try them later, but only if I can make the cheap ones work reliably. I will certainly try the Schottky diodes though!
 
@splaz: I'm getting PCBs made by a small one-man shop the does great prototyping boards (two-layer, plated through holes, no solder stop or silkscreen though) at really good prices. Will try the flexible regulators with a possible revision B of my board!

 

[yawg3d]

Thanks to everyone who helped! I finished the PSU and an Apheared 47 today. Have some pics :
 
 

 

 
... and both worked just fine from the start!

 

[holland]

Quote:

@splaz: I'm getting PCBs made by a small one-man shop the does great prototyping boards (two-layer, plated through holes, no solder stop or silkscreen though) at really good prices. Will try the flexible regulators with a possible revision B of my board!

 
share your source?  are you in the US?  etching boards is fun....the first few times, it just gets tedious after.

 

[yawg3d]

I got the boards at pcb-devboards.de. I think Mr Ulrich will ship to the US if you ask him, but I'm not sure whether it's economically sensible to do so. For prototypes, the prices are hard to beat (except mayber if you have them done in China) - I found a few similar offers for one-sided boards, but noone else offers affordable 2-sided boards with plated-through holes.

 

[00940]

Thanks for that website. Prices are indeed very nice for dual-sided pcbs.