Power supply and GND

[aballen]

Hey guys I'm back from vacation and getting ready to build this.  Still don't understand the 1000uf capacitors though.  I've seen them in other places, but not with a good reason, in fact I found an instructable that states they are not needed, but they look cool so he put them in.

Can anyone explain why they are needed?  I would hate to omit them if they are really needed, but if there is no good reason for them, I'm not putting them in.

[Cliff Schecht]

Those caps are there for a few reasons, some not so obvious. The obvious reason is to filter out any noise from the linear regulator, especially the 780x series which don't have very good line/load regulation (meaning they only remove so much ripple on their own). Capacitors placed in the shunt configuration (i.e. a capacitor to ground) act as a filtering element to remove additional ripple that the regulator doesn't squash. While the datasheets typically recommend a relatively minimal value that yields acceptable performance in most applications, we are actually a somewhat specialized application. Modern guitar pedals and amps have a LOT of gain that will amplify any noise introduced from the power supply. The easiest way to buck this noise is at the source which in our case is the power supply. Larger capacitances will filter out more noise which means less crap gets into your pedals. I think I use about 470uF after a linear regulator and my power supply doesn't have any noise problems at all (now my pedals OTOH.. ).

The not so obvious reason to use this capacitor is to keep the linear regulator stable. Many regulators will go unstable (oscillation, weird latch-ups, smoke, etc..) without some sort of capacitance at the output. The easy way to stay out of trouble here is to check the datasheet. If they don't specifically say that their device is stable with no input/output capacitors then I assume it isn't and use at least the recommended minimum values (not that it costs much more to go overboard here).

[aballen]

If this helps filtering out noise, and keeps the power clean, well then I want to do it, the whole point of this power supply is that each jack will be isolated and clean.

Looking at other examples here http://www.generalguitargadgets.com/projects/24-power-supplies

The only one that uses the big 1000uf cap is the "ultra clean" power supply, and it appears to only use one on the input side of the regulator.

So should I do two 1000uf caps, instead of the small ones recommended for ripple protection or in addition to them(or just one)?

I never really expected this design of a power supply to be so involved.

Again, really appreciate the feedback.

[Cliff Schecht]

1000uF really is overkill. I even wrote that in my last post but I guess deleted it before I posted it for whatever reason. After a linear regulator anything above 100uF really isn't noticeable IME.

Power supply design, like all analog design, is a very detail oriented process. You have to understand every part of the system on all levels in order to make the best design choices. Nowadays they make it very easy with ready-to-go power solutions on a chip but things can get hairy quickly when you add multiple output voltages, control circuitry and more power handling.

Of course what we do isn't exactly science and isn't quite art, it's a happy medium that leaves a LOT of room for creativity. I've met guys that I consider real artists when it comes to analog design but it's pretty funny how they are in awe of the cool stuff that we do with analog. Last summer during a coop with some very talented IC designers (guys with many more years of experience than myself), I put a big smile on their faces when I brought a theremin into work, plugged it into a little 5W tube amp I built and let them wail away for a while.

[tempus]

While we're talking about noise contribution from the PS, word has it that using an adjustable (LM317 for example) is much quieter than the standard 78** series. I know that in mic preamps I've built using 317/337 has been quieter to my ears (I don't have the test equipment to back this up though), and they discuss it quite a bit over at diyaudio, where they're looking for super quiet specs.

Not sure if that's something we as musicians should be too concerned about about or if our amp noise swamps any PS noise from our pedals, but it is pretty easy to use the 317s and I don't think there's a much bigger parts count.

There's also the clean up shunt design here: http://www.wenzel.com/documents/finesse.html although there's some debate over how well it works in the real world.

[Cliff Schecht]

The LM317 series regulators have better line/load regulation, better accuracy and better inherent ripple rejection than the 780x series of regulators. They also have a lower overall noise floor. The 317 really is a better part and is what I use in my pedalboard power supplies. It's not very hard to get the standard circuit built and working quickly and the noise floor of the power supply is definitely good enough for rock and roll!

[PRR]

> don't understand the 1000uf capacitors

We only use the "Peak" of the AC wave. 80% of the time, the load runs on energy stored in the first capacitor.

So it has to be big-enough.

As a _general_ guide, think 1uFd per 1mA, or 1,000uFd per Ampere of load.

In lower-voltage systems, you often want more.

You are hoping to start with 11VAC and regulate to 9V DC. The 11V AC would give 15.5V with perfect rectifiers and infinite capacitor. Real rectifier loses a couple volts. Practical-size caps sag. The 9V rectifier typically needs 3V extra so it can shave-off to 9V exact.

You need a good 12V DC at the end of each half-cycle when the cap has sagged.

Duncan PSD is handy for such predictions. You need some estimate of transformer and capacitor resistance; the defaults are for high-voltage work.

Duncan needs a load. If you need about 12V raw at 0.3A, that's 40 ohms. (Somehow I typed 50 in the simulator so these sims are optimistic.)



The 300uFd cap sags to 9.3V. You can NOT get regulated 9V from that.

The 1,000uFd cap sags to 11.8V.

The 2,000uFd cap sags to 12.3V. This will work.

The 1,000uFd cap "might" work _if_ the "11VAC" runs high or the load is less than 300mA. (But it is bad practice to rely on "if"..... what if you play a gig with low wall voltage? What if you get a new pedal that sucks a full 300mA?)

[brett]

Hi
+1 for the 1000uF per amp rule (works for power amp supplies as well as pre-regulation).

General comment: It is interesting how terminology causes confusion. Many countries have an Earth wire for AC (which is connected to THE Earth), and Grounds (positive, negative, whatever) for regulated supplies and other stuff (not necessarily connected to the Earth). Simple, descriptive, cost-free and strangely not adopted in some countries. A bit like SI and metric.
cheers

[PRR]

> terminology causes confusion

And it's not a national thing.

There are MANY things we call "ground" (or "earth"). They may even connect. They are not the same!

Offhand:
Safety Ground - ensures your metal case is at a reasonably safe potential.
Garbage Ground - diverts buzz/RFI around sensitive systems
Power Common
Signal Common
Power-Signal Common

[Cliff Schecht]

On complicated mixed signal board with analog, digital and switching power supplies onboard I usually use at least three separate ground symbols (AGND, DGND, PGND) to help differentiate what components are associated with which part of a circuit. All of them usually connect together either on a dedicated ground layer (four layer boards) or at a single grounding point that is very close to the bulk decoupling capacitors "-" terminal (1-2 layer boards) depending on how many layers are needed and what the project budget looks like.

Then if mains wiring is involved you also have the earth ground to deal with. The proper way to wire an "earth" ground (the green wire from your standard black IEC power cable) is to drill a hole very close to the power inlet, solder the earth ground to a solder lug and securely attach said lug to the chassis. This is what UL and CE labs expects to see anytime they open up and inspect equipment with 3-wire mains power. Lock-tite on the fastening screw isn't a bad idea (this is what most big manufacturers do) but I don't bother, just crank that screw in very securely.

[aballen]

Now that is a great explanation.  Thanks guys.

Of course I have another question.  If I go with the 2000 uf cap, will that compensate for me going with a 7809?  I can go with a 317, but its pretty much back to the drawing board if I do that... and I have never worked with the 317 before.

[tempus]

I'm not sure what you mean about the 2000 uF cap and 9v. According to PRR's calculations (which are reliable - this guy knows his stuff) the 2000uF cap is ideal for 9v.

Using a 317 is almost as easy (and almost the same thing) as using a 7809. The only difference is that you have to set the output voltage with a pair of resistors. Check out this page:
http://www.electronics-lab.com/articles/LM317/

In your case, you would replace the 5K pot with a fixed 1.5K resistor and your output should be 9.06V (according to the calculator on the page). You can also bypass the 1.5K resistor with a 10uF or so cap to give even better noise performance. You can keep the input to output diode the way it is.

[aballen]

Ok guys I think I got it... completely reworked.  The few additional parts make the layout a little trickier so I made one circuit, which I will duplicate 8 times across my board for eight ultra clean isolated outputs.  I'm thinking I can do eight across, or 4 across, stacked if it will save space, and I left room for a heat sink, in case I need it(don't think I will, but just to be safe, there is space).

Here is my circuit, I think I got it all right... 2200 uf cap on the input side, switch to the LM317, two extra resistors, an extra cap on the output side for even more filtering.


Not sure if my cap values are all right, but I think there is room on the board for different values, making space for the 2200uf was key.
Here is the layout


It is getting a bit silly, but all of my pedals are DIY, so I want my power to be.... once power is worked out, I'll DIY a nice pedalboard too.

I'm hoping for another round of feedback on the new design... so let me know what you guys think... I'm ready to etch, but I need to order some big caps too, so there is no crazy rush.

Really appreciate all the feedback I'm getting here guys.  Great stuff!

[tempus]

Looks pretty good. I'm not sure how far your regulator is going to be from the 2000uF cap at the rectifier. but if they're fairly close I don't think you really need the 100 uF cap on V_in What I would do is throw a .01 - .1uF cap from V_in to ground and V_out to ground. The bigger caps filter out low frequency stuff and smooth out the ripple, and the lower value caps do a better job of filtering out the higher frequency stuff. Some designers even call for 3 sets of caps - 1 for low (big electros) 1 for medium (around 0.1uF) and 1 for high (0.01 or smaller). I think I would probably use something a little higher than 1uF at the output - maybe more like 10uF.

[Cliff Schecht]

In the schematic just posted, I would swap C2 and C4. Also a 0.1uF right on the pin for Vout is never a bad idea.

Also, some general Eagle advice, you don't need to put junctions on single connections. If you can grab a part and the wire moves with it, you are golden (meaning you have proper connectivity). Junctions can force connectivity but aren't necessary. If a wire isn't grabbing go to "move", click the part and drop it in the same place. The only place you need junctions is where two or more wires connect (wires include pins and such). If you don't place a junction on every wire junction, Eagle's ERC will bitch at you (not a bad thing).

[Jdansti]

I'm by no means a PS expert, so I'll pose this question to those who have experience and knowledge in that area.  Would it be a good idea to have a switch on the "hot" AC leg (as opposed to the neutral leg) and a fuse?  Maybe these features are already intended off board.

[PRR]

> if they're fairly close I don't think you really need the 100 uF cap on Vin

Correct. In the old days we sometimes had the raw power in one rack and the regulator in another rack. Then you need a little 100uFd *at* the regulator to steady it. If they are less than a foot apart, one good input cap also steadies the regulator's input side.

Old-old large caps didn't work good for high frequency (and these regulators are high-frequency chips even if we only see 120Hz and DC on them), so we used a small fast part-uFd cap to steady the high-frequencies. With modern large electrolytics (even to 2000uFd), you probably don't need to have a 0.1uFd.

[aballen]

Who knew a simpl question about ground would turn in to this.

Ok so to answer one question, yes I will have a switch off board, with a fuse.

Here are my latest
1) dropped the small cap on the input side to a smaller one
2) added a .01 uf cap on the output side
3) replaced the 1 uf cap with a .1uf cap.

Can I go with film caps on the smaller ones, or do they need to be polarized?





Are we getting there?

[Cliff Schecht]

Almost, but I'm not sure why you dropped the cap value after the regulator so much. They recommend at least 10uF here IIRC and I prefer closer to 100uF. Remember this capacitor not only filters out supply ripple but also any uncorrelated noise that is added from the regulator itself. Linear regulators use large transistors as either a series (pass) or shunt element which will always have a certain amount of noise associated with it. Take a peek at the datasheet, they even specify a certain noise performance for this device with the recommended capacitances. More capacitance at the output means a lower cutoff frequency for the output filter and better rejection of unwanted noise from the regulator.

[aballen]

I dropped the input cap because I thought you were suggesting that when you said switch C2 and C4.  I've bumped it back up to 100uf.

Seems like I'm getting close.  Here is the latest.... also, is a non polarized cap ok for the small values.  I have tons of film box caps in .1 and .01 and I think I'm going to have a hard time finding a .01 alum electrolytic.