Voltage Regulator Design

I agree with all the illustrious posters here that layout is very probably the problem. You don't say what particular variant of the part you are using, which will make a difference (obviously) in the layout.

I have some generic rules for power layout (all available in app notes and many of which are in your datasheet). Every power supply design is different, and as noted, autorouters are not the optimal tool (to put it politely) for the task.

1. Keep the high current path separate from your control signals. This includes the ground return!
2. Put a void (all layers) under your inductor. I have had the wierdest things happen, and my safe method is to ensure there is no intereference with the magnetic field. I don't care how well a manufacturer says the inductgor is shielded - if it's not a toroid there *will* be some vertical (through the board) magnetic flux and putting a plane underneath will interfere with it, effectively changing the magnetics (even toroids have issues, but most of the flux is contained withing the ring)
3. Route your feedback and sense signals in a shielded fashion. Using a

4-wire style sense is best - never use the output plane for sensing. I have seen supplies that do work ok, but I prefer not to risk it.

1. Take the output to plane (with a nice large chunk of copper from the output of the inductor as close to the output of that inductor as possible.

2. If you have a device with a ground / power pad, punch vias through it (lots of them) for both thermal and electrical reasons. If you use the LLP-14 package, for instance, the pad is ground and should be punched to a plane and extended to the ground pin on the package for the shortest possible path for power ground (the switches are inside the device) to plane.

For the specifics here, I strongly suspect you are violating rules 1 and 3 (which are, of course related).

Without seeing a physical layout it's hard to be certain, but switching noise on the control / feedback signals is the most common failure I know of in switchmode layout.

Hope that helps

PeteS

Thank you guys for the suggestions regarding PCB design...

I did some searching online and it seems that someone else had a similar problem with this chip and their problem was that their input cap was too far from their chip, causing spikes on the input voltage. I had checked the voltage at the input to the board, but not at the input to the chip (I'm a dummy). I checked it at the chip and sure enough, every time the FET was turning on, there was a ton of noise. I put a cap directly across the input and ground to the chip that wasn't working right, and sure enough, it worked perfect. The 2200uF input cap was about 2in away, and I guess that is enough to mess with it. Constraints on the board size and enclosure dimensions forced me to put components in places I normally wouldn't have put them, but somehow I will have to change that now. The reason the 9V side worked, presumably, is because that cap was able to be within 0.5in of the chip. I never would have guessed that 2 inches of 100 mil trace could cause such serious noise!

The inductors that I am using are both toroid inductors and are in fact located as far as possible from any other components and especially sensitive connections such as the feedback. I like the idea of a void near them to make it even better shielded though, thank you.

So, until my next problem with this board, it seems that I'm doing alright.

I will take your suggestions and try to implement them into this design to make it even better and of course remember them for the future.

I would love to learn more, though, can anyone recommend a good website that outlines PCB design techniques, or a link to a PDF, or a good book that I can order to learn the best way(s) to design PCBs? I really want to get better at this and I'm a real newbie when it comes to it - I'm only learning as I go - and from my mistakes!

Some of the best power supply layout guides are from the controller manufacturers. Linear Tech has some truly excellent application notes (although their parts tend to be expensive). TI has made great strides and now produces very high quality application notes for their power range. IR does some pretty good app notes and tech briefs (slanted at their FETs and drivers, of course, but worthwhile nonetheless).

National, whose part you are using, also have decent app notes (they used to do a 3 book set of application notes for various parts - maybe that's available electronically nowadays)

I would head over to the sites and browse around.

For some interesting commentary and articles on the analog world from Bob Pease (staff scientist at National), head over to

(main page at

)

I don't know if there's a specfic book that addresses SMPS layouts (it's covered in a lot of texts somewhat tangentially, but I am not sure if there's a dedicated one).

Cheers

PeteS