SparkFun Forums

[Hengy]

I just finished my first PCB design in Eagle. I pretty much figured out the software myself, referring to the Sparkfun tutorials when I was stuck.

So here is my design. I made most of the parts custom. It is a power supply for a hybrid opamp/tube/MOSFET preamp/headamp. The goal is lowest noise possible. I tried to implement a star-ground system as much as possible after the last filter capacitor, C2. IC1 (LM317) will drive the noise sensitive audio circuitry, and IC2 (7812), the tube heater. The rectangle thing between the input terminal and the filter capacitors is a toroidal inductor.
The files are here: http://matthengeveld.com/np100v12/power_supply.zip



Would anyone be willing to take a look at my design, and check for mistakes, things I've missed, or things I could improve on? I plan on using BatchPCB. All suggestions are be greatly appreciated!

Thanks, Hengy

[Ross Robotics]

At first glance, it's good. But pretty sure someone with more experience with op-amps will chime in.

Could you post the schematic to see if your connections are correct?

[Hengy]

Hey codlink,

Thanks for your reply.

I have this made on a breadboard, and there is very little noise. My biggest fear is that I will get noise from some rookie mistake designing the PCB.

My concern is, "if I send this to BatchPCB as is, will it work?" :think:

I have double checked all the connections, and it also simulated perfectly in LTSpice. I am not concerned about that. The schematic is in the zip file.

Thanks, Hengy

[leon_heller]

There are a couple of acute angles between tracks. They should join at right angles.

[rrpilot]

Are you doing anything with the second layer? I would put a big solid ground plane on the other side.

[Garth]

Leon's right. You should never have angles of less than 90 degrees. If all the etchant doesn't get washed out of the corner thoroughtly, it will continue to corrode the metal.

I would put .1uF monolithic capacitors from each pin of the LM317 to ground, as close to the pins as practical, for better stability, noise, and regulation. And yes, the ground plane idea is advisable. That doesn't mean that every single ground connection goes immediately to the plane, but grounding and unwanted coupling and noise is too complex to discuss here.

[Hengy]

Thanks for the posts.

I will fix the acute angles.

Garth, I have 0.1uF ceramics beside the output connections. I have tested this circuit on a breadboard, and it works perfectly. I can only assume it will perform the same or maybe better on a PCB (of course, as long as I design the PCB right!).

As for the ground plane, I have looked at both a ground plane and a star-ground approach. The 7812 on the left will be feeding a rather noisy circuit, which I know gets through the reg. despite the noise suppression. I think that using a ground plane might end up coupling some of that noise to the LM317. Also, from what I have read, star-grounding everything to the last cap in the filter stage does a good job of moving stray current to the caps and not the rest of the circuit.

Writing this, I just realized that you might mean something else.. You said that not every single ground connection should go to the plane. So, do you mean that there should be a ground plane, connected at a point - either the entry point of ground, or the last cap in the filter stage, and no connections to any components? Sort of like EMI shielding?

If you could explain that a bit more, i would really appreciate it!

Thanks, Hengy

[Garth]

I can kind of see what you are trying to do there. As far as noise issues go, the 7812 and LM317 are not very high-speed devices, and they certainly don't give the complications of switching regulators. For your application, what kind of noise are you especially trying to get away from?

I know you're not doing a switching regulator, but depending on what you're powering that has you concerned about noise, the following might be helpful.

Grounding, shielding, and ground planes are a complex subject, and what applies in high-impedance, low-frequency circuits like audio is not the same as what applies in low-impedance, high-frequency circuits like RF, digital, and switching regulators. What probably escapes the hobbyist most is that in the latter, the ground return current does not follow the shortest & widest path, but instead, mutual inductance makes it travel directly under the trace whose current it is returning, taking on the shape of the trace, so there is equal an oposite current flow in the two directions, effectively making it a balanced circuit. This minimizes inductance as well as unwanted stray coupling. For this reason, copper pours and ground fills do not qualify as a ground plane, as they do not provide a continuous return path under a trace.

As for the star, the center of a star should be the place that matters most, like the ground pin of a switching regulator IC, or of a data converter. Having the center at some irrelevant place out in the middle of nowhere is generally not very helpful. Even in the case of just feeding power, it should be the power source and be well bypassed to ground right there.

If there are separate planes on a board separating something like digital from analog circuits to keep digital noise out of the analog, and the grounds are joined in one place, any traces that go from one circuit to the other should cross the barrier only at the place where the planes are joined, not out where there's a cut between the planes, and there should be a bypass capacitor on the power right there at the junction.

When I design switching regulator circuits, I always add ferrite beads at the input and output to help keep the high-frequency noise out of other circuits. Layout for switching regulators is very critical.