Dual layer PCB with second layer being only ground

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
After my PCB design for class D failed miserably, i noticed that it would have probably worked good if the ground plane i tried to do (and ended up being nothing but trouble) would have been a dedicated ground plane. And i'm thinking to do just that - leaving the PCB design as-is, just using thicker traces and a dedicated ground plane on the top layer. The design is all thru-hole btw. I can do SMD but since i have all the parts and they have tested working on a stripboard (and not working properly on my current PCB), and just got a little PCB drill which is really handy, i thought it'd be fun to do thru-hole. Improvising vias with a piece of wire going from wherever needed to the ground plane won't be too hard, and it sure is gonna look funky with all that copper on top.

Has anyone done this type of board before and has useful insight about it? If not, i'll probably make it anyway to see what gives.
 
Last edited:
Hello,

Ground planes are mandatory for class D amps. IR has some application notes on IRS2092 describing the importance of ground planes and how they must be designed on different stages of a class D amp.
Attached you will find an example.

regards,
savu
 

Attachments

  • bottom.jpg
    bottom.jpg
    284.3 KB · Views: 524
  • silk.jpg
    silk.jpg
    472.2 KB · Views: 461
  • top.jpg
    top.jpg
    344 KB · Views: 476
Last edited:
I've done loads of boards of all kinds. A board that doesn't work as-is is unlikely to spring to life just because you pour a ground plane over one side. Sort out the return paths on the board, keeping them as short as possible, and keeping (powerful) switching signal returns away from and definitely not crossing less powerful analog returns before you pour the ground plane. Ground planes can make you lazy in your layout if you are incautious.

w
 
always ask Where is the current flowing?

when sticking to just 2 layers you should consider added power bus bars and jumpers as options that can improve electrical performance

especially with mixed signal, switching circuits the circuit connection's parasitics and mutual coupling are major determinants of the product performance

if you're not going for at least 4-layer PCB then you need to abandon the idea that you can get all of your connections on the 2-layers at the same time as controling the coupling and parasitics
 
Hello,

Ground planes are mandatory for class D amps. IR has some application notes on IRS2092 describing the importance of ground planes and how they must be designed on different stages of a class D amp.

regards,
savu

My PCB worked fine without a ground plane.
I kept things as close as possible to the 2092.
Loads of careful decoupling.
Plenty of decoupling around the 12vregulator.
 
A class D amplifier may appear to work fine but still suffer from self disturbance at high output currents and extreme duty cycles.

Decoupling becomes useless when only a few cm of PCB tracks are added between the capacitors and the load. A ground plane reduces PCB track inductance considerably and allows for longer PCB tracks (and makes decoupling more efficient even when distances are short). On low voltage supplies, I use local decoupling with high value (1uF) SMD capacitors to ensure that the resulting LRCRLRCR..LRCR string is reasonably damped (or resonates at frequencies too low to cause trouble, (the Cs are the SMD caps, the Ls come from PCB tracks joining them, and the R comes both from caps and PCB tracks).
 
Last edited:
A class D amplifier may appear to work fine but still suffer from self disturbance at high output currents and extreme duty cycles.

The problem with a 2092 circuit is much of it doesnt rely that much on ground.
The output mosfets switch from b+ to b- and vice versa.
So I make the b+, gnd and b- tracks as thick as possible.
Add to this careful decoupling and it works fine without a power plane.
 
Last bit of FeCl3 + cold = it took FOREVER to etch. It's about done now, i'll be drilling it and populating it to see what happens. I'm thinking of building a heater thingy with a little motor to shake the container. This should get stuff done in a reasonable amount of time.

I used to etch that way. You'll _love_ the following way.

I put on rubber gloves and use a shallow plastic food-storage container and add 1 part Muriatic Acid (the common kind that's sold in hardware stores, or for swimming pools, which is actually 28% Hydrochloric Acid) to 2 parts Hydrogen Peroxide (the common "household" 3% type that's sold in drugstores and pharmacies).

This etchant can etch a 1-oz board in about five minutes, at room temperature, with gentle mechanical agitation. And it's almost transparent.

I mix it in a shallow plastic food-storage container and wear rubber gloves so I can use a balled-up paper towel to gently wipe the surfaces of the board, as it etches, which seems to speed up the etching time significantly.

Caution: The concentrated acid's fumes would probably be very bad to breathe, or to have around metallic items. And the acid would probably be very bad to get onto anything else that's not plastic.

Caution: Add the acid to the peroxide, NOT the other way around!
i.e. FIRST put the two parts of "3% Hydrogen Peroxide" into the plastic container. THEN add the one part of "28% Hydrochloric Acid".

Another good thing about it is that the ingredients are all available LOCALLY, whenever I need them, and are very inexpensive.

Cheers,

Tom
 
Last edited:
If you think about it, you could make "home-made" multi-layer boards fairly easily, by just using a stack of boards.

Imagine how much better your layout could be if you had an entire layer each for ground, V+, and V-!

You might have to drill some large holes if you really needed soldering access on an inner layer after stacking, or etch away a little copper in places where you needed wire pass-throughs. And you would probably have to assemble and solder them in a certain sequence. But you could still get many of the advantages of having, say, four layers (except for some of the inter-layer capacitance, extremely-short vias and other inter-layer connections, and a few other niceties). But remember that you can get extremely thin board stock, if needed.

I guess maybe a two-sided board on top, for the signals/components layer and the ground layer, with two inverted singled-sided boards underneath, for V+ and V-, might work OK.

If you planned it well-enough, maybe you could even drill holes in the lower-layer boards so that solder pads from the top layer(s) could be recessed and the boards could lie flat against each other, in which case they could probably even be glued together.

Or, you could even use a two-sided board for signals and components and (possibly) accept that there would be small spaces between boards and have either two (w/one 2-sided) or three (all 1-sided) more boards for the gnd, V+, and V- layers.

I must confess that I haven't tried it yet. But I've wanted to. I've done lots of two-sided boards at home and alignment is not a problem unless the boards are so long that my old HP Laserjet 4's slightly-inconsistent positioning in the direction of paper travel causes problems. But that doesn't usually cause any problems unless the boards are more than about 8 inches long.
 
Read about that way, i think i'll just stick to FeCl3 as i tend to spill it everywhere.

Anyway, my new board didn't work either. I'll keep trying. If it worked on an ugly stripboard wired mostly with thin jumper wire, i'm definitely doing something wrong. After all, most SMPS boards are single layer thru hole and they work just fine at higher powers than my amp does.

Stacking boards to make multilayer sounds cool, but it would be too much trouble IMO.
 
It eventually turns out there was nothing wrong with my layout. It was all about decoupling capacitors... or lack thereof. I had 100nF decoupling caps on all parts - opamp, comparator, drivers, mosfets. I had also left spots for rail decoupling caps and output caps for the voltage regulators. Only... i forgot to fit them. Took me a while to realize it. :smash:

I disconnected the gate driving caps and started probing around with my scope. The comparator was still oscillating even without the final stage working. There was about 70mV ripple on the positive 12v rail, 4 volts on the negative... Whaaaaaaaa???? FOUR VOLTS??? The 7912 was oscillating.

I fitted the 1uF output capacitor for the 7912 and it went away like magic. I reconnected the gate driving capacitors. There still was trouble with the pmos gate drive - it would swing back and forth several times during the cycle. Actually the problem wasn't with the gate drive - it was the power rail swinging down! :eek: 47uF caps on the rails and it's gone. The 6800uF main caps are near the transformer, i'm using a mains transformer for testing. A meter of wire between caps and PCB sure makes funky stuff happen.

I still have a little problem - the n-channel mosfet is taking too long to turn off, but i'm sure i just need to adjust the resistors in the driver circuit. The waveforms look clean now.
 
Have a good look at the board and see if you can see an error. One good way of doing this is to write a netlist or export it from your schematic capture software and then check it pin by pin. If you export it you have to be careful you're not exporting an error you made in the schematic.

You can often make an incorrect board work by breaking tracks (if necessaary) and running wired connections. These are called 'white wires' in professional circles. White wire is in fact often used. You can even drill a hole in the board to pass a wire from one side to the other. Run the white wires neatly in straight lines with right angle bends and glue them down with hot glue if you want to keep the board.

w

Oh, I see you posted while I was writing...
 
Okay, here's material for laughs.

I traced the slow turnoff (and the fact that the mosfets were still heating like crazy) to a shorted 1N4148 in the pmos drive. I replaced it. Afterwards BOTH waveforms were looking "slow" and things got even more toasty. Which led to the original mistake. That was... wait for it... i had wired the 1N4148s backwards. :eek: Cross conduction ahoy! I fixed that and the amp is now working fine. The heatsink is still getting *a little* warm, but that's the dead time resistors requiring some tweaking, the residual output is a pure sinewave, dead time is zero. :)

Thank god for the lightbulb before the transformer otherwise i would have been looking at a pile of fried mosfets instead of just heating up. ;)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.