Capacitor Layout

.

There will be a lot of peak current in the junction of all the ground leads. Try to keep that layout tight and concise.

JR

I can't tell much from that picture...

I would orient the capacitors so all the ground leads are physically close to each other for short, low resistance interconnection.

JR

80hinhiding said:

In the attached graphic I just made, which layout/wiring is the best to use in a power supply?  And why?

Click to expand...

It's as hard to answer generally as a question like, "What is the best wine to pair with chicken?" You just need to follow JR's advice, and try to minimize ground impedance.  You really need to include the rectification into the picture as well.

As a large generalization, it's best to orient a bank of filter caps so the connections that all share a common reference are all clustered together, as that will generally minimize impedance.  Other than that, it's impossible to answer without a complete schematic.

80hinhiding said:

In the attached graphic I just made, which layout/wiring is the best to use in a power supply?  And why?

Adam

Click to expand...

The way the caps are connected together is not as important as the way they are connected to the rest of the circuit; you have to understand that "ground" has a resistance that allows rectifying currents to develop parasitic voltages and that these parasitic voltages should not be allowed to mix with the signal flow. See attached. As JR mentioned, keep all connections short and solid.

abbey road d enfer said:

The way the caps are connected together is not as important as the way they are connected to the rest of the circuit; you have to understand that "ground" has a resistance that allows rectifying currents to develop parasitic voltages and that these parasitic voltages should not be allowed to mix with the signal flow. See attached. As JR mentioned, keep all connections short and solid.

Click to expand...

Chiming in here to agree with this wholeheartedly. These rectified, signal induced power supply currents, in a circuit powered by split, bipolar supplies, can cause fault currents that, if not handled through ideal PCB foil layout, can induce signal related voltages that can cause the circuit as a whole to greatly exceed the distortion of the amplifier ICs used in the circuit.

An amplifier powered by split supplies will generate power supply currents in each supply that need to be re-combined away from sensitive circuit nodes, so that the voltages developed by these essentially rectified signal currents, do not find their way back into the actual signal path of the circuit. I have seen many examples of circuits built with simple, sometimes single sided PCB layouts, that have resulted in these rectified signal induced power supply currents being added to sensitive circuit nodes through simple PCB foil issues, and added to the circuit output. The Ashly SC-50 is a good example.

The easiest way to avoid this is to not be cavalier about power supply 'bypasses' and 'ground', and instead consider these power supply bypass current sources to be sources of rectified signal injection. If these power supply currents can be isolated from the ground references of the circuit, and at least re-combined so that they become much more benign 'simple signal currents', then the PCB layout will deliver much better performance.

This process and perception is probably the best answer to 'what is the best geometry of power supply bypass capacitors'.

When covering old ground my posts tend to be a little brief.

Layout of PS reservoir caps is a good object lesson about fundamentals.

A few givens...

-all PCB traces and wires have finite resistance so current flows will generate voltage drops (ohm law E=I/R).

-charging capacitors from transformer windings with diodes result in very non-linear currents. High peak current for small fraction of mains waveform.

Optimal PS design uses a combination of both brute force (making runs short and fat), and being smart about where the current is flowing. Small voltage drops in series with the charging capacitors are inconsequential, but these currents and voltage drops in series with a clean ground are important, so try to visualize the flowing current like water in pipes, and keep the sewage out of the drinking water. .

If you imagine every trace/wire like a small resistor that can help with understanding which ones matter and which don't as much.

Hope this helps.

JR

Hi,
this is more tube amp oriented but same principles apply:

http://www.valvewizard.co.uk/Grounding.pdf

80hinhiding said:

I hate to admit it but I'm still confused.  Are you saying not to use too much capacitance directly after the diode full wave bridge

Click to expand...

NO, use as much capacitance as you need/want to keep ripple voltage within desirable range.

rectifier, and to keep the junctions a specific physical distance from things like the regulator? 

Click to expand...

It is not about distance per se, while resistance of PCB traces or wires increases linearly with length.

The concept of keeping the crap out of the ground is not really clear.  I understand we're filtering out frequencies and creating a DC waveform for the sensitive audio circuits, and we don't want other junk showing up there.

Click to expand...

The crap needs to go down the sewer, the confusing thing about "ground" is that it must both be dirty and clean (I know that isn't helping).

I've noted to keep my connections short and layout tight, to minimize resistance.

Click to expand...

The current flow when charging reservoir caps is mainly at the peak voltage of the mains waveform. This current will develop a voltage across even short traces/wires and this voltage is the all too familiar hum/buzz we hear from poorly designed products. This current is unavoidable, and larger caps will generate higher peak current, but smaller ripple voltage so a common tradeoff.

This current (like sewage) is always with us, and can't be avoided, but we must keep it in the correct pipes. This same current (and small voltage drops, will be in the wires coming from the transformer too.  These current induced voltage drops will be small in the context of capacitor ripple voltage so inconsequential. This means we will have small voltage drops all over the place... It is good practice to keep traces fat and short to keep these voltages small, but more importantly we need to grab the ground (0V) reference from only one arbitrary point in the ground node.  Voltages are relative, so the small drops will appear superimposed on top of the reservoir caps where they are inconsequential.

If there is zero current flowing in the path connected to this ground reference there will be no additional voltage drop errors.

Hope this helps.

JR

Edit: I'm reading this document right now: http://www.valvewizard.co.uk/Grounding.pdf

A

Click to expand...

80hinhiding said:

This is quite interesting. Hmm
So would two tinned bus wires, a smaller circle inside a bigger circle be of any benefit?  The smaller circle for ground reference, one point only in the center of them all.  Positive lead of cap to the bigger circle bus, negative lead to the inner circle bus.

Click to expand...

It doesn't matter as much where the one ground point is, only that there is only one point.

Is that taking the idea too literally?

Click to expand...

it was meant literally (while the sewage allusion is figurative).

Should I shield the reservoir bank from the rest of the power supply with some copper?

Click to expand...

probably not... you are confusing magnetic or electrostatic interference that travels through the air...  Layout is all about conducted interference through electrical connections.

JR

A

Click to expand...

Of those figures (assuming these are showing "mechanical" connections), I'd use 1A or 2, with the bridge rectifier hooked to one side, and the circuitry being powered hooked to the other side. Traditionally on schematics the transformer and rectifier on the left. Power and signals tend to go from left to right.

John Roberts covered this, but perhaps it will help if I say it from another direction. Schematic wires and connections are all assumed to have zero resistance, but of course real wire and PCB traces don't.

This is fine for many circuits, but not for some others. It doesn't account for the high peak current and resultant voltage drops that tend to happen in power supplies. The rectifiers only charge the caps for maybe 10 percent of the power cycle, so to make up for the (for this argument) constant discharge of the load, the current between the rectifier and the caps will be 10 times the load current during rectifier conduction. This current can cause a significant voltage drop along the wires between the rectifier and the capacitors. Connect your power and ground to the powered device across the caps, not across the + and - output of the bridge rectifier.  This makes the ripple voltage higher at the bridge.