PCBs and Grounds

[Tim Shoppa]

"Motorboating"? Is that some sort of ripple?

It's a form of ripple induced by the output stage load that causes
the whole circuit to oscillate. Imagine a
comparator that drives a medium-current load like a relay. The voltage
rises above the threshold (which comes form a voltage divider from the
power supply), the comparator turns the output on, and the load pulls
the power supply downa fraction of a volt. The threshold (because it
comes from a divider on the power supply) now drops, the comparator
turns off, and the cycle repeats. Frequency is often in the 10-50 Hz
range (hence the "motorboat").

It's not as big of a problem today with the widespread availability of
stiff voltage regulators being so cheap and the usual isolation of "power and
logic" supplies from "analog" supplies, but it happens, especially when
electrolytics dry out and no longer do their job. I stated
the problem in terms of droop on the power line but with non-zero
impedances on ground connections, "ground bounce" as a result of loads turning
on and off can induce a similar thing.

It can also happen with pure analog circuits but generally the
designers of these use topologies that are nearly insensitive to this
oscillation mode. Start cascading a couple of common-emitter amplifiers
though and you'll run into it.

Tim.

 

[John Larkin]

Thanks, Jim. Looks *very* much applicable!

Be careful of the Ultracap stuff; most of it is nonsense, including
this particular paper. Their PCB impedance calculators seem to use
grossly simplified equations, and often deliver absurd, even negative,
trace impedance values.

John

 

[Paul Burridge]

Be careful of the Ultracap stuff; most of it is nonsense, including
this particular paper. Their PCB impedance calculators seem to use
grossly simplified equations, and often deliver absurd, even negative,
trace impedance values.

Oh bugger. All this contradictory information isn't conducive to
trying to learn this complex subject. Even people who seem to know
what they're talking about frequently disagree with each other on this
group, I've noticed. :-(
BTW, a scan of the new PCB I've made for that notorious Field-Strength
meter has finally appeared on ABSE. I've tried as best I can to
implement the suggestions I've been given in this thread. I did try to
post the original artwork before proceeding to etch the board, but
connection problems thwarted me (as usual). I therefore can only hope
that this finished board is acceptable to the cognecenti here. I don't
fancy having to make a third one!

 

[Walter Harley]

[Motorboating] can also happen with pure analog circuits but generally the
designers of these use topologies that are nearly insensitive to this
oscillation mode. Start cascading a couple of common-emitter amplifiers
though and you'll run into it.

ISTR running into that problem with older audio power amp ICs, like the
LM383. Attention to layout and good bypassing was the recommended solution.

 

[Jim Meyer]

Be careful of the Ultracap stuff; most of it is nonsense, including
this particular paper. Their PCB impedance calculators seem to use
grossly simplified equations, and often deliver absurd, even negative,
trace impedance values.

John

The preparatory theory looks good to me. Is it only the
calculator that gives you heartburn? If the theory is all nonsense,
could you point out what you consider the biggest error they make?

Jim

 

[John Larkin]

On Thu, 22 Jan 2004 07:56:14 -0800, John Larkin


Oh bugger. All this contradictory information isn't conducive to
trying to learn this complex subject. Even people who seem to know
what they're talking about frequently disagree with each other on this
group, I've noticed. :-(

Just listen to me and you'll always be OK.

Use Appcad if you need to do trace impedances.

BTW, a scan of the new PCB I've made for that notorious Field-Strength
meter has finally appeared on ABSE. I've tried as best I can to
implement the suggestions I've been given in this thread. I did try to
post the original artwork before proceeding to etch the board, but
connection problems thwarted me (as usual). I therefore can only hope
that this finished board is acceptable to the cognecenti here. I don't
fancy having to make a third one!

The ground pours look nice. On a single-sided board, a 'perfect'
layout is often topologically impossible, so a signal or power jumper
here and there allows the ground plane to have priority.

John

 

[John Larkin]

The preparatory theory looks good to me. Is it only the
calculator that gives you heartburn? If the theory is all nonsense,
could you point out what you consider the biggest error they make?

Jim

They are just assuming that a simple Spice model of paralleled caps
accurately represents a real PC board. In real life, traces, planes,
and parts change things a lot, and especially lower the SRFs and
theoretical Qs. In real life, hardly anybody arranges bypasses in
incremental capacitance steps to evenly spread the SRFs! Traces do
that for you already.

This sort of bypass analysis is common, even by people who should know
better. Spice produces such nice authoritative graphs!

I've done a bit of experimenting on multilayer boards; I sometimes
design in an SMA connector so I can TDR the bare and loaded boards.
The plane capacitance is itself the best high-speed bypass, and is
ignored by most people. The plane-plane capacitor is also fairly
lossy; I've observed no resonant effects using surface mount caps to
bypass planes, and actually no capacitor position sensitivity either.

John

 

[Paul Burridge]

Just listen to me and you'll always be OK.

Oh, right.

The ground pours look nice. On a single-sided board, a 'perfect'
layout is often topologically impossible, so a signal or power jumper
here and there allows the ground plane to have priority.

Thanks. That's what I've done; there are two jumpers on this board to
preserve the GP in one piece. At least that's *one* thing I've guessed
right!