I'm looking at some PA amp designs (300-500W).

Amp layouts often use or specify twisted power wiring to
partly cancel magnetic fields. Would a co-axial power cable
(if available, otherwise, two pairs of wire twisted together)
improve on this, since a twist has a limited pitch?

Another related question. In high-power amps, a layout
technique referred to as "distortion cancelling" is being
used. Again, this aims to reduce the coupling of magnetic
fields from high-current PCB traces back into the input
stages. I understand the principle, but what's the actual
technique - what geometries are effective? Is it just a
matter of minimising the loop area, and if so, could the
loop traces be run on opposite sides of the PCB to do that?
Can you point me to any tutorial-type resources on this
type of layout?

Clifford Heath.

Clifford Heath wrote:

Amp layouts often use or specify twisted power wiring to
partly cancel magnetic fields. Would a co-axial power cable
(if available, otherwise, two pairs of wire twisted together)
improve on this, since a twist has a limited pitch?

No. Electrostatic shields don't do anything against magnetic fields,
which are the main problem at those distances and frequencies. Twisting
the conductors is very effective (and dates back at least to the era
of filament wiring).

Another related question. In high-power amps, a layout
technique referred to as "distortion cancelling" is being
used. Again, this aims to reduce the coupling of magnetic
fields from high-current PCB traces back into the input
stages. I understand the principle, but what's the actual
technique - what geometries are effective? Is it just a
matter of minimising the loop area, and if so, could the
loop traces be run on opposite sides of the PCB to do that?
Can you point me to any tutorial-type resources on this
type of layout?

That just sounds like reasonable PC board layout precautions more
than anything else. Doug Self's series of articles on amplifier
design in Wireless World touches on it, but it's not one of the
more significant distortion sources. It might be in an amp with
transformer coupling, though.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Scott Dorsey wrote:
No. Electrostatic shields don't do anything against magnetic fields,

I wasn't thinking of electrostatics, but magnetics. A coaxial
cable should have almost zero magnetic field in any direction
and scale, even less than twisted pair. Shouldn't it?
Or is that true, but the effect is too small to matter?
The way I figure it, if you have a 1" twist pitch, then
any component within 1/2" which is sensitive to pickup will
see significant magnetic field despite the twist, which shouldn't
be the case with co-ax. Obviously it's best to keep such high
current wiring away from such components, but anyhow...

That just sounds like reasonable PC board layout precautions more
than anything else. Doug Self's series of articles on amplifier
design in Wireless World touches on it, but it's not one of the
more significant distortion sources. It might be in an amp with
transformer coupling, though.

Thanks for that. I'm thinking of the more recent Silicon Chip
magazine amps, which claim much lower distortion than earlier
amps (like ETI466, PM300, etc) due to better PCB layout. I don't
have the text of the mag articles where they argue this, and in
any case I recall it was a discussion on principle, not "this
track is laid here rather than there for this reason".

Clifford Heath.

"Clifford Heath"

Another related question. In high-power amps, a layout
technique referred to as "distortion cancelling" is being
used. Again, this aims to reduce the coupling of magnetic
fields from high-current PCB traces back into the input
stages. I understand the principle, but what's the actual
technique - what geometries are effective? Is it just a
matter of minimising the loop area, and if so, could the
loop traces be run on opposite sides of the PCB to do that?


** Any loop areas associated with the input signal, input ground and
feedback ground need to be kept small as possible.

The problem is with class B output stages where the DC supply current to
the output devices very distorted ( half sine shape) and harmonics can
become injected in the loops mentioned - which ruins the THD specs
particularly at the higher audio frequencies.

Just as undesirable is the case of a stereo amp where the same distorted
current injection transfers from one channel to the other - producing
distorted crosstalk.

Close up those loops and keep the high current DC wiring well away from
them and you can beat it.





........... Phil

On Wed, 25 Aug 2004 12:32:20 +1000, Clifford Heath
wrote:

Scott Dorsey wrote:

...

That just sounds like reasonable PC board layout precautions more
than anything else. Doug Self's series of articles on amplifier
design in Wireless World touches on it, but it's not one of the
more significant distortion sources. It might be in an amp with
transformer coupling, though.

Thanks for that. I'm thinking of the more recent Silicon Chip
magazine amps, which claim much lower distortion than earlier
amps (like ETI466, PM300, etc) due to better PCB layout. I don't
have the text of the mag articles where they argue this, and in
any case I recall it was a discussion on principle, not "this
track is laid here rather than there for this reason".

There's the 'principle" that PCB traces have non-zero resistance.
Google for Douglas Self's website. ISTR that's one of his eight things
that cause distortion is bad layout, specifically where the feedback
point is placed - he has schematics where he says you run the feedback
path from 'here' (the speaker terminal) rather than 'here' (the
junction of the output transistors), even though 'electrically'
they're the same, but on a practical PCB and chassis, they're not.
He'll want to sell you his book, too, which is likely to be
interesting as well.

Clifford Heath.

-----
http://mindspring.com/~benbradley

Ben Bradley wrote:
On Wed, 25 Aug 2004 12:32:20 +1000, Clifford Heath
wrote:


Scott Dorsey wrote:


...


That just sounds like reasonable PC board layout precautions more
than anything else. Doug Self's series of articles on amplifier
design in Wireless World touches on it, but it's not one of the
more significant distortion sources. It might be in an amp with
transformer coupling, though.

Thanks for that. I'm thinking of the more recent Silicon Chip
magazine amps, which claim much lower distortion than earlier
amps (like ETI466, PM300, etc) due to better PCB layout. I don't
have the text of the mag articles where they argue this, and in
any case I recall it was a discussion on principle, not "this
track is laid here rather than there for this reason".


There's the 'principle" that PCB traces have non-zero resistance.
Google for Douglas Self's website. ISTR that's one of his eight things
that cause distortion is bad layout, specifically where the feedback
point is placed - he has schematics where he says you run the feedback
path from 'here' (the speaker terminal) rather than 'here' (the
junction of the output transistors), even though 'electrically'
they're the same, but on a practical PCB and chassis, they're not.

The flip side might be stability.

He'll want to sell you his book, too, which is likely to be
interesting as well.


Clifford Heath.


-----
http://mindspring.com/~benbradley


--
The e-mail address in our reply-to line is reversed in an attempt to
minimize spam. Our true address is of the form .

Phil Allison wrote:
** Any loop areas associated with the input signal, input ground and
feedback ground need to be kept small as possible.

Thanks Phil, I thought it was that simple.

On the other hand, ISTR that SC played with the shape and position of
the traces around the OP transistors to reduce or cancel the magnetic
fields, perhaps by running the collector and emittor traces equidistant
from a ground trace so that current outbound to any OP transistor and
current inbound each cause magnetic fields due to their ground-loop
shape, and the two fields cancel so neither reaches the input stage in
any quantity.

The principle is sound, but the recent (Jan 04) design doesn't use the
obvious method (outgoing and incoming traces on the same side above a
ground plane), and I wondered if anyone could explain the rationale for
what they *did* do. Or if my proposal is correct either...
Maybe SC just wanted to avoid a double-sided PCB?

I think I just explained to myself what it was I was originally asking
about. Hope it makes sense to someone else also.

Thanks to Ben also. I had already come across the concept of carefully
placing the feedback takeoff point, it's not what I had in mind, though
obviously important.

I feel more confident about proceeding with my high-power bass guitar
amp design now.

Clifford Heath.

"Clifford Heath"
Phil Allison wrote:
** Any loop areas associated with the input signal, input ground and
feedback ground need to be kept small as possible.

Thanks Phil, I thought it was that simple.



** The idea is simple - but maybe not so simple to achieve when you have
to build a low THD amp.

The PCB pattern and cabling layouts can go through many iterations before
the desired specs are reliably met by a product - the proof is a good
looking ( mainly 2H and 3 H ) residual on a scope after filtering out the
fundamental at say 6 kHz - and that displacing the various input and high
current cables slightly does not harm this.

With a stereo amp one can just look at the HF crosstalk on a scope for any
signs of harmonics - they have the same cause.




........... Phil

Ben Bradley wrote:

On Wed, 25 Aug 2004 12:32:20 +1000, Clifford Heath
wrote:

Scott Dorsey wrote:

That just sounds like reasonable PC board layout precautions more
than anything else. Doug Self's series of articles on amplifier
design in Wireless World touches on it, but it's not one of the
more significant distortion sources. It might be in an amp with
transformer coupling, though.

Thanks for that. I'm thinking of the more recent Silicon Chip
magazine amps, which claim much lower distortion than earlier
amps (like ETI466, PM300, etc) due to better PCB layout. I don't
have the text of the mag articles where they argue this, and in
any case I recall it was a discussion on principle, not "this
track is laid here rather than there for this reason".

There's the 'principle" that PCB traces have non-zero resistance.

Very much so !

I find it easier to consider traces as being resistors. That way, the
sources of error are easy to spot.


Google for Douglas Self's website. ISTR that's one of his eight things
that cause distortion is bad layout, specifically where the feedback
point is placed - he has schematics where he says you run the feedback
path from 'here' (the speaker terminal) rather than 'here' (the
junction of the output transistors), even though 'electrically'
they're the same, but on a practical PCB and chassis, they're not.

The example of the positioning of the feedback node is an excellent one. I
recently had occasion to ask a pcb layout guy to move this point for
precisely the reasons you state.

Good layout also defines and constrains current paths - especially
important with high currents.


Graham

"Clifford Heath" wrote in message


I'm looking at some PA amp designs (300-500W).

Amp layouts often use or specify twisted power wiring to
partly cancel magnetic fields. Would a co-axial power cable
(if available, otherwise, two pairs of wire twisted together)
improve on this, since a twist has a limited pitch?

The *signal* on a power cable can be thought of as being balanced.
Therefore, the use of twisted pair is indicated.

Another related question. In high-power amps, a layout
technique referred to as "distortion cancelling" is being
used.

AFAIK, the most obvious example of this principle was the input wiring on
the original DH-200 Mosfet power amp, where routing this lead in accordance
with the directions reduced THD @20 KHz by at least 50%.

Again, this aims to reduce the coupling of magnetic
fields from high-current PCB traces back into the input
stages.

In the case I'm thinking of, the coupling was manipulated seemed to be
electrostatic coupling. I think that either kind of coupling could be either
controlled or avoided to minimize distortion.

I understand the principle, but what's the actual
technique - what geometries are effective?

I can tell you how to minmize coupling, but minimizing coupling need not be
the best approach. If some specific kind of coupling is required, then the
means used are going to be specific to the particular design.

Is it just a matter of minimising the loop area, and if so, could the
loop traces be run on opposite sides of the PCB to do that?

That could be a means that would be used to minimize coupling. However,
small circuit traces on the same side of the board can be closer together
than the thickness of the board.

Can you point me to any tutorial-type resources on this
type of layout?

http://www.zilog.com/docs/z8/appnote...mizing_emi.pdf

http://www.lcr-inc.com/whitepapers/E...c_Coupling.pdf

http://www.chrontel.com/pdf/an36.pdf

http://www.rfm.com/products/apnotes/an42.pdf

http://www.maxim-ic.com/appnotes.cfm/appnote_number/735

http://www.elmac.co.uk/c_e_layt.htm

Clifford Heath wrote:
Scott Dorsey wrote:
No. Electrostatic shields don't do anything against magnetic fields,

I wasn't thinking of electrostatics, but magnetics. A coaxial
cable should have almost zero magnetic field in any direction
and scale, even less than twisted pair. Shouldn't it?

No. Why would putting an electrostatic shield around a wire do
anything about the magnetic leakage issues? Remember, you are talking
about something that isn't operating as a transmission line, so all
of the constant impedance stuff is basically out the window. It's just
a wire with an electrostatic shield since you're way below a full wave
length.

Thanks for that. I'm thinking of the more recent Silicon Chip
magazine amps, which claim much lower distortion than earlier
amps (like ETI466, PM300, etc) due to better PCB layout. I don't
have the text of the mag articles where they argue this, and in
any case I recall it was a discussion on principle, not "this
track is laid here rather than there for this reason".

What does that mean? Lower THD doesn't really matter much by itself.
You care more about the kind of distortion than the amount of distortion
most of the time these days.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

In rec.audio.pro Scott Dorsey wrote:

: No. Why would putting an electrostatic shield around a wire do
: anything about the magnetic leakage issues? Remember, you are talking
: about something that isn't operating as a transmission line, so all
: of the constant impedance stuff is basically out the window. It's just
: a wire with an electrostatic shield since you're way below a full wave
: length.

If the shield is the return path for the current in the wire, the magnetic
fields of the wire and shield exactly cancel outside of the shield; all of
the B field is completely enclosed by the shield -- assuming perfectly
centered wire in a perfectly cylindrical, uniformly conducting shield...

Twisted pair is usually used because it is almost as good for most situations
and a lot cheaper.

Bob Miller

wrote:
If the shield is the return path for the current in the wire, the magnetic
fields of the wire and shield exactly cancel outside of the shield;

Thanks for the confirmation Bob - this is exactly what I was thinking
about. Glad someone got it, and I'm not yet completely out of touch :-).

Twisted pair is usually used because it is almost as good for most
situations and a lot cheaper.

Twisted four (two hots and two grounds in a square x-section with each
pair diagonally opposite) should be even better than twisted pair
(nearly complete cancellation because the AB on this side is cancelled
by the BA on the other) but still at a fairly low cost... no?

BTW, are you the same guy that did atomic clocks for HP?
I remember admiring posts about that a few years back.

Clifford Heath.

PS: Thanks Arny, I'm checking out the DH-200 now.

Clifford Heath wrote:

wrote:
If the shield is the return path for the current in the wire, the magnetic
fields of the wire and shield exactly cancel outside of the shield;

Thanks for the confirmation Bob - this is exactly what I was thinking
about. Glad someone got it, and I'm not yet completely out of touch :-).

Twisted pair is usually used because it is almost as good for most
situations and a lot cheaper.

Twisted four (two hots and two grounds in a square x-section with each
pair diagonally opposite) should be even better than twisted pair
(nearly complete cancellation because the AB on this side is cancelled
by the BA on the other) but still at a fairly low cost... no?

That sounds similar to 'starquad' cable used typically for microphones operating
in noisy environments.

Graham

In rec.audio.pro Clifford Heath wrote:

: Twisted four (two hots and two grounds in a square x-section with each
: pair diagonally opposite) should be even better than twisted pair
: (nearly complete cancellation because the AB on this side is cancelled
: by the BA on the other) but still at a fairly low cost... no?

Yes. But all of this starts to run afoul of electrical codes for
flexible cables. You have to have a ground in there, usually. Cable
which is 4-wire plus ground can be had (for a price -- it gets used
for some "wye" three-phase power applications), but the fifth wire will
distort the "star-quad" arrangement. It also may be difficult to safely
wire two conductors into each connection point on the connectors.

: BTW, are you the same guy that did atomic clocks for HP?
: I remember admiring posts about that a few years back.

Naah. I do 6.25GHz serial HW today, have done graphics engines and
IEEE floating-point math hardware in the past...

Glad you recognized the "Bill and Dave" association in my (munged) email
address ;-)

Bob Miller
Agilent Technologies