EMC issue with single board computer

[Mike Warren]

Is that a tradename ? Over here in the UK we used 'Zintec' from
British Steel but the Asians don't seem to have any direct
equivalent. Either that or the purchasing guys don't know about it.
The phosphating can be an utter PITA.

http://www.bluescopesteel.com.au/go/product/zincanneal-g2s-steel

I can easily connect my DMM to the case and get 0 ohms. No scratching
the surface is required.

-Mike

 

[Eeyore]

http://www.bluescopesteel.com.au/go/product/zincanneal-g2s-steel

I can easily connect my DMM to the case and get 0 ohms. No scratching
the surface is required.

That's the good stuff then !

Graham

 

[PeteS]

That's the good stuff then !

Graham

Something I didn't ask earlier; do the AC97 signals come *off* the
board? If so, you'll have stubs and they will be an absolute b*tch to
deal with.

If they *do* come off the board, put a terminator on the line to ground
(try 47 in series to a 1k paralleled with 100pF to ground to start
with. You might get away with something simpler). I had precisely this
issue when doing some early prototyping with a premade board.

Cheers

PeteS

 

[Mike Warren]

Something I didn't ask earlier; do the AC97 signals come *off* the
board? If so, you'll have stubs and they will be an absolute b*tch to
deal with.

Thanks Pete. Unfortunately the signals do not go to any headers.
I've pretty much resigned myself to stopping the emissions at
each wire. Very slow work and there will be unpleasant side effects
to deal with. :-(

-Mike

 

[Michael A. Terrell]

Thanks Pete. Unfortunately the signals do not go to any headers.
I've pretty much resigned myself to stopping the emissions at
each wire. Very slow work and there will be unpleasant side effects
to deal with. :-(

-Mike

What does the OEM say about your problems? You may not be the only
one having trouble with their board. They may have to redesign them to
make them pass the required tests if they expect to sell a lot of
boards.


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

 

[Eeyore]

What does the OEM say about your problems? You may not be the only
one having trouble with their board. They may have to redesign them to
make them pass the required tests if they expect to sell a lot of
boards.

Good advice. They may even know of a fix.

Graham

 

[Terry Given]

What does the OEM say about your problems? You may not be the only
one having trouble with their board. They may have to redesign them to
make them pass the required tests if they expect to sell a lot of
boards.

why dont you look closely at the offending board, figure out how its
emitting the mess, and fix it? slugging oscillator outputs with caps to
reduce slew rate is brutal, but effective. Of course if its a crap
layout there wont be much you can do, but another rough trick I have
used is to make up a "ground plane" pcb - 0.8mm FR4, single sided, solid
copper. put holes in it for all the component legs to poke thru, place
it on the bottom of the PCB (Cu facing outwards), and solder all the 0V
connections to it.....

a good way to tell a crap layout from a good one is to hold the pcb up
to the light. If you can see through it, its probably crap. Although
just because you cant see through it, doesnt mean it isnt crap (I've
seen numerous atrocious 4,6,8,10 and 12-layer PCB layouts).

Cheers
Terry

 

[Mike Warren]

What does the OEM say about your problems? You may not be the only
one having trouble with their board. They may have to redesign them
to make them pass the required tests if they expect to sell a lot of
boards.

Their response was to send me their CE and FCC test reports showing
it had passed.

-Mike

 

[Mike Warren]

why dont you look closely at the offending board, figure out how its
emitting the mess, and fix it? slugging oscillator outputs with caps
to reduce slew rate is brutal, but effective. Of course if its a crap
layout there wont be much you can do, but another rough trick I have
used is to make up a "ground plane" pcb - 0.8mm FR4, single sided,
solid copper. put holes in it for all the component legs to poke
thru, place it on the bottom of the PCB (Cu facing outwards), and
solder all the 0V connections to it.....

a good way to tell a crap layout from a good one is to hold the pcb up
to the light. If you can see through it, its probably crap. Although
just because you cant see through it, doesnt mean it isnt crap (I've
seen numerous atrocious 4,6,8,10 and 12-layer PCB layouts).

There are very few places to attach a shield since the PCB is so
dense. Anyway, I need a way that is practical for production. We
won't be buying a high enough quantity to make the manufacturer
redesign.

I'm currently waiting on a board from another supplier to see if
it will do the job.

I can get the emissions down a few dB by doing multiple mods
to the board and it would probably pass but these are not practical
for production.

-Mike

 

[Eeyore]

Their response was to send me their CE and FCC test reports showing
it had passed.

In what configuration ?

Graham

 

[Eeyore]

There are very few places to attach a shield since the PCB is so
dense.

It's not about a 'shield' !

Anyway, I need a way that is practical for production. We
won't be buying a high enough quantity to make the manufacturer
redesign.

I'm currently waiting on a board from another supplier to see if
it will do the job.

I can get the emissions down a few dB by doing multiple mods
to the board and it would probably pass but these are not practical
for production.

I suspect your background is primarily audio ??? Like myself. I can see where
you possibly haven't grasped the subtleties of EMC. I was lucky to have an tutor
early on back in ~ 1987.

There is truly a very different approach required to EMC compared to classic
audio. I've heard comments like some of your own only too many times.

The thing is, if you get the concept right, it's actually not *that* difficult
to tame. I'm concerened you may do another iteration and find it no better.

Fancy offering me a paid vacation ?

Graham

 

[Mike Warren]

I suspect your background is primarily audio ???
Yes.

There is truly a very different approach required to EMC compared to
classic audio. I've heard comments like some of your own only too
many times.

Are you talking about the shield or ground plane needing to be bonded to
ground everywhere, much like a ground plane on a PCB? If so, I can't do
that here as there is very little ground available to solder to. I have
actually
done what Terry suggested a couple of days ago but could only get 3
connections to ground. Of course, it made very little difference.

The thing is, if you get the concept right, it's actually not *that*
difficult to tame. I'm concerned you may do another iteration and
find it no better.

I'm concerned about that too. :-(

Fancy offering me a paid vacation ?

I would. Just have to convince the powers that be. ;-)

How about some private tuition over phone/email? If you think
that's viable, email me with your rate.

-Mike

 

[Eeyore]

Are you talking about the shield or ground plane needing to be bonded to
ground everywhere, much like a ground plane on a PCB?

At least in part but not exclusively.

If so, I can't do
that here as there is very little ground available to solder to.

A common situation !

I have actually done what Terry suggested

Terry's good !

a couple of days ago but could only get 3
connections to ground. Of course, it made very little difference.


I'm concerned about that too. :-(


I would. Just have to convince the powers that be. ;-)

I'll bet !

How about some private tuition over phone/email? If you think
that's viable, email me with your rate.

I'll certainly follow up on this. I'm intruiged by what you're doing as it looks
like it may overlap with something I have in mind too.

The email address in my headers works btw !

How many hours ahead of London are you ? What format are your schematics / pcb
files in too ?

Graham

 

[Mike Warren]

The email address in my headers works btw !

I have just sent an email to the hotmail address.

-Mike

 

[Terry Given]

Mike Warren wrote:




At least in part but not exclusively.





A common situation !





Terry's good !

whats the emoticon for blushing?

I'll bet !





I'll certainly follow up on this. I'm intruiged by what you're doing as it looks
like it may overlap with something I have in mind too.

The email address in my headers works btw !

How many hours ahead of London are you ? What format are your schematics / pcb
files in too ?

Graham

Cheers
Terry

 

[PeteS]

whats the emoticon for blushing?


Cheers
Terry

One thing for you to check before going too much further - I have had
some very noisy devices that have relatively highspeed clocks and D-A
and A-D converters onboard.

Look carefully around the AC97 device and check the decoupling for it's
analog side. I've had excessive noise from system clocks 'leak' through
the converters into the analog system, and then it has a really nice
big antenna, exacerbating the problem.

If you have a schematic for the board, check what decoupling they used
for the analog side - a common problem is not to provide a separate
power rail into the audio section (I derive mine from the primary rail
with a ferrite bead and fairly hefty decoupling).
Failure to provide a separately decoupled supply for the analog section
_could_ lead to this issue.
If the noise is getting in by way of the analog power, you might have a
chance to suppress it near the device.
Even if there's separate decoupling, providing a cap _across_ from one
power domain to the other can suppress a lot of radiated hash.

Cheers

PeteS

 

[Mike Warren]

One thing for you to check before going too much further - I have had
some very noisy devices that have relatively highspeed clocks and D-A
and A-D converters onboard.

Look carefully around the AC97 device and check the decoupling for
it's analog side. I've had excessive noise from system clocks 'leak'
through the converters into the analog system, and then it has a
really nice big antenna, exacerbating the problem.

If you have a schematic for the board, check what decoupling they used
for the analog side - a common problem is not to provide a separate
power rail into the audio section (I derive mine from the primary rail
with a ferrite bead and fairly hefty decoupling).
Failure to provide a separately decoupled supply for the analog
section _could_ lead to this issue.

The analog supply has 2 10uF ceramics and 2 100nF ceramics on it. The
digital supplies have one 10uF and 2 100nFs. All are fed through
ferrite beads.

If the noise is getting in by way of the analog power, you might have
a chance to suppress it near the device.
Even if there's separate decoupling, providing a cap across from one
power domain to the other can suppress a lot of radiated hash.

I'll give that a try.

-Mike

 

[Michael A. Terrell]

Their response was to send me their CE and FCC test reports showing
it had passed.

-Mike

That attitude would make me look for another supplier.


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

 

[Eeyore]

That attitude would make me look for another supplier.

They may not have fully understood the question / reason for it and might
actually have thought they were being helpful !

Especially likely if both parties don't share a common first language.

Maybe Boki can act as interpreter ? It's time he contributed.

Graham

 

[Mike Warren]

They may not have fully understood the question / reason for it and
might actually have thought they were being helpful !

Especially likely if both parties don't share a common first language.

That's quite possible. I asked where the connection was between AGnd
and Gnd when ferrite bead FB41 was removed and they answered that it
would be OK for me to sort out the ferrite bead if I wanted.

-Mike