EMC issue with single board computer

[Mike Warren]

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.

Hmm. That didn't come out quite right. This is exactly what I said:

"I tried removing FB41, which the circuit diagram shows as the only
connection between GND_AUD and ground, but I still read 0 ohms
between the 2 grounds. Is it possible that there is connection between
the 2 grounds inside the chip?"

-Mike

 

[PeteS]

Hmm. That didn't come out quite right. This is exactly what I said:

"I tried removing FB41, which the circuit diagram shows as the only
connection between GND_AUD and ground, but I still read 0 ohms
between the 2 grounds. Is it possible that there is connection between
the 2 grounds inside the chip?"

-Mike

Your board vendor probably followed the reference design, which I just
downloaded. They do something which has caused holy wars in this group:
split grounds.

I would put money on grounds being internally connected in the device,
incidentally.

The datasheet is very sparse about power and ground requirements and
doesn't clearly state anything at all with reference to grounds.

FB41 is obviously supposed to couple the grounds, but a bead across the
split grounds will simply isolate the grounds for highspeed signals,
rather than couple them. The reference design shows a short between the
grounds (although it should really be the same plane, imo).

I suggest a zero ohm link across the FB41 position, and because it's a
highspeed problem, perhaps even 0.01uF across the grounds as well.

Cheers

PeteS

 

[Eeyore]

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.

Hi Mike. I didn't receive your mail btw.

Graham

 

[Eeyore]

Hmm. That didn't come out quite right. This is exactly what I said:

"I tried removing FB41, which the circuit diagram shows as the only
connection between GND_AUD and ground, but I still read 0 ohms
between the 2 grounds. Is it possible that there is connection between
the 2 grounds inside the chip?"

Very likely indeed.

Graham

 

[Eeyore]

Your board vendor probably followed the reference design, which I just
downloaded. They do something which has caused holy wars in this group:
split grounds.

I would put money on grounds being internally connected in the device,
incidentally.

The datasheet is very sparse about power and ground requirements and
doesn't clearly state anything at all with reference to grounds.

FB41 is obviously supposed to couple the grounds, but a bead across the
split grounds will simply isolate the grounds for highspeed signals,
rather than couple them. The reference design shows a short between the
grounds (although it should really be the same plane, imo).

I suggest a zero ohm link across the FB41 position, and because it's a
highspeed problem, perhaps even 0.01uF across the grounds as well.

" perhaps even 0.01uF across the grounds as well. " ????????

Graham

 

[PeteS]

" perhaps even 0.01uF across the grounds as well. " ????????

Graham

On some unusual layouts, I have found (as a practical matter) that a
zero ohm link doesn't always give a solid ac path (internal self
inductance perhaps - a lot depends on construction). A small cap seems
to help on some occasions.

Cheers

PeteS

 

[Eeyore]

On some unusual layouts, I have found (as a practical matter) that a
zero ohm link doesn't always give a solid ac path (internal self
inductance perhaps - a lot depends on construction). A small cap seems
to help on some occasions.

A capacitor across a zerohm link ? You're kidding ?

Graham

 

[PeteS]

A capacitor across a zerohm link ? You're kidding ?

Graham

Depends on the grounds - in this case we don't know the shapes,
separation or anything else.

I had a case with a LM73 where it was sitting on top of a ground plane
with high density currents. To make it behave, I decoupled that part of
the ground plane to a different part of the same plane (but separated
by a small void) and *it worked*.

Strange but true.

Cheers

PeteS

 

[Mike Warren]

Hi Mike. I didn't receive your mail btw.

Oh. Was I supposed to remove the "REMOVETHIS." bit?

I did remove it and it didn't bounce.

I'll try again.

-Mike

 

[Eeyore]

Oh. Was I supposed to remove the "REMOVETHIS." bit?

LOL !

I did remove it and it didn't bounce.

I'll try again.

I'll look forward to receiving it. :~)

Graham

 

[Mike Warren]

I'll look forward to receiving it. :~)

I've sent another email. If you don't get this one please send
something to miwa at csas dot net dot au and I'll reply to it.

-Mike

 

[jasen]

Eeyore wrote:

On some unusual layouts, I have found (as a practical matter) that a
zero ohm link doesn't always give a solid ac path (internal self
inductance perhaps - a lot depends on construction). A small cap seems
to help on some occasions.

what happens if you replace the cap with a second link?

 

[Eeyore]

I've sent another email. If you don't get this one please send
something to miwa at csas dot net dot au and I'll reply to it.

Not here yet ! You did delete the dot after REMOVETHIS ?

Graham

 

[Mike Warren]

Not here yet ! You did delete the dot after REMOVETHIS ?

Yes.

It does begin with rabbitsfriendsandrelations, doesn't it?

-Mike

 

[Eeyore]

Yes.

It does begin with rabbitsfriendsandrelations, doesn't it?

It certainly does !! ;~)

Graham

 

[Mike Warren]

It certainly does !! ;~)

And if it finishes with @hotmail.com then both emails were sent to the
right address.

I also sent the emails from 2 different accounts in case hotmail didn't
like one of them. Maybe hotmail doesn't like Australia.

I once had an email from Norway take 6 months to get to me.

-Mike

 

[Eeyore]

And if it finishes with @hotmail.com then both emails were sent to the
right address.

I also sent the emails from 2 different accounts in case hotmail didn't
like one of them. Maybe hotmail doesn't like Australia.

Maybe. If it's not here soon I'll mail you.

I once had an email from Norway take 6 months to get to me.

You beat me there. Just one month in my case.

Graham

 

[John Woodgate]

what happens if you replace the cap with a second link?

It short-circuits the first one, of course.

 

[John Woodgate]

In message
<44e79d21$0$494$61c65585@uq-127creek-reader-03.brisbane.pipenetworks.com.
au>, dated Sat, 19 Aug 2006, Mike Warren

I once had an email from Norway take 6 months to get to me.

You should have de-selected 'hiking mode'.

 

[PeteS]

what happens if you replace the cap with a second link?

The only time I tried (with the wierd issue on the LM73) a pretty hefty
link did not work as well as a cap. As I had a solution and time was
short I did not follow up any further at the time.

That said, I would expect it to ease the situation, of course. My
suggestion of the cap is only because I have seen it help, even where
one might think there should be no effect. Indeed, a zero ohm link may
help the situation sufficiently to cure the problem. If not, it's
something to try that may or may not help and won't take long to test.

Cheers

PeteS