Can placement of frame ground to ground connections cause a resonance?

[KJ]

Can the physical placement of frame ground to ground connections cause
a resonance that would allow EMI to radiate at or around those resonant
frequencies?

I have a rectangular multi-layer board (16"tall x12" wide overall)
which connects mechanically to the metal frame along the two 16" sides.
The top and bottom layers of the board are frame ground planes which
connect directly to the mounting holes along the two sides.

Around the perimeter of a rectangular sub area (15" tall x 8" wide) are
32 .01 uF bypass caps that connect between digital ground planes and
the frame ground planes. I followed all the good EMI design rules and
such and of course things still ended up radiating out the I/O cables
anyway so I clamp on the ferrites to pass certification and move on.
The only problem frequency was the 4th harmonic of the oscillator on
the board.

I got to wondering though if the physical placement of the connections
between digital and frame ground might have been the cause of the
troubles. The straight line path between any two connection points
would seem to define a possible 'half wavelength' that would have nodes
at the connection points. Taking the x,y coordinates of each of the 32
connection points and computing what the offending frequency would be
resulted in a list of 512 frequencies, the lowest seven of which were
below my 4th harmonic.

So did the location of these 32 connection points then create a sort of
bandpass filter that allowed all my 4th harmonic stuff to leak right
out? If so, just what is the mechanism for it to do so?

As far as I can tell, I just might have found an interesting
coincidence for this particular board between 'half wavelengths defined
by my frame to digital ground connections' and the problem frequencies
that showed up in testing and haven't got any theory that would connect
them. I was planning on testing the hypothesis somewhat by adding some
connections between digital and frame ground in a few places in the
'middle' of the rectangular area to see if that snuffed out the problem
at the 4th harmonic since now the longest 'half wavelengths' would be
roughly half the size they are now which would mean I'd be good up to
the 8th harmonic.

If the digital to frame ground connection locations as I've described
them a contributor to the problem, and if so, why? Does anyone have an
explanation for why the current would be doing this? Or was this just
like I said, an interesting coincidence for this particular board?

KJ

 

[Rene Tschaggelar]

Can the physical placement of frame ground to ground connections cause
a resonance that would allow EMI to radiate at or around those resonant
frequencies?
Yes.


I have a rectangular multi-layer board (16"tall x12" wide overall)
which connects mechanically to the metal frame along the two 16" sides.
The top and bottom layers of the board are frame ground planes which
connect directly to the mounting holes along the two sides.

Around the perimeter of a rectangular sub area (15" tall x 8" wide) are
32 .01 uF bypass caps that connect between digital ground planes and
the frame ground planes. I followed all the good EMI design rules and
such and of course things still ended up radiating out the I/O cables
anyway so I clamp on the ferrites to pass certification and move on.
The only problem frequency was the 4th harmonic of the oscillator on
the board.

I got to wondering though if the physical placement of the connections
between digital and frame ground might have been the cause of the
troubles. The straight line path between any two connection points
would seem to define a possible 'half wavelength' that would have nodes
at the connection points. Taking the x,y coordinates of each of the 32
connection points and computing what the offending frequency would be
resulted in a list of 512 frequencies, the lowest seven of which were
below my 4th harmonic.

So did the location of these 32 connection points then create a sort of
bandpass filter that allowed all my 4th harmonic stuff to leak right
out? If so, just what is the mechanism for it to do so?

As far as I can tell, I just might have found an interesting
coincidence for this particular board between 'half wavelengths defined
by my frame to digital ground connections' and the problem frequencies
that showed up in testing and haven't got any theory that would connect
them. I was planning on testing the hypothesis somewhat by adding some
connections between digital and frame ground in a few places in the
'middle' of the rectangular area to see if that snuffed out the problem
at the 4th harmonic since now the longest 'half wavelengths' would be
roughly half the size they are now which would mean I'd be good up to
the 8th harmonic.

If the digital to frame ground connection locations as I've described
them a contributor to the problem, and if so, why? Does anyone have an
explanation for why the current would be doing this? Or was this just
like I said, an interesting coincidence for this particular board?

Much is possible with microwave frequencies. You could also introduce
absorbers, caps of say 1nF with 50 Ohms in series. While it isolates
DC, it poses 50 Ohms at frequencies above say 100MHz.

A standard is also to use metal cases for the pcb.

Rene

 

[Rich Grise]

Can the physical placement of frame ground to ground connections cause
a resonance that would allow EMI to radiate at or around those resonant
frequencies?

I have a rectangular multi-layer board (16"tall x12" wide overall)
which connects mechanically to the metal frame along the two 16" sides.
The top and bottom layers of the board are frame ground planes which
connect directly to the mounting holes along the two sides.

Around the perimeter of a rectangular sub area (15" tall x 8" wide) are
32 .01 uF bypass caps that connect between digital ground planes and
the frame ground planes. I followed all the good EMI design rules and
such and of course things still ended up radiating out the I/O cables
anyway so I clamp on the ferrites to pass certification and move on.
The only problem frequency was the 4th harmonic of the oscillator on
the board.

I got to wondering though if the physical placement of the connections
between digital and frame ground might have been the cause of the
troubles. The straight line path between any two connection points
would seem to define a possible 'half wavelength' that would have nodes
at the connection points. Taking the x,y coordinates of each of the 32
connection points and computing what the offending frequency would be
resulted in a list of 512 frequencies, the lowest seven of which were
below my 4th harmonic.

There should be no current, AC or DC, between digital ground and frame
ground. To get the digital noise out of the frame ground, look at loops.
Where does the power supply current flow, and where can you short the AC
component to tighten the loop?

Good Luck!
Rich

 

[Ken Smith]

Can the physical placement of frame ground to ground connections cause
a resonance that would allow EMI to radiate at or around those resonant
frequencies?

Yes it can. You don't even really need the system to resonate for it to
radiate. All you have to do is cause a large enough AC current to flow in
the case.

I have a rectangular multi-layer board (16"tall x12" wide overall)
which connects mechanically to the metal frame along the two 16" sides.
The top and bottom layers of the board are frame ground planes which
connect directly to the mounting holes along the two sides.

So you've created something like a 12" long folded dipole with the drive
point on top. Do you have metal blocking the 16" edges of this shape?

I'd suggest that you try isolating one 12 inch side of the PCB. If the
radiation decreases greatly, this is the way it is getting out.

The ideal situation is to have a PCB in completel metal box with all
connections to the outside world in one place and the PCB grounded right
at that place. This prevents currents from flowing in the case.

[...]

I got to wondering though if the physical placement of the connections
between digital and frame ground might have been the cause of the
troubles. The straight line path between any two connection points
would seem to define a possible 'half wavelength' that would have nodes
at the connection points. Taking the x,y coordinates of each of the 32
connection points and computing what the offending frequency would be
resulted in a list of 512 frequencies, the lowest seven of which were
below my 4th harmonic.

You never said what the frequency is.

Remember that the length of a wavelength in free space is not equal to its
length in some funny structure with FR4 and sheet metal. The wave length
in the product will be shorter than the wavelength in air.

them. I was planning on testing the hypothesis somewhat by adding some
connections between digital and frame ground in a few places in the
'middle' of the rectangular area to see if that snuffed out the problem

There is another reason to add a middle mounting point. If this thing has
to withstand vibration, you want another mounting point not exactly in the
middle.

 

[KJ]

Ken,

Thanks for your insights, a few follow ups below.

Yes it can. You don't even really need the system to resonate for it to
radiate. All you have to do is cause a large enough AC current to flow in
the case

Understood about the 'large' enough currents. My basic question though
is that given the X,Y locations of each of the tie points do these
points define resonant frequencies that would cause the board to
radiate more at and around those frequencies than at others?

So you've created something like a 12" long folded dipole with the drive
point on top. Do you have metal blocking the 16" edges of this shape?

Not sure what you mean by 'metal blocking'. The metal comes up the two
16" sides with five mounting screws on each side. Across the top and
bottom of the board (i.e. the 12" direction) there is no metal.

I'd suggest that you try isolating one 12 inch side of the PCB. If the
radiation decreases greatly, this is the way it is getting out.

Given the description above I think I already am isolated, or missing
on what you're suggesting.

Remember that the length of a wavelength in free space is not equal to its
length in some funny structure with FR4 and sheet metal. The wave length
in the product will be shorter than the wavelength in air.

Correct, speed of light divided by sqrt(Er) for the FR4 material. As
an example, the (X,Y) coordinates of two of the 32 points are (2.035",
15.515") and (8.675", 0.245") which results in a straight line distance
of 16.65" or 42.3 cm. Ballparking Er as 4.1 then the speed would be
3E10/sqrt(4.1) = 1.48 E10 m/s. Given that, I compute the wavelength
that corresponds to 42.3 cm as being 350 MHz. Since the tie points are
the nodes, it would seem to be that a 350/2 = 175 MHz standing wave
resonance could possibly exist.

Anyway, that was how I did the calculations, and when I sort the list
of 512 such frequencies that resulted from the placement of the 32 tie
points, 170.8 MHz is the lowest frequency and the list extended quite a
ways up. While I didn't have any frequencies that exactly matched the
4th harmonic of the oscillator I'm presuming that this isn't a terribly
discriminating filter and that each of those 512 frequencies really
moosh out into a range that probably overlaps with the next one
somewhat.

My oscillator was 44.2 MHz, the 4th harmonic being 176.8 MHz, the two
nearest possibly resonant frequencies from my calculations are 175.15
and 177.20 MHz. Given that there are approximations and assumptions in
the calculations I wouldn't expect to find a point that exactly
matches. I just found it to be an odd coincidence (and therefore maybe
not actually a coincidence) that the 3rd harmonic at 132.6 did not
cause problems (it's way below 170 MHz) and yet the 4th harmonic for
some reason did. So if this is not just a coincidence, then what is
the mechanism in play here? Why am I getting AC current at these
frequencies in my 'case' (i.e. the frame ground layers on the PCB that
tie to the physical metal frame)?

There is another reason to add a middle mounting point. If this thing has
to withstand vibration, you want another mounting point not exactly in the
middle

Actually the board is rather thick (.125") for mechanical stiffness, it
is in a somewhat vibrating system (i.e. a stationary machine that has
motors for moving stuff around). This board has 11 connectors of
various species all oriented along one of the 16" sides, that need to
be removable for servicing.

The 'middle point' connection that I was talking about also would be a
connection to the top and bottom PCB layers not to the physical metal
frame since the physical metal frame only runs along the two sides.
These two PCB layers spread over the PCB surface and make contact with
the mounting holes and then to the metal frame as well as pressing
against the metal along the two sides.

The ideal situation is to have a PCB in completel metal box with all
connections to the outside world in one place and the PCB grounded right
at that place. This prevents currents from flowing in the case.

The metal box is not all that ideal, it's an assembly problem,
potential airflow problem and generally speaking is more of a means to
contain emissions that have not been properly controlled in the first
place. In any case, the top/bottom frame ground PCB layers that tie to
the physical metal frame where it is available are an attempt to form
such a Faraday cage. It's not quite the perfect cage since there are
obviously cut outs for where the parts exist but since the parts are
small so are those cutouts. All in all though not that much unlike the
commercial cages that one can purchase for such applications.

KJ

 

[KJ]

Rene,

I'm kinda missing how increasing the impedance by adding a fixed series
resistance would help. Seems like that would make the problem worse.

Generally speaking, the 'standard metal case' contains what was not
controlled properly in the first place. My top and bottom PCB layers
being copper should make for a reasonably good Faraday cage for
containment.

I'm trying to understand the mechanism that must exist that is allowing
the higher frequencies AC to possibly be flowing in these layers and
radiated out. At least that's my hypothesis since clamping the ferrite
cores around the I/O cables did the trick and most (but not all) of
these connectors were shielded and tied to frame ground where they
connected to the board. Sniffing around on the board itself didn't
light up anything to indicate that there were any of the obvious things
going on (i.e. signals crossing plane cuts, gross impedance mismatch,
etc). I'm surmising that I'm dumping high frequency AC on to 'frame
ground' and trying to understand why in this case and does it have to
with the relatively large area and the physical distance between 'frame
ground to digital ground' tie points.

KJ

 

[Ken Smith]

Ken,
[....]

Yes it can. You don't even really need the system to resonate for it to
radiate. All you have to do is cause a large enough AC current to flow in
the case

Understood about the 'large' enough currents. My basic question though
is that given the X,Y locations of each of the tie points do these
points define resonant frequencies that would cause the board to
radiate more at and around those frequencies than at others?

Yes, the placement of the connections can effect the radiation. It
increases it at frequencies where the RF signal can find a resonant
circuit with one part exposed to the outside. Consider this end view:


=================================== Sheet metal
! ! <....... Screws
================== <..... PCB


Now look up on the roof at a TV antenna. If you imagine a break in the
middle of the PCB, and some twin lead, you will see how the two structures
look a lot alike. In this case the resonance that would be "found" by the
RF would be more the width of the sheet metal and less the distance
between mounting holes.

Not sure what you mean by 'metal blocking'. The metal comes up the two
16" sides with five mounting screws on each side. Across the top and
bottom of the board (i.e. the 12" direction) there is no metal.

So it is like by ASCII art above and if you look at it edge on, you can
see the hole between the PCB and the chassis without looking through
metal. RF works a lot like light. It can "shine" out this opening. If
you plug the hole with something conductive (eg: sticky copper tape), you
may find the RF doesn't get out as much.

Given the description above I think I already am isolated, or missing
on what you're suggesting.

Doesn't the PCB electrically connect the mounting screws on the two
sides, together? I thought it did. I suggest you have it not do this.

Correct, speed of light divided by sqrt(Er) for the FR4 material.

That isn't the only mode. Consider the gap between the PCB and the sheet
metal. That will have its own speed of light. If will form yet another
bunch of resonances.



[...]

Anyway, that was how I did the calculations, and when I sort the list
of 512 such frequencies that resulted from the placement of the 32 tie
points, 170.8 MHz is the lowest frequency and the list extended quite a
ways up. While I didn't have any frequencies that exactly matched the
4th harmonic of the oscillator I'm presuming that this isn't a terribly
discriminating filter and that each of those 512 frequencies really
moosh out into a range that probably overlaps with the next one
somewhat.

Also remember that the calculation was based on estimates and
simplifications. You have components on the PCB and the actual bit of FR4
vs the ideal one and the inductance of the mounting holes and etc all
trying to make nonsense of any calculations. As such, I'd take any number
that comes out as close as a suspect and try to defeat that resonance to
see if it is the real trouble maker.

[...]

Why am I getting AC current at these
frequencies in my 'case' (i.e. the frame ground layers on the PCB that
tie to the physical metal frame)?

If it was me, I'd just assume that Maxwell hates me.

[....]

The 'middle point' connection that I was talking about also would be a
connection to the top and bottom PCB layers not to the physical metal
frame since the physical metal frame only runs along the two sides.
These two PCB layers spread over the PCB surface and make contact with
the mounting holes and then to the metal frame as well as pressing
against the metal along the two sides.

So there is no sheet metal near the surface of the PCB then?

The metal box is not all that ideal, it's an assembly problem,

Trained ants :>

potential airflow problem and generally speaking is more of a means to
contain emissions that have not been properly controlled in the first
place.

Once you make some RF, it is "emitted". If you can't prevent its creation
you have to stop it somehow. The first line of defence in the layout of
the PCB. You'd like it to trap the RF on the conductors it is supposed to
flow. Adding things that convert the RF to heat can also be helpful.
"Source termination" on fast signals lowers the amount of RF put onto the
trace and thus the amount that can get away.

 

[Fred Bloggs]

My oscillator was 44.2 MHz, the 4th harmonic being 176.8 MHz, the two
nearest possibly resonant frequencies from my calculations are 175.15
and 177.20 MHz. Given that there are approximations and assumptions in
the calculations I wouldn't expect to find a point that exactly
matches. I just found it to be an odd coincidence (and therefore maybe
not actually a coincidence) that the 3rd harmonic at 132.6 did not
cause problems (it's way below 170 MHz) and yet the 4th harmonic for
some reason did. So if this is not just a coincidence, then what is
the mechanism in play here?

It is a coincidence, the 4th harmonic has to 80dB down on even the
signal line. How and where are you measuring this so-called radiation?