Ground plane under crystals

[Richard H.]

Should the can on a crystal be tied to ground, or left floating?

I usually see ground planes under crystals, sometimes with solder mask,
sometimes intentionally bare/tinned. If the can should be grounded, it
seems sloppy to rely on contact instead of a joint, so why the bare pad
instead of solder mask? At <50MHz, would mask really make a difference
in the RF absorption?

Of course, the can is quite noisy on a 'scope, so it'd seem sensible to
ground it. What's the best approach?

As a point of reference: http://www.ecsxtal.com/pdf2/hc-49us.pdf
I see they offer a version with a ground pin. Maybe that's the solution
to my quandry, but I don't see it done commonly.

Thanks,
Richard

 

[Larry Brasfield]

Should the can on a crystal be tied to ground, or left floating?

If you care about close-in phase noise, ground it.
Otherwise, if you will be happy with any frequency
that falls within the xtal tolerance, don't bother.

I usually see ground planes under crystals, sometimes with solder mask,
sometimes intentionally bare/tinned. If the can should be grounded, it
seems sloppy to rely on contact instead of a joint, so why the bare pad
instead of solder mask?

The bare pad does not make much sense when you
think about it. So I hesitate to guess the "why".

At <50MHz, would mask really make a difference in the RF absorption?

It will certainly force non-contact, at least under any
normal vibration. It would slightly reduce capacitance
between the can and ground (assuming the pad is
grounded). This, in turn, will make the capacitance
between the terminals and ground slightly lower and
potentially much more stable. I doubt that loss in the
dielectric will have any significance, at any frequency
that crystals are ordinarily operated.

Of course, the can is quite noisy on a 'scope, so it'd seem sensible to
ground it. What's the best approach?

I would ground it unless it did not matter.

As a point of reference: http://www.ecsxtal.com/pdf2/hc-49us.pdf
I see they offer a version with a ground pin. Maybe that's the solution
to my quandry, but I don't see it done commonly.

It commonly does not matter. Does it matter to you?

 

[legg]

Should the can on a crystal be tied to ground, or left floating?

I usually see ground planes under crystals, sometimes with solder mask,
sometimes intentionally bare/tinned. If the can should be grounded, it
seems sloppy to rely on contact instead of a joint, so why the bare pad
instead of solder mask? At <50MHz, would mask really make a difference
in the RF absorption?

Of course, the can is quite noisy on a 'scope, so it'd seem sensible to
ground it. What's the best approach?

I've seen crystal oscillators stopped (temporarily) or staggered by
noisy switch closure or other randomly pulsed EMI, when the case was
not grounded. I ground them, even if just with a solder blob - this is
easier if the ground plane has no mask.

RL

 

[Phil W]

Soldering it makes for good mechanical stability anyway.

PW

 

[Richard H.]

It commonly does not matter. Does it matter to you?

I don't know yet. I strive to ask "why", rather than ignoring a common
practice, or copying it blindly.

It would seem to have benefits either way, but casual contact seems very
haphazard and soldering the can doesn't seem right. I suppose I'll get
the flavor with the ground pin and remove all doubt about whether it is
/ should be grounded.

Thanks to all for the comments!

 

[Dave VanHorn]

The reason that I put it there, is to form a shield for the crystal leads
which run back to the processor. On my boards, you'll see that this ground
under the crystal is an isolated finger, and touches nothing except the
crystal caps, and the nearest uP ground pin. From there, it joins system
ground. I normally don't connect the cans to the plane, in fact I use
insulators under the cans to make sure nothing makes any unapproved
connections.

 

[Nicholas O. Lindan]

I see they offer a version with a ground pin.

Grounding the can is a generally good idea -- as it is
a bad idea to have a large metallic object that is not
tied to a known voltage.

The extra pin also serves to hold the crystal in place.
An unsupported through-hole crystal will otherwise put a
mechanical cantilevered load on it's leads when the board
is subject to vibration/getting dropped.

A bare-wire strap around the crystal is often used in
lieu of the third pin.

I don't know of any reason for a ground plane under the
crystal. TTBOMK it is a flourish added by the layout guy.
It also saves on etchant.

 

[Meindert Sprang]

It would seem to have benefits either way, but casual contact seems very
haphazard and soldering the can doesn't seem right. I suppose I'll get
the flavor with the ground pin and remove all doubt about whether it is
/ should be grounded.

I have never done it and recently, one of my boards was tested for IEC60945
(very stringent EMC test). It came out right.
Millions of receivers/transmitters have been produced in the past were
crystals were plugged in sockets. No ground connection. Many of the boards I
see have no ground connection. The only ones that have, are the ones where
the crystals are mounted horizontally. Here it is just done to prevent them
from moving under high G loads.
Oh, and SMD crystals do not even have the possibility for a ground
connection.

And based on the nature of a crystal (low frequency compared to it's
simensions and high Q) I'd like to thing they don't radiate at all.

Meindert

 

[Richard H.]

A bare-wire strap around the crystal is often used in
lieu of the third pin.

Indeed - I'd forgotten about that, and a couple boards I've just grabbed
from the junkpile also tie this strap to ground.

The particular crystals I'm using are low-profile cans, so they install
vertically instead of bending the leads. But the strap-around-the-can
seems to endorse the practice of grounding the can.

Thanks!

 

[Richard H.]

The reason that I put it there, is to form a shield
for the crystal leads which run back to the
processor.

Interesting. I presume you've got some pics on your site, so I'll go
check this out. Thanks!

I normally don't connect the cans to the plane, in
fact I use insulators under the cans to make sure
nothing makes any unapproved connections.

OK, I'll bite - why do you want the can floating instead of grounded?
(What insulators do you use? And in addition to the solder mask, or
only on bare boards?)

I'd think grounding might help EMI emissions, but it seems you're
concerned about causing circuit problems? Perhaps when the "noise"
being grounded is 0 or 180 degrees to the crystal signal? (Speculating
here... my knowledge of crystals and harmonics is limited.)

Thanks!

 

[Dave VanHorn]

OK, I'll bite - why do you want the can floating instead of grounded?
(What insulators do you use? And in addition to the solder mask, or
only on bare boards?)

I don't know where we get them, but they are little sheet insulators with
holes for the leads.
I don't want the osc pads shorting to the can. That would make for a lot of
antenna, I'd think.
I'm not worried about the can itself. The signals on the two leads are
nearly the same amplitude, and 180 degrees apart, so in such a small area,
they should cancel nicely.

 

[Richard H.]

Millions of receivers/transmitters have been
produced in the past were crystals were plugged in
sockets. No ground connection. Many of the boards I
see have no ground connection. The only ones that
have, are the ones where the crystals are mounted
horizontally. Here it is just done to prevent them
from moving under high G loads.
Oh, and SMD crystals do not even have the
possibility for a ground connection.

All excellent points. Evidence suggests that bare pads under crystals
and/or grounding of restraining straps may be overkill, precautionary,
or maybe just superstition. And some folks go out of their way to avoid
it. Very curious.

And based on the nature of a crystal (low frequency
compared to it's simensions and high Q) I'd like to
thing they don't radiate at all.

In the end, this is probably the key - whether these emissions are even
significant, to the circuit itself or for EMI. And with SMD crystals,
there's really no question about how it should be done.

Thanks!

 

[Geo]

I don't know of any reason for a ground plane under the
crystal. TTBOMK it is a flourish added by the layout guy.

If the crystal is mounted horizontally then I place copper on the top layer to
stop me (or the autorouter) using the space for anything else.

Geo

 

[Richard H.]

I don't want the osc pads shorting to the can. That
would make for a lot of antenna, I'd think.

That's actually what led me to start wondering here, after a recent
batch of boards that didn't have solder mask under the can.

With solder mask, are the extra insulators really needed, or do you work
much with bare boards? (i.e., concerned that vibration would damage the
mask underneath?)

I'm not worried about the can itself. The signals
on the two leads are nearly the same amplitude,
and 180 degrees apart, so in such a small area,
they should cancel nicely.
[...]

Noise?

Again, maybe a moot issue, or a bad example I've run across... on a
recent board, I happened to probe the can with my scope and noticed a
lot of signal on it. Apparently it's not strong enough to be an EMI
issue, but it got me to thinking about grounding the can.


BTW, I didn't find any images on your site that showed your approach to
crystal ground pads. Can you provide a link?

Thanks!

 

[Spehro Pefhany]

That's actually what led me to start wondering here, after a recent
batch of boards that didn't have solder mask under the can.

With solder mask, are the extra insulators really needed, or do you work
much with bare boards? (i.e., concerned that vibration would damage the
mask underneath?)

For typical low-profile through-hole crystals, I use a keepout larger
than the crystal can in the footprint so that no copper ends up under
the crystal itself (on the top layer) other than the pads. I like to
see a ground plane under the crystal on some other layer. The mask
would "probably" be okay if it visually looks okay with no copper
edges showing through on the sides of traces, and if vibration is that
bad, the crystal leads will probably fail first.

If there are no stand-offs under the crystal, it's possible for solder
to come up through the holes and form a short to the crystal can. It
probably would be a good idea to keep the pad size minimum on the top
layer consistent with the design rules, and larger on the bottom if
desired.


Best regards,
Spehro Pefhany

 

[Richard H.]

If there are no stand-offs under the crystal, it's
possible for solder to come up through the holes
and form a short to the crystal can.

Hmmm. Good point, above & beyond all the discussion about grounding.

It probably would be a good idea to keep the pad
size minimum on the top layer consistent with the
design rules, and larger on the bottom if desired.

I've wanted this for other reasons, and it seems to be the one thing I'd
like that Eagle doesn't do yet... for some reason, it makes THT pads the
same size on top and bottom.

Thanks!

 

[Dave VanHorn]

With solder mask, are the extra insulators really needed, or do you work
much with bare boards? (i.e., concerned that vibration would damage the
mask underneath?)

Depends on the pad geometry, and the way the bottom of the crystal is
shaped.
The insulators make sure it will never be a problem.

Again, maybe a moot issue, or a bad example I've run across... on a
recent board, I happened to probe the can with my scope and noticed a
lot of signal on it. Apparently it's not strong enough to be an EMI
issue, but it got me to thinking about grounding the can.

I've used that point to check oscillator function.
In essence, you're capacitively coupled to the oscillator, on both sides, so
whichever side is stronger (output pin) dominates. It's probably best to
ground the can, but then you need a manual solder blob, or a three-pin
crystal with can ground pin. I haven't seen it be a problem.

BTW, I didn't find any images on your site that showed your approach to
crystal ground pads. Can you provide a link?

I don't have a convenient way to get there from my cad software.