Repost ---> PCBs and Moisture?

[EdV]

sorry to repost but I thought mine may have got lost in the sea of
spam:

We got a batch of boards built and two of them show a curious
tendency(consitent) to shut down the crystal oscillator at -20 C. We
also notice that the capacitance of the xtal caps changes more on the
"bad" PCBs at -20C than the "good" boards.

1. Moisture in PCB Layers that didn't get baked out?
2. Uncured epoxy betweem layers?
3. Residual flux?


#3 I could address with vapor degreasing or other cleaning. #1 and 2
would mean baking at 100C or so which I guess is ok to see if it has
an effect.


Any thoughts, insights or expereinces as always are welcome.


Ed V.

 

[PN2222A]

We got a batch of boards built and two of them show a curious
tendency(consitent) to shut down the crystal oscillator at -20 C. We
also notice that the capacitance of the xtal caps changes more on the
"bad" PCBs at -20C than the "good" boards.

1. Moisture in PCB Layers that didn't get baked out?
2. Uncured epoxy betweem layers?
3. Residual flux?

Why do you suspect the PCB?
First, rule out failures due to the capacitors.
If the caps used by your vendor have a different tempco than you
specified,
then they may cause the oscillator to stop.
Measure the caps individually, outside of the PCB / oscillator
environment.

If that doesn't answer your question, measure the characteristics of the
crystal over temperature (good ones and bad ones).
Also determine the gain of your amplifier stage over temperature.

regards
PN2222A

NPN (Is = 14.34f Xti = 3 Eg = 1.11 Vaf = 74.03 Bf = 255.9 Ne = 1.307 Ise
= 14.34 Ikf = .2847 Xtb = 1.5 Br = 6.092 Isc = 0

Ikr = 0 Rc = 1 Cjc = 7.306p Mjc = .3416 Vjc = .75 Fc = .5 Cje = 22.01p
Mje = .377 Vje = .75 Tr = 46.91n Tf = 411.1p Itf = .6

Vtf = 1.7 Xtf = 3 Rb = 10)

 

[EdV]

Why do you suspect the PCB?
First, rule out failures due to the capacitors.
    If the caps used by your vendor have a different tempco than you
specified,
    then they may cause the oscillator to stop.
    Measure the caps individually, outside of the PCB / oscillator
environment.

If that doesn't answer your question, measure the characteristics of the
crystal over temperature (good ones and bad ones).
Also determine the gain of your amplifier stage over temperature.

regards
PN2222A

NPN (Is = 14.34f Xti = 3 Eg = 1.11 Vaf = 74.03 Bf = 255.9 Ne =1.307 Ise
= 14.34 Ikf = .2847 Xtb = 1.5 Br = 6.092 Isc = 0

Ikr = 0 Rc = 1 Cjc = 7.306p Mjc = .3416 Vjc = .75 Fc = .5 Cje = 22.01p
Mje = .377 Vje = .75 Tr = 46.91n Tf = 411.1p Itf = .6

Vtf = 1.7 Xtf = 3 Rb = 10)

The design engineers have swapped all of the clocking components from
two boards that work properly through temperature with two boards that
don't. The "temperature intolerance" follows the PWBs not the
components. This is also the first time we have used this board
vendor(Malaysia). The previous builds(here in town) did not show this
problem.

Ed V.

 

[EdV]

CustomPCB by any chance?

Maybe. Have you used them before? I did a "Reply to aauthor" earlier
and forgot my "deja"/google id mail account is not reachable. Sorry
for the misdirection.

Thanks,
Ed V.

 

[whit3rd]

The design engineers have swapped all of the clocking components from
two boards that work properly through temperature with two boards that
don't.  The "temperature intolerance" follows the PWBs

Do you have a guard ring around the sensitive input?

On-chip oscillators usually self-bias with a high impedance
in feedback to the oscillator 'input' pin from the output pin,
and a small amount of leakage current (either due to surface
cleaning or buried layers) is not unexpected. A printed
guard ring (even if it's only a broken ring, I.E. a C) will handle
the surface leakage, and you can replicate the guard in
any buried layers easily enough.

It isn't the board that's changing at low temperature, it's the
self-bias source (which is often just like a leakage current).

 

[PN2222A]

Why do you suspect the PCB?
First, rule out failures due to the capacitors.
If the caps used by your vendor have a different tempco than you
specified,
then they may cause the oscillator to stop.
Measure the caps individually, outside of the PCB / oscillator
environment.

OK, look for over-etched traces or fractures within the PCB.
Measure the resistance of the specific traces around the clock
oscillator.
The copper has a higher coefficient of expansion than the glass/epoxy.
it _could be_ that a trace is going open as the copper shrinks.

I'd be watching the suspect PCBs alone as they're chilled.

Good luck!

regards
PN2222A

 

[EdV]

Do you have a guard ring around the sensitive input?

On-chip oscillators usually self-bias with a high impedance
in feedback to the oscillator 'input' pin from the output pin,
and a small amount of leakage current (either due to surface
cleaning or buried layers) is not unexpected.  A printed
guard ring (even if it's only a broken ring, I.E. a C) will handle
the surface leakage, and you can replicate the guard in
any buried layers easily enough.

It isn't the board that's changing at low temperature, it's the
self-bias source (which is often just like a leakage current).

I checked the layout and there are guard traces leading from the xtal
ground plane to the uC xtal pins. It is also laid in accordance with
the uC manufacturers recommendation. The manufacturer is also pretty
baffled by this one.

All of the components from a "bad" PWB; uC, xtal and caps work on a
"good" board and all the same components from a "good" board don't
work on a bad board. (at -20C it works at room fine)

Their technician also said that the capacitance at the xtal caps
increases more on a bad board than a good board. Sorry I didn't get
the numbers.

Thanks for the suggestions,

Ed V.

 

[Archimedes' Lever]

sorry to repost but I thought mine may have got lost in the sea of
spam:

We got a batch of boards built and two of them show a curious
tendency(consitent) to shut down the crystal oscillator at -20 C.

Place the clock circuitry on a small daughterboard, encapsulate it, and
attach to main circuit by way of pins.

We
also notice that the capacitance of the xtal caps changes more on the
"bad" PCBs at -20C than the "good" boards.

Some caps could have hygroscopic surfaces, and going below dew point
makes the surface accessible to water. Also, many SMD caps are attached
to the PCB via a spot of glue. Said glue spot typically does not
displace all the space under an SMD part, therefore, you are left with
another place for accumulations to occur. Another point for vacuum
encapsulation, which displaces all voids allowing for little change.

1. Moisture in PCB Layers that didn't get baked out?

Both pre and post assembly. You also have to be careful that your fab
house is laminating your boards with very little moisture around.

2. Uncured epoxy betweem layers?

Shouldn't be. You tried different PCB strata types... have you tried a
different fab house?

3. Residual flux?

That depends on both your assembly process, and your choice of soldering
system, and most importantly, your post assembly cleaning process.

#3 I could address with vapor degreasing or other cleaning. #1 and 2
would mean baking at 100C or so which I guess is ok to see if it has
an effect.

100 C bake for an hour is a good thing. You can also place it in a
vacuum for a while to release any trapped moisture in the PCB strata as
well.

Any thoughts, insights or expereinces as always are welcome.

If it is a small board, as opposed to making the daughterboard, one
supposes that one could encapsulate your entire assembly. If it a larger
board, then the daughterboard suggestion is the best I could come up with
for temperatures that low.

 

[EdV]

Yes, I was going to say that our (entirely unsatisfactory) experience withthem
was unrelated to your problem until I saw PN2222A's post.  One of the problems
we experienced was severe over-etching to the point of losing continuity on 12
thou traces on both straight sections and 45deg bends.

Thanks for the info.

 

[EdV]

OK, look for over-etched traces or fractures within the PCB.
Measure the resistance of the specific traces around the clock
oscillator.
The copper has a higher coefficient of expansion than the glass/epoxy.
it _could be_ that a trace is going open as the copper shrinks.

I'd be watching the suspect PCBs alone as they're chilled.

Good luck!

regards
PN2222A- Hide quoted text -

- Show quoted text -

Thanks!