Maker Pro
Maker Pro

Soldering bypass caps to QFP possible?

J

Joerg

Jan 1, 1970
0
John said:
YIKES!

What WERE they thinking?

You have your own R&D plus manufacturing so designs have to exceed the
Larkin quality threshold before they go into production. So you know
it's all good. But when you are consulting stuff like this is a fairly
common thing you'd see :-(

Sometimes I have clients who want a new design from scratch. Others come
with existing ones that plain don't work. Consulting is like architects
and building contractors in one person. We get to design/build someone's
dream home but we also have to restore flood or fire damage. In many
cases we have to serve a company specialized in their field but not
electronics. And if problems creep up they often don't after some
substantial field exposure. For example storms only happen during a very
short season and the usual ESD pistol tests at TUEV or UL won't reveal
too many problems up front (it passed all those).

Definitely make a small pc board to glue on top of the chip, with
little wires hanging down, to solder to the ic pins.

Sort of like this one:

ftp://66.117.156.8/883A.jpg

I might have to :-(
 
J

Joerg

Jan 1, 1970
0
Fred said:
Joerg a écrit :

Use 0204 format caps.
For ex:
http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=490-4310-2-ND

Plus they'll be much easier to solder.



Ouch!
I never tried it but maybe this works:
- just scratch the varnish on the pins top surface (let it stay between
the pins, as a regular solder mask would do on a PCB)
- apply a *tiny* blob of solder paste across both pins
- pick the cap and roughly center it on the blob
- gently hot air reflow all this.

-see what happened...

Yes. I want to also try that with a 01005 cap since it would nicely
"ride" the gap. Some paste, some flux, then heat.
 
J

Joerg

Jan 1, 1970
0
Baron said:
I don't know if you have any space where you can go right through the
board and put caps in the holes. I've done this with decoupling caps
on microwave amps.

The top is full of traces but there aren't any useful GND vias anywhere
near anyhow.
 
J

Joerg

Jan 1, 1970
0
dalai said:
Un bel giorno Joerg digitò:


If the pin are adjacent, it should be quite easy and clean if you use hot
air and very small capacitors.

Anyway, if you really need 0.1 uF caps, it's not necessary to solder them
directly on the pins. It's enough to solder them below 2-3 cm from the
chip. There is a very nice application note on the Xilinx site about bypass
caps:

http://www.xilinx.com/support/documentation/application_notes/xapp623.pdf

It isn't, under one condition: If the pins are via'd right to the
planes. In this case they weren't.
 
J

Joerg

Jan 1, 1970
0
Leon said:
There used to be a company here in the UK called Precison Repair
Services that could probably fix them, but they don't seem to be
around any more. They've done some comparatively simple rework for me
in the past, but some of the things they used to do were amazing.

We'll have to find a place like that, preferably somewhere in the US.
 
S

Spehro Pefhany

Jan 1, 1970
0
Nice! Unfortunately this already is a fine-pitch part, a 100-QFP with
0.5mm pitch. To avoid this capacitor stuff I could also muffle things
from an ESD/RFI point of view if the front panel overlay in front of a
large LCD was conductive. But it ain't. Placing another (conductive)
film below would work but its glue side would face up and eventually
stick to the current overlay. IME that turns yucky really quick.

Looks like one of those up the creek situations.

I would have no problem with doing one or a few for field tests, but
actually shipping a whole bunch of those things with that kind of
rework.. especially if the environment is not benign in the extreme
(which the conformal coating would tend to indicate is NOT the case)..
they should evaluate the cost per failure (from this point in time,
water's already under the bridge) compared to redoing the boards
salvaging whatever makes sense. The end users may be quite patient
with a sure fix, but if they are shipped something else, which then
fails, it may be the last time your client hears from them (maybe
indirectly they might hear from their legal dept, but probablyh not
AP). That's more of a business decision.. cost and risk management.

If you absolutely have to-- maybe strip the conformal coating, glue an
0805 or 0603 part to the top with a similar kind of cement as is used
for gluing SMT parts to boards (Locktite XXX) and run a couple of fine
wires (with some give in them) a few mm to the pins. Inspect the hell
out of it, then conformally coat again to glob the wires and cap down.
It still won't be as good as having a proper layout.


Best regards,
Spehro Pefhany
 
J

Joerg

Jan 1, 1970
0
Spehro said:
I would have no problem with doing one or a few for field tests, but
actually shipping a whole bunch of those things with that kind of
rework.. especially if the environment is not benign in the extreme
(which the conformal coating would tend to indicate is NOT the case)..
they should evaluate the cost per failure (from this point in time,
water's already under the bridge) compared to redoing the boards
salvaging whatever makes sense. The end users may be quite patient
with a sure fix, but if they are shipped something else, which then
fails, it may be the last time your client hears from them (maybe
indirectly they might hear from their legal dept, but probablyh not
AP). That's more of a business decision.. cost and risk management.

It's not just that, also turn-around time. A re-layout and new run
realistically takes many weeks. Rework is actually quite common. When I
took apart an aircraft (!) radio from a mainstream manufacturer in that
market I was initally shocked by the amount of rework in there.
Especially when considering the production volume. Then a service tech
told me that this is quite "normal".

If you absolutely have to-- maybe strip the conformal coating, glue an
0805 or 0603 part to the top with a similar kind of cement as is used
for gluing SMT parts to boards (Locktite XXX) and run a couple of fine
wires (with some give in them) a few mm to the pins. Inspect the hell
out of it, then conformally coat again to glob the wires and cap down.
It still won't be as good as having a proper layout.

Either that or a 01005 cap soldered directly.
 
L

linnix

Jan 1, 1970
0
I would have no problem with doing one or a few for field tests, but
actually shipping a whole bunch of those things with that kind of
rework.. especially if the environment is not benign in the extreme
(which the conformal coating would tend to indicate is NOT the case)..
they should evaluate the cost per failure (from this point in time,
water's already under the bridge) compared to redoing the boards
salvaging whatever makes sense. The end users may be quite patient
with a sure fix, but if they are shipped something else, which then
fails, it may be the last time your client hears from them (maybe
indirectly they might hear from their legal dept, but probablyh not
AP). That's more of a business decision.. cost and risk management.

I agree. Fixing the layout is the best thing to do. 2500 boards are
peanuts, unless they are big mult-layers boards.
If you absolutely have to-- maybe strip the conformal coating, glue an
0805 or 0603 part to the top with a similar kind of cement as is used
for gluing SMT parts to boards (Locktite XXX) and run a couple of fine
wires (with some give in them) a few mm to the pins. Inspect the hell
out of it, then conformally coat again to glob the wires and cap down.
It still won't be as good as having a proper layout.

The next best thing to do is to solder an identical QFP lead-frame on
top of it, with the passive components custom bonded to the lead-
frame. How many pins are there for the existing QFP?

We can build the custom QFP for you. You don't even have to send us
your boards.
 
J

Joerg

Jan 1, 1970
0
linnix said:
I agree. Fixing the layout is the best thing to do. 2500 boards are
peanuts, unless they are big mult-layers boards.

4-layer but some pricey stuff on there, LCD and such.

The next best thing to do is to solder an identical QFP lead-frame on
top of it, with the passive components custom bonded to the lead-
frame. How many pins are there for the existing QFP?

100 pins.

We can build the custom QFP for you. You don't even have to send us
your boards.


Possibly my client might take you up on that :)
 
L

linnix

Jan 1, 1970
0
4-layer but some pricey stuff on there, LCD and such.



100 pins.

I was hoping for 48 pins, since we are ordering that. I'll check if
they have 100 pins in stock. If not, usually they need to make 10,000
pcs (any pins multiple of 4). With some arm twistings, they might do
5,000.
Possibly my client might take you up on that :)

Just one cap? Or do you want to add more parts in it.
 
J

Joerg

Jan 1, 1970
0
linnix said:
I was hoping for 48 pins, since we are ordering that. I'll check if
they have 100 pins in stock. If not, usually they need to make 10,000
pcs (any pins multiple of 4). With some arm twistings, they might do
5,000.


Just one cap? Or do you want to add more parts in it.

Well, I am really up the creek now. The cap helps but not a lot (because
they also split the ground off). The next available node for either GND
or VCC is miles away :-(
 
B

Ben Jackson

Jan 1, 1970
0
Yes. I want to also try that with a 01005 cap since it would nicely
"ride" the gap. Some paste, some flux, then heat.

Better still, you might be able to put that on the pads instead of
balancing it on the pins.

You mentioned the distance to the nearest ground... Does this micro
have multiple grounds that are not tied under the chip? You might have
to tie them together if you don't want it to reset when there's a storm!
 
J

Joerg

Jan 1, 1970
0
Ben said:
Better still, you might be able to put that on the pads instead of
balancing it on the pins.

You mentioned the distance to the nearest ground... Does this micro
have multiple grounds that are not tied under the chip? You might have
to tie them together if you don't want it to reset when there's a storm!

Yep, unfortunately that's the situation :-(

Going for the system solution now (shielding around the whole board). It
looks much easier to retrofit.
 
J

Jim Thompson

Jan 1, 1970
0
Yep, unfortunately that's the situation :-(

Going for the system solution now (shielding around the whole board). It
looks much easier to retrofit.

At GenRad, I always asked for (and got) a clear space for the
analog... and I did the layout. Digital guys have no concept of
ground or shielding or coupling :-(

...Jim Thompson
 
J

John Devereux

Jan 1, 1970
0
Joerg said:
Yep, unfortunately that's the situation :-(

Going for the system solution now (shielding around the whole
board). It looks much easier to retrofit.

I wonder what is the interference path, (and why can't the static gun
test it?).

- Conducted interference from the wires entering the case?

- Or radiated from nearby wires that themselves have conducted
interference on them?

- Or direct EM radiation from the lightning?
 
J

Joerg

Jan 1, 1970
0
Jim said:
At GenRad, I always asked for (and got) a clear space for the
analog... and I did the layout. Digital guys have no concept of
ground or shielding or coupling :-(

Us consultants often have to pick up the pieces afterwards and try to
save the day :-(
 
J

Joerg

Jan 1, 1970
0
John said:
I wonder what is the interference path, (and why can't the static gun
test it?).

- Conducted interference from the wires entering the case?

- Or radiated from nearby wires that themselves have conducted
interference on them?

- Or direct EM radiation from the lightning?

Number three ;-)
 
J

John Devereux

Jan 1, 1970
0
Joerg said:
Number three ;-)

That's interesting. I can see that happening with cell phones held
near - but for some reason I would never have expected a direct EM
effect from lightning. Unless very close of course!
 
J

Joerg

Jan 1, 1970
0
John said:
That's interesting. I can see that happening with cell phones held
near - but for some reason I would never have expected a direct EM
effect from lightning. Unless very close of course!

Not on this product, but I've seen other cases where traces have been
vaporized, SOT23 diodes that had no plastic housing anymore, etc. "But
the lightning only struck a fence post 20ft away and the circuit had not
connection to anything ..."
 
T

Terry Given

Jan 1, 1970
0
Joerg said:
Not on this product, but I've seen other cases where traces have been
vaporized, SOT23 diodes that had no plastic housing anymore, etc. "But
the lightning only struck a fence post 20ft away and the circuit had not
connection to anything ..."

in the late 90s our machine shop moved across the road (we expanded).
The IT department asked R&D what to do re. ethernet, and was told "fiber
optics". But that cost a few k, so they strung an ethernet cable up on
the power poles, beside the phone lines. About 1km away (across the road
from the company our former R&D mgr and senior engineers set up)
lightning hit a ground-mounted transformer, and made a hell of a mess.
the resulting EMP snotted about a dozen ethernet cards, and fried the
brain operating an old turret-punch that had been upgraded from relay
control to PC control. Ignoring the turret punch, it cost $3k to fix all
the other PCs. The turret punch, being customised about a decade earlier
by nobody knows who, was essentially unfixable. When the IT department
queried R&D they got a great deal of laughter, and "WE TOLD YOU SO".
After this little event, they bought the thousand bucks worth of fiber
optics. I believe the turret punch got sold as scrap. oddly enough,
nobody got fired.

Cheers
Terry
 
Top