Curious use of exposed solder ball areas on PCB

[lscw]

I cracked open my old broadband router to have a peek inside and saw these unusual (to me) areas of exposed solder balls on the back side. Are they EMI shielding? If so, why not just copper pour?

Incase anyone is interested the router is a Sagemcom 2704N, P/N 253610462
The chip (glued!) under the aluminium heatsink is a Broadcom BCM6318 xDSL SoC, the one under the ceramic heatsink is a Broadcom BCM43217 WLAN jobby.

[perieanuo]

it's for heat radiation, cooling purposes

[golden_labels]

Recently there was another topic with a similar pattern and the conclusion was it’s for increased current capacity, with the pattern itself used to accomodate for wave soldering.

[lscw]

Thanks for pointing that out.

I see that's a more conventional PCB trace tinning but in this example they are part of the ground plane.

Here's another image of it, with the piece of tape removed

[T3sl4co1l]

I would assume it's to give something for that EMI tape to stick onto.  It's conductive EMI tape, right..?

Tim

[lscw]

I'm not sure what the tape is for, it isn't at all conductive according to my DMM. The same stuff is seen on the top side covering some of the RAM pins

[T3sl4co1l]

Well... there are passives under there, I suppose it would be rather irresponsible to use conductive tape, wouldn't it.   Still guessing it has some sort of shielding or absorbing effect; but clearly it's also not making electrical contact to ground.  Dunno.

Ed: oh weird, they've also got 0-ohm jumpers going ground to ground.  Even though those grounds look to be well stitched.  I wonder just how many layers they got away with, here?  And a few are non-pop, R842, R848 for instance.  The exposed ground areas really only seem to correlate with heat dissipation areas, like around the regulator chip (U305?, the thermal vias by R842).

Oh weird also: what's the gray connector?  It looks to be one pair, transformer and capacitor coupled, and low impedance all the way up to the SoC.  Was thinking at first, that transformer might be PoE, but no...  Wait, is that what a DSL coupling network looks like?  ISDN..?

Tim

[lscw]

Yeah, they're certainly going belt and braces with those 0 Ohm jumpers. I figured maybe to reduce ground loop albeit very slightly, what with the upper one at R846 being near the end of the trace they're bridging. At least I can't think of another reason.

It's just a two layer board. U305 and its two inductors, ses I see that now. I wonder if maybe R842 could have been for a thermal conductor - the route back from south side of that to the negative pin on the electrolytic cap is 10 times more convoluted than any difference that would save.

The upper left area corresponds with the WiFi chip. I don't know if the designer left contingency for soldering a can over that area on the top side or if that exposed tinned copper perimeter is just for added EMC?

It's an ADSL/ADSL2+ modem router.  Not much in english comes up for that isolation transformer, nor for the UMEC UT20B20S equivalent on the router below.

Incase you're interested, the ADSL pair comes in to the two outer pins on the primary side of the transformer and, as you say back out to the film cap.
On the secondary side the signal traces come from pins 1 and 4, each through a series 630R resistor (R422/419), 14nF caps (C444/441) and into the SoC.
After coming out of R422/419 the signal traces also go into the two 42nF caps (C434/463, labels just out of sight), 72R resistors R421/420 and into the SOIC Broadcom 6301KSG ADSL line driver. The datasheet for that is conspicuously unavailable.

I was too curious not to take a look inside a Sky Q (SR115) ADSL2+ router I also had lying around... more exposed solder balls. Again these are integral to the ground plane but this area of ground plane has a corresponding identical one on the other side. So I guess these have a completely different purpose?

[T3sl4co1l]

Aha!  Haven't taken apart a DSL modem so that explains that.  Coupling network sounds pretty normal, that'd be AC line coupling, more coupling, bias and termination.  Oh, not quite normal, that might be a duplexing transformer actually, that would explain the two pairs coming off it (and separate driver).

Only two layers?  That's insane!  Well, those jumpers probably are needed then!

This board, those two solder-studded pads and the "hi pot" mark beside them, I wonder if that's a conspicuous test pad?

Still wouldn't make sense to expose so much area on the other for testing, and in such sensitive areas.

Tim