DSPRelated.com
Forums

[Fwd: Re: mounting DSP C6711 in socket]

Started by Jeff Brower December 25, 2004
Shazia-

> It is 256 BGA package ( 1.27mm)

Is this Rev D C6711? Or earlier? You should be using at least Rev D, as
previous versions are not recommended for new designs. Rev D has a 272
pin BGA package, same as C6713.

I don't have any experience with BGA socket. I might guess it would go Ok
if the quality of the socket is good and the ball grid array is no more
than 2 mm extra distance from the board pads.

But one thing I would suggest you check carefully when you first debug
with a socket is Vcc core noise and heat dissipation. The D rev has 16
Gnd pins in the middle of the chip, in addition to various Gnd pins
interleaved amoung the outer rows. Initially, with Rev A-B, those Gnd
pins were not there -- TI had to add them to achieve the desired clock
rate + Vcc core ripple specification + heat dissipation. So if you are
using revision less than D with a socket, this could be an issue.

Why not build up 2 PCBs for debug -- one with socket and one without?

-Jeff

> Jeff Brower <> wrote:
> Shazia-
>
>> We r customizing the DSP board for our application. We have designed
the board and now we need to order for PCB. I need to know that, will
it make any difference, if we mount the DSP in a socket soldered on
baord instead of soldering the DSP directly on board????
>> Please get back with me ASAP.
>
> What is the DSP package type?
>
> -Jeff




Shazia,


Jeff Brower <j...@signalogic.com> wrote:


Shazia-

> It is 256 BGA package ( 1.27mm)

Is this Rev D C6711? Or earlier? You should be using at least Rev D, as
previous versions are not recommended for new designs. Rev D has a 272
pin BGA package, same as C6713.

I don't have any experience with BGA socket. I might guess it would go Ok
if the quality of the socket is good and the ball grid array is no more
than 2 mm extra distance from the board pads.

But one thing I would suggest you check carefully when you first debug
with a socket is Vcc core noise and heat dissipation. The D rev has 16
Gnd pins in the middle of the chip, in addition to various Gnd pins
interleaved amoung the outer rows. Initially, with Rev A-B, those Gnd
pins were not there -- TI had to add them to achieve the desired clock
rate + Vcc core ripple specification + heat dissipation. So if you are
using revision less than D with a socket, this could be an issue.

Why not build up 2 PCBs for debug -- one with socket and one without?

-Jeff

> Jeff Brower wrote:
> Shazia-
>
>> We r customizing the DSP board for our application. We have designed
the board and now we need to order for PCB. I need to know that, will
it make any difference, if we mount the DSP in a socket soldered on
baord instead of soldering the DSP directly on board????
>> Please get back with me ASAP.
>
> What is the DSP package type?
>
> -Jeff

Jeff has some good suggestions.  I have used sockets with 6211, 6711, and 6713 devices [including mounting sockets on TI's assembled 6711 DSK boards <this is not for the 'faint at heart' or impatient>] successfully.  The only problems that I have experienced have been manufacturing related [99% of the BGA contacts were soldered instead of 100%].  This is usually caused by a job that is only one or two boards and the 'assembler/manufacturer' not setting up a proper solder profile for the socket being used.

My advice is to try not to use socketed DSPs as part of a plan.  Normally solder the DSP and use vias to make all non power/grd signals visible on the opposite side of the board.  If you can, build 3 initial bring up boards [though two will work] - never use one unless you are very experienced with the hardware.  If boards are large and expensive, only stuff the 'heart of the system' on a couple of boards to get them up and running.

Good Luck,

mikedunn