Possible to Change Processor Chips in BBB?

Hi Stephen,

I saw this post that may help - they suggest to use the AM3358, which is what BBBs will switch to eventually too (it doesn't have EtherCAT). Farnell sell the AM3358 in single quantities. Removing a BGA while keeping it intact and then soldering it back in sounds like a very skillful thing to do but some people may have managed it with practice.

shabaz wrote:

 

Removing a BGA while keeping it intact and then soldering it back in sounds like a very skillful thing to do but some people may have managed it with practice.

removing the BGA is the easy part, but you're going to lose the existing solder balls when you do.  There is expensive equipment that's capable of re-balling one-off devices, and while I'm sure there will be someone out there who's botched up something to do it with some cardboard, double sided sticky tape and a toaster oven you're quickly approaching the ridiculously pointless. 

BGA's simply aren't a good choice unless you have all the proper equipment and a proper manufacturing setting. Even there, they won't be wanting to have to remove one and put it on a different board, too much hassle.

Gerald's response in that thread is interesting, according to the datasheet the 3352 doesn't simply lack EtherCat, it lacks the PRUs altogether!

selsinork wrote:

 

Gerald's response in that thread is interesting, according to the datasheet the 3352 doesn't simply lack EtherCat, it lacks the PRUs altogether!

With neither EtherCAT, PRUs nor SGX, I wonder if TI is trying to position the AM3352 as a microcontroller rather than an applications processor.  There's a bit of a performance gap between a Cortex-M4F and any Cortex-A, so perhaps the AM3352 fills a fairly important underpopulated niche.

And it may not cost much at microcontroller volumes with so many things pruned off its die.

The difference in part numbers is a silicon revision issue.

The B revison (latest) on the TI data sheet and listed as "no stock" in TI's website is the one you would want.

The parts Digikey have in stock are the initial release (two silicon revisions out of date).

You can plough through the errata on TI's website and check if you can live with the bugs.

If you are planning  a product you should establish some dialogue with a fully franchised TI disti (or possibly direct with TI tech support) and find out what parts you can get and when.

My advice (having had my own fingers burnt a few times and seen others go down completely (anyone remember Nascom and the memory chip shortage that killed them)) is to have the chips in my hand before I lay out the board.

I wouldn't design a product on the basis of recovered chips (at least not if any of my own money or reputation was at stake).

MK

Haha, nice coincidence.  Last night I wrote:

With neither EtherCAT, PRUs nor SGX, I wonder if TI is trying to position the AM3352 as a microcontroller rather than an applications processor.  There's a bit of a performance gap between a Cortex-M4F and any Cortex-A, so perhaps the AM3352 fills a fairly important underpopulated niche.

 

And it may not cost much at microcontroller volumes with so many things pruned off its die.

And today I see on Olimex's blog that "AM3352-SOM module with 1Ghz Sitara Cortex-A8 SoC prototypes are ready" !!!  A 1k volume price of 26 euro is mentioned.

Notice that on their AM3352-SOM-EVB evaluation baseboard spec it says "Dual Ethernet 100MB", which means that the SOM should be perfect for router/firewalls as an alternative to the hack-WRT-router approach.  The AM3352 might as well be considered TI's low-end comms processor.

Morgaine.

Those Olimex AM3352 modules do look great and very compact. According to the comments if some customer places an order of more than 1000 of a processor variant, then they may stock that variant in their online shop  too. I hope that happens.