Cisco Sintered BeO (Ceramic) BGA Switch Processors

Gold Refining Forum

Help Support Gold Refining Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

cosmetal

Well-known member
Supporting Member
Joined
Apr 1, 2017
Messages
445
Location
Sacramento, CA USA
Ok . . .

I consider myself to be above average in data mining, but I need you GRF chip experts help with this one.

I have a Cisco 6500 Network Switch (circa 1999) that I have been slowly depopulating. The processors found on its boards have me stumped. They are:

1 ea. Pinnacle 06K0596 PQ processor - white ceramic shield between the chip and BGA balls.
1 ea. Pinnacle 09K5063 PQ processor - lavender ceramic shield between the chip and BGA balls.
1 ea. Cafe2 09K4130 PQ processor - white ceramic shield between the chip and BGA balls.
1 ea. Centauri 06K0017 PQ processor - white ceramic shield between the chip and BGA balls.
1 ea. Titan 29L7370 PQ processor - white ceramic shield between the chip and BGA balls.

They all have their chips embedded atop of a white and lavender ceramic heat shield that I assume is a sintered Be0 powder. I can not find anything online regarding their composition. The only thing that keeps popping up about them is how secretive Cisco is about their chips.

Has anyone ever tried to process these type of Cisco BGA chips? If yes, care to share your insights into how to process them without turning the BeO into powder and your yield if any?

Many thanks!

James
 

Attachments

  • Cisco Processor - Board BGA.jpg
    Cisco Processor - Board BGA.jpg
    352.6 KB · Views: 786
  • Cisco Processor - Chip BGA.jpg
    Cisco Processor - Chip BGA.jpg
    349.9 KB · Views: 785
  • Cisco Processor - Chip Exposed.jpg
    Cisco Processor - Chip Exposed.jpg
    260.7 KB · Views: 786
  • Cisco Pinnacle Processor - White.jpg
    Cisco Pinnacle Processor - White.jpg
    274.7 KB · Views: 786
There is no technical reason to use BeO in a switch. If they did it would create a lot of different environmental hazards, especially for service technicians. I think it's safe to assume that the white ceramic is alumina substrate used to mount the chip.

BeO has one distinctive characteristic, it is a very good heat conductor and if it was to be used it would sit between the chip and the cooler, not between the chip and the circuit board.

BeO is also quite light, about 75% of alumina. Not that you would immediately react on the heft of the ceramic, but enough that you could measure density with a pocket scale and some water in a glass.

The easiest way to decide if it is BeO is to take a similar thick piece of aluminium and an ice-cube. Put the ice-cube on one side of the aluminium and your finger on the other and feel how cold the aluminium feels. Do the same with the ceramic substrate. BeO conducts heat almost as good as pure aluminum so there shouldn't be a big difference in the feeling of the cold, while alumina ceramics is ten times better on insulating against heat. That should be easy enough to feel with the fingers.
You could do the same test by heating one corner of an aluminium piece and the substrate.

Göran
 
g_axelsson said:
There is no technical reason to use BeO in a switch. If they did it would create a lot of different environmental hazards, especially for service technicians. I think it's safe to assume that the white ceramic is alumina substrate used to mount the chip.

BeO has one distinctive characteristic, it is a very good heat conductor and if it was to be used it would sit between the chip and the cooler, not between the chip and the circuit board.

BeO is also quite light, about 75% of alumina. Not that you would immediately react on the heft of the ceramic, but enough that you could measure density with a pocket scale and some water in a glass.

The easiest way to decide if it is BeO is to take a similar thick piece of aluminium and an ice-cube. Put the ice-cube on one side of the aluminium and your finger on the other and feel how cold the aluminium feels. Do the same with the ceramic substrate. BeO conducts heat almost as good as pure aluminum so there shouldn't be a big difference in the feeling of the cold, while alumina ceramics is ten times better on insulating against heat. That should be easy enough to feel with the fingers.
You could do the same test by heating one corner of an aluminium piece and the substrate.

Göran

Goran,

Thank you for your informative answer. I am particulary glad for the "trade tips".

Best,
James
 
kernels said:
Short answer is that it is obviously a 'flip chip' style package and thus practically worthless for PM recovery.

Kernels,

I agree that they could be worthless for PM recovery. But, I have to respectfully disagree with you're "obviously a flip chip" assessment. :D According to Solid State Technology http://electroiq.com/blog/2002/06/bga-csp-and-flip-chip/

"The term BGA covers a wide range of package types. In this article, BGA refers to a 35-mm or larger device with 760-µm solder balls. The term CSP describes devices with 250-µm solder balls and an interposer layer between the die and solder balls. The overall package size of a CSP is typically no larger than 1.2 times the size of the silicon. Approximately 100 names are used by package manufacturers to describe various CSP designs, and many have “BGA” in their designation names. Flip chips are bare silicon devices that have solder bumps in the 75-µm range (Figure 1)".

BGA - CSP - Flip Chip Comparison.gif
Figure 1. Typical BGA, flip chip and CSP structures.

Quite frankly, I am not sure what these chips are. That is why I started this post. Not a "classic" BGA, CSP or Flip Chip. Too custom. These chips are over 35mm and have very large solder balls. But, they don't have an epoxy package covering them. The Pinnacle type were the only ones that had a large aluminum heat sink on top of the complete package. Then the printed aluminum shield attached directly to the silicon die. The die is attached to the large alumina ceramic (not BeO) shield which has the large BGA type solder balls.
Pinnacle_Board - Front1a.jpg
Processors on Riser Board 1a.jpg
Have you any experience with these type of Cisco chips?

Has anyone on GRF ever tried to process these type of Cisco chips?

As far as PM recovery, I'll see. As part of my refining education, I will be working on them as so little has been written about them from the recovery standpoint. So little information.

Best,
James
 
If you want to know what is inside you need to crush one and have a look.
I belive mine is very similar to yours. First photo is with heatsink removed.
On second photo you can clearly see bonding between pads and silicon dye.
I don't know what metal is, maybe silver, maybe tin, but is definitely not gold.
 

Attachments

  • DSCN1794.jpg
    DSCN1794.jpg
    409.8 KB · Views: 685
  • DSCN1796.jpg
    DSCN1796.jpg
    134.8 KB · Views: 685
The bonding between the alumina substrate and the die is most probably small solder balls... maybe 75 um in size... what I would call a flip-chip. :D
A flip-chip on a BGA made with a ceramic substrate.

Interesting to see the internal layering with via holes and dark conductors in several layers.

To be fair, the white ceramic might be something else than alumina, there are many different ceramic substrates that has been used over the years. But even if it isn't alumina the same reasoning is true for other ceramics too.
Anyone wanting to read more about the ceramic substrate can google LTCC (Low Temperature Co-fires Ceramics) or just check this website for a brief introduction.
http://www.elite-advanced.com/low-temperature-co-fired-ceramic-ltcc/
or wikipedia.
https://en.wikipedia.org/wiki/Co-fired_ceramic

Göran
 
When the die is soldered directly to a substrate, we tend to call that 'flip chip' mounting. It doesn't matter whether that substrate is ceramic (like your chip), fiberglass (like modern north-bridge chips) or straight onto a PCB. At the end of the day, there are no bond-wires, only a tiny amount of ENIG under the solder balls, so practically worthless for PM recovery.

But don't trust me, I'm just some random guy on the internet, process a batch of them.
 
It is done in similar way than multi layer PCB. I found similar structure also in ceramic CPUs (486, Pentium).
In ordinary PCB or even "green fiber flip chips" layers are connected with copper vias, but in ceramic ICs some metalic substrate is printed on ceramic surface and in holes connecting different layers any silicon dye. In hybrid circuits I heard they some times use silver and palladium for printing on ceramic.
I don't know what is material in ceramic ICs. Maybe this could be source of silver everyone finds when processing ceramic CPUs.
 
shmandi said:
If you want to know what is inside you need to crush one and have a look.
I belive mine is very similar to yours. First photo is with heatsink removed.
On second photo you can clearly see bonding between pads and silicon dye.
I don't know what metal is, maybe silver, maybe tin, but is definitely not gold.

Shmandi,

Thank you for sharing your pictures.

Did these chips also come from a network switch?

Best,
James
 
kernels said:
When the die is soldered directly to a substrate, we tend to call that 'flip chip' mounting. It doesn't matter whether that substrate is ceramic (like your chip), fiberglass (like modern north-bridge chips) or straight onto a PCB. At the end of the day, there are no bond-wires, only a tiny amount of ENIG under the solder balls, so practically worthless for PM recovery.

But don't trust me, I'm just some random guy on the internet, process a batch of them.

Kernels,

Please don't take my response as being critical of your answer. I, was, and still am, confused regarding the whole BGA, CSP and Flip Chip definitions found online.

I suppose that an industry guideline for packaging definition goes "out the window" when it comes to custom chips. To me, the chips I found on the Cisco 6500 Switch are obviously custom Cisco creations. Of which, there is very little information available and, of which, Cisco must be proud. Hence, their custom individualized tradenames printed on their top aluminum heat sinks/shields.

Cheers,
James
 
cosmetal said:
Did these chips also come from a network switch?

Best,
James
I think they came from some older HP server board. They come in different shapes and sizes and also purple colour.
 

Attachments

  • DSCN1804.jpg
    DSCN1804.jpg
    198.5 KB · Views: 594
Yes, like the Cisco ones they are a flip chip design. - I use flip chip to differentiate between old core style processors which contain bonding wires and the surface mounted core of the flip chip which has no bonding wires. It fits 99% of the types out there so it's valid. Using the exception to say that the vastly applicable rule is incorrect isn't good analysis.
 
All,

So far, I have processed these Cisco chips as follows:

1.) Heated all solders balls with hand-held butane culinary torch and scrapped to remove.
2.) Hot HCl soak until all remaining solder and organic residue was removed.
3.) Hot wash with boiling DH2O.
4.) Using the culinary torch, removed the top aluminum heat sink/shield from the die and its ceramic support. Dies remained stuck on their substrates.
5.) Broke chips into smaller pieces and trimmed ceramic material from around the die pieces.
6.) Placed die pieces in one beaker and remaining ceramic pieces in another beaker.
7.) Hot HNO3 soak.
8.) Cooled and let pieces continue the soak overnight.
9.) Hot wash with DH2O.
10.) Photographed results.

Questions:

ENIG, gold brazing or other?
How can the die be removed whole from the ceramic substrate? I only had one fall off completely.


I would like to try future processing on other ceramic BGA/Flip Chips without breaking the dies. The epoxy cement used on these Cisco chips won't budge. I don't want to apply too much heat in fear that I could vaporize the gold.

I know that the dies on plastic-backed BGA/Flip Chips come off with incineration. But, obviously, I can't incinerate ceramic substrates. Can I?

Any answers, comments or suggestions are always greatly appreciated!

James
 

Attachments

  • Cisco Die Face4_1.jpg
    Cisco Die Face4_1.jpg
    1.8 MB · Views: 575
  • Cisco Die Face9_2.jpg
    Cisco Die Face9_2.jpg
    1.6 MB · Views: 574
  • Cisco Die Face5_3.jpg
    Cisco Die Face5_3.jpg
    2.3 MB · Views: 574
  • Cisco Die Face4_4.jpg
    Cisco Die Face4_4.jpg
    2 MB · Views: 575
  • Die Comparison2_5.jpg
    Die Comparison2_5.jpg
    670.4 KB · Views: 575
anachronism said:
Yes, like the Cisco ones they are a flip chip design. - I use flip chip to differentiate between old core style processors which contain bonding wires and the surface mounted core of the flip chip which has no bonding wires. It fits 99% of the types out there so it's valid. Using the exception to say that the vastly applicable rule is incorrect isn't good analysis.

It fits 99% of the types out there so it's valid. Using the exception to say that the vastly applicable rule is incorrect isn't good analysis.

A very good point, Jon.

Thank you for your feedback!

Cheers,
James
 
Are those pictures of the bottom of the die ? In that case, a lot of die have a Gold-ish color to them, while not being Gold. The ENIG plating would have been on the solder-ball side of the BGA chip, but there is probably a tiny amount of ENIG under the die too. The ENIG that was there has likely been dissolved into the solder balls when they were reflowed (melted in production).

Braze is something completely different, more like Gold based solder. It is used on non flip-chip ceramic chips to attach the die to the ceramic, but then you would have bond wires on the other side of the die to electrically connect the die to the pads on the ceramic.

I think you may be wasting your time, but it is your time to waste.
 
kernels said:
Are those pictures of the bottom of the die ? In that case, a lot of die have a Gold-ish color to them, while not being Gold. The ENIG plating would have been on the solder-ball side of the BGA chip, but there is probably a tiny amount of ENIG under the die too. The ENIG that was there has likely been dissolved into the solder balls when they were reflowed (melted in production).

Braze is something completely different, more like Gold based solder. It is used on non flip-chip ceramic chips to attach the die to the ceramic, but then you would have bond wires on the other side of the die to electrically connect the die to the pads on the ceramic.

I think you may be wasting your time, but it is your time to waste.

Kernels,

"Are those pictures of the bottom of the die?"

Yes, they are.

And, you, and others, are right about there being no appreciable gold ENIG let alone, gold braze, on the dies or within the chips.

I ran the dies in a hot A/R solution for an hour or so and tested with SNCl2. I am happy to report that the result was negative for gold.

Why would I be happy? Because I have a result that I can share about a specific type of custom IC from Cisco for which very little information is available. I pursued this inquiry because of that lack of available data and, also, in the hope that being a trademarked telco IC from a leader in network technology, it could have been different from other flip chips.

"I think you may be wasting your time, but it is your time to waste."

Not a waste at all. I enjoyed it and it was part of my continuing education.

Peace all,
James
 

Attachments

  • Cisco Dies in Test Tube.jpg
    Cisco Dies in Test Tube.jpg
    1.5 MB · Views: 537
Back
Top