328 TQFP

Interesting message received from Mouser today:

You have purchased the following part number(s) from Mouser Electronics within the last two years. (yeah, like last week!) The manufacturer has published a Product Change Notice. Atmel ATMEGA328P-AU PCN: Change from Gold to Copper Bonding Wire (between the die and the external legs presumably?)

Wonder what impact that will have on anything? New datasheet out too, with a different look: http://www.atmel.com/Images/doc8271.pdf Took a quick glance thru Section 29, electrical characeristics, nothing jumps out.

Huh. Well that is interesting to know, not sure what a person would do about it. Also not sure of the pros and cons Cu vs Au, certainly not current in that area although I did bond just a few dies back in the day.

This was likely because of cost, not performance.

Copper is more conductive, but more importantly it is much cheaper than Gold. However, copper is a more difficult to bond during manufacturing because it requires an inert atmosphere and is harder (stiffer) than Gold.

As manufacturers learn to optimize the copper bonding process, the cost benefit gets even higher.

and except for possible thermal issues with the bond totally the same electrically... I do hope they have the bonding right...

Doc

Docedison: and except for possible thermal issues with the bond totally the same electrically... I do hope they have the bonding right...

Doc

Hmm, you bring up an interesting question about thermal issues Doc. Copper is more thermally conductive than gold, 401 W/(m·K) vs. 318 W/(m·K). This by itself would be a potentially good thing, meaning heat can flow out of the die a little faster. However, the rate of thermal expansion is also slightly higher, 16.5 µm/(m·K) vs. 14.2 µm/(m·K) (both @ 25 °C). Expansion and contraction due to heating and cooling will put mechanical stress and strain on the joints, and it will be proportional to the rate of thermal expansion. Still I don't know if the higher thermal expansion will matter enough to be a problem, even at the small scale of die-to-package wire bounding. On the other hand, I don't know enough to rule it out either.

After scanning through both Sections 29 and 30 (Section 30 has the typical characteristics for the different part numbers), I haven't noticed any significant changes. Since the datasheets don't seem to have changed, we'll just have hope that before making this change to their manufacturing process Atmel did adequate testing to ensure this new bonding technique wouldn't have an appreciably negative effect on their ICs' reliability or performance.

It would be nieve to think that someone at ATmel (or virtually another company dealing with parts rooted in metallurgy) didn’t consider the CTE differences between Al and Au.

(I would seriously doubt that Atmel is the first company to do this. By now, it's probably a well-established and tested process, involving mostly "buy a new wire-bonder for the fab" sort of modification. (or "move this part to the fab with the new wire-bonder.")

OTOH, I'm still pissed at Intel for that whole Flash memory fiasco, when they moved production to a new fab and the chips DIDN'T WORK. (This was back at the dawn of flash memory. It's hard to believe that now you can go out and buy 16GBytes of flash an at office-supply store for about $10.)

[quote author=James C4S link=topic=121820.msg920089#msg920089 date=1347337062] It would be nieve to think that someone at ATmel (or virtually another company dealing with parts rooted in metallurgy) didn't consider the CTE differences between Al and Au. [/quote]

I'm sure more than one person considered the issue, what takes significant time and resources is doing the literature research and/or experimental validation necessary to deal with it. In other words, I don't doubt the technical compentency exists at ATmel to execute this process change properly, but we (outside the company) have no way of knowing if management allowed that to happen. This is especially true because this change was apparently to achieve lower costs. That isn't in and of itself a bad thing, but history is rife with examples of people and companies being penny wise and dollar foolish.

hope that before making this change to their manufacturing process Atmel did adequate testing to ensure this new bonding technique wouldn't have an appreciably negative effect on their ICs' reliability or performance.

I would rather think that Atmel is 'running' on the evident success of others, time will tell though. Al... is interesting in of itself as it is a problem in IC fab and the inclusion of this metal always brings up questions in my mind as it is always an issue at high temp's because of it's mobility... under the right conditions it will wander about.

Doc

Docedison: I would rather think that Atmel is 'running' on the evident success of others, time will tell though.

It's certainly possible that just they are relying on the success of outside organizations without performing any investigation or verification, like jumping on the metaphorical bandwagon without really knowing where it's headed. Though I really hope those in charge at ATmel were smarter than that. In any case, as you mentioned, all we can do is wait and see the results regardless of the details that went into this decision.

As an update to this two-year-old thread, I just got notification from Mouser that the PU (DIP) package ATMegas and ATTtinies are going to copper bind wires as well. Hopefully this will make the chips at least a penny cheaper…

Buy in bulk, makes them a lot cheaper. 9% (10 lot), 37% (25 lot), 41% (50 lot), 43% (100 lot) less. No sales tax from Mouser too, just shipping.

If you're worried about $0.01 in cost, you aren't buying from catalog distributors like Mouser anyway.