First, some apologies to all of you in the packaging industry: I find packaging one of the least interesting, hardest-to-spin-into-an-engaging-story corners of the semiconducting industry. But Eoplex CEO Arthur Chait stopped by IEEE Spectrum yesterday to show me what he expects will be a revolution in chip packaging, and I must admit it has me thinking differently. Eoplex, developer of a 3D printing technology used to make high-end cellphone antennas and other small, complicated contraptions, really might make packaging sexy.
Here’s a little background on Eoplex, because you’ve probably never heard of it. A California startup, Eoplex has come up with a combination of secret sauces and manufacturing techniques that lets it print sub-micron size voxels of stuff to mass produce 3D objects. After some simple, but secret, processing this stuff turns into metal, ceramics, and empty spaces. The result can be miniature machines with moving parts, metamaterials-enabled multi-function antennas, piezoelectric powered energy harvesters, coin-sized hydrogen fuel cells, pretty much anything
As amazing as all that stuff is—and the menagerie of micromachines Chait carries with him really is amazing—finding a way to make money off of it isn’t as simple as it might seem. They tried energy-harvesters for car tire pressure sensors, but automakers have scaled back on their plans for that. They’ve also made some devilishly complex smart phone antennas, but, because only the highest-end phones need them, that business has proved smaller than expected, says Chait.
But this time Chait thinks Eoplex has hit it big. There’s a type of chip packaging called QFN (for Quad Flat No leads) that’s all the rage for packaging chips for mobile devices. The key to QFNs are the lead-frames, delicate spider-works of metal laid out by the dozens on a rather expensive tape. The lead frames are made by etching away a film of metal from the tape, using various nasty chemicals. What you’re left with is a set of leads that look like q-tips surrounding a central slab of metal that conducts heat away from the chip. The long end of the q-tip is just to anchor it to the frame; it doesn’t have an electrical purpose. In the packaging process the chips are stuck to the lead frame, and delicate wires are connected between the chip and the leads on the tape. That familiar black plastic is then flowed over the chip, and then the frame, chip, and plastic are peeled from the tape. Finally, the packaged chips are diced up, tested, and shipped.
Sounds fine, right? Well there are a few hitches. First, in order for the leads on the lead-frame to be mechanically stable enough, they are all connected along the edges. That means that until the packages are diced up they’re all shorted to each other and they can’t be tested en masse. Second, the dicing itself has to go slower than you might expect in order to avoid smearing the delicate leads and accidentally shorting some. Third, even after dicing, those connection points leave a lot of wasted metal in the package (remember the stick end of the q-tip?), with all the parasitic capacitance and other electrical unwanteds that implies. Fourth, because of those q-tip sticks, you can’t fit more than 2, maybe 3, rows of leads at each edge of the chip. And fifth, did I mention the tape is surprisingly expensive?
So the ideal QFN package would not be made using a chemical etch (so it’s a relatively green process), have no extra metal (so it would have better electrical performance, so chips could be tested before dicing, and so dicing would be quicker), as many rows of leads as you want (so you could get data on and off the chip faster), and no tape (so it would be cheaper). That’s just what Eoplex has been peddling to major chip makers and packaging houses this season, according to Chait.
Instead of starting with a film of metal on tape, Eoplex prints a layer of recyclable steel alloy and then tops it with the electrically important parts of the leads (the tip, not the stick) and the thermal pad—both made of silver.
What’s really cool is that the 3D printing is so finely controlled. In order to make the interface between the silver and the steel strong enough to survive processing, but weak enough so that all the tiny leads stayed with the chip when its time to peel everything away, they had to engineering the interface down to the micrometer-level. Using a proprietary pattern of alternating steel and silver voxels they managed to get the interface just right. I peeled a set myself [see the photo, above] and there wasn’t a single lead missing.
Eoplex has its sights on about $700 million worth of the QFN market, which could grow to $1.3 billion by 2014. By then, he says, the process might be used in other packaging forms such as ball grid array, meaning a $4 billion market.
Although they’re in discussions and testing with some real industry heavyweights, Eoplex is going to need some luck to capture such fortunes. Then again, they’ve already had some. They discovered the secret voxel pattern that’s key to the QFN product by accident. And, according to Chait, that might be good for keeping competitors at bay
“The barrier to entry with luck is a lot higher than with hard work,” he told me. In other words it takes only a few scientists and engineers to stumble on something amazing, while it might take hundreds to engineer it on purpose.