Black box flight recorder - location signal?

Maybe for little chips. It's the CTE mismatch that has set the 2-cm die size limit for yonks.

Cheers

Phil Hobbs

All of the hordes of talking monk... er, heads discussing the loss of this plane seem to assume that the "ping" will travel directly from the beacon to the receiving equipment, and that the received signal's bearing is a direct pointer to the beacon.

Now, I know little about real-life SONAR, but it seems like every book or movie about submarines has one hiding from someone else behind a "thermal incline" or something similar which makes two-way SONAR detection difficult or impossible. If the plane is underwater, aren't there thermal effects and currents which might refract the signal?

If so, could their effects be large enough to throw off the searchers?

Jes' curious.

Frank McKenney

In practice all you can really hope to do is head in the direction of increasing intensity and cross your fingers. Once you have a couple of ships in the area getting a signal then you might do a bit better.

Just staying in the zone that can hear the signal seems to be a challenge although they have acquired it for a couple of hours.

Yes. You can even have what are in effect acoustic mirrors if there are suitable temperature inversions curving the sound down again.

A related problem in the air sometimes focusses wind turbine noise down onto a few unlucky folk because of wind shear in the boundary layer.

I expect in 4km of water against a rugged sea bed you could get all sorts of unhelpful focussing effects and multipath echoes.

Please kill file me. I'd appreciate it.

I'm not sure what CTE has to do with it. There are die attached schemes that allow all sorts of expansion. Glass bead or silver glass die attach for instance. If you could make the chip, the package engineers will solve the problem of how to package it.

Note the effect of wafer defects depends on the current flow of the device. Lateral current flow devices are more tolerant than vertical current flow devices to wafer defects. That is one of the reasons that you don't find LSI bipolar chips, even if heat wasn't a problem. Trilogy was doomed from the start, but the venture capitalist don't know semiconductors. ;-) VCs want moon shots, not more of the same old tech. But there are real life barriers.

Now if you invented some cheap laser annealing system to "heal" the wafer, then maybe chip size could increase.

I used to work in the packaging department at IBM Research. Even with the best available underfills, temperature gradients and CTE mismatch led to reliability problems with the corner C4s on chips larger than 15 mm, and 20 mm was the outer limit. It was a bit easier in the PbSn days.

IIRC memory chips usually have redundant rows so that dead ones can be turned off without hurting the yield. Multicore processors are much the same--you can get 8-core or 12-core versions, and it's the same silicon in many cases. The only difference is that cores that flunk get turned off.

Cheers

Phil Hobbs

There are SCRs and diodes the size of your head. As I recall, the hockey puck construction goes something like, tinned copper end plates, molybdenum bonding pads, silicon die/wafer, ceramic surround.

SCRs and diodes, of course, being much more tolerant of things than fine scale VLSI. I'm not sure if they're making IGBTs that big yet; even the "smaller" ~1000A modules usually use smaller dies bonded together, though that may just be for the "high speed" ones.

Tim

I don't deal with those devices directly, but of course, we all have the interwebs. Going though the ABB website, there are devices in the 10mm to

I'd say 20mm on a side is a big ass chip. Doing a quick check, a haswell is about 13mm on a side.(well if were square)

Redundancy is one way to get around yield issues. Those RAMs used to have laser fuses. I haven't kept track of that technology. But I was at a static RAM company and back in the day they would test the wafers, map out the repairs on a per water basis, zap the fuses, then retest the wafers. The zapper did no testing itself. It just read the plan off a disk.

The wiki has a line that could use more explanation: "Amdahl eventually declared the idea would only work with a 99.99% yield," which wouldn't happen for 100 years.

The wiki says Trilogy was going for 2.5 inches on a side. Yikes!

So $230 million in 1980. If I used the CPI table correctly, that is a factor of 3 in today's dollars. So that would be $690 million. That is a lot of VC, but not Facebook scale.

Not sure if anyone else has posted it.....

I found a video of a black box pinger teardown:

Jump to about the 3:00 mark for the pinger

They are way ahead of you. The current technology is the SSFDR (solid state flight data recorder).

These use flash memory. I wouldn't be surprised if it contains a generic, large sized SD memory card, ruggedized, but readable using the standard SD interface and file system.

On a sunny day (Sat, 12 Apr 2014 08:45:08 -0700) it happened "Paul Hovnanian P.E." wrote in :

Yep, nice work!