Found via Slashdot: Tesla rebuts NYT article on Model S range. With detailed data.
I've worked on various projects in the past where having some amount of datalogging built into the product would have helped greatly with diagnosing problems, whether inherent in the product or caused by customer misbehavior.
The last few embedded controllers I've designed have had somewhat minimal event logging: basically, storage for a few hundred abnormal events, with minimal detail, plus cumulative statistics.
One of the current projects is an expansion of this, to storing hundreds of thousands of events, with greatly increased detail - basically, attempting to cover the entire working life of the product (which is likely to be many years).
This capability is somewhat constrained by the need to operate over an extended temperature range, and tolerate severe vibration, etc., so just using an SD card for storage isn't a realistic option. Fortunately, there are now I-temp SPI flash chips of decent capacity available, with rumors of automotive-spec parts in the foreseeable future.
There was a suggestion of including voice recording, to capture the last few seconds of audio before a major fault - "Dmitri! Hold my vodka and watch this!" - but doing that in any useful way would have been too much trouble, for basically a joke capability. (Anyway, the board gets conformal coated, which wouldn't do the microphone any good.)

Hmmm, a NYT reporter with a script (aka, agenda) who won't let the facts interfere with a good story. I blame the Tesla guy for even considering them a real journal anymore. I'm glad he was able to call the journo on his lies and post it all for everyone to see, but he has no one but himself to blame for not researching the journo ahead of time and for trusting the NYT to do honest reporting. I suspect, having said all of that, a significant amount of his market does read that rag and consider it actual reputable, accurate, news reporting. And one does have to go where the business is.
I suspect that putting in decent logging capabilities is easier in an automobile than a small circuit board. I'll be curious to hear how your adventures in logging go with your future products.
Posted by: Marcus Erroneous | Friday, 15 February 2013 at 06:25
Yes, the automobile logger has to be much easier; one can shoehorn in a few cubic inches of highly sophisticated controller and storage. When the entire product is (waves hands madly) maybe 1000 cubic inches and already tightly packed, it's more of a challenge.
The dedicated logging module currently in the works is a 1.75 × 2.75" board, which will be maybe 3/16" thick when assembled (not counting the connector). It could have been smaller, but this is assumed to be Small Enough for the mission at hand, and I tossed in some bonus monitoring inputs and a footprint for a cage-type microSD socket in addition to the big SPI flash chips.
There's another main logic board in the works, and I found enough room for one of the big SPI flash chips on that one, so, once the firmware is developed, it'll have lifetime logging capability.
We're still working out exactly what conditions get recorded, and at what intervals, such that we maximize useful information when a unit is returned while still having 10 power-on years' worth of long-term storage. I suspect there'll be factory-configurable filtering, as well as options for dividing up the available memory among different types of log data.
Posted by: Eric Wilner | Friday, 15 February 2013 at 06:48