(June 2005) New file server
Just a quick note to say that I'll be trying out a new fileserver for this site soon, so don't be surprised if it goes off-line for a while...
And, amazingly, it seems to have worked. Chalk one up for Linux/SuSe9.3... Now to have a play with PHP, etc ;)
(Dec 2004) The Disco Dance Floor Tile Mystery
After a while even incinerating petrol stations can become a little boring (photo), so tiring of the high life we decided to turn our attentions to illuminating disco dance floors, and other such lighting applications. After developing a range of illuminated tiles and controllers, etc, LedTech (now renamed as Visive) sit back in the comfort of a job well done as the first few installations go without a hitch...
Then, out of the blue, the 'phone rings. Disaster! On one dance floor about fifteen tiles mysteriously stopped working overnight. No apparent cause, so despite both suffering from the lurgy our fearless heros Ian and Richard leap into action and head off down south to inspect the problem... After a night of debauchery in Manchester Crem is unfit to attend, something for which he later thanks the gods...
On arrival and inspection of the tiles the mystery deepens - they appear to have rotted away. Tracks are completely corroded, leaving behind a few browning lumps and yellow stains. (Click on the images to expand.)
Various theories are evolved. Richard pokes at the brown lumps and wonders if they can be the result of some electrochemical corrosion. He suggests they could be oil, sniffs at them but has a cold so doesn't notice anything. Ian wonders if they're rat shit.
Anxious phone calls to Crem establish that he hasn't got a clue either, but favours the rat-shit theory on the grounds that rats are intelligent enough to express their opinions when it comes to flashing dance floors...
Then, out of the blue, a passing workman reveals something that the management of the disco probably wanted to conceal - "Did you know the main sewer backed up last night? Major flood, covered the dance floor with crap..."
"Yep. Totally flooded the cellar, had dyno-rod in with pressure jets to wash it out. Surprised you can't smell it..."
"Oh" [Richard looks at end of finger] [Ian looks at Richard] [Everyone else looks relieved]
Ah, the joys of electronics... So, that's not a warranty repair after all then, is it?
(1 Nov 2004) Lurgy Strikes
Entire factory shut down by mystery virus, urgh.
Picture of Richard suffering can be found (here), still not a pretty sight.
Quite a lot of the last month has been spent here chasing problems with Atmel's Mega AVR processors. We've had a batch of AtMega128's (date-code 0403) where even though they're rated to work at 16MHz over an industrial temperature range (-40C..+85C), about two thirds of the ones we've used fail at below +40C, with approximately 10% failing at room temperature (20C).
The main problem is that the devices cannot correctly read data from the on-chip Flash memory. I've written some test code which can be used to check and detect this problem and it is available from the download page here now.
This problem is being discussed, somewhat inaccurately, on the AVR freaks website: hereWe've had to recall various designs that use AVR's from this batch and test them. So far, out of ~70 tested only ~20 pass at 40C. The ones that fail have to be replaced, and frankly this is a bit of a pain in the arse.
Thankfully we have Herve, who can be relied upon to remove devices with surgical precision and delicacy... Pictures of the artist at work can be found (here)
Stop press: Ze Herve has developed an entirely new technique (tm), which involves heating the PCB flat on the desk and then picking the AVR off the boad by dabbing at it with some sticky-tape on the end of a finger. (lazy bastard).