Be careful handling computer racks

[Tony Duell]

> 
> Tony Duell wrote:
> > Off-topic, but modern consumer electronics (particularly the cheap 'own 
> > brands' tends to be assembled in thin sheet-metal cases with edges that 
> > will give you a nasty-ish cut if you catch them. 
> 
> I've stopped using the really cheap PC cases. The Antec 300 I'm using 

I wasn;'t just thinking of comptuers. Things like VCRs, DVD players, etc, 
have nasty sheet metacl covers now that draw blood. 

> now seems to have had the metal edges rolled over at least once, which 
> pretty much leaves it with smooth rounded edges that have the cutting 
> ability of a wet paper towel.
> 
> > Modern components shouldn't get hot enough to burn you :-).
> 
> Exceptions: power transistors (esp. BJTs). SCRs. Big diodes. Well, just 
> about anything that's handling more than about ~10W of power really...

Hmm.. I prefer it if even those are cool enough to touch without a nasty 
burn. I always feel that too-hot == unreliable.

A diode should be dropping 0.7V or so. For that to dissipate 10W, it must 
be carrying about 14A. That's not small. Ditto for thyristors. 

> 
> That said an un-heatsinked 7805 running at 1A output with a 12V input 
> will run surprisingly hot, too, and that's only 7W. I've seen 7805s hit

True. The only ime I run a 7805 without a heatsink is when a 78L05 would 
have done :-), but I'm using the larger-cuirrent version becasue I had 
one in the junk box.

>  > Glassfets are another matter.
> 
> Mmm, vacuum tubes. "After turning off main power, wait 15 minutes for 
> valves to cool."

That sounds about right. Big output tetrods/pentodes get very hot...

> 
>  > And an IC that has developped internal shorts can get hot
> > enough to burn you, some DRAMs were prone to this. 
> 
> A few old 1k*4 SRAMs did this too. 2114 anyone?
> Those things were nasty little buggers. If it wasn't a bad bit, it was a 
> dead bitline, address line or just a dead chip shorting the power bus. 
> Must be one of the least reliable SRAM chips ever made.

Oh, you've noticed that too. I don't like shotgu debugging as you know, 
but if I see 2114s in a device and have unexplained problems, I often 
replace them with known-good ones just to see. And most of the time that 
cures the problems.

My HP9836CU would randomly fail to detect the intenral FDC card during 
the power-on tests. It turned out (and this is not in any manual that I 
can find) that the test for the presense of an FDC was to read and write 
the sector buffer RAM on the board. And this RAM was a pair of 2114s (it 
could also be 2112s, in the former case only 1/4 of each chip is used). 
Guess what...

> 
> > Yes, I was thinking of bruns from the soldering iron, or something heated 
> > by it Most of the time it's when I am soldering a piece of wire to a 
> > large-ish metal object (pin of a 4mm plug) and I don't let it cool for 
> > long enough before taking it out of the vice.
> 
> I seem to do that with startling regularity... Somehow the concept of 
> objects touched by soldering irons also being hot hasn't quite sunk in yet.

It's suprising how long a metla pin will stay hot after soldering.

-tony