Don't forget about overload. A semiconductor fails soon after its rating (for your TRIAC, maybe 8-16A for 16ms), whereas a mechanical relay will carry fault current (100s A) and probably still open again. (If the contacts weld, it might not reopen on its own, but a firm jolt, or a little poking with a screwdriver if the contacts are accessible, ought to do.)
The on-off ratio of a mechanical relay is unbeatable.
The risetime (namely for a mercury-wetted reed type) is amazing, fractional nanoseconds. Semiconductors have only been able to beat that relatively recently (since the 70s?).
Contacts are ludicrously nonlinear beasts. It should be no surprise that common relays make so much trouble with RFI and transients.
I don't think anyone's ever constructed a SPICE model of a relay, to a sufficiently accurate degree that it could be recognized as such; that way lies madness (and convergence errors)!
Tim
--
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com
"Julian Barnes" wrote in message
news:noq270$79e$12@dont-email.me...
> In a (very) recent project, I have successfully used a tiny triac[1] in a
> mains voltage switching role. 40 years ago I'd have used a relay for this
> and I can't think of any alternative in those days. Now I've used this
> thing the size of a small pea and it doesn't even have a heatsink tab of
> any kind for mounting yet it can switch up to 600V at 800mA. This to me
> (having been absent from progress for extended periods) is like black
> magic. OK, so the old mechanical relays provided secure galvanic
> isolation which a triac can never approach I'm guessing, but aside from
> that, is there any role for the mechanical relay in modern electronic
> design?
>
>
> [1]http://www.farnell.com/datasheets/84063.pdf