> curves. The typical is 5mA (what you say you are supplying; 1.0 on the
> normalized curve) the maximum is 10mA, for the MOC3011. That's at 25'C (it
> takes more current at low Ta). That's before aging. Plenty of app notes on
> this,
> you need a *lot* of margin to do this right, probably 15-20mA nominal.
> I didn't work out your resistor value tolerance, Vf range and power supply
> tolerance and Vol range but those are factors too. MOC3012's have 50%
lower
> guaranteed trigger current.


Things you all already knew, (didn't you?) :-)

- You MUST always design using worst case specs for all components and
parameters . (unless you don't care whether the equipment always works).

- Typical values are fine for getting eg probable current drain of a batch
of IC's BUT typical values MUST NEVER be used for design.

- When multiple component and parameters are involved you MUST use the worst
case values of all of them simultaneously when designing (unless you don't
care ... etc).

For instance, as Spehro suggests -

Say you have

- a PORT that has a 4.9 volt typical and 4.5 volt guaranteed high output
level.
- a 470r resistor +/- 10%
- An opto coupler with typical / guaranteed trigger currents of 5 and 8 mA.
- Optocoupler diode voltage of typical guaranteed 2.0/2.5 volts at rated
working current.
- etc , but that will do

(Haven't got to load drive requirements, and more)

Using typical values, opto drive current = V/R
= (Vport - Vopto drop)/R
= (4.9-2)/470 = 6.2 mA
This exceeds the typical 5 mA drive requirement.
It has a safety margin (apparently of 6.2/5 ~= 1.25

The design will work (sometimes)(maybe).

Using worst case values we get -

(4.5-2.5)/(470*1.1) = 3.87 mA

This will not meet the 5 mA typical drive requirement, let alone the worst
case 8 mA worst case requirement.
It fails by a factor of 8/3.87 =~ 2:1

The "real" design and the typical design are different by a factor of about
2 x 1.25 = 2.5 times.

While these figures are cooked it shows how far off the mark one can be with
figures which are apparently OK to use.

NEVER USE TYPICAL FIGURES FOR SPECIFIC DESIGN DECISIONS.
(with the proviso that it may be OK to do so for statistically based aspects
of a design if you understand the implications).



    Russell McMahon

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.