>>>> In practice you get the superb CES2310 which is not 
>>>> sold by
>>>> is in a SOT23 package

>>>> 30V
>>>> 4.8A continuous.
>>>> 1.25W

>> Really? 1.25W in an SOT23 pkg?  For how long?

> Less time than it takes to read this message?

Well, maybe a little longer.
That answer is (almost) funny but not as educational as it 
could be.

That figure is an obvious extrapolation of the other 
available data and like many other specs needs to be 
understood in context. And questioned carefully (like any 
other spec) if you intend to rely on it.

Working backwards, they claim 100 C/Watt thermal resistance 
junction-ambient on fr4 < 5 seconds.
At 1.25 W that's a 125 C rise.
Note that Rth is very much manufacture dependant and can 
vary widely with practices. Some manufacturers make an art 
form out of coaxing extra performance out of packages which 
are apparently similar to other low performers. An example 
is Zetex's E-line TO92 like devices which claim a 1.25W Pd, 
achieved, as well as may be, by an unusual leadframe design 
and attention to packaging minimisation.

Tj operating is 150C (which value you don't want to use if 
you value longevity) so at 25 C ambient and the 100m thick 
piece of copper from another recent post you can run this 
baby at 1.25 Watt. The 5 seconds they mention is far (far 
far) longer than the junction thermal capacity alloweth so 
they are really saying that a large copper PCB area will 
allow you to keep the junction inside bounds at (say) 25 C 
ambient at 1.25 Watts for 5 seconds.

This tells you that, assuming they know their stuff and are 
telling you the dinkum good oil that the bond wires and 
general construction is up to this level of performance and 
that getting the heat out [tm] is the main factor.

I, like BB, would look askance at any attempts to run the 
device at this level for "any time" [tm]. However, given 
what I know of it otherwise, the guys (and or gals) who 
built it know their stuff.

This is by far the best spec I could find in an 
available-in-China MOSFET in this general class apart from 
those from Zetex, who I was sure would rise to the 
challenge, but would charge a fair price. As I didn't want 
to pay a fair price I didn't consider Zetex.

In operation at somewhere under 500 mA peak and a gate drive 
in the 2,5V - 3 V range and several 100 kHz switching and 
large on cycle % the CES2310 has a real world Rdson of as 
close to zero as matters. The SOT23 on a scrap of vero with 
attached leads smiles gently at the load.

At rated continuous current of 4.8A and Vgs of 10V and Rdson 
claimed of 34 milliohm (no footnotes except "typical, not 
production tested))  you'd have a power dissipation of under 
0.8 Watts. So, while real-world Rdson, footnotes or no, will 
be greater than claimed, it's probably [tm] possible to run 
this FET at full rated current, given maximum gate drive, at 
OK ambient temperatures on a semi infinite PCB heatsink.

Nobody would be advised to try this in production equipment 
(but he will do it anyway, that being how nobody behaves) 
BUT the point is that the spec, like most of the others does 
appear to represent a data point on the outside edge of the 
reality envelope, and not one far outside it, as may be 
thought at first glance.

An important lesson is that specs must be understood, read 
in context and conjunction with other specs and even then 
taken with care and sometimes a grain of salt, even when 
true.

I assume that BB will agree.


        Russell







-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist