My responses are below, preceeded by ***

JaMi Smith
Optical Crossing Inc.


-----Original Message-----
From: Abd ul-Rahman Lomax [mailto:[EMAIL PROTECTED]] 
Sent: Wednesday, January 23, 2002 9:57 AM
To: Protel EDA Forum
Cc: JaMi Smith
Subject: Re: [PEDA] Looking for a resistor supplier.

At 06:04 PM 1/22/2002 -0800, JaMi Smith wrote:
>Now you really are looking for problems in your design!
>
>You should NEVER include solder plate in your CURRENT calculations

[...]

this is correct, but I'm not quite sure why the comment about "now you
are 
really looking for problems...."

No one suggested considering solder plating, as far as I noticed.
Certainly 
I didn't. However, there is no harm considering solder plating, provided

that one uses the correct values for its resistance. I'd expect,
however, 
that it would be not worth the effort.

*** I included it as a conversation starter, because, yes, you did
infact include it in your comment, almost it seemed to me as if you had
not realized it, which is exactly why it should be discussed in this
forum, since it is something that so very often is overlooked.

>,
>since
>Standard 60/40 solder only carries about 16% the amount of current as
>copper, and that 16% is a very rough number that actually varies quit a
>bit itself.

Sure, but there are also plenty of other variables; if one is making an 
estimate, one might as well toss in a correction for solder plating, if 
there is to be any. Either (a) the resistance will be trimmed anyway or
(b) 
there will be software compensation so the actual value is not critical.
In 
this latter case, one is simply aiming for a reasonable value.

*** #1, an unintended error of 8% far exceeds what any good designer
wants to throw into any equation. #2, there is no known correction for
an unknown variable, i.e.: I obtained the 16% number years ago from a
scientist at Multicore, and he said that it was good for a specific
composition of solder only, and that other solder compositions could
vary up to an additional 5%. ALTHOUGH REALISTIC, THE 16% NUMBER IS ONLY
APPROXIMATE AND SERVES TO SHOW EXACTLY WHY THE SOLDER SHOULD BE EXCLUDED
FROM THE EQUATION - PRECISELY BECAUSE IT IS SO MUCH LESS THAN COPPER AND
ALSO BECAUSE IT IS SO VARIABLE BY ITSELF. #3, Why on earth do you want
to complicate the problem by adding a process that is the equivalent to
an old SELECT IN TEST procedure to an otherwise simple design that you
are going to produce 100 to 200 of (if I remember correctly), a
procedure which by itself would cost as much in technician time man
hours as buying a whole reel of 1800 parts for each ot the 100 to 200
boards would.


However, the solder plating will not make a huge difference, so an
urgent 
warning to avoid including it is a bit overstated.

*** Assuming 1/2 ounce copper and the same additional thickness in
solder, would yield you a home made "sense resistor" with the value only
about 58% of what you would have thought you would have had if you
included the solder in the equation and calculated for a total thickness
of 1 oz. "a bit overstated" ??????????? I don't think you quite
understand the scope of the problem . . .

>Virtually ALL of the CURRENT calculators or tables or graphs which
>determine the amount temperature rise above ambient for a given width
>and thickness of copper trace, are ALL based on the old MIL STD 275
>Graphs which specifically tell you NOT to include the solder thickness
>in the calculations for precisely this reason.

I would not use the term "reason" in respect to those old tables. I've 
never been convinced that they were soundly based, I've never seen 
experimental verification. Are the tables conservative (i.e., actual 
temperature rise will be expected to be *lower* than stated) or are they

predictive (i.e., one would expect experimental evidence of temperature 
rise to be approximately what is given in the tables)?

*** You are obviously not either an electronics technician nor an
engineer. I let my commenting stop at that poinr, and simply say that
you should ask an engineer about the problem.

>CALCULATE THE COPPER THICKNESS ONLY!
>
>Any calculations to determine the drop in milliohms for a given width,
>thickness, and length of copper trace, should likewise not include
>solder, for precisely the same reason.

No, this is not true. One should not use solder plating thickness for
the 
temperature rise tables because they were designed not to use it. There
is 
*no* harm in compensating for solder plating thickness; however, one
should 
know what variables are involved: reistance variations based on mixture,

process variations in thickness, edge effects, etc. In fact, I'd
recommend 
at least looking at the effect of the solder plating, because it *will* 
affect the final resistance, the only question is how much it will
affect it.

*** #1, Wrong! Go back and read NOTE 7 on the old MIL STD 275 "charts"!
#2 If you could accurately determine what the exact characteristics of
the specific solder that you were using you might begin to have an
argueable point, except for 2 problems, the first of which is that you
are only going to gain an additional 16% (approximate) in current
carrying capacity, and you are forgetting the most important thing about
solder, which is, that it melts, and that when the current gets high and
the solder gets hot then at some point it will melt and reflow , which
means that your calculated extra current carrying capacity just flowed
away . . . . . #3, You have obviously never seen a copper trace get hot
enough to burn off the solder plating on it, have you? i.e.: under
certain circumstances your extra solder plating is going to go up in
smoke, and when it does, I hope you weren't really counting on the
solder to carry any extra current, because when the solder is gone, then
your trace will most certainly become a fuse.

More to the point, one may wish to make a board depending on trace 
resistance SMOBC (solder mask over bare copper), because the copper 
thickness and plating may be better controlled. Further, wider/longer 
traces would be preferred to thinner/shorter traces, because the process

variations in trace width would have less of an effect.

*** So . . . What has this got to do with whether or not to include
solder in your current calculations?

>In either case, you should not include solder plate in either the
>calculations or the design itself. In other words, any solder plating
>will screw up the actual values. This means that for this type of
design
>you would need to soldermask the specific trace in question, or
>otherwise prevent it from being plated with solder.

Obviously an unmasked trace, covered with solder, which might be quite 
thick, would be unsuitable for use as a controlled-resistance trace. But

the norm would be that such a trace would be solder-masked anyway. Once 
again, the effect of having solder on the pads will be small if the
trace 
is relatively long.

*** So . . .

If this is a board which is going to be produced in very small
quantities, 
it may be more cost-effective to use a commercial part that would not
need 
trimming, etc. Even in larger quantities, if trimming can be avoided by 
using a commercial resistor, and the resistor is inexpensive enough, it 
might be cheaper. Note, however, that simulating a Kelvin resistor on a 
circuit board would be easy, if the application requires that kind of
accuracy.

*** The part cost $11.80 for 10 from Digikey, that is monumentally
cheaper than any of the other processes that you have discussed, and
exponentially less complicated.


For lowest cost, I'd go for software compensation....

*** Duh! . . . are you now going to include a micro in this design just
to compensate for all of your other "design features" that you would now
have us incorporate, or were you instead going to subject each bare
board to a complex and expensive analysis performed by some super
automated testing system which needed some test engineer to operate it
and a week to program it. Hello! 

And for accuracy, one could include a temperature sensor sitting on the 
trace and temperature compensate as well. Or a test current periodically

injected might be even better, it would deal with all kinds of
variations 
in one fell swoop.

*** Another $10.00 circuit just to watch the extras that you are going
to add to the original design, and gee, a precision thermistor to boot,
or maybe we could just use another "current sense" circuit to sense the
current in the original "current sense" circuit, but then again, that
might need another "current sense" resistor.

*** The amount of time that has been wasted in this exchange is
phenomenal, but I have persisted for the very same reason that I wrote
the first response, which is that I believe that there are many people
on this list that did not know that they should NEVER INCLUDE THE SOLDER
PLATING IN THE CURRENT CALCULATIONS. After this exchange, they will
never forget it.

*** The bottom line is this, don't count the solder, and the minor
benefit that you will gain from it will always be in your favor.

*** For those of you that have the capability, go out to the lab and
take a well protected variable supply and run some current thru a piece
of solder, and see what it will do compared to a nice copper conductor
of the same size. 

*** Knowledge, Logic, and Reason, have no substitutes. js


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to