At 06:21 AM 7/31/2002 -0400, Frances Wheeler wrote:
>Question to smartest of smartest designers out there:

She's too busy to read this list, so the rest of us will have to do.

>Copper pours of this size are poured last because they are time
>consuming. The pours can take 4 hours, and even longer if they
>  are not right the first time.  Question to any of the best out
>  there.....can we avoid a copper pour and merge a gnd layer to
>  the top?

Suppose the worst designer reading this list knows how to accomplish the 
task? Should he keep it to himself?

To the question. First of all, the amount of time that a copper pour takes 
depends heavily on the settings. Since this is a pour for a no-signal outer 
layer (a very useful stackup technique, BTW), the pour might be a bit less 
complex than it would be otherwise. I'd use a relatively large pour track, 
perhaps 10 mils or maybe even larger. Pour pitch should be set to zero. 
Protel interprets this as a command to separate the tracks by exactly the 
track width. I.e., a maximum efficiency pour. It is generally not necessary 
to pour in more than one direction.

If one desired pour track to pass between pins with a narrow width, it is 
not necessary to set the entire pour to that width. Instead. Make a 
pass-through pattern for a part and copy that pattern over the part with 
the pour already done. Protel will assign these copied tracks the proper net.

Don't you love it when you ask a question and the answer comes back, "Wrong 
question!"? So, here is an answer to the question itself, just in case that 
pour is just plain too big to be practical:

I was a Tango DOS user, and Tango did not have a pour facility, so I used 
film merge techniques. In fact, I literally wrote the book on this for 
Accel. I'll modify the process to match what I would do with Protel:

First, route the ground (if that's the net being used) with track entirely 
on the layer to be poured. You may be able to use the autorouter to do 
this; since there are no signal tracks on the layer (other than via 
fanouts), the routing is trivial. This track is going to become the ties on 
the thermal reliefs around each pad, so you might want to extend the track 
beyond the pad. I could give a way to speed up this process as well, but 
that is not important in understanding how to make the merge. This process 
will satisfy DRC (with, perhaps, appropriate top layer width settings). It 
will be plotted in the positive.

For the negative, you can't use a Protel inner plane for reasons that have 
already been given: the plane is created based on specified clearance from 
holes, and SMT pads don't even have holes, plus the thermal reliefs, if you 
specify them, will be based on relieving a round hole and pad, the external 
pad shape is ignored. What you want is a plot of the pads and vias, albeit 
expanded by a given clearance. You could use a padmaster plot, but there is 
no provision in such a plot for expansion. You could provide the 
photoplotter people with expanded flash definitions, but I'd rather see all 
of what is necessary incorporated into the board file itself, plus, 
perhaps, plot definitions (in the CAM Manager.)

What works is a solder mask plot, since it is pad-based, and I think 
expansion can be controlled layer-wise. Paste mask would only clear the SMT 
pads. You may need to set up separate design rules for plotting the blowout 
negative and for plotting the actual solder mask negative, and enable the 
appropriate rule before generating the corresponding plot. I don't see a 
way around that, perhaps someone can think of one. Typically, on an SMT 
job, the solder mask settings would be substantially smaller (say, 2 to 4 
mils) than what one would use for a plane blowout. (at least 10 mils)

If you want to add blowout areas to the ground plane, I'd suggest using a 
mechanical layer.

When the films are merged, the sequence of plotting is crucial. The 
negative must be plotted first to subtract copper, then the positive, with 
the connection tracks, must be added to it. I gave simple written 
instructions to the photoplotters and they never got it wrong. (Like Plot 
FileA negative, add Plot B positive.) I don't think that Protel allows us 
to control the plot file names; that is unfortunate, I've seen a number of 
frustrating situations arise out of this. Sure it is easy to rename the 
files, but the goal is a design that, next time, will photoplot the same 
without any special manipulation, i.e., opportunity for error).

CAMtastic, I think, can be used to correctly view such merged plots. I used 
to use Tango itself, which, like Protel, imports its own gerber, to view 
the merges, but using a CAM program is a little better.

Now, there is a complication. What about fanout track or other non-ground 
track on the layer? Using global edits, all non-ground track on the layer 
should be selected and copied to the blowout mech layer and then expanded 
appropriately. This part of the process is not DRCd, so it should be done 
very carefully. However, it is correct by design, i.e., if you do the 
simple process correctly, the non-ground track will all be clear of the plane.

Note that the same process can be used to split the merged plane, with only 
a little extra attention required because the "non-ground track" criterion 
suggested for copying the blowout track to the blowout layer is no longer 
sufficient, each section of the split plane must be separately treated. 
There is a little more room for error here; if one neglects to copy a track 
that was not part of the net for the plane area in which the track is 
located, that track will short to the plane in the final merged plot. So 
one would want to take special care that every such track has been blown 
out. This is not hard to do, however, using selection and viewing 
techniques that are not difficult to work out.

It is very easy to edit a design set up for merge plots, since there is no 
pour to complicate and slow things down. And DRC still checks connectivity 
completely, the only hazard being unwanted shorts to the plane areas, 
which, again, can be visually checked, completely, with ease. (If the 
tracks that must be clear, i.e., blown out, are highlighted and nothing 
else, and the blowout layer is viewed on screen such that it obscures 
anything underneath, any visible selection shows an error.)

Abd ul-Rahman Lomax
PCB design, consulting, and training
Protel EDA license resales
Easthampton, Massachusetts, USA
(413) 527-3881, efax (419) 730-4777

* Tracking #: 64590AB6FE87524E9D92AA3FC50AD5EC47795393

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
* To leave this list visit:
* Contact the list manager:
* Forum Guidelines Rules:
* Browse or Search previous postings:
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to