paper rocks

For doing stuff like that, I configure sets of layers relevant to each combination. So, "Top Components" = top layer (and solder and silk) and inner layer 1 (usually GND). "Bottom Components" = bottom layer (and stuff) and inner layer (N-2) (VCC on a 4 layer board, GND or whatever with more layers).

For inner signal layers, I grab those, plus the adjacent plane layers. Just whatever's facing.

I know PADS has layer settings, although I don't recall if you can configure sets, or if you're eternally ticking layers on and off one at a time. (In Altium, it's two clicks.)

If you have big holes in a plane layer (so a signal layer "sees through" to the ones below), it would be reasonable to include the "visible" layer in one of these sets. It would also be preferable to fill in, or patch over, that hole, rather than leave it blank, of course.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Ages ago, when PCBs were TAPED, copies, etc were made using ColorKey (3M i believe). ColorKey came in many colors and was transparent; perhaps that could be used instead of paper - allowing one to see more layers.

you can also print to the "trasparent paper" (sorry don't know the name in english) that was used for presentations before Powerpoint.

Bye Jack

Color seems to "hide" some errors from me. I make a monochrome or grayscale PDF and things that need fixing jump out at me.

snipped-for-privacy@gmail.com wrote on 6/8/2017 8:00 AM:

You mean Viewgraphs? I still call them that. I guess Powerpoint is one brand name substituted for another.

--

Rick C

The power pours are colored by net name/voltage. That helps me see which ones need to be connected. I prefer to connect everything with one solid pour, but sometimes you need to run a fat trace to connect separated patches of power pour. There's some famous map-making theorem that limits how many countries can be connected on a map or something.

PCB layout is astoundingly complex. Initial parts placement dominates subsequantial signal and power routing and thermals. It's more fun if you have fast controlled-impedance single and differential signals on multiple layers with termination and crosstalk concerns. If I were in prison or something, and had nothing else to do, I could easily spend a couple of years optimizing this one. Maybe even get it down to 8 layers. I'd need a good color printer in my cell. Solitary would be best.

After I change the shape of a pour, I have to run the PADS pour manager to redraw it all, and that takes a most of a minute. Then I run the connectivity and clearance checks, which are fairly fast, to see what I may have broken. I must have done that loop a couple of hundred times by now on this board.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

"Or something"??? You are thinking of how many colors are needed so no two countries that share a border (other than the point of a vertex) will have the same color. The answer is four. The point is this is independent of shape of the countries. It is essentially a topological math problem unlike the issues involved in making good power plane pours.

I remember dealing with a layout specialist who had learned electronic issues as rules of thumb. One was to never make a power connection long and skinny. My design had a power sense circuit which was fed through a series resistor. This circuit was on the other end of the board from the power pour area and I wanted to connect it via a long thin trace like any other logical trace. He vehemently objected explaining how bad it was to do that. I realized there was no point in trying to reeducate him so I had to think for a minute until I realized it would not violate his rules to move the damn resistor over to the power pour and run *that* trace long and skinny. He didn't like it, but he couldn't object. In the end we worked well together because I would cooperate with him rather than just give orders.

--

Rick C

I prefer wysiwyg prints of gerbers, scaled up, if necessary.

Needs to be actual (all) copper representation, which, strangely enough, is extremely difficult for some CAD programs to produce. It spends much rendering images on the screen - none of which are real dimensions. Something to do with the width of a line....

....which explains to me why so many artifacts show up unexpectedly, on random layers, when checking.

RL

RL

IMO the color rendering of layout software leaves things to be desired. Some packages can't do transparent at all (I never understood why not), others are ok but it becomes murky when there are more than five layers turned on. Why can't there be a "space hovercraft" function that lets one "fly" through the layers in 3D? That would also be fun during a coffee break, doing high-speed fly-throughs, maybe with a crash-and-burn feature if you bang into a via.

However, we shall be thankful for CAD and modern printers. I remember how it was during my dad's time. A huge layout was on transparent film on the ground, lit from below, people had to remove their shoes and then crawl across it to discuss and make changes. One had to make sure to wear socks without holes in them.

--
Regards, Joerg 

http://www.analogconsultants.com/

Printing out layers of a CAD drawing would lose so much important info for me that I never do that. Also, I don't own a printer, hate the paper stuff. Paper is only good as food for worms, really great, rich compost.

I use the screen because it is full of much more info than you can get on paper. I never work on large boards, so the image on the screen, even zoomed out fully is still a 3, 4 or 5x of a real board. I just can't see the advantage of printing something like a board layout when the screen is so useful.

I totally agree that a lot of improvement could be made to the rendering of layers. Even something as simple as grouping layers for changing views would be enormously useful. A fly through could be useful, but I think other things that are easier to implement could be more useful.

--

Rick C

Friday I was glad we still have a non-electronic newspaper. The blow-off hose on my fermenter for Belgian Tripel plugged up and phhhsssseeeee ... POOF. So I stuffed crumpled newspaper around it until I could fashion this:

I never built up any piece of plumbing this fast.

Depends on how complex and tight things are. My first hybrid was so dense that the CAD was not allowing most positionings because it signaled interference all the time. So we printed and cut out all models for parts and shoved them around by hand. Big boss came in. "Hey, this is not kindergarten!" :-)

The mechanical guys have that. Just imagine you could do this in a layout:

--
Regards, Joerg 

http://www.analogconsultants.com/

Uh, you are aware this is not real time video, right? They make these images one frame at a time offline and then put them together to make a video. At least I've never seen a system that could render a scene this complex in real time.

I worked for a company (for a short while until they ran out of money) that in the early 80's had a machine that would let you render a wire frame image (like the space shuttle which is much less complex than this pumping station which uses full shading) and rotate it in 3D in real time from a joy stick. All discrete logic TTL at that time. We were working on an ECL version when they pulled the plug. It was a small startup funding by the 20 year old lead engineer's father... until he couldn't get venture capital and the 8 or so employees (some less than a month like me) were laid off. Goofy as hell. Why hire if you may be laying off in a month!???

--

Rick C

Many CAD tools today support 3D models at least, and customizable design rules in case that's not good enough.

Altium can do that; it's had that for years, as evidenced by Dave's old intro:

You place keyframes with time durations between each, then hit playback, then render into a video format.

You can get a proper 3D view of the stackup or routing as well (though it's finicky because, imagine this: the board is kind of thin to fly into..).

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

I have seen this sort of fly-through live in a 3D CAD. You do need an Intel i7 or similar horsepower processor and/or a good graphics chip. SolidWorsk can do it, for example:

We have used it during online conferences to dive into innards of systems together, without being at the same location. My job was to watch out for and point out EMC concerns.

That's a tad different these days :-)

In the world of start-ups many folks go pedal-to-the-metal all the way to the end. Even if the painful end is in plain sight. Like approaching a red traffic light at 100mph.

--
Regards, Joerg 

http://www.analogconsultants.com/

I used to do marine automation, control systems for ships, so spent some time in shipyards and out on sea trials. Avondale had a "lofting room" where metal sheet patterns were drawn 1:1, paper on the floor, and of course people walked on the drawings. A robotic follower thing would cruise the inked lines, and in another building a flame cutter would slice out the pieces from a big sheet of steel. A bunch of such pieces would be welded together to make, for example, the sorta spherical underwater bulbous bow.

Pretty cool for 1970's tech.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

A Gerber file makes copper pours by scanning a small dot onto film. The tighter the features (real tight for a BGA) the smaller the required dot, and the bigger (2nd power) the Gerber file. When I do a layout, I draw the pour outlines, and I have to run the "pour manager" to see the fill, essentially computing the Gerber scan. That's slow, and I can't move that scan; I have to turn off the fill and go back to the pour outline, to move anything. It's hard to visualize the actual pour from the outlines (impossible, actually) so tweaking the pours is a slow iterative process. I don't know if other PCB CAD programs work the same way.

The pour manager is slow but very intelligent, providing clearances from other nets, filling nooks and crannies, pretty well resolving pour overlap conflicts. But you've got to check it carefully.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

It's a good low-tech solution; looking at screens for too long is fatiguing, and you can easily fatigue yourself right into making mistakes.

There's some site for a C/C++ static-code-analysis tool where they've shown it's output for some of the "Big Name" open source software codebases that they've run through it. An large number of errors it detects they mention were likely due to fatigue, stuff like (just an example:

for (size_t i=0; i < MAX_SIZE; ++i) { big_array[i + 1] = other_array[i]; big_array[i + 2] = other_array[i + 1]; big array[i + 3] = other_array[i + 2]; ... big_array[i + 32] = other_array[i + 31]; big_array[i + 1] = other_array[i]; big_array[i + 1] = other_array[i]; big_array[i + 1] = other_array[i]; etc. }

where they just copied and pasted and then got tired and forgot to fill out the pattern.

I print pieces of layouts to PDFs so I can mark them up for the layout engineer. Sometimes I print the PDFs to paper so I can review them with him but usually not. Printing the entire board would be useless, if it could be done.

I always wondered why I couldn't define "views" (groups of layers). It would be trivial to do. I'm always turning on and off layers to find some wire. The flying thing would be great. Spotting the reference plane would be easy trivial.

I sure have. There isn't a huge amount of data there.

The fastest computers of the '80s were nothing like what is commonly available for laptops, today.