Route while Placing or only afterwards?

[tom66]

Whenever I watch Dave's videos e.g. on the lab supply project I notice he generally tends to "route-while-u-place", which is mostly what I do. But, I talked to an engineer at the company I worked at and he did not like the idea. He seems to claim the one true way is to place everything down (the whole board), with minimised ratlines, then route it. That way, you know it will all fit before you route it. Conversely, my argument is by routing as you place, you will know if you have space for the traces and how your ground fills and power planes are going to work.

Note that the main difference between us was he is a professional electronics engineer mostly working on 4-6 layer BGA boards whereas I am a hobbyist/student and almost always do 2 layer boards because they're cheap to prototype.

Another thing I tend to do, which is a mix of the two methods, is to place a large block of components, fit it on the board, route that portion and repeat until the board is filled. I tend to keep the routing in that particular portion as self contained as possible.

What are the opinions of others? Especially those who work in the industry?

[mikeselectricstuff]

Placement is by far the most important thing to get right - time spent here will be saved while routing. Place all the larger parts and spend time optimising for easy routing. If necessary test-route parts to check density.
Much of the smaller stuff can be placed later as position won't be as critical, although it's a good idea to place decoupling caps on QFPs after placing the part so you don't suddenly find you have no room for them.

[AndyC_772]

I never start routing a board until after it's placed. It's a guaranteed way to waste time re-routing parts of a design over and over again for no real benefit IMHO. By all means place components with due consideration to where critical traces will go, but there's really nothing to be gained by actually putting them in until placement is complete.

With experience, you'll get a feel for how close together components can be placed while leaving room for tracks between them. Usually the primary constraints on a design are the components themselves and the physical outline of the PCB, and that will dictate how densely packed the board ends up being. If there's not much room for tracks then you can always increase the number of layers; on a really dense design it'll be vias, not traces, that you run out of room for.

[Someone]

Routing independent subcircuits such as switching supplies, rf, and precision analog sections ahead of other placement can be useful to maintain their performance but will use more board space in the final design, then move those around as needed to fit amongst the remaining loose parts. Usually not worth it on boards smaller than eurocards.

[marshallh]

I generally make a first-pass best effort on placement, then start routing the most critical (i.e. maybe wide bus fpga > ddr2 DQ banks)
Sometimes there will not be room to break things out and I can move a bit without breaking other things.

Guys who lay out PCBs for breakfast, lunch and dinner will be able to place everything 99-100% right the first time.

[KJDS]

It's been a long long time since I've done a professional layout.* These days I just sit next to a layout guy and we'll run through the placement, I explain the critical bits, hand over the real slartybartfast drawings and the components get placed. Sometimes I then need to go talk to the mechanical engineer and get some more board space, or move some connector positions just to make things fit.

It's very rare that routing will start until all the components are placed. The only time I could see it happening is on a board that had a lot of repeat circuit blocks, eg a sixteen channel mixing desk. Constrain the board space for the sixteen channels, place and fully route one then repeat, otherwise it's always finish the placement before routing.

*I did a few layouts on a contract recently, but these were very very RF heavy boards, RF power transistor test circuits with just the RF impedance transformers and the associated bias circuits. I went through a stage of making them look like space invaders.

[codeboy2k]

I haven't done anything complex in about 10 years now, but when I did, I placed absolutely everything first, and I spent a lot of time here optimizing it... sometimes 3 days up to a week with very little routing being done, if at all; it really depends on the board size and complexity here. Even after that, I generally end up about 80-90% right, and have to move a handful of components around anyways during routing.  Spending the time up front to place efficiently will always route faster with less frustrating times when you have to move a whole section to get something in that you missed..

During the placement process, I often will route parts in my head as I place, so that I can see and get a feel for what the flow be like. This usually works for me, but clearly gets harder on more complex boards, and harder to keep track of everything.

Most of my experience on these larger boards has been years in the past, on multibus , sbus and mbus boards.  I haven't done much of anything complicated and dense lately; all my current projects have been small with a handful of components or simply embedded software development work so I am rusty these days.

[David_AVD]

I often place and route some (mechanically or electrically) critical sections first, then go onto the less important (can go anywhere) bits after.  It depends on how much I'm trying to pack in the given space.  I think everyone develops their own style and as long as it works for you, go for it!

[AlfBaz]

Another thing to watch out for is underestimating space requirements of a groups of traces such as buses. You may leave a corridor and get them to fit only to find they're too close and they couple. Nothing worse than having to rip shit up

[Rerouter]

i prefer to break my layout into modules, collect a functional blocks parts together to work out minimized and easy escape routing of power and signal, route the critical parts of the module then start tying modules together..

generally i will need to play a bit or shift modules a little but i never find myself having to do major re-route (even if the name may suggest otherwise  )

[AlfBaz]

High pin density parts with heavy power supply decoupling requirements are always a pain for me. Place the decoupling caps close and getting your signals out becomes difficult, get the signals out and then your caps are miles away from the chip

[Dave]

Most of the time, I roughly position the components and begin routing. I then make minor changes to the positions and orientations of the components as I route. It has worked for me so far.