Can you explain why? I am not trying to start a fight - I actually just don't understand what will be gained by filling vias.

I know for a fact that the slight convexity or unevenness resulting from it will most likely ruin your alignment, and make the BGA hard to position. And you will never know for certain if some balls didn't seat right.

You do not gain anything in routing density. So what's the draw?

EDIT: if you use a hotplate, heat will transfer directly through the via faster than other, un-via'd pads, resulting in uneven heating...

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut

Consider me a newbie at @ BGA SMT. As I said before, I would like direct powers/ground pins to go direct to the plane. This is just the start...

Yes, you had stated this very excellent observation before, hence my statement about "sanding down the board with the finest grit sandpaper"

This is the whole point, I would think you do gain more maneuvering around the power and ground pins if they are directly under the BGA pins, without the branch in the middle of four pins.

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

I still don't think you will gain anything in terms of density. Regardless of the via position - centergrid or offset, you can still route only one wire between the vias. Connecting power vias 0.3mm away from the pad does not present problems. Sanding the board will remove the soldermask which is necessary both for easy placing, providing some slack in positioning by guiding melting balls to the center of the pad, and avoiding shorts.

I consider myself a rebel, but the idea of soldering on hand-filled vias after sanding, scares the pants off me. Do let us know how it goes - who knows, I've been very wrong before.

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut

Well I won't continue to argue with the one who inspired me to finally dive into BGA! I'm sure I have more to learn more so than you. But one thing: don't you see if the power via's were direct underneath the FPGA pins, there would be more routing ability for the surrounding signal pins?

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

No arguing here. I don't want to discourage in any way. Try it by all means.

It's hard for me to remember the details, but a couple of years back when I started on the BGA kick I was appalled at the dogbone patterns and the offset vias. I remember wanting to place vias under the pins myself thinking it would help.

To answer your question: consider a real breakout pattern, like the one I posted on the previous page. Note all the vias. Now, in your mind, move all of them a little down and to the right. Do you gain anything from that exercise? No - a 1mm grid is a 1mm grid, whether it's under the pad or in between the pads. You are still limited to 1 wire between pads/vias no matter where.

Consider the best-case breakout pattern. The first row goes straight out. The second row sneaks between the first row pads. The third row has nowhere to go but down and out. In your scenario, it goes straight down with pad/via combo. The fourth row sneaks in-between. That's it! There is no way to route anything else! In the normal scenario, third row goes to the side to vias and down and out. Fourth row goes to the side, down and out in between. Four rows is all you get with 2 layers. Either way. Nothing gained except a lot of problems.

Anyway, sorry to be belligerent. I hope you don't really consider this an argument. I have a lot of respect for you and your work, and feel very protective. But do try it - maybe you are thinking of something I am missing!

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut

I hear you. Consider this a discussion then, no arguing I'm looking at a 16x16 1mm matrix of a 256-pin Spartan 6 BGA, most of the power pins are in the center of this device. I need to get my screen capture utility back on my computer! I made the initial footprint for the XC6SLX25-3FTG256C

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

Although the pattern will not be significantly affected whether the via is in the pad or not, a significant issue that's not been discussed is how to prevent the wicking of the solder ball down through the via. It may be possible to fill the via, use a smaller drill size, etc. to prevent this from happening. Most of these solutions just cost more.

For small BGA packages with pitch size less than 1mm, it may not be possible to avoid the via in pad technique. I've consciously avoided the via in pad technique except for some packages where the vias (in a pad) connect a ground/power pad into a power plane to improve heat transfer.

I have noticed that many 0.8mm pitch micro BGA (uBGA) packages try to alleviate this problem by putting signal pads in strategic locations so the fanning them out is possible without resorting to a via in pad approach. Many of these packages do not use a full grid like the 256 and 484 BGA packages. This frees up space for signal fanout in the interior of the part.

_________________
Michael A.

Here is my old BGA256 test board (do not use this layout directly, there are errors). I didn't bother breaking out more than 2 rows, as I was constrained by board size already. Note the brown power distribution ring on the back. If you needed to break out 2 more rows, you could easily make the ring smaller and closer to the center, and have spokes lead to the power pins from this ring. It all depends on where the pins are, but you probably get the idea.

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut

You don't have to worry about melting points and gravity. The capillary action is much stronger than gravity, and it's okay if the solder melts again when soldering down the BGA.

Unless your PCB manufacturer allows 0.1 mm trace/clearances, so your third row can sneak past the second. BTW: this is a 1 mm pitch BGA.

I've received all the parts for my hot plate construct. I'm waiting on a local machine shop to drill the 6" deep 3/8" holes for the heater elements. I've measured the width of the elements @9.45mm. While I could've gone for a 10mm hole size to accommodate thermal expansion, it would've left .5mm of air gap and non-ideal heat transfer. I chose a 3/8"(~9.53mm) hole for each element.
Tomorrow I'll get the aluminum block back. Then I'll have to drill and tap holes for the TC and the 4 'legs'. BTW, the TC will not be deep in the center of the block, it only protrudes about 1/2". I'll mount it on an outer edge close to the top.
My first victim will be a rev. h Parallel Video Board, where there is 100-pin, 144-pin, and 48-pin .5mm QFP packages.

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

I don't think you'd have to worry about thermal expansion. The block expands just as much as the heater elements.

One thing I do worry about, since I like to build things that last, is the issue of dissimilar metals. The element shells are stainless steel, and the block is 6061 aluminum. Over time, the heating elements may fuse to the block which would not be desireable if one of them needed to be replaced. I was thinking of a thin coating of thermal paste, like the kind used when mounting a heat sink on an IC. But with .08mm gap, I don't think this type of thick coating will work.



EDIT: .08mm, not .08"!

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

We did exactly what you're talking about regularly at a place where I worked in the mid-1980's, although I myself never had to remove them, so I don't know if there was any issue of them being stuck and needing replacement and not wanting to come back out after the goop had hardened or turned chalky. It might be a good idea to make the hole extend (maybe at a much smaller diameter) all the way to the opposite side of the block so there's a way to push the heaters out in addition to pulling from their wires, if they're the ones I'm thinking of. OTOH, I'm sure you won't put anywhere near as many hours a hear on them as we did, and they probably won't burn out and need replacement. They will probably be more likely to burn out if you run them without the goop so they have to get a lot hotter to transfer the same amount of heat across the gap.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Thank you for reminding me! My holes are a tiny bit loose from multiple drill passes (my little drill press does not allow for enough travel). I was planning on using paste, but completely forgot. I thought about wrapping aluminum foil to make the elements fit tighter, but grease is probably better.

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut