I wonder why octagonal IC packages in the LQFP times have never been a thing.
In my opinion they would allow easier pcb routing, they take less space and weight for the same die size, shorter leads to the die compared to square packages as well.
Pick and place machines wouldn't have a problem with them, and EDA software is fine with 45° routing anyway, so it seems like a logical thing to do.
Yet I've never seen one.
Any idea why?
I don't think you'd get any advantage. If you look at a square and an octagon with the same perimeter, the octagon does not look great. On top of the that, you normally lose a bit of space at the corners (compare PLCC and TQFP packages) so I think you'd lose a lot with double the corners.
I think it would make sense from a PCB designer point of view, however production wise it is not the best idea.
Comparing a square and octagonal TQFP package with the same area, shows that the ammount of pins stays equal without any problems. And surely, routing a PCB could be a lot easier with this!
But, to provide another viewpoint:
The production of octagonal dies is less efficient than for square dies. So if the chip designer wants to make his chip as small as possible, this is not going to help... The die inside the package would still be square to make production cost effective.
And of course, square dies are easily sawn by vertical and horizontal lines, whereas ocotagonal shapes need another cutting method...
Traditionally new packages are developed for better manufacturing yields or higher density. So if there would be any advantage, it would have to be in that primarily. I don't think that manufacturing yield is a reason.. still the same pin technology, pin pitch etc. Density is pins/area -> so a given package should have more pins on it to make sense.
If you consider a square and an octagon with the same overall width/height (e.g. the octagon fits inside the square), you could calculate the perimeter of both packages. I assume the perimeter as a direct indicator of how many pins can fit on a package.
For a rectangle the perimeter is 2*w + 2*h. For an octagon the perimeter is 8 times the length of a side. The length of a side for an octagon is 0.4142*w. (source: wikipedia).
Substitute x=y=1cm, you get 4*1=4cm perimeter for a square and 8*0.4142*1=3.3136cm for an octagonal. That's 17.16% loss in perimeter length, or 17.16% less pins on the same width/height package.
That makes sense, because you're basically "cutting a corner" 4 times. So I can understand why these packages don't exist.
To add a new perspective consider this: modern day ICs often include a lot of digital circuitry. This circuitry is described in a hardware description language which is converted via a toolchain into standard cells, which are connected by an autorouter. These tools are designed for rectangular structures and the geometry is easier in those cases. Also its easier to cut wafers into rectangular pieces, just think of PCBs they are also mostly rectangular. Of course the chips are (mostly) bounded to the package, but still their geometry and relative size matters (differing angles and lengthes of bonding wires).
Additionally you should consider pins close to a vertex, their internal connection would be pretty close to each other (same for squared packages - you do not have pins close to the vertex) and by using an octagonal package, you would increase the relative part of the circumference blocked due to that issue.
In the end I guess they are more complicated to produce for various reasons and give the user little advantage.
