-
In no particular order:
Filtering inside has the same effect on through motor traffic, but you move the difficult stuff away from the main/cell boundary street junctions.
Fewer drivers will drive fast through the cell from the other side because the distance they have to drive inside it is increased completely unnecessarily. You need to make the distances that have to be driven inside cells from all boundaries roughly even. My standard example where this has not been done well is in Lower Clapton north of Powerscroft Road--filtered on the north, east, and south sides, so all drivers have to come in from and go out via the west side (Clapton Pond). It's a pretty long cell east-west.
At its worst, placing filters on the cell boundary can cause drivers accessing a part of the cell they have to spend a long time driving through the cell to behave like rat-runners at the (unfiltered) entry junctions, too--turning fast and injudiciously. If you filter inside instead, you get rid of crashes at the junctions almost completely. It's the bad turning manoeuvres that cause most of them, including the vast majority of crashes involving cyclists (left hooks, right turns across the path of a cyclist on the cell boundary street).
Normal-looking junctions with the cell boundary streets that mean drivers on the boundary streets have to drive more cautiously.
More active side street junctions with turning movements (no banned turns at all) mean less motor traffic capacity and less severance (easier to cross) on cell boundary streets. Edge filtering treats the cell boundary as the enemy. It isn't, and the interaction with side streets is vital for good functioning of boundary streets. Traffic engineers are usually instructed to at least maintain motor traffic capacity, and as having fewer active junctions does that, they often favour edge filtering, but this generally results in schemes of poor quality. It can kill a lot of natural functions of street corners stone dead, e.g. behind the guard railing that's often been placed there in the past. The junction dies and all that happens there is through traffic flowing past.
More routes into and out of the cell, greater chance of achieving 'same way in, same way out'.
I did a quick filtering pattern without edge filtering here in the area between Stratford and Leyton:
https://www.lfgss.com/comments/15505024/
It's a point that's surprisingly difficult to get across, and even harder to get people to implement non-edge filtering.
Oliver Schick-
Thanks for the detailed response. I see what you're saying and it includes a few things I hadn't considered.
The main issue I have in this specific example is the lack of turning areas if cells are filtered in the middle. I'm not sure how this is normally avoided.
Also, increased delivery driver movements due to having to enter and re-enter a cell multiple times to access different ends of the street.
These are the two plans being considered although it seems that the clear preference is for edge filtering.
aggi
Interesting you mention this, the proposed Bruce Grove LTN has filters on the edge of cells and some large cells.
What is the downside of filtering on the edge of cells? It seems much more intuitive than filters in the middle of cells.