Diverging Diamonds

The study of British and Irish roads - their construction, numbering, history, mapping, past and future official roads proposals and general roads musings.

There is a separate forum for Street Furniture (traffic lights, street lights, road signs etc).

Registered users get access to other forums including discussions about other forms of transport, driving, fantasy roads and wishlists, and roads quizzes.

Moderator: Site Management Team

Post Reply
User avatar
Stevie D
Member
Posts: 8000
Joined: Wed Dec 20, 2006 17:19
Location: Yorkshire

Re: Diverging Diamonds

Post by Stevie D »

Mark Hewitt wrote: Mon Aug 13, 2018 15:04Looking at the approaches. Should the rightmost lane be arrowed as ahead and right, given that you need to move into the lane to the right? I guess it's possible to do a u-turn here, without encountering any lights, but traffic also merging to turn right may not be expecting that. In fact that quite short weaving space between traffic turning right off the mainline and traffic turning right onto the mainline may be an issue?
I would not permit U-turns on the side road at a DDI. It would require a very sharp turn (typically an angle of at least 120° or maybe more), which would mean U-turning traffic would need to slow right down and that would impede the flow through the junction. Facilitating that turn could be confusing for drivers who are already worried about whether they will be able to navigate the junction successfully. And, depending on the requirements of the junction, the left-turn motorway exit slips might be signalised to tie in with the crossovers, so U-turning traffic would come into conflict with traffic turning left off the mainline with a green light.

If there is a real need to allow U-turns for traffic on the side road(s) then a U-turn loop could be added before the junction where it won't cause problems with the junction itself.

In order to reduce the scope for confusion, I would avoid a right-turn arrow on the approach to the crossover, and only introduce it on the bridge/underpass itself. That is no different (and arguably less confusing) than the current preferred approach of not allowing right-turn arrows to be used on the approach to any roundabout.
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

nowster wrote: Wed Aug 15, 2018 14:14
ais523 wrote: Wed Aug 15, 2018 14:00 I suspect the major disadvantage is that it can't easily be changed to a trigger-based (rather than timing-based) light sequence without throwing all the timings off and causing vehicles to get stuck inside the junction...
The two phases aren't "fill" and "release" as the default but "left-to-right" and "right-to-left". Traffic getting stuck in the middle will only ever encounter one red light.
"Fill the bridge then empty the bridge" applies to phasing of standard diamonds with finite right-turning slots, not to DDI's, as pointed out by nowster. With basic timings, you would expect some straight-through vehicles to be trapped within the intersection, but not many, and only tail-enders of the platoon. Therefore, for any straight-through user, there would only ever be one stop at a red light, and often none.
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

Bryn666 wrote: Wed Aug 15, 2018 15:45 See attached for a demo of who goes and when.
Incorrect. You have shown a 'fill and then empty' sequence. That is NOT the DDI sequence. In a DDI, a traveller going (let's say) eastward will experience two green signals in succession during phase 1 (or he may have turned left or right onto the motorway). During phase 2, both of those signals facing him will be on red, stopping him at either the first (western) crossover point, or the second (eastern) crossover point.
User avatar
Bryn666
Elected Committee Member
Posts: 35755
Joined: Thu Jun 20, 2002 20:54
Contact:

Re: Diverging Diamonds

Post by Bryn666 »

A later post shows the sequence you refer to.
Bryn
Terminally cynical, unimpressed, and nearly Middle Age already.
She said life was like a motorway; dull, grey, and long.

Blog - https://showmeasign.online/
X - https://twitter.com/ShowMeASignBryn
YouTube - https://www.youtube.com/@BrynBuck
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

Bryn666 wrote: Sat Aug 18, 2018 11:45 A later post shows the sequence you refer to.
Sorry - I missed the later post (I'm on holiday and wifi here is a bit dodgy). But of course your second diagram shows the ACTUAL sequence used in DDIs. If that is not used then much of the DDI efficiency is forfeited. And, as far as i know, no DDI is operated in the fill-then-empty mode. The key point is to obtain 2-phase operation and eliminate turn-blocking, which the DDi does. The SPUI, another good design, also eliminates turn-blocking, allowing lengthy phases, but it remains 3-phase.

There could conceiveably be traffic situations, mainly very quiet off-peaks, where your first mode would be useful at a DDI, depending on which turn you intend. I would think that it would not be hard to detect such situations and switch the mode automatically.
User avatar
Bryn666
Elected Committee Member
Posts: 35755
Joined: Thu Jun 20, 2002 20:54
Contact:

Re: Diverging Diamonds

Post by Bryn666 »

Peter Freeman wrote: Sun Aug 19, 2018 01:34
Bryn666 wrote: Sat Aug 18, 2018 11:45 A later post shows the sequence you refer to.
Sorry - I missed the later post (I'm on holiday and wifi here is a bit dodgy). But of course your second diagram shows the ACTUAL sequence used in DDIs. If that is not used then much of the DDI efficiency is forfeited. And, as far as i know, no DDI is operated in the fill-then-empty mode. The key point is to obtain 2-phase operation and eliminate turn-blocking, which the DDi does. The SPUI, another good design, also eliminates turn-blocking, allowing lengthy phases, but it remains 3-phase.

There could conceiveably be traffic situations, mainly very quiet off-peaks, where your first mode would be useful at a DDI, depending on which turn you intend. I would think that it would not be hard to detect such situations and switch the mode automatically.
Yes, it all depends on if turn flows are more dominant than straight ahead flows. A good UTMC system should be able to switch between staging plans as far as I can tell.

The only thing stopping these being trialled in the UK is the fear of doing anything different by the innovation stifling (unless it involves multi coloured cones) policy makers.
Bryn
Terminally cynical, unimpressed, and nearly Middle Age already.
She said life was like a motorway; dull, grey, and long.

Blog - https://showmeasign.online/
X - https://twitter.com/ShowMeASignBryn
YouTube - https://www.youtube.com/@BrynBuck
User avatar
Hugo Nebula
Member
Posts: 536
Joined: Sat Apr 10, 2004 19:04
Location: Zone 5

Re: Diverging Diamonds

Post by Hugo Nebula »

Bryn666 wrote: Wed Aug 15, 2018 10:17
Herned wrote: Wed Aug 15, 2018 08:26 That's the problem, they only work with 4 arms, and presumably need fairly balanced traffic flows across the different routes. I can't think of many 'pure' junctions which would be suitable
M60 Junction 7 is the perfect test site for one.
That's as maybe, but the problem with this junction isn't the junction itself, but getting clear of it. At very busy times in the morning, the northbound traffic that's held up by that useless gyratory before Stretford Mall backs up to and onto the junction. In the evening, this junction works fine, as all traffic using it is queued on Chester Street waiting to cross the Edge Lane junction.
User avatar
Bryn666
Elected Committee Member
Posts: 35755
Joined: Thu Jun 20, 2002 20:54
Contact:

Re: Diverging Diamonds

Post by Bryn666 »

The gyratory could be eliminated and run as a two stage fork junction - indeed there are plans to get rid of it but no real detail exists yet. Right turns southbound are accommodated by Kingsway next to the Stretford Mall.

Harder to sort out the Sale side.
Bryn
Terminally cynical, unimpressed, and nearly Middle Age already.
She said life was like a motorway; dull, grey, and long.

Blog - https://showmeasign.online/
X - https://twitter.com/ShowMeASignBryn
YouTube - https://www.youtube.com/@BrynBuck
ais523
Member
Posts: 1139
Joined: Tue Sep 08, 2015 19:52
Location: Birmingham

Re: Diverging Diamonds

Post by ais523 »

Re fill/release versus left/right: the issue with left/right is that traffic can't flow in both directions at once, so the total capacity for straight-on movements is less than 50% of what it would be with a straight road.

I guess I envisaged building a fill/release diverging diamond large enough that if you reached it during a "fill" phase, the lights would have changed to "release" by the time you reached the second set of lights. That means that the phasing would consist of an alternation between two short phases, timed so that reaching the junction at green means that you'd have a green light all the way through, and if you reached it at red, then (due to the short phases) it would turn green fairly quickly, meaning in light traffic there'd hardly be an obstruction to your journey. (Note that in order to avoid a gridlock situation, this model would require "release" to become longer relative to "fill" the more queuing traffic there was – fairly easy to implement with sensors, but which means that the odds of waiting would become higher in long queues. That's not too surprising, and probably applies to any junction design.)

Left/right might potentially be more resilient against gridlock (but I'm not sure about that), and has the property that it can use much longer phases (which no longer have to be proportional to the size of the junction), although the larger the junction becomes, the more time is lost when changing from one phase to the other (unlike the fill/release model, in which the phase change loses a small constant amount of time so long as the lengths of the phase are synchronized to the size of the junction). It doesn't seem to have any other major advantages, though; in particular, it seems very hostile to traffic turning right off the free-flowing road (which will have to stop at red lights twice and can only advance at the transition from one phase to the next, meaning that a lot of stacking space is needed inside the junction to stop that movement backing up onto the main road). Compare to fill/release, where right-turning traffic off the main road gets a free run during "release" phases, which can be lengthened if necessary to clear any queue. (Left-turning traffic off the main road gets a free run during "fill" phases instead, but if that ends up becoming a problem due to the need to lengthen "release" phases, you could simply add a freeflow left to solve the problem.)

As such, I'm mildly surprised that left/right has become the dominant model. I'm guessing it's a consequence of these junctions normally being built fairly small, so that the fill/release phase timings would have to be too short to be usable, and thus left/right was the only remaining option?
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

nowster wrote: Tue Aug 14, 2018 12:12
Peter Freeman wrote: Tue Aug 14, 2018 11:54 BTW, we are discussing the DDI here as if it's a theoretical proposal and as if a decision is pending on whether to try it out. In fact we're far down the track. It's been done. It's been sorted. It works.
Where "here" is Australia? There are none currently in Britain or Ireland.
The "here" in my post meant "on Sabre".
Last edited by Peter Freeman on Fri Aug 24, 2018 01:07, edited 1 time in total.
User avatar
nowster
Treasurer
Posts: 14803
Joined: Tue Aug 31, 2004 16:06
Location: Manchester

Re: Diverging Diamonds

Post by nowster »

Peter Freeman wrote:
nowster wrote: Tue Aug 14, 2018 12:12
Peter Freeman wrote: Tue Aug 14, 2018 11:54 BTW, we are discussing the DDI here as if it's a theoretical proposal and a decision is pending on whether to try it out. In fact we're far down the track. It's been done. It's been sorted. It works.
Where "here" is Australia? There are none currently in Britain or Ireland.
The "here" in my post meant "on Sabre".
Yet not in Britain or Ireland. The acronym SABRE has a meaningful expansion...
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

Herned wrote: Wed Aug 15, 2018 08:26 That's the problem, they only work with 4 arms, ...
Indeed they do only work with four arms, and should be restricted to such. For more than four arms, in my view, a dumb-bell is best (better than a two-bridge roundabout).
Herned wrote: Wed Aug 15, 2018 08:26 ... and presumably need fairly balanced traffic flows across the different routes.
No, they don't. The DDI is most suited to locations with high turning flows. A site with predominantly straight-through traffic on the crossing road is better off with a standard diamond since that can have a straight-through duty ratio up to 100%, as opposed to the DDI's 50%.
Herned wrote: Wed Aug 15, 2018 08:26 I can't think of many 'pure' junctions which would be suitable.
If a conversion of a four-arm, two-bridge roundabout to DDI is feasible (and I haven't yet seen a good footprint overlay to prove that that's the case), then I can think of many UK sites with the required "purity".
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

Stevie D wrote: Thu Aug 16, 2018 11:18If there is a real need to allow U-turns for traffic on the side road(s) then a U-turn loop could be added before the junction where it won't cause problems with the junction itself.
If you use GE or maps to look at the two major implementations in the middle-east (Saudi Arabia and Abu Dhabi), you'll see that pre-DDI U-turns are provided exactly as you suggest.
Last edited by Peter Freeman on Wed Oct 16, 2019 14:10, edited 1 time in total.
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

nowster wrote: Fri Aug 24, 2018 00:39 Yet not in Britain or Ireland. The acronym SABRE has a meaningful expansion...
Yes, I thought that might have been what you meant. Accepted :oops:
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

ais523 wrote: Thu Aug 23, 2018 03:30 Re fill/release versus left/right: the issue with left/right is that traffic can't flow in both directions at once, so the total capacity for straight-on movements is less than 50% of what it would be with a straight road.
- which is why the DDI is best suited to sites where most crossing-road traffic wishes to interchange with the motorway rather than to proceed straight ahead. If the opposite is the traffic desire, then use a standard diamond.
ais523 wrote: Thu Aug 23, 2018 03:30 I guess I envisaged building a fill/release diverging diamond large enough that if you reached it during a "fill" phase, the lights would have changed to "release" by the time you reached the second set of lights. That means that the phasing would consist of an alternation between two short phases, timed so that reaching the junction at green means that you'd have a green light all the way through, and if you reached it at red, then (due to the short phases) it would turn green fairly quickly, meaning in light traffic there'd hardly be an obstruction to your journey.
A sensor-triggered immediate green phase at quiet times could achieve minimal delay for arriving traffic even in the (standard DDI) left/right model. This might be quite normally implemented, since my assumption has always been that sensor-triggered phase changes are employed at DDIs rather than simply dumb timing.
ais523 wrote: Thu Aug 23, 2018 03:30 (Note that in order to avoid a gridlock situation, this model would require "release" to become longer relative to "fill" the more queuing traffic there was – fairly easy to implement with sensors, but which means that the odds of waiting would become higher in long queues. That's not too surprising, and probably applies to any junction design.)
Yes, but such asymmetry is not as easy to achieve as in a standard diamond. In a DDI, even with fill/release operation, the crossover signal must be red for east when it is green for west, so one direction cannot be accorded green benefit without depriving the other direction. On the other hand, using left/right operation, a dominant right turn onto the motorway (let's say southwards) can be routinely accorded a longer eastbound green phase. The only vehicles to suffer from this would be westbound, and that's probably OK if they're not the dominant flow.
ais523 wrote: Thu Aug 23, 2018 03:30 Left/right might potentially be more resilient against gridlock (but I'm not sure about that), …
The design is gridlock-immune owing to the absence of turn-blocking.
ais523 wrote: Thu Aug 23, 2018 03:30 …, and has the property that it can use much longer phases (which no longer have to be proportional to the size of the junction), although the larger the junction becomes, the more time is lost when changing from one phase to the other
I can't see why you say this. The phase change duration (unproductive time) is constant in the left/right model. It occurs simultaneously at each crossover signal. Some traffic gets 'trapped' on the bridge, which is OK.
ais523 wrote: Thu Aug 23, 2018 03:30 in particular, it seems very hostile to traffic turning right off the free-flowing road (which will have to stop at red lights twice
Yes, if the end of the off-ramp has signals
ais523 wrote: Thu Aug 23, 2018 03:30 … and can only advance at the transition from one phase to the next, meaning that a lot of stacking space is needed inside the junction to stop that movement backing up onto the main road). Compare to fill/release, where right-turning traffic off the main road gets a free run during "release" phases, which can be lengthened if necessary to clear any queue.

(Left-turning traffic off the main road gets a free run during "fill" phases instead, but if that ends up becoming a problem due to the need to lengthen "release" phases, you could simply add a freeflow left to solve the problem.)
I may be misunderstanding you, but in a DDI standard model all four left turns get a free run, or at worst a give-way, constantly, except for pedestrian/cyclist requests.
ais523 wrote: Thu Aug 23, 2018 03:30 As such, I'm mildly surprised that left/right has become the dominant model.
I apologise for saying this again, but left/right has not "become the dominant model". It is, and AFAIK always has been, the standard and only DDI model.
Last edited by Peter Freeman on Wed Oct 16, 2019 14:19, edited 2 times in total.
someone
Member
Posts: 512
Joined: Mon Apr 24, 2017 10:46
Location: London

Re: Diverging Diamonds

Post by someone »

ais523 wrote: Thu Aug 23, 2018 03:30I guess I envisaged building a fill/release diverging diamond large enough that if you reached it during a "fill" phase, the lights would have changed to "release" by the time you reached the second set of lights. That means that the phasing would consist of an alternation between two short phases, timed so that reaching the junction at green means that you'd have a green light all the way through, and if you reached it at red, then (due to the short phases) it would turn green fairly quickly, meaning in light traffic there'd hardly be an obstruction to your journey. (Note that in order to avoid a gridlock situation, this model would require "release" to become longer relative to "fill" the more queuing traffic there was – fairly easy to implement with sensors, but which means that the odds of waiting would become higher in long queues. That's not too surprising, and probably applies to any junction design.)
Maybe I am misunderstanding a whole lot of stuff, but surely to have green all the way though, the other direction has red all the way through. So this is not fill/release but just left/right with short phases, unless the lights are changing so fast it is more like a disco.

Would not the increased amount of stopping and restarting that comes with the necessary very short phases increase congestion, or at they least just remove the net benefit. Depending on the volume of traffic you may have to stop and start several times just to reach the first set of lights, whereas a left/right design means you may have to wait much longer, but one you get moving you will be straight through. I also know I would find the former far more annoying.

And how can a release phase be longer than the fill one when they occur simultaneously for each direction? If eastbound is releasing, westbound is filling. But it would have a longer phase for that than it has for releasing, which obviously will be wasted as there is no extra space for filling. That may have value for times with a dominant flow, but again then it just turns the whole thing into a form of left/right.
ais523 wrote: Thu Aug 23, 2018 03:30Left/right… in particular, it seems very hostile to traffic turning right off the free-flowing road (which will have to stop at red lights twice and can only advance at the transition from one phase to the next, meaning that a lot of stacking space is needed inside the junction to stop that movement backing up onto the main road). Compare to fill/release, where right-turning traffic off the main road gets a free run during "release" phases, which can be lengthened if necessary to clear any queue.
And I do not undertand this either.

Under the left/right model the right turning traffic is released onto the bridge (so that capacity is not being wasted) when the left side has a green light. Sure, that does means if it hits a red light on the slip road then it will hit the second one on red too.

But under the fill/release model the bridge will be filled leaving traffic on the slip road having to wait for a gap to merge. It will likely only allow one or two vehicles out at a time in the brief period when releasing and before the filling traffic gets a green light. Although under your description of both lights being green, such a time would not exist.

And if traffic is light enough that the junction would never fulling fill, negating this problem, then you may as well just stick with a conventional diamond anyway.

Have I completely misunderstood all of this? Because I am seeing no benefit to this fill/release model.
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

someone wrote: Fri Aug 24, 2018 16:17
ais523 wrote: Thu Aug 23, 2018 03:30I guess I envisaged building a fill/release diverging diamond large enough that if you reached it during a "fill" phase, the lights would have changed to "release" by the time you reached the second set of lights. That means that the phasing would consist of an alternation between two short phases, timed so that reaching the junction at green means that you'd have a green light all the way through, and if you reached it at red, then (due to the short phases) it would turn green fairly quickly, meaning in light traffic there'd hardly be an obstruction to your journey. (Note that in order to avoid a gridlock situation, this model would require "release" to become longer relative to "fill" the more queuing traffic there was – fairly easy to implement with sensors, but which means that the odds of waiting would become higher in long queues. That's not too surprising, and probably applies to any junction design.)
Maybe I am misunderstanding a whole lot of stuff, but surely to have green all the way though, the other direction has red all the way through. So this is not fill/release but just left/right with short phases, unless the lights are changing so fast it is more like a disco.

Would not the increased amount of stopping and restarting that comes with the necessary very short phases increase congestion, or at they least just remove the net benefit. Depending on the volume of traffic you may have to stop and start several times just to reach the first set of lights, whereas a left/right design means you may have to wait much longer, but one you get moving you will be straight through. I also know I would find the former far more annoying.

And how can a release phase be longer than the fill one when they occur simultaneously for each direction? If eastbound is releasing, westbound is filling. But it would have a longer phase for that than it has for releasing, which obviously will be wasted as there is no extra space for filling. That may have value for times with a dominant flow, but again then it just turns the whole thing into a form of left/right.
ais523 wrote: Thu Aug 23, 2018 03:30Left/right… in particular, it seems very hostile to traffic turning right off the free-flowing road (which will have to stop at red lights twice and can only advance at the transition from one phase to the next, meaning that a lot of stacking space is needed inside the junction to stop that movement backing up onto the main road). Compare to fill/release, where right-turning traffic off the main road gets a free run during "release" phases, which can be lengthened if necessary to clear any queue.
And I do not undertand this either.

Under the left/right model the right turning traffic is released onto the bridge (so that capacity is not being wasted) when the left side has a green light. Sure, that does means if it hits a red light on the slip road then it will hit the second one on red too.

But under the fill/release model the bridge will be filled leaving traffic on the slip road having to wait for a gap to merge. It will likely only allow one or two vehicles out at a time in the brief period when releasing and before the filling traffic gets a green light. Although under your description of both lights being green, such a time would not exist.

And if traffic is light enough that the junction would never fulling fill, negating this problem, then you may as well just stick with a conventional diamond anyway.

Have I completely misunderstood all of this? Because I am seeing no benefit to this fill/release model.
Someone: No, you have not misunderstood it. Your comments are correct. Anyway, in a DDI, fill/release mode is NOT used. It would make no sense and would wipe out most of DDI's great benefits.

In the standard left/right mode it is possible to think of certain circumstances, especially involving dominant right-turning movements, where a longer green phase for one direction ( let's say 60% / 40% ) might be of benefit. I don't know whether this has been used anywhere: it would not be difficult to invoke, as required, under detector loop control.

By the way, for those still confused, there are good simulations to be found on the web, especially on Youtube. They're mainly from state transport departments in the USA.
ais523
Member
Posts: 1139
Joined: Tue Sep 08, 2015 19:52
Location: Birmingham

Re: Diverging Diamonds

Post by ais523 »

someone wrote: Fri Aug 24, 2018 16:17 Maybe I am misunderstanding a whole lot of stuff, but surely to have green all the way though, the other direction has red all the way through. So this is not fill/release but just left/right with short phases, unless the lights are changing so fast it is more like a disco.
I'm envisaging the junction as being large enough that traffic takes a substantial time to get from one set of lights to the other. During a "fill" phase, the lights to exit the junction are red (in both directions), and to enter are green (from the minor road), red (from the major road's right turns). Traffic enters the junction, in both directions, and as the start of the queue starts to reach the end of the junction, the lights change; so the traffic that entered the junction has a free run out of the other side. This also gives a green light to the major road's right turns, which join the back of the queue of traffic leaving the junction.

Fill/release is substantially different from left/right as the settings on the lights are rotationally symmetrical (as opposed to left/right where a 180 degree rotation changes red lights to green lights and vice versa).
And how can a release phase be longer than the fill one when they occur simultaneously for each direction? If eastbound is releasing, westbound is filling. But it would have a longer phase for that than it has for releasing, which obviously will be wasted as there is no extra space for filling. That may have value for times with a dominant flow, but again then it just turns the whole thing into a form of left/right.
No, with fill/release phasing, if eastbound is releasing, westbound is also releasing. If you leave a "release" phase indefinitely, no new traffic can enter the junction from the minor road, and the junction will effectively form a pair of freeflow rights for the major road.
ais523 wrote: Thu Aug 23, 2018 03:30Left/right… in particular, it seems very hostile to traffic turning right off the free-flowing road (which will have to stop at red lights twice and can only advance at the transition from one phase to the next, meaning that a lot of stacking space is needed inside the junction to stop that movement backing up onto the main road). Compare to fill/release, where right-turning traffic off the main road gets a free run during "release" phases, which can be lengthened if necessary to clear any queue.
And I do not undertand this either.

Under the left/right model the right turning traffic is released onto the bridge (so that capacity is not being wasted) when the left side has a green light. Sure, that does means if it hits a red light on the slip road then it will hit the second one on red too.
Under the left/right model, right-turning traffic off the main road ends up on the "closed"/non-flowing half of the junction. That means that each full cycle of the junction's two phases can only allow an amount of right-turning traffic through equal to the number of vehicles that fit inside the junction. In other words, you need to cycle it quickly to avoid blocking the right-turners. This means that a slow-cycling model can be dangerous if there's a lot of right-turning traffic, as queuing on the major road is something that it's normally important to avoid.
But under the fill/release model the bridge will be filled leaving traffic on the slip road having to wait for a gap to merge. It will likely only allow one or two vehicles out at a time in the brief period when releasing and before the filling traffic gets a green light. Although under your description of both lights being green, such a time would not exist.
With fill-release timing, the right-turning traffic off the main road joins the back of the queue for straight-on traffic during the release phase, and continues through the junction behind it. The release phase can continue until the queue clears, if necessary to prevent the main road becoming blocked.

I guess there's a bit of symmetry here: with left/right, straight-on traffic on the minor road in one of the two directions can be given a freeflow by holding the lights, but this blocks right-turning traffic in both directions; a full phase cycle is needed for the
right-turning traffic to get through the junction (and it can do so without stopping only if the cycle time is comparable to the size of the junction, thus likely fairly short as these junctions are designed with a short distance between the sets of lights). Meanwhile, fill-release requires a full phase cycle for straight-on traffic to get through the junction (and it likewise can do so without stopping if the "fill" phase length matches the size of the junction), whereas right-turning traffic can be given a freeflow by holding the lights.

However, the difference is that with left/right, you can't get all the movements working well simultaneously. Changing from a left phase to a right phase (or vice versa) loses the amount of time it takes the junction to clear, taking the capacity below 50%, so you want the junction to be small so that it clears quickly. That means that right turns off the major road are almost impossible, as right-turning traffic needs to fill the junction during one phase and is released during a different phase. You could try to clear that via making the release phase for the right-turning traffic longer, but that's the fill phase for traffic turning right from the other direction. So if you have a lot of right-turning traffic coming from both directions, you need to change phases quickly (and more quickly the smaller the junction is), meaning you lose much of the benefit of your straight-on freeflow.

With the fill/release model, you have a rotational symmetry in the junction: both straight-ons work simultaneously during the fill/release transition, so although the capacity is again 50% (you have to wait for a fill phase before you can enter the junction), increasing the size of the junction doesn't actually reduce the capacity at all (it just changes the optimal cycle time). Right-turning (off the main road) traffic can only move during a release phase, and follows on behind the straight-on traffic, so in times of heavy right-turning you need to make the release phase longer relative to the fill phase; but this doesn't cause any awkward blockages (the only blockage you get as a consequence of that is for traffic entering the junction from the minor road, which is held entirely behind the junction on the minor road and thus is queuing somewhere where it's reasonably safe for it to queue). And of course, that's allowing both right turns simultaneously, so you never have to worry about forming a queue onto the main road in one direction to clear the queue from the main road in the other direction; because a release phase in one direction is a release phase in the other direction, you never have to pick a direction to favour.
Peter Freeman
Member
Posts: 1390
Joined: Thu Aug 30, 2012 07:52
Location: Exits 9 & 10, M1 East, Melbourne, Australia

Re: Diverging Diamonds

Post by Peter Freeman »

ais523:

I can see that your fill/release mode of operation would work, and under certain circumstances could even work better than the (actually implemented) left/right mode. However, it ends up having many characteristics of normal diamond intersections. Fortunately it still does have the DDI benefits of (a) reducing the number of 'points-of-conflict' (= safer) and (b) larger radius at right turns, resulting in smoother and higher-speed flow which in turn increases capacity (a notable advantage also of SPUIs, by the way).

Your point about DDI being hostile to right-turners from the off-ramps, potentially leading to mainline queuing on the major road, is true at sites that have signals at the ends of off-ramps. Many examples do not. In practise it doesn't seem to be a problem.

I noted in another post (the other thread, perhaps) that there is much variation in the detailed design, especially regarding how the on- and off-ramps join. For example, merge versus lane-add; acceleration lane or only a give-way line; deceleration lane or simple diverge; signals or not; etc.

When I was inspecting the many MO and KS implementations, I observed, at a couple of locations, that the signal change was not simultaneous at the two crossovers. I didn't have time to watch for long enough to analyse what the strategy was, and the roads weren't busy then, but it may be that there is an element of fill/release there. That is, a 'hybrid', but definitely not pure fill/release. It would be interesting to have input from a USA traffic engineer. Can Missouri-based Scott comment? Alternatively, on the DDI 'official web site' there is a question submission facility.

https://divergingdiamond.com/

Whatever signal strategy is adopted, there is no doubt, from empirical evidence, that DDI's work exceptionally well.

Just two comments on your last post -
ais523 wrote: Sat Aug 25, 2018 07:56 I'm envisaging the junction as being large enough that traffic takes a substantial time to get from one set of lights to the other.
This is not the case at most current American sites. A DDI could be built that way (and a few are), but it would lose one of the DDI's other advantages: small land-take.
ais523 wrote: Sat Aug 25, 2018 07:56 However, the difference is that with left/right, you can't get all the movements working well simultaneously. Changing from a left phase to a right phase (or vice versa) loses the amount of time it takes the junction to clear, taking the capacity below 50%, so you want the junction to be small so that it clears quickly.
In pure left/right operation the junction doesn't clear: some straight-ahead vehicles are trapped on the bridge, but relatively few. Therefore there is virtually no unproductive time and the duty ratio is very close to 50%. If more is required for straight-ahead traffic, then don't adopt DDI.
User avatar
Mark Hewitt
Member
Posts: 31412
Joined: Fri Dec 17, 2004 12:54
Location: Chester-le-Street

Re: Diverging Diamonds

Post by Mark Hewitt »

Having had to visit the Metro Centre twice over the weekend (don't ask!)

I wonder if the Metro Centre junction which is a signalised diamond would actually be a good candidate here. It's very busy for a single overbridge. Traffic can queue back for a long distance on the A1, so much so it has it's own dedicated queue lane. The majority of traffic is headed n/b and turning right into the complex. But it does also serve local roads to the South.

https://goo.gl/maps/HyjaNPAfPEk
Post Reply