The thing with discharges is, they are usually very tight with thermal balance. Having two arc tubes close to each other means the rated power would have to differ quite significantly whether the other tube is lit or not. And even whether the other tube is above or below it. Mainly for the lower wattage tube the difference would be rather large.
Yes, the incandescent filaments are relatively sensitive as well, but they are way farther separated relative to their dimension. The usual 2cm distance for a 1mm diameter filament spiral is 20:1, that ensures pretty well at least some isolation.
So for an equivalent with 1cm diameter arctubes it would mean the arctubes would have to be 20cm apart. Not that practical within one bulb. And you may get a bit better with just two lamps side by side.
 

The 2 bulb solution sounds good.

Sylvania briefly made a TwinArc sodium that contained one 35W and one 70W arc tube connected in parallel.  One of the arc tubes had a lower pressure filling so it would always strike first, I cannot remember which but do have an example of this lamp.

By operating the lamp on a circuit having two ballasts and a changeover switch which extinguished one arc before re-starting the other, it was possible to dim the light output while maintaining high efficacy and same colour.

It was not successful though, because SON lamps can be dimmed fairly well just by under-running a lamp on a lower current ballast.  Of course lamp efficacy does drop and the colour rendering is poor when dimmed, but for streetlighting customers who wanted energysavings from dimming, most simply used a small additional ballast connected via a relay, and took that out of circuit when they wanted to dim a lamp e.g. after midnight.

Sylvania briefly made a TwinArc sodium that contained one 35W and one 70W arc tube connected in parallel.  One of the arc tubes had a lower pressure filling so it would always strike first, I cannot remember which but do have an example of this lamp.

By operating the lamp on a circuit having two ballasts and a changeover switch which extinguished one arc before re-starting the other, it was possible to dim the light output while maintaining high efficacy and same colour.

It was not successful though, because SON lamps can be dimmed fairly well just by under-running a lamp on a lower current ballast.  Of course lamp efficacy does drop and the colour rendering is poor when dimmed, but for streetlighting customers who wanted energysavings from dimming, most simply used a small additional ballast connected via a relay, and took that out of circuit when they wanted to dim a lamp e.g. after midnight.

Individual dimming of single arc tubes could be a way to do this, but you would have to consider the side effects of doing that.

James How was the control circuit for that system protected from going "out of sync" with the lamp state ?

As in

Intended :
 - 1st strike to 70W arc tube, with power to 70W ballast

 - at midnight quick interruption then power restored to 35W ballast, the cold arc tube is 35W and it starts



Malfunction :
 - 1st strike to 70W arc tube, with power to 70W ballast

 - At some point before midnight somebody bump into the pole, kicking out the arc, then the lamp restrike into 35W arc tube, still on 70W ballast

(lets say the controller is made so it can detect power interruptions, so it will "understand" lamp cycling as result of line voltage glitch. So only mechanical shock to the lamp remains a possible reason)

 - Lets say the lamp can recover from this malfunction : The 35W arctube overheat on 70W ballast and go out on its own, then the lamp restrikes back into the 70W arctube

 - If the change to 35W by the gear was at the time before the lamp recovered, or if it was bumped again on 35W gear, then it will restart into 70W arc tube on 35W gear, and this is a stable working point - from there it will be stuck on all night with the 70W arc tube partially warmed up on 35W gear

That situation sounds like a recipe for lamp control disaster!

I would guess the synchronization was not the biggest problem - the situation described by Ash could be easily treated by disabling the ignitor after the lamp ignites and reenable it only when the controller decides so. That means whenever the lamp will self extinguish, the ignitor remains disabled until the situation is detected by the controller, so the controller changes it's con figuration and only then reenables the ignitor and so ignites the other lamp. Still there could be some rarer cases of lost sync which would not be solved by this, but there would be other methods (the disabling ignitor came just after barely five minute thought, the engineers had for sure way more time to solve it)...
Another five minutes: The higher power tube is designed for lower OCV than the lower power one (because the higher power one is supposed to strike first, so have lower ignition voltage), the ballast for the higher power one has insufficient OCV to run the lower power one, so when it happens, the lower power one extinguish, allow the higher power arctube to ignite and so the lamps resynchronize to the ballast.

The main problem would be just the plain performance on at least one power level:
Normally the arctube design is optimized for the best system efficacy on the given mains voltage. That means whatever way you deviate from the standard parameters, you most probably end up with lower efficacy and/or shorter life. This includes the resulting ignition voltage.

For the system above to really work, the ignition voltage have to differ with sufficient margin even with uneven wear (one tube lights longer hours than the other one). That means quite significant difference between the tubes.

But that means, only one of them could be at the efficacy optimum, so the other one will perform way worse than it's standard counterpart of the same wattage. With that you can easily loose 20..30% of efficacy.
So if you compare that to the dimming system (where the full power setting corresponds to the optimum), the performance would not be that much worse, while you will suffice with rather standard, so cheap, lamp (and no synchronization problems at all), you just need an extra series dimming choke or capacitor (depends on the ballast type - series reactor, HX and mag-reg means extra choke, CWA extra capacitor).

I think the not so good performance and the high cost (proprietary solution, so no competition, lower volume lamp production,...) were the main factors, which make that two arctube dimming system not flying on the market...

I can't imagine many spots where a dimming HID would go.

I can't imagine many spots where a dimming HID would go.

Dimming HPS are quite common on the main roads in and around bigger cities. When there is still dense traffic, it operates at full power (here the most frequent are 150W HPS on about 8m poles, spaced about 16m or so). When the traffic ceases, the lights are dimmed down to about 50% output (around 60% power or so). The energy saving is part of the reason, the other part is the greater comfort and safety when driving out where the roads are not lit at all (not that big illumination/darkness shock for the drivers, so helping the safety).
Usually the systems consist of a supply voltage reduction transformer with many taps and controller for the whole stretch of road it is serving. Some are pure time preprogrammed, some are remote controlled from a city traffic management dispatch center (they are overseeing the complete traffic, it's flows through the city and so adjusting the road crossing signal timing, managing smooth ways for emergency services and control the lighting).

Quote from: wattMaster on August 16, 2016, 09:24:09 AM

I can't imagine many spots where a dimming HID would go.

Dimming HPS are quite common on the main roads in and around bigger cities. When there is still dense traffic, it operates at full power (here the most frequent are 150W HPS on about 8m poles, spaced about 16m or so). When the traffic ceases, the lights are dimmed down to about 50% output (around 60% power or so). The energy saving is part of the reason, the other part is the greater comfort and safety when driving out where the roads are not lit at all (not that big illumination/darkness shock for the drivers, so helping the safety).
Usually the systems consist of a supply voltage reduction transformer with many taps and controller for the whole stretch of road it is serving. Some are pure time preprogrammed, some are remote controlled from a city traffic management dispatch center (they are overseeing the complete traffic, it's flows through the city and so adjusting the road crossing signal timing, managing smooth ways for emergency services and control the lighting).

If it's internet connected, there's the possibility of hackers messing around with the lighting.

I think historically these systems have their own separate networks.
And yes, everything is possible to hack, but what would the hacker get? Few seconds longer/shorter red signal? Dimmed lights? Moreover it would have to mess up with the system for long time continuously to have any effect and that time the operators will know well there is something happening and so send the metro-police to control it manually on the site if really needed.
Don't forget that remote control has the ability to just change the preferences and advance/delay the timing a bit, so nothing really safety related.

I think historically these systems have their own separate networks.
And yes, everything is possible to hack, but what would the hacker get? Few seconds longer/shorter red signal? Dimmed lights? Moreover it would have to mess up with the system for long time continuously to have any effect and that time the operators will know well there is something happening and so send the metro-police to control it manually on the site if really needed.
Don't forget that remote control has the ability to just change the preferences and advance/delay the timing a bit, so nothing really safety related.

I'm thinking of something like the hacker making the streetlights have a X-Mas/Christmas light twinkle to them, or make them act like they are all cycling.

I'm thinking of something like the hacker making the streetlights have a Christmas/Christmas light twinkle to them, or make them act like they are all cycling.

That is physically not possible. The controllers autonomously control the 15 minutes stabilization after power up and then the power ramp down rate to make sure the lamps operate properly. So if you will commands it very fast up/down or so, they will just stay at full power (or maximum one step below; the individual steps are just few percent, so not noticeable at all).

Quote from: wattMaster on August 16, 2016, 11:22:19 AM

I'm thinking of something like the hacker making the streetlights have a Christmas/Christmas light twinkle to them, or make them act like they are all cycling.

That is physically not possible. The controllers autonomously control the 15 minutes stabilization after power up and then the power ramp down rate to make sure the lamps operate properly. So if you will commands it very fast up/down or so, they will just stay at full power (or maximum one step below; the individual steps are just few percent, so not noticeable at all).

Well, it would be much slower. But it would be quite annoying.

With the old system maybe so, but the newer the systems get, the more "smart" features developers push into them. I am pretty sure you can do some on/off epilepsy with the newest IOT controlled LED lights, that dont have any significant inherent "time constant" limitation

Sounds like disaster waiting to happen. Do we really need "smart" streetlights? I would be fine with streetlights with a motion detector.