Hacker News Reader

Comment by bradley13 | Hacker News Reader

Right. They're going to get all the streetlights in an entire city to blink in unison. Sure they are.

Anyway, in most cases the LEDs don't run at 100%. In our area a, I happen to know that they are between 29% and 40%. That dimming is accomplished by blinking. Which would conflict with the idea of having them blink synchronously.

Much simpler solution: ensure the LEDs ghts are directed exclusively downward, and have only as much illumination as needed.

Even better: turn the darned things off when they are not needed - why illuminate empty streets?

View on HN

I'm not sure I'd be so quick to dismiss the feasability, at least not based on your reasoning so far.

FM Radio stations have managed to accomplish getting the speaker in your car radio to "blink synchronously" according to a complex signal pattern in the exact same way as millions of other radios simultaneously.... for about a century now. As long as all streetlights are in locations with good radio reception (which, since they're all on streets and are tall, is likely), it would not be that hard to implement in principle a central FM or other radio signal that oscillates according to the expected blinking. Lights that are engaging in dimming could multiply another signal by the central signal to diminish it even more.

Okay, let's do some math. Very rough napkin numbers, but it'll get us close.

A 150Hz blink rate means a period of 1/150s, or ~6.7ms.

The speed of radio signals through air near the surface is a bit lower than c, let's call it 299,250 km/s. [0]

When you multiply those numbers you get ~1500km. That means each 'on' and 'off' signal is 1500km long, moving near c. More specifically, the signals are concentric rings around the transmitter 1500km in width.

Let's look at London. I picked it at random, but it just happens to have a land area around 1500km^2. Assume the transmitter is in the center of the perfectly circular city, that means it takes 3.3ms for each 'on' and 'off' signal to cover the entire city. Each on and off pulse is just about exactly long enough to cover the entire city.

All that together means that at any given moment 50% of the lights in the city are on, and the other 50% are off. That's about as far from synchronized as you can get, really.

Also, light is slightly faster than radio in these conditions[1]. The light reaching the center of the city won't be pulsing in the same phase as the signal you're sending out. It will also be blurred and muddied due to the difference in signal speed. You probably wouldn't see much flicker in the light at all, as 50% of the lights are always on. On average, you receive 50% of the light you would otherwise, but that value will never reach 0. There won't be any periods of total darkness where you can make an observation, as the article suggests.

Moral of the story: light (and radio) is much, much slower than most people realize.

[0] https://ieeexplore.ieee.org/document/170108 [1] I couldn't find a confident number in my 16 seconds of searching. Seems like just a few percent faster, but the real number doesn't matter here.

Bonus: you can synchronize the streetlights exactly, but it's much more expensive and complex than a single radio transmitter. You need to have every single light contain a receiver with an accurate clock that synchronizes with a geosynchronous satellite with an atomic reference clock. Search for "GPS disciplined oscillator" if you're curious.

>Let's look at London. I picked it at random, but it just happens to have a land area around 1500km^2

Um am I really really tired and not reading correctly, or is this confusing linear area with square area? London doesn't stretch from New york to Florida.

186000 miles/second * pulse of .003 = 558 miles ~= 900 km

I knew I fucked something up. Don't do math while smoking, kids.

Either way, propagation delay is a huge problem with synchronizing systems. It's not something you usually think about

Dont you have propagation delay in reverse too? Let's say you do turn off all the lights at the exact same time. The light hitting the telescope also travels at the speed of light.

There’s engineering feasibility, and then there’s feasibility. Is it any less feasible to operate fewer or dimmer street lights? How about directing light more efficiently? This is a problem of bureaucratic and societal engineering.

I was really just exploring a hypothetical. This whole idea of synchronizing lights is pretty dumb. It won't work, and even if it did it's still a bad solution.

Better solution is less lights, or turn them off at night or any of the other solutions we've been talking about for 30 years. Remember when street lights were killing sea turtles?

I’m not a physics person, but couldn’t this be easily solved by adding a programmed delay that corresponds to the distance from the transmitter? Street lights aren’t known to migrate.

In principle, yes. But not easily.

The physics of radio wave propagation are surprisingly complex. Buildings and objects in the path between transmitter and receiver can cause various interference effects that basically slow down the signal. I'm glossing over a lot, but it would take significant effort to precisely measure the exact delay between any one receiver and the transmitter. It's much more complicated than just the distance, and each one has to be measured individually.

I'm not an expert in RF, but I think this concept of a single synchronizing transmitter just doesn't work in the real world. You'd probably build a cell network with many transmitters covering small areas. Each transmitter has the hardware to synchronize with thr others, and the lights just have a simple circuit slaved to the transmitter clock.

There's probably more clever ways to synchronize signals over long distances, but RF is black magic to me.

Wavelength of FM radio is about 3m, so you only have to be a few centimeters shifted to be seeing an entirely different phase and be out of sync.

If the LEDs blink at 150 Hz, then a single on-off cycle has a "wavelength" of 2000 km.

The wavelength of the control signal doesn't seem relevant.

Plus you just add a position based delay. This is quite the bikeshed!

How does the LED know its own position? At that point you might as well abandon FM and use GPS.

Send a few bytes of data onto the SD card on lamp installation. If GPS can do it all in one then that is probably better given how cheap the units are these day.

You must not be familiar with multipath interference. Radio isn't quite as synchronous as you believe it to be.

And of course the light itself is also undergoing multipath distortions like the radio signal. And for astronomy, the changing atmosphere might be the most important secondary reflection. And the multipath arrangement between emitter and observer is different for each pair.

Could you make a dark period long enough to mask the multipath variance among emitters and observers? Or would this dark period be so long as to cause annoying flicker?

> Right. They're going to get all the streetlights in an entire city to blink in unison. Sure they are.

The article states they use GPS for synchronization. I wish they went into it further, but a number of affordable chipset can do emit PWM at the proposed frequencies, globally aligned in time to PPS(every receiver computes the same top of second), then the telescope exposes during the blanking time.

On a related note I have a 1MHz GPS time source in my office that is synced to a 10MHz OCXO(more common for test gear) so my equipment is phase synchronized to GPS. It's really spooky to sync the oscilloscope to one time source and probe another time source with the scope and watch how their phases don't perceptibly drift from each other.

Yes exactly, aim them downward and turn the unused ones off, which is most of them most of the time, and the problem will be solved for everyone. The reason we should do it is that it will save money and power for everyone, the night sky is just icing on the cake.

Getting lights to blink in unison is a neat idea, and technically not hard nor too expensive for new lighting. Personally I think the biggest problem with it is that cities aren’t going to agree to overhaul all existing outdoor lighting only for the sake of people using telescopes. Even if there was widespread buy-in this is a project that could easy take 50 years or more, and cities might be reluctant to commit to this idea right when technology is rapidly changing.

Precisely. Why go for the complex, expensive, and fragile solution when better shades (and maybe proximity/motion sensors) would do the job?

Where I live they installed street lights that are a lot dimmer by default, but they are connected and have motion detectors. Whenever someone comes near, the light gets brighter and so do the others in direction of travel. You kind of get a light wave traveling with you and afterwards they dim again.

The mayor then got complaints that the street lights are not as bright as mandated by some regulation (when measured with an empty road in their dimmed state ...).

> Even better: turn the darned things off when they are not needed - why illuminate empty streets?

That is a way better idea than the one in the article, for sure. So many benefits. Would need the motion sensing technology to be top notch tho. Need to illuminate for pedestrians as well. Or perhaps illuminated sidewalks, or smaller lights for pedestrians.

Some of the newer street lamps in my area have smaller light for pedestrians! They're mounted at human height and shine down on the sidewalk.

Motion sensing, I find, similar to LED technology, is often used more for increasing lighting levels rather than decreasing. My pet peeve are people who have motion sensors that activate their 20k lumen floodlights based on motion on the public sidewalk. Cities should not allow those.

> Motion sensing is often used more for increasing lighting levels rather than decreasing

Isn't that just a matter of perspective [whether the presence of motion is used to increase lighting or the absence of it is used to decrease]?

No it's a horrible idea, imagine the lights turning on and off as people pass. No!

Indeed that would be terrible, especially in dodgier areas. Not only you're advertising your position to potential attackers, but you're also less likely to see them until they're very close.

I believe relatively dim, warm coloured lights are a much better idea.

Last sentence is a fine idea, it is also possible to move toward developing a society where “potential attackers” is a thought much more distant in people’s minds.

> it is also possible to move toward developing a society where “potential attackers” is a thought much more distant in people’s minds

Amen, but alas we're not there.

In any case, threats don't always come from purposeful attackers. It's useful to see animals, vehicles and bikes with no light, etc.

>Amen, but alas we’re not there

What is troublesome is that many places, that many of us probably hail from, “were there” not so long ago, and it was destroyed by duplicitous, malicious political and cultural actors.

Having “safe streets” is not a pipe dream. Indeed not having them is a symptom of people trying to destroy our societies.

You could see the attackers just as well as they see you. I meant because of the annoyance that it would create.

The problem with warm lights is that they are much harder to see in the dark. Maybe dim lights around 505nm would be a better idea.

Thinking like an attacker: find an ambush spot - probably near the trigger-point of the motion-sensor; hold very still; light goes off; victim approaches; ???; profit.

I agree with you, though: lights blinking off and on all the time are not great. I drove somewhere in Europe where they did that with streetlights on major roads, and found it startling, long past the point at which I'd have assumed I'd become used to it. It wasn't the ones that were responding to me, it was the ones that responded to traffic on the opposite side of the road - they'd flash on in my peripheral vision and make me jump. Did not like.

> You could see the attackers just as well as they see you.

That's only if they're moving within the sensors that switch on the light.

> The problem with warm lights is that they are much harder to see in the dark.

Not sure what "in the dark" means here. Light by definition is easier to see in the dark than it is in light as it has less light to compete with.

In any case, I find warm lights much easier. Cold lights are too much sensorial information for me. In any case, street light at night shouldn't be meant to allow you to see everything like it's a sunny day. It's supposed to give you enough light that you can see your surroundings and potential threats around you. For that, warm lights should be more than enough. If you need more light clarity, light your own phone torch or something.

>If you need more light clarity, light your own phone torch or something.

Good point! In this case of nighttime lighting infrastructure it has become so vastly overbuilt as if to save people “the trouble” of these simple targeted solutions.

I mean they need to be pretty bright to be visible at all. Notice how red leds still look dim at night, but green blue are shining in the dark room. The eye can't adapt well to dim warm light.

And that is the other reason. Those would still make nights look like nights, they should look sort of moonlight like with reasonable brightness.

I'm still unsure what you're advocating for.

Dim, but cool coloured lights? That's certainly better than bright cool ones but I still find them very annoying and overstimulating at night.

The only thing I can't do with dim, warm lighting is read small text from something other than a backlit screen. For everything else, it's perfectly fine. And if I do see myself having to do that, I'll just add a spot of light to what I'm trying to read, problem solved.

The problem is that warm lights just look dimmer than they are in reality, you need objectively much brighter warm light to see anything at all. 505nm is a commonly available wavelength that is very near the eye's peak sensitivity in the dark, so a very low powered light would provide sufficient light, and it wouldn't be more overstimulating than moonlight unless it's way too bright. The only drawback it seems to have is that it's rather close to the O-III astronomical filter, I don't know how much it matters.

I think their point is that all the wavelengths that make the lights most-useful from a safety perspective are also those which interfere with seeing the stars and whatnot.

Low pressure sodium is useful for safety (you don't need color vision to see the shadow of an attacker, a fallen branch, shiny reflective shards of a glass bottle, or a pothole), efficient (Lumen/W), and extremely tame for star-observing (look through a narrow-band filter to block the two very thin very close spectral lines (so close you need decent instruments to see that its not just a single one)). The downside is that you don't get color vision, because the light is so close to monochromatic that you might as well think of it as truly monochromatic for human vision purposes.

Ok, but that's where I'm confused. You don't need to see everything in full colour and detail to be safe. You need to be able to make out shapes, movements and shadows.

A few years ago I lived in a neighbourhood where all the street lights had a dark yellow tint, very warm coloured. They weren't all that bright either. I could see a person a block away with ease.

Again, that was far from ideal to read a book, but why would I do that out on the streets at night?

Even worse is Why pay to occasionally faintly illuminate the bottoms of airplains ever?

I it find harder to see objects under street lights than a person with a head lamp or a car, everything just blends into one color

In Europe the entire mains grid is synchronised - the smallest deviation in fase & frequency is the same. It’s fairly straightforward to sync them with this distributed reference clock.

Edit: https://www.mainsfrequency.com/

The frequency observed across all loads in the system might be very consistent, but the phase is absolutely not. Different types of transformers introduce different phase shifts, so this can't work.

Why not, just sync them to the AC power grid they're all powered from.

Different types of transformers introduce different phase shifts.

Couldn't you still account for offset?

Maybe? I guess you could "calibrate" it with GPS. I don't know how consistent the phase at a specific location really is in practice on long timescales.

> They're going to get all the streetlights in an entire city to blink in unison. Sure they are.

Thre are a super simple and effective distributed algorithms for this, if each lamp can see another few lamps. They don't even need a good clock, just a relatively stable oscillator within a few percent of a standard frequency, a light sensor and an integrator.

I agree that it's better to just put hoods on the lamps to direct the light to only where it's needed, but distributed synchronization is a solved problem.

I find it difficult to believe that a city-wide distributed synchronization algorithm could actually work. How do you deal with the photo-electronic-photo propagation delay, and ensure that each sensor is paying attention to the correct emitters, possibly including itself?

Even if you somehow got it to work, one guy with a hacked flashlight could throw the whole thing into chaos.

Dimming is usually done by pulse width modulation. So you can definitely get them to blink at the same frequency, you just change the pulse width.

They need phase locked too. Edit: article says they're using GPS, probably using PWM-aligned-to-PPS.

Reply on news.ycombinator.com