LED Traffic light flash

[Chuck Olson]

A few months ago I proposed a circuit to run LEDs on 120VAC 60Hz line using
a capacitor for the current-limiting element, and putting two LEDs in
parallel, oppositely polarized to conduct that current (maybe for
night-light use). The forward drop of one would keep the reverse voltage low
for the other. One of the comments made concerned the surge of current that
would be conducted at the moment of turn-on due to the higher frequency
content of the turn-on wavefront. Could this be the reason for the
brightness flash of red LED traffic lights at turn-on? Certainly the LED
array would have to plop right in where the incandescent lamp was, powered
by 120 VAC. But it's hard to imagine all the rather large, high-voltage
capacitors needed to do the current limiting on what I estimate may be 200
or more LEDs. One comment I just read, concerning the flash, said it's due
to poor regulation of the DC supply. But that implies the switching takes
place between the DC supply and the LEDs, and that's hard to imagine, as
well. Even worse, the use of capacitors on a large scale would wreak havoc
with the utility's power factor, not to mention the arcing of switch
contacts (but maybe the arcing is already taken care of since the old
incandescent lamps also produced a surge of start-up current) . So do any of
you know FOR SURE just how these LED traffic lights are powered, and why
they flash brighter at turn-on?

Thanks,

Chuck

 

[John Fields]

A few months ago I proposed a circuit to run LEDs on 120VAC 60Hz line using
a capacitor for the current-limiting element, and putting two LEDs in
parallel, oppositely polarized to conduct that current (maybe for
night-light use). The forward drop of one would keep the reverse voltage low
for the other. One of the comments made concerned the surge of current that
would be conducted at the moment of turn-on due to the higher frequency
content of the turn-on wavefront. Could this be the reason for the
brightness flash of red LED traffic lights at turn-on? Certainly the LED
array would have to plop right in where the incandescent lamp was, powered
by 120 VAC. But it's hard to imagine all the rather large, high-voltage
capacitors needed to do the current limiting on what I estimate may be 200
or more LEDs. One comment I just read, concerning the flash, said it's due
to poor regulation of the DC supply. But that implies the switching takes
place between the DC supply and the LEDs, and that's hard to imagine, as
well. Even worse, the use of capacitors on a large scale would wreak havoc
with the utility's power factor, not to mention the arcing of switch
contacts (but maybe the arcing is already taken care of since the old
incandescent lamps also produced a surge of start-up current) . So do any of
you know FOR SURE just how these LED traffic lights are powered, and why
they flash brighter at turn-on?

---
Since it's imperative that LED arrays for traffic-light use exhibit
extremely good reliability, I doubt whether the apparent increased
brighteness flash during turn-on is really attributable to a current
surge, unless it's within the spec's for the lamp. If I had to guess
(which I'm doing) I'd think that it was due to the eye's AGC and its
response to the sudden change in intensity.

I don't know how traffic light LED arrays are designed, but I design
largish LED arrays and the way I take care of turn-on surges is to
make sure there _are_ none, by driving the array with the output of a
full-wave bridge tied directly across the line. No series cap (with
its attendant problems) in the line, and no filter cap on the output
of the bridge, since you can't see a 100/120Hz ripple anyway. I get
rid of transients with large series opposed Zeners across the mains
side of the bridge, and size the series current limiting resistor for
each string to limit the current through the string to the specified
operating current at high line and take the brightness hit at low
line.

 

[Bob Stephens]

If I had to guess
(which I'm doing) I'd think that it was due to the eye's AGC and its
response to the sudden change in intensity.

That brings to mind a question that has interested me for a while. I know
nothing about optics, so please bear with me. Do people who wear the
polarized? eyeglasses which adapt to ambient light and become darker in
bright sunlight weaken the eyes' 'AGC' response? It seems to me that if you
always walked around in mild, filtered light, your eyes would become lazy
and less able to react to a sudden intense busrt of light like a flashbulb.

Anybody know anything about this?

Probably should have OT'd this...


Bob

 

[John Smith]

A few months ago I proposed a circuit to run LEDs on 120VAC 60Hz line using
a capacitor for the current-limiting element, and putting two LEDs in
parallel, oppositely polarized to conduct that current (maybe for
night-light use). The forward drop of one would keep the reverse voltage
low
for the other. One of the comments made concerned the surge of current
that
would be conducted at the moment of turn-on due to the higher frequency
content of the turn-on wavefront. Could this be the reason for the
brightness flash of red LED traffic lights at turn-on? Certainly the LED
array would have to plop right in where the incandescent lamp was, powered
by 120 VAC. But it's hard to imagine all the rather large, high-voltage
capacitors needed to do the current limiting on what I estimate may be 200
or more LEDs. One comment I just read, concerning the flash, said it's due
to poor regulation of the DC supply. But that implies the switching takes
place between the DC supply and the LEDs, and that's hard to imagine, as
well. Even worse, the use of capacitors on a large scale would wreak havoc
with the utility's power factor, not to mention the arcing of switch
contacts (but maybe the arcing is already taken care of since the old
incandescent lamps also produced a surge of start-up current) . So do any
of
you know FOR SURE just how these LED traffic lights are powered, and why
they flash brighter at turn-on?

Thanks,

Chuck

All states, AFAIK, require that LED traffic signals have a 0.9 power factor
or better. This comes from the ITE (Institute of Transportation Engineers,
http://www.ite.org/) specifications. Hence, all manufacturers, AFAIK, use
PWM or PFC ICs to meet the specifications. The flash on turn-on is indeed an
overshoot in the power supply usually due to improper compensation or
topology. I speak from experience and the company for which I once worked
will remain nameless.

Usually, the 120 VAC is applied to a circuit containing a PFC (Power Factor
Correction) IC with a transformer and, after rectification, puts out 30 to
100 VDC with a high power factor input. The LEDs are usually in
series/parallel strings of a sufficient quantity to meet the specifications
for light intensity. The total LED current is sensed and regulated. Lenses
may be used to focus the light, especially if the quantity of LEDs is small.
You may have noticed the sharp reduction off-axis brightness of some of
these signals.

Controllers at intersections use solid-state switches. The ones with which I
was familiar contained zero-crossing optical isolators to drive the triacs.

There is a lot more to it than this, but this may already be more than you
wanted to know.

John

 

[John Fields]

There is a lot more to it than this, but this may already be more than you
wanted to know.

 

[John Smith]

Well, let's see...

The current for the red LEDs have to just about double at high temperature
because of the drop in efficiency and Vf of the LEDs. But the green LEDs
required only about a 10% increase, as I recall. Yellow LEDs act about like
red ones and blue ones act about like green ones.

An early design I inherited was a boost PWM circuit with a string of LEDs
adding to about 300 V. There was a FET switch in series with the string of
LEDs which opened when no AC was applied. Why? Well, the filter cap was
large enough to cause the LEDs to stay on for awhile after AC was removed.
You can imagine what that does to traffic if it is a green signal. Worse
yet, the series FET switch failed a couple of times. It turns out that green
LEDs are so efficient that the signal may glow for minutes and be very
visible at night under a shorted switch condition. Very bad to have an
intersection with both directions of traffic seeing green. I learned
something very important from this. Always build a switching supply with a
transformer. That way, if your switch fails, output goes to zero. This is
true as well for getting 5V from a 12V battery for a computer. If you use a
series switch and it fails, good by computer. Yes, I know one can provide a
crowbar and fuse, but I don't like crowbars for reasons I don't want to get
into. My preference is for the power supply output to go dead for any power
supply fault I can think of.

Traffic signals have a CMC (Conflict Monitor Circuit) built in (this might
be specified by the ITE). Its function and its name have a distant
connection. Traffic controllers, as I said, use solid state switches
consisting of triacs and optos. If a triac fails shorted (or leaky), you've
got a problem with a signal not being able to turn off (a conflict). So, the
controller has a monitor which measures the voltage going out to the signal
and, if the SS switch is supposed to be off but the voltage is actually
above a threshold, it stops all normal signal sequencing and puts the
intersection in flash (flashes red in all directions simultaneously). In the
early days, the snubbers around the triacs allowed relatively hefty leakage
currents to flow. No problem for a 135 Watt incandescent to suck it up, and
it would keep the voltage low when the triac is off. Not so with an LED
signal. The nature of an LED signal is such that it will usually allow
enough voltage to build up at it terminals for the controller's monitor to
throw the intersection into flash mode. So, a circuit is added can draw the
snubber's current and hold the voltage to just a few volts unless the full
line voltage is applied.

And then there is the use of the CMC to force the controller into flash mode
if a certain percentage of LEDs fail. We used a cheap uC to monitor several
strings of LED currents and then disabled the CMC or, better, blow the line
fuse which accomplishes the same thing. I think the specs say that the
signal must appear as something like 500k Ohms or more for this fault.

We also had an autodimming option. We actually put a phototransistor in the
signal looking out the front. It caused a reduction in LED current at night
for two purposes. First, green signals are very bright to the eye at night.
They are blinding. This improved on that a lot. Second, it saves about 40 or
50% on power consumption. The spec said that all the signals in an
intersection must be coordinated, that is, dimmed at the same rate and time.
I didn't do that because there was no communications system. But, I learned
that nobody could tell the difference by eye. We always stated that we did
not meet the letter of the specifications. I don't think the end user cared.
But, I don't think many municipalities ordered it.

Our pedestrian Walk/Don't Walk Countdown signal had an 8405 in it. The micro
was used to learn the walk and don't walk cycles and then provide a
numerical display to show how many seconds remained to get out of the
crosswalk. It had to compensate for aperiodic sequences. Did you know that
policemen may be given a pushbutton to plug into an intersection controller
to manually control the walk/don't walk cycles? Usually near a stadium or
auditorium. I never saw this, but I was told that it was a possibility.

I've run down.

John

 

[John Fields]

Well, let's see...

The current for the red LEDs have to just about double at high temperature
because of the drop in efficiency and Vf of the LEDs. But the green LEDs
required only about a 10% increase, as I recall. Yellow LEDs act about like
red ones and blue ones act about like green ones.

An early design I inherited was a boost PWM circuit with a string of LEDs
adding to about 300 V. There was a FET switch in series with the string of
LEDs which opened when no AC was applied. Why? Well, the filter cap was
large enough to cause the LEDs to stay on for awhile after AC was removed.
You can imagine what that does to traffic if it is a green signal. Worse
yet, the series FET switch failed a couple of times. It turns out that green
LEDs are so efficient that the signal may glow for minutes and be very
visible at night under a shorted switch condition. Very bad to have an
intersection with both directions of traffic seeing green. I learned
something very important from this. Always build a switching supply with a
transformer. That way, if your switch fails, output goes to zero. This is
true as well for getting 5V from a 12V battery for a computer. If you use a
series switch and it fails, good by computer. Yes, I know one can provide a
crowbar and fuse, but I don't like crowbars for reasons I don't want to get
into. My preference is for the power supply output to go dead for any power
supply fault I can think of.

Traffic signals have a CMC (Conflict Monitor Circuit) built in (this might
be specified by the ITE). Its function and its name have a distant
connection. Traffic controllers, as I said, use solid state switches
consisting of triacs and optos. If a triac fails shorted (or leaky), you've
got a problem with a signal not being able to turn off (a conflict). So, the
controller has a monitor which measures the voltage going out to the signal
and, if the SS switch is supposed to be off but the voltage is actually
above a threshold, it stops all normal signal sequencing and puts the
intersection in flash (flashes red in all directions simultaneously). In the
early days, the snubbers around the triacs allowed relatively hefty leakage
currents to flow. No problem for a 135 Watt incandescent to suck it up, and
it would keep the voltage low when the triac is off. Not so with an LED
signal. The nature of an LED signal is such that it will usually allow
enough voltage to build up at it terminals for the controller's monitor to
throw the intersection into flash mode. So, a circuit is added can draw the
snubber's current and hold the voltage to just a few volts unless the full
line voltage is applied.

And then there is the use of the CMC to force the controller into flash mode
if a certain percentage of LEDs fail. We used a cheap uC to monitor several
strings of LED currents and then disabled the CMC or, better, blow the line
fuse which accomplishes the same thing. I think the specs say that the
signal must appear as something like 500k Ohms or more for this fault.

We also had an autodimming option. We actually put a phototransistor in the
signal looking out the front. It caused a reduction in LED current at night
for two purposes. First, green signals are very bright to the eye at night.
They are blinding. This improved on that a lot. Second, it saves about 40 or
50% on power consumption. The spec said that all the signals in an
intersection must be coordinated, that is, dimmed at the same rate and time.
I didn't do that because there was no communications system. But, I learned
that nobody could tell the difference by eye. We always stated that we did
not meet the letter of the specifications. I don't think the end user cared.
But, I don't think many municipalities ordered it.

Our pedestrian Walk/Don't Walk Countdown signal had an 8405 in it. The micro
was used to learn the walk and don't walk cycles and then provide a
numerical display to show how many seconds remained to get out of the
crosswalk. It had to compensate for aperiodic sequences. Did you know that
policemen may be given a pushbutton to plug into an intersection controller
to manually control the walk/don't walk cycles? Usually near a stadium or
auditorium. I never saw this, but I was told that it was a possibility.

I've run down.

 

[Harry Dellamano]

Ok John Fields, this is for you. You remember a while ago when I stated that
LEDs run in voltage mode have some advantages over constant current mode and
you called me every name in the book. Well this is why some traffic lights
flash.
As stated the LEDs must be Power Factor Corrected (PFC) and act like a
incandescent lamps which has a power factor of one. The PFC unit runs in
voltage mode and if the LEDs are to be driven by a constant current, another
driver and control loop must be added. The LEDs must turn on like
incandescent lamps and that is quite fast. At 60 cycles it is difficult to
get the current loop up and settled in 75mS with no overshoot, hence the
light flash. I wrestled with this problem for some time until I relised if I
threw out the current loop section and ran them in voltage mode, it would
follow the AC line at turn on with no overshoot and 1/2 the parts. Also the
temperature / intensity problem is greatly reduced.
Oh yes the LEDs must be matched in the array but when you buy them in large
batches, they come in different bins (voltage grades).
Come to think of it, I think only Tony Williams did not call me a SOB.
Thanks Tony.

Now JF, spank me with your keyboard.
Harry

 

[John Smith]

On Sat, 13 Nov 2004 00:21:19 GMT, "John Smith"

I made a mistake. It was not an 8405 but an 8055 in that pedestrian signal
of the last paragraph. Some other products (LED message signs) used up to 50
or so Cypress 8-bit micros.

Best compliment I've had in a long time and the only compliment ever from a
respected authority. Thank you.

John

 

[Jeff Stout]

In our system, it is the controller which dims the intersection. The
controller
would pulse the field relays (solid state relays) at the zero crossing to
change
the intensity of the lights. The controller keeps track of the time of day
and
knows when it is night and day. Our controller even kept track of daylight
savings time and leap year and leap day and would change its behavior
depending on time of day or the day of week.

But getting back to the light dimming issue. Dropping half sin waves as
a method of dimming the intersection really does a number on the NEMA
specification. NEMA specifies that a phase (a signal light) is on when the
when the voltage is more than such and such RMS and on for more than
T milliseconds. What does that specification mean when every third half
sin wave is missing? The RMS voltage at night is now lower than the
specification, but all the lights can be seen on the intersection.
Questions
like that drove me nuts when I designed our first Conflict Monitor.

For those who don't know, a Conflict Monitor monitors the actual voltages
on every signal light at the intersection (not what the controller says they
should
be, but what they actually are) and looks for conflicts. An example of a
conflict
is if all green lights in all four directions are green. A second example
is if all
the lights (red, yellow and green) in a single direction are on at the same
time.

Jeff Stout

Well, let's see...

[snip]



We also had an autodimming option. We actually put a phototransistor in the
signal looking out the front. It caused a reduction in LED current at night
for two purposes. First, green signals are very bright to the eye at night.
They are blinding. This improved on that a lot. Second, it saves about 40 or
50% on power consumption. The spec said that all the signals in an
intersection must be coordinated, that is, dimmed at the same rate and time.
I didn't do that because there was no communications system. But, I learned
that nobody could tell the difference by eye. We always stated that we did
not meet the letter of the specifications. I don't think the end user cared.
But, I don't think many municipalities ordered it.

They probably didn't order it because dimming is usually done in the
controller.

[snip]

 

[John Smith]

In our system, it is the controller which dims the intersection. The
controller
would pulse the field relays (solid state relays) at the zero crossing to
change
the intensity of the lights. The controller keeps track of the time of
day
and
knows when it is night and day. Our controller even kept track of
daylight
savings time and leap year and leap day and would change its behavior
depending on time of day or the day of week.

But getting back to the light dimming issue. Dropping half sin waves as
a method of dimming the intersection really does a number on the NEMA
specification. NEMA specifies that a phase (a signal light) is on when
the
when the voltage is more than such and such RMS and on for more than
T milliseconds. What does that specification mean when every third half
sin wave is missing? The RMS voltage at night is now lower than the
specification, but all the lights can be seen on the intersection.
Questions
like that drove me nuts when I designed our first Conflict Monitor.

Hi, Jeff -

I am aware of this method of dimming. We sent an engineer to investigate
problems at an intersection after the signals had been changed from
incandescent to LED types. I don't remember the exact complaint, but I think
there was flickering of the lights, perhaps some lights refused to even
light up. LED signals regulate the current through the LEDs so that they can
meet specifications at low or high line voltage as well as at temperature
extremes. By trying to regulate the light level with line voltage, you get a
confused regulator in the traffic signal. It doesn't like anything but a
constant line voltage and, regardless of what the line voltage is doing, it
will try to hold a constant LED current even with line fluctuations.

The problem was fixed by disabling the controller's dimming capability. They
later installed the signals with built-in dimming and everything was fine.

For those who don't know, a Conflict Monitor monitors the actual voltages
on every signal light at the intersection (not what the controller says
they
should
be, but what they actually are) and looks for conflicts. An example of a
conflict
is if all green lights in all four directions are green. A second example
is if all
the lights (red, yellow and green) in a single direction are on at the
same
time.

Jeff Stout

[snip]



We also had an autodimming option. We actually put a phototransistor in the
signal looking out the front. It caused a reduction in LED current at night
for two purposes. First, green signals are very bright to the eye at night.
They are blinding. This improved on that a lot. Second, it saves about 40 or
50% on power consumption. The spec said that all the signals in an
intersection must be coordinated, that is, dimmed at the same rate and time.
I didn't do that because there was no communications system. But, I learned
that nobody could tell the difference by eye. We always stated that we
did
not meet the letter of the specifications. I don't think the end user cared.
But, I don't think many municipalities ordered it.

They probably didn't order it because dimming is usually done in the
controller.

[snip]

Just because ours had a problem, doesn't mean other manufacturer's LED
signals won't work with controller dimming. I would urge careful evaluation
with LED type signals if the controller is allowed to do the dimming. At the
very least, contact the LED signal manufacturer and ask about this function.
A lack of communication may cost someone a lot of money for air fare, auto
rental, and motel.

John

 

[John Fields]

Ok John Fields, this is for you. You remember a while ago when I stated that
LEDs run in voltage mode have some advantages over constant current mode and
you called me every name in the book. Well this is why some traffic lights
flash.

 

[Jeff Stout]

I can understand how the design of an LED signal would more difficult if the
controller is dropping half sin waves periodically. But then I realized
that the controller can be programmed to not dim the intersection at night.
But then, the dimming problem remains.

Jeff Stout

 

[John Smith]

I can understand how the design of an LED signal would more difficult if
the
controller is dropping half sin waves periodically. But then I realized
that the controller can be programmed to not dim the intersection at
night.
But then, the dimming problem remains.

Jeff Stout

Not just difficult. I would think that designing a current-regulated power
factor controlled power supply would be next to impossible where the input
voltage has missing half cycles arbitrarily. But then there are a number of
brilliant design engineers in this group who could probably do it with their
eyes closed.

The problem disappeared when the controller was programmed to not dim and
the signals with the dimming option were installed. Everybody was happy.

John

 

[David Lesher]

Why the concern/spec over PF? Can you even notice that load on the
grid when you have 100's of gas-discharge street lamps running?

 

[Rich Grise]

Why the concern/spec over PF? Can you even notice that load on the
grid when you have 100's of gas-discharge street lamps running?

The I^2*R losses in the wires are very real, indeed. And more than
they would need to be with the power factor corrected.

Cheers!
Rich