Control/Dimming LED dimming

[JohnDAz]

we recently purchased a set of colorado batten 144s to light up our cyc. While they provide considerable flexibility when it comes to color wash, I can't get them to fade up smoothly. Regardless of which dimmer curve I use or how slow the fade-in time is, there is always a point where they transition from being off to bumping on - albeit at a low level but it is a noticeable bump when the rest of the stage is dark. I did an experiment where I set the intensity at different levels and slowly raised up blue to see at what point it turned on. With the intensity set at 100% blue bumped on at 0.4% according to my ETC Ion; at 50% it was on at 1%; at 25% intensity it popped on at 2%, at 10% it was 4%, and at 5% blue bumped on at 9%. It always bumped on at the same apparent intensity. When I set blue at 100% and rolled up the intensity the colorados bumped on at roughly 0.4%. The question is - is that just the physics of the LEDs? Is there an input threshold at which they go from being off to being on rather than the smooth fade up one gets as an incandescent element heats up? Fade-outs haven't been a problem. If I was just bashing lights up and down for rock and roll or dance competitions it wouldn't be a problem but as a designer I value the potential for subtlety more than flash.

What is the collective experience and/or wisdom out there?

 

[church]

LEDs are constant voltage/current devices; meaning to achieve the desired colour and intensity they are supplied from a fixed voltage and a current regulator maintains a fixed current. Dimming is usually achieved by changing the ratio of the time the LEDs are turned ON and the time they are turned OFF. The human eye does the rest. The problem with the dimming step is dependent (mostly) on: the switching speed of the switch in each LED string usually a FET; the clock speed and refresh rate of the control chip that controls the FET switch: the number of steps available in the control register that determines how many steps are available to vary the ON/OFF ratio; and the response time of the LEDs. Note to avoid the human eye perceiving this switching as flickering the switching speeds must be significantly faster than the response time of the human eye - must be highher than 200Hz and needs to be higher still to avoid problems with cameras. Within these limitations you also have to factor in trying to match the dimming curves with traditional systems.

It is possible to dim LEDs by varying the analogue voltage but this will change the colour. I cannot think of a stage LED fixture where the colour of the LEDs changes when they dim. This is one of big advantages for me of this technology.

TRIAC control of LEDs is also available - normally for situations where the LED is replacing an incandescent lamp, National Semiconductor make control devices for this application which allows them to work with standard architectural controls.

I own two different types of LED fixtures and they have excellent performance. The newest fixtures have much better dimming than the first ones purchased 3 years ago and reflect the improvement in technology. I use both fixtures alongside tungsten fixtures. the older LEDs have a just noticeable step when turning on or off I always use them in a way that minimises this limitation while making the best use of their strengths i.e. saturated.

I have only used the Colorados twice I did see what you describe just not as pronounced. Sometimes a different board can minimise this problem as well.

A couple of links

http://www.ti.com/lit/an/snva605/snva605.pdf
http://www.st.com/web/en/resource/technical/document/application_note/CD00185256.pdf

 

[JD]

One other idiosyncrasy is DMX itself. You basically have 256 steps of resolution. Step #1 is about .4% !
With incandescent lamps, the filament warm up time at low settings covers this nicely. High $ LED units actually ramp from DMX 0 to DMX 1 to simulate the turn-on. An incandescent lamp is not visible at .4% but an LED will be, which makes things worse, especially if they are facing you.

The only work-around is to cue them late on an up-feed and hope that the other stage lights distract from the bump.

 

[JChenault]

It is possible to dim LEDs by varying the analogue voltage but this will change the colour. I cannot think of a stage LED fixture where the colour of the LEDs changes when they dim. This is one of big advantages for me of this technology.[/url]

I know that your,attachments mention this, but I have never noticed,a color change when battery powered LED lamps start getting low. Is it possible your source is in error, or that the change is not at all similar to incandescent?

 

[JChenault]

One other idiosyncrasy is DMX itself. You basically have 256 steps of resolution. Step #1 is about .4% !
With incandescent lamps, the filament warm up time at low settings covers this nicely. High $ LED units actually ramp from DMX 0 to DMX 1 to simulate the turn-on. An incandescent lamp is not visible at .4% but an LED will be, which makes things worse, especially if they are facing you.

The only work-around is to cue them late on an up-feed and hope that the other stage lights distract from the bump.

I believe the high quality stage units also use a curve that matches the response of the human eye. Ie if it sees a D!X value of 0x01, it puts out much less than 0.4% voltage, at the top of the curve ( say from 254 to 255) the change in voltage is much greater. Incidental your conventional dimmers do this as well.

We are also starting to see 16 bit dimmers. These tend to just give you 65535 steps of intensity, but I don't believe they give you a "realistic" dimmer curve so that 90% to 95% is as much change to the eye as 5% tp 10%

 

[derekleffew]

... I cannot think of a stage LED fixture where the colour of the LEDs changes when they dim. This is one of big advantages for me of this technology. ...

Some people happen to like the effect of incandescent amber-drift, and for those designers, ETC has thoughtfully included it as an optional mode in the Desire series.
From ETC Products - Desire D40:

The additional Red Shift option allows the Desire fixtures to mimic a tungsten fade, ...

JohnDAz, it's simple really: the more expensive the LED fixture, the more thought and care has gone into the dimming curve, particularly noticeable at the low end. One of the primary reasons, aside from intensity, a $2500 units costs more than a $500 unit.


Did gafftapegreenia ever publish the results of his "LED shootout"? I seem to recall he analyzed at length various fixture's dimming curves.

 

[dj41354]

The question is - is that just the physics of the LEDs? Is there an input threshold at which they go from being off to being on rather than the smooth fade up one gets as an incandescent element heats up? Fade-outs haven't been a problem. If I was just bashing lights up and down for rock and roll or dance competitions it wouldn't be a problem but as a designer I value the potential for subtlety more than flash.

The "pop on" problem you're talking about isn't with the LED's in the fixture, it's with the electronics that control them. Most LED fixtures use PWM to control the brightness of the LED's. As was stated, PWM is a scheme where the LED's are turned "on" and "off" fast enough that you can't really tell they're turning on and off at all (there are some good reasons this scheme is used instead of simply controlling the current in the LED's.. but that's another subject). Again as was stated, for this to be "fast enough to not see", the LED's need to make (at least) 200 "on/off" cycles in one second (5msec per "on/off" cycle). If the "on" time and the "off" time are equal, the LED's are being driven at half power.. make the "on" time longer than the "off" time and the LED's look brighter.. make the "on" time shorter than the "off" time and they look dimmer.

Here's where it gets a little technical, but it's important to understand the terms "period" and "resolution" as they relate to PWM. The 5msec on/off cycle time of the LED's is called the "period" of the PWM implementation. The amount of on time compared to off time is called "duty cycle", and is normally expressed as a percentage ratio of the on time to the "period" (on time/period X 100). So for example, 50% duty cycle is when on time and off time are the same (on time is half the period)... 100% duty cycle is when the LED is on for the complete cycle (never turns off).

The "resolution" refers to how finely the 5msec "period" can be sliced up by the electronics to do the on and off timing. Microprocessors are typically used for these circuits, and as such, "resolution" is determined by the "number of bits" dedicated by the microprocessors internal hardware for this task. 8bits and 10bits of "resolution" are fairly common, and represent 256 slices and 1024 slices respectively. Here's the important part... The dimmest "on" setting for the LED, is when one(1) of the "resolution" slices turns the LED on, and the rest turn the LED off... so with 8bits of "resolution" and 5msec of "period", that's 1/256 of 5msec, which is about 0.02msec "on" and 4.98msec "off". With high powered LED's, that's more than enough "on time" for the LED to cast considerable light. A shorthand way to think about this, is to divide the maximum lumen output of the fixture by the "resolution", and the result is the lumens generated when the fixture is at it's lowest "on" setting. So if your starting out at 3000 lumens, with 8bits of resolution, the minimum "on" setting is something like 10-12 lumens (3000/256).. that's a pretty considerable "pop on". It's better at 10bits of resolution.. where the minimum setting is in the range of 3 lumens, but even that will be seen to "pop on", especially if there are multiple fixtures being used. It turns out that to get smooth dimming at the bottom and a minimum amount of "pop on", you really want 12bits (4096 slices) of resolution or better. It also turns out that you won't typically find this spec on a mfgr's data sheet.

 

[JD]

LEDs do have quirks, but all can be worked around by having a fixture with the proper software and power supply. That usually means that you will get what you pay for. The only quirk which is a stopper has to do with the monochromatic nature of the LEDs themselves. It produces a very jagged spectrum. This is why amber and white LEDs have been added to the RGB mix. In actuality, there is no such thing as a white LED. These are actually LEDs that contain phosphor to convert and broaden their output.

Still, as an LD, you must look at the light thrown from a fixture and decide if it is pleasing to your eye. Again, more expensive fixtures will usually be more pleasing. To date, it is still not as natural as an incandescent source. (IMHO)

 

[church]

I know that your,attachments mention this, but I have never noticed,a color change when battery powered LED lamps start getting low. Is it possible your source is in error, or that the change is not at all similar to incandescent?

It does happen but it is unlikely that you will see for a battery powered device due to the voltage and current constraints associated with the power source. You will note different results using a power supply with a regulated output voltage and a constant current source as is frequently found in LED fixtures.

It is most noticeable if you have two Identical LED sources side by side - if you increase the power supply voltage on one source you can see a change but you probably would not notice it on a single fixture without the comparison.

This seems strange because coloured LEDs are bandgap devices which means that the colour of the light they produce is dependant on the impurities used in the doping of the "p" and "n" type semiconductor so you would not initially expect the wavelength of the emitted light to change. In practice when you increase the forward bias voltage across the LED you also increase the power disipation in the LED which increases the pn junction temperature. As the pn junction temperature increases the bandgap energy of the semiconductor decreases changing the wavelength which is visible as a shift in colour. There are also other effects that happen as the junction temperature changes and you have to look at Shockleys equation for this there is a very readable explanation on the DigiKey website at the link below.

Driving LEDs: How to Choose the Right Power Supply - Lighting Solutions | DigiKey

 

[JD]

church is right, not much of a shift on the monochromatic ones (RGBA.)
The shifts I have noticed happen on the white LEDs. The source color (blue/UV) probably does not change much, but the way the phosphate in the LED reacts to it appears to. In most cases, the "white" balance shifts upward in color temperature. The phosphorous used is actually a mix of several phosphates. The sensitivity of the different compounds is not the same, so as the exciter dims, the balance changes.

Again, on RGBAW units, it is possible for the manufacturer to predict this shift and compensate with the other diodes in much the same way that software can be used to give RGBAW units an artificial "incandescent" dimming color curve.

 

[Esoteric]

One other idiosyncrasy is DMX itself. You basically have 256 steps of resolution. Step #1 is about .4% !
With incandescent lamps, the filament warm up time at low settings covers this nicely. High $ LED units actually ramp from DMX 0 to DMX 1 to simulate the turn-on. An incandescent lamp is not visible at .4% but an LED will be, which makes things worse, especially if they are facing you.

The only work-around is to cue them late on an up-feed and hope that the other stage lights distract from the bump.

This is what I have run into. There is no way around this problem that we have discovered on the lower priced units. Even on the best LED units out there if you look hard enough, you will see this issue (although 99,999 out of 100,000 will never see it on a VLX while 1 out of 3 will see it on your typical $300 LED PAR). Our Visioneer products use a 32 bit dimming curve to get around this (and it works very well), but the off side is that because of the resolution involved, it throws off your fade timings (a 5 sec fade on your console now takes longer than 5 seconds).

Mike

 

[JChenault]

This is what I have run into. There is no way around this problem that we have discovered on the lower priced units. Even on the best LED units out there if you look hard enough, you will see this issue (although 99,999 out of 100,000 will never see it on a VLX while 1 out of 3 will see it on your typical $300 LED PAR). Our Visioneer products use a 32 bit dimming curve to get around this (and it works very well), but the off side is that because of the resolution involved, it throws off your fade timings (a 5 sec fade on your console now takes longer than 5 seconds).

Mike

Wait. This is not making sense to me.

If the software running the console is semi reasonable, the time to fade should be independent of the number of bits.

Or do you mean the Visioneer creates internal values adds a few ms to the fade time? ( kind of like an incandescent filament)

 

[Esoteric]

Wait. This is not making sense to me.

If the software running the console is semi reasonable, the time to fade should be independent of the number of bits.

Or do you mean the Visioneer creates internal values adds a few ms to the fade time? ( kind of like an incandescent filament)

Yes, in 32 bit mode it adds values to the fade time internally at the unit. Once you figure out how it works, it isn't so bad, but it almost appears random at first how much time it add to the fade (because it is dependent on the number of steps that must be crossed and the amount of time of the original fade). This has the effect of making beautiful fades, but at seemingly random fade times (although they are not random, there are just many variables). Of course you can operate the unit in conventional mode which presents you with an identical dimming curve to Elation, Chauvet, and Blizzard units (at lower cost though). Or you can assign the 32 bit dimming to 1/4 of the curve while leaving the other 3/4 conventional.

The original issue is simply what you get at that price point. We are always brutally honest with our clients about facts like that (along with the CRI of sub $1000 LEDs, their tendency to overload cheap single CCD cameras, etc).