100W LED Cob bead

[gyro_gearloose]

10 bit gives you 1024 steps (so a switch frequency of about 1kHz?)

You’d need a switching frequency somewhat higher than that. Image a 10-bit 50% PWM signal at 1khz. It will be on for half a second, and off for half a second. You’d need a switching frequency hundreds of times higher than that if you don’t want to see the LED flickering.

[alistermorton]

OK.

 

I need to do some reading up on how this is done, I think.

 

EDIT: Ah, light breaks (as it were).

[Shez]

EDIT: Also - the RC4 dmx2dim (for example) does very smooth dimming of LEDs - much better than the cheap dimmers you get on eBay, yet it's being fed the same DMX data. I'm guessing there's some additional slew rate limiting going on with the pulse width transitions.

 

The SX Series of products now includes 19-bit dimming, which is eight times finer than 16-bit with a whopping 524,000 steps of resolution. For users, this means exceptionally smooth LED dimming, especially at the bottom end of the curve, which can be problematic. All SX dimmers also feature dimmer curve crossfading to each individual dimmer; this enables users to tweak the dimming curve 256 steps for each separate dimmer output. Dimming is suddenly as smooth as silk.

 

Interesting!

[richardash1981]

This may be more complex that it first seems, LED chips need a constant current power supply, or for dimming a variable but regulated current.

As a cheap implementation One of these (loads sellers, also found on Amazon Marketplace etc.). They run on an (unregulated) DC input and provide the regulated variable current output.

 

I've not tried one (because I was driving resistor-ballasted LEDs, so I got the voltage output version) but based on that I would expect a respectable (but not stellar) result - the first step is fairly dim, and smooth from then up if your DMX refresh rate is high enough - 10Hz isn't, because you can see the jumps between refreshes unless you move the fader very slowly!

[Gerry]

 

As a cheap implementation One of these (loads sellers, also found on Amazon Marketplace etc.).

These atren't suitable, as I stated in my original post.

They only output 350/700Ma. Need an output of 3A to run a 100W LED Cob bead.

Cheers

Gerry

[alistermorton]

EDIT: Also - the RC4 dmx2dim (for example) does very smooth dimming of LEDs - much better than the cheap dimmers you get on eBay, yet it's being fed the same DMX data. I'm guessing there's some additional slew rate limiting going on with the pulse width transitions.

 

The SX Series of products now includes 19-bit dimming, which is eight times finer than 16-bit with a whopping 524,000 steps of resolution. For users, this means exceptionally smooth LED dimming, especially at the bottom end of the curve, which can be problematic. All SX dimmers also feature dimmer curve crossfading to each individual dimmer; this enables users to tweak the dimming curve 256 steps for each separate dimmer output. Dimming is suddenly as smooth as silk.

 

Interesting!

 

Even the previous series were good.

 

I've been impressed with the service from Theatre Wireless, too. The head honcho called me from the states on his mobile to confirm the pinout of the DMX out port (I couldn't find it documented and needed to make a 3.5mm to XLR lead that afternoon) after I emailed them, instead of trying to sell me a prebuilt lead.

[gyro_gearloose]

 

As a cheap implementation One of these (loads sellers, also found on Amazon Marketplace etc.).

These atren't suitable, as I stated in my original post.

They only output 350/700Ma. Need an output of 3A to run a 100W LED Cob bead.

Cheers

Gerry

 

One of these? They aren't cheap, and they don't appear to dim all the way down to 0%

[Gerry]

 

 

One of these? They aren't cheap, and they don't appear to dim all the way down to 0%

Correct not cheap also not DMX controlled but I could add a DMX decoder to provide the "potentiometer voltage"

Cheers

Gerry

[gyro_gearloose]

You could, but the driver cannot dim all the way down to 0%. It needs a relay on the incoming mains supply to cut the power to acheive complete darkness.

[boatman]

The problem with cheap LED controllers is that they are only 8-bits and any LED run at a PWM ratio of 1/256 is still quite bright. You need to translate the 8-bit DMX value into a 16-bit PWM value using a square law like this (((DmxValue + 1) ^ 2) - 1). That will translate 0-255 into 0-65535 and give you a proper dimming curve.

 

This would be ideal but very few drivers will run or switch fast enough to do 16 bit PWM. You generally have to fudge it by messing around with the bit periods so that the lower-order bits are very short and the higher order bits are long, which gives you smaller dimming steps at the bottom.

 

The ATmega328 (as used in Arduino) running at 20MHz can produce two channels of true 16-bit PWM at 305Hz in hardware. It works perfectly.

[musht]

 

I'm aware of the LM2596 type boosters but they give a settable fixed output for a variable input so as I increase the DMX signal to a DC output above about 4V DC the full 36V DC is output from the booster.

Cheers

Gerry

 

2596 is a buck , step down, trick is usually to integrate the PWM signal to change the voltage appearing on the Vref the regulator is working to:

 

http://www.hackvlog.com/2015/08/how-to-control-lm2596-buck-converter.html

 

http://www.analog.com/en/analog-dialogue/articles/convert-a-buck-regulator.html

 

Been meaning to try out the XL4015 , also available for less than you could post it for in UK, higher voltage range and current than 2596.

[timsabre]

The ATmega328 (as used in Arduino) running at 20MHz can produce two channels of true 16-bit PWM at 305Hz in hardware. It works perfectly.

 

Yeah, but you can't pwm a constant current switcher with it, because the switcher has a minimum "on" time which is a lot longer than 1 bit period. So nothing happens at the low end dim levels where you were hoping for the nice smooth dimming, until the pwm on time exceeds the switcher's on time.

[musht]

The ATmega328 (as used in Arduino) running at 20MHz can produce two channels of true 16-bit PWM at 305Hz in hardware. It works perfectly.

 

Yeah, but you can't pwm a constant current switcher with it, because the switcher has a minimum "on" time which is a lot longer than 1 bit period. So nothing happens at the low end dim levels where you were hoping for the nice smooth dimming, until the pwm on time exceeds the switcher's on time.

 

Yeah, but, leave the switcher free running and alter what it sees via its Rsense pin.

[alistermorton]

So, can I make sure I've got this clear in my mind:

 

For small loads, where a resistive dropper is acceptable and a constant voltage source can be used, it is relatively trivial to stick a power mosfet in the circuit and switch it at whatever rate and m-s ratio is need to get nice smooth dimming, but for bigger loads, where a resistive dropper would be unacceptable, it's necessary to use a constant current (or at least actively current limited) source, and switching this kind of source is more problematic, due to the settling time needed for the regulation.

 

Is that right?

 

I've also got a few old minims kicking around that would be good experimental mules for LED but it's looking as if the dimming is (as ever) going to be more problematic than I'd first envisaged.

[timsabre]

leave the switcher free running and alter what it sees via its Rsense pin.

 

Doesn't work. If the PWM pulse is shorter than the switcher's minimum "on" time then you're stuffed. Most switchers run at 500KHz-1MHz so at 200Hz pwm the best you can hope for is 12 bit (4096 steps) and in real life it's even less. One trick is to dynamically alter the drive current of the switcher so once you get down to a low dim level, you switch to say 100mA output and then PWM that, however it is very hard to get a smooth changeover.

 

 

Alister, yes that is right. Even a linear constant current regulator like an LM317 does not like being switched faster than about 250KHz.