Jeff Rule Build Log

Hello World!

My name is Jeff and I have been working with electronics and arduino’s for 5+ years. I am interested in starting to design my own boards to add arbitrary functionality to my electronic projects. You can see some of the work I have done working with my partner Nick on his Art at the link below. The art sculptures use sound, light, magnetics and mechanical movement all driven by arduino controllers.

The project I need to do next is a power supply and dimmer for some LED lights. We are working with these LED filaments that come in the edison style bulbs and they take a lot of voltage to work with. Much larger than normal LED projects. Each led COB (filament) consumes 60-70 volts DC and consume between 5 to 15 mA of current. It is very difficult to find this kind of power supply off the shelf. Then we want to do the typical and be able to dim them with high precision. This implies a dimmer that can handle 130+ volts. I want it in a pretty compact form as it will be embedded in a art sculpture. All these components to my experience are not available to work at these high voltages off the shelf. Most of that stuff is in the < 40 volt range. The filaments have 24 LED’s per filament and consum 1w / filament. The are consuming 0.15mV of current so that works out to about 2.8V / LED and that seems reasonable in the world of LED’s. Below is a photo of one of the filaments just barely lit (@60v) so you can see the individual LEDs.

A photo of the LED filaments I am working with.

Normally they drive these filaments from a rectified mains voltage with only a simple linear current in series with them. I am talking about the filament bulbs. The problem with this is that it is difficult to dim and have quite some flicker. They only start to glow when the rectified mains voltage reaches 60V or more which causes a lot of dead range with your dimmer. If you want to have accurate dimming, you need a high voltage power supply with some PWM control. If it is a once off project and you want it quick, you could buy a dimmable filament bulb, disassemble it and hook it to your filaments. This you should be able to dim with an external dimmer. Either way you are going to end up with a power supply which is higher than 60V so take safety in consideration.

1 Like

By high voltage do you mean higher than the approx 140+ VDC you normally get when rectifying 120V ac. I think that is just enough to drive them but does not leave any room for extra capacity if I would want it. But, I don’t think I need it.

Thank you so much for the input. Yes I am well aware of the voltages are quite high here. We did do one piece where were cannibalized the power supplies directly from the light bulbs. I also used a 120V AC zeroCross dimmer but that only gives 60 divisions of dimming (based on the model of 60hz for power) and creates an unacceptable ‘chunky’ dimming pattern, not enough divisions to create a smooth dimming pattern.

The lights are part of an fine arts sculpture so we have limited space and we are using a lot of the filaments (126). So the voltage part of the small boards in each bulb is correct but I feel I am certainly going to need to beef up the current. Those bulbs don’t have the correct components to drive that many filaments, at least that is what I assume. Now we typically will not be driving all bulbs at the same time, they are wired up into 27 individual segments (so we will need 27 individual dimmers) the final effect is to have have the light dance around the sculpture. If we do turn them all on, it would only be briefly as it would just be to much to look at.

Yes I am very concerned about the width of the dimming window for the voltage range. It is pretty small, so I think I would need to go with at least a 12bit pwm signal to give myself a range of values in that 60 to 70V range we need go produce light for a single filament or 120 - 140? range for two filaments in series (which is what we have configured it as) if it were linear 12bit should give me around 500 pwm values between 60 and 70v but it is not. So I don’t think I will know until I get something on the workbench to play with. I might need a 13 or 14 bit value PWM signal to do proper dimming. And I frankly don’t know much about the hardware necessary to generate a pwm signal in the first place. So much to learn and try!

The artists is going to be making any number of different sculptures using these filaments (a series) so the knowledge we gain here on driving and diming these filaments will be re-used in future projects.

Be careful about driving with ac. Most LEDs can handle only limited reverse voltage

Yes, understand. I was not clear. We used an AC dimmer between the 120V mains and the DC power rectifier (120V AC -> 140V DC). The LEDs were only driven off of power after the AC was rectified.

Finished all of Shine, just waiting on parts to assemble. Starting on Getting to Blinkly.

With voltage control you are working on the knee curve, so difficult to get the resolution you need and it will not be tolerant to voltage knee spread

Use current control instead, then an 8bit PWM is ok

Also, the knee voltage changes a lot as a function of temperature

Sounds like some good advice. Where can I look for examples of devices that use current control? I am not very familiar with this. Except of course in the context I think of power supplies that are current controlled. Each of these filaments draws 1W of power @ 130-140V DC and that works out to roughly 15mA of current. Can that be subdivided enough to get a dimming effect. Just putting that out there to make sure we are talking about the same thing. I will be driving between 4 to 8 filaments per dimming circuit so 4W to 8W being dimmed.

If you want to use the SP16CPC26 you need to get the supply voltage right otherwise you might run into dissipation problems. Say the knee is 60V at 15mA and your supply voltage is 63V. This means the IC takes the other 3V. Now if you set the current at 15mA (Full on) it means you get worst case : 16 outputs all pulling 15mA at 3V which comes down to: 3V x 0.015 x 16 = 0.72W. Now I see this IC has an exposed pad to improve cooling. For some reason I can not find the allowed dissipation of this package. I see there is a thermal shut down on board. But under normal circumstances you want to avoid hitting that limit of course.
However in case you run over the limit a solution could be to add an extra IC and spread the outputs over more ICs. As you can daisy chain them this can be easily implemented.
In case you do not want to use this IC and uses “hard” switches with a current regulator I recommend: the NSI45015 from On-semi. It is small 2pin SOD-123 type component. These you must put than in series with every filament. The problem is that you either need a micro controller with enough pins matching the number of filaments or you must us some other serial to parallel type IC. So in terms of part count this will incur more components and cost I think. I think the SP16CPC26 type solution is a more elegant and cost effective solution.

You probably won’t see many current control offline converters

I would use a FSL300 series buck converter, but instead of feeding the output voltage back to the floating switch, you feed the signal from an amplifier that senses the current into GND


If I have time, I will do a simulation tomorrow