Low voltage DC Dimmer ( for Aussies and kiwis )

[mrpackethead]

I'm tinkering with a dc dimmer. Everything we use down here is low voltage (12-48VDC) the most common seems to be 24V.. Most of the strings come with a transformer, and run on AC, but the strings work equally as well on DC. Anyway, i think that running everything on low voltage dc, makes a lot of sense, and makes control, and safety all very easy.. ( no need to mess with any mains voltages )..

The design in my head is an 8 channel unit, dmx control on a small pic..PWM output from a PIC micro/dmx is well established.. ( thanks JEC ala the pixel projects ).. *I was about to go and prototype the attached output circuit, and i thought that perhaps i should see if anyone can make some intellegent comment or improvement about it..*

Instead of fusing the output, i'm thinking that i'd like to put some active protection into the circuit.. Shoudl the current flowing across the sense resistor exceed some threshould, the npn transistor should take the input to the fet low, and hold it low.. shutting the fet off.
Choice of component values will need to be approprate of course. This shoudl protect things, if the output is shorted etc.. Perhaps a fuse as well?

End design idea is that i can stick the whole thing into a peice of pvc pipe.. 24V DC in, dmx in/out, and 8 "squid tails" for the outputs to the lights.

I'm thinking of a very low cost FET, it was designed for automotive use.. about 0.50 each.. An entire board and componets should be in the order of ~$16.00, or about $2/channel..


Your intellegent and freindly comments are very welcome and encouraged.

[tconley]

Have you looked at this one?

http://diylightanimation.com/index.php?topic=337.0

[Macrosill]

That is a nice breakout board to convert a pc psu to a usable psu for testing and such with actually converting it. I personally like the old style conversion with banana plugs on the psu case. Its one less peice of equip, pcb, to short out by accident.

[joneslights]

I could use a low voltage dimmer for some lights that I have here.

[mrpackethead]

well the circuit works mostly, i need to make some allowance for the in-rush current of lamps.. it was triggering my over currnet sensors.. Some integration of the currnet over time, might help but it ups the complexity..

Hey the fets are cheap enough anyway if you blow one or two.. though its a pain to replace them.


I was quite sure of teh relevance of the ATX power supply conversion, other than it probably woudl make a good supply of DC power, if you wanted to use it that way. Unforuantly most of them have 12V outputs and most of the lights we have are 24V..

anyway im thining of makgin somwthing that shoudl work anywhere betten about 9V and 48V, and im not stickign a power supply on it, which means it shoudl be pretty flexible for anyone to stick whatever they like on it.. ( hey you could run it off a ligether adaptor in your car!! ).. blinky on wheels!

[P. Short]

Depending on the value of the resistors R2, R3 and R4, it strikes me that the over-current transistor will oscillate on and off. Basically, the over-current turns the transistor on, which shuts the FET off, allowing the transistor to turn off and the FET to turn back on.

[mrpackethead]

thanks Phil, i think you're right, i need to stick the scope on this and have a look and see what happens.. I really need some kind of latch to hold the transistor on..

The other alternative, is that i feed this signal back to the PIC as an interupt.. it could simply shut down the impacted channel, untill its power reset... I'd still put a fuse on board i think...

what do you think?

[P. Short]

This is outside an area that I've spent much time on, and I'm beginning to suspect that I was wrong in what I wrote above. It's been many decades since I've looked at the linear behavior of FETS, it's time to go back and read the relevant chapter in AOE.

[mrpackethead]

I think you've set a world record. Someone on DIYC saying I might be wrong :-) its the mark of an expert. Anyway, all things aside...

It won't be at least till the weekend for me to get a chance at looking at this.. I might see if i can model it in spice though...

This is a first time trying to do something like this for me.. I think its time to go back to fundementals and think long and hard.

I'm still quite tempted to take some feedback to the pic, and shutdown the channel. maybe even put a 'short' led on the board to indicate a short condition.

[gorehound]

I ran this in multisim using random values for the cap and resistors and just put an LED on the output and it blinks. I used a function generator for the PWM. I have no clue what I am doing, I just like to simulate circuits to see what they do so go easy on me.
http://i74.photobucket.com/albums/i251/gore_hound/dimmer.jpg

[mrpackethead]

Try

C1 = 100pF, R1 = 10k, R2 = 0R17, R3 = 100, R4 = 10k.

r2 is the sense resistor.. In this case i'm looking to shut the fet down if theres more than 4A flowing.. You need to develop about 0.7V across it, in order to turn the transistor. With 4A flowing across a 0R17 resistor you get your .7V

Then instead of using a Led, stick a 'load' on the output of varying sizes.. you'll need to use something thats going to draw more than 4A, and then see if the circuit shuts down.

[mrpackethead]

Hi All, i made some progress on this today.. ( another wet miserable day here ), and i put together a bread board circuit for this, and it works nicely.. I poked and prodded with the scope, and also modeled it on the computer, and now understand whats going on at the limits of operation..

I've decided that i'm going to make this board as simple as possible, and only put the bare minimum of components into it, making it as flexible as possible. I'll be designing it with DMX in mind, but i think it will only be a software change to make it REN compatiable.

* Power supply will be 12-48VDC, ( match to your lights ).
* 8 Channels output, each good for up to 4A per channel.
* DMX input.. Starting address will be coded in the firmware
* output connectors will be screw terminals, for attachement to 'squid tail'
* input power connector will be screw terminal.
* serial input will be a 2 x rj45 connectors ( in and out )
* current limiting circuit for channel protection, with fuse on input
* will be designed to be installed in pvc pipe, for outdoor installation.

Hang five for a few days.


Pending the wife allowing i might even draw up the circuit today, and get started on a pcb.

[mrpackethead]

Small modification to improve reliability sent me back to the drawing board.. Thanks to JEC and Phil for some inspiration.. I had a situation, where the main switching fet would have been in some situations operating in a linear mode.. ( ie, half switched on ).. Thats a good way to make them hot!.. Also, using the transistor to switch on the over current protection, requires .7V to be developed on the sense resistor.. Not really a desirable situation, as potentially, this could mean several watts needing to be dropped, and 5W resistors are not only large, but they are also expensive.. So i've changed the current sense, to use a LM339 Compartor.. I only need to drop about .1V across the sense resistor now, meaning, i'm only looking at about .3W to drop.. Im tempted to see if i can drop down to .05V, though it might suffer from reliablity..

The compartor is set up as an inverting output.. Ie, provided the current flowing is below the shutdown threshold, the output will be high. That output holds a small signal fet on.. ( think of a kind of enable ).. THe PWM signal flows through that fet.. So, if the enable fet is off, the main fet is off, as well.. The protection circuitry, adds a few components to the build, but mostly some resistors, and its only a few extra cents..

The problem is that by not having the protection circuit there, that if you short out the output, theres a good chance you'll fry the fet, even if you have a fuse, as the fuse wont act fast enough..

Circuit diagrams, and scope traces coming soon.

[mrpackethead]

Circuit diagram for the dimmer section..

Some tweaking of R6 and R3 may be required to get the high currnet set point right, but its working well now for me.. I'm intending to set the high current set point at 3A.. however the fets are good through to 16A...

Won't be too far away, i'll have the full schematics finalised for the controller board.

[mrpackethead]

Well, i finally got round to finishing up the schematics, and pcb.. Please see attached.. It all takes time, and i've had a very busy time at work.. the prototype worked nicely, and now i'm going to run off a beta run of boards.. I'll get about 15 boards. here's hoping that i didn't make any ugly mistakes on the pcb.

Just a heads up on the functionality and design.

(a) Power input can be somewhere between 9 and 48V DC
(b) control is via DMX input.. (rj45 connectors are pinned out the same as JEC's pixels)
(c) Dimming control is via PWM modulation.
(d) The fets are good for 16A each. The tracks on the board are probably not.. they could thickened up i guess if you wanted more current capability..
(e) each channel features over-current protection.. This cuts in very quickly in the event of a over-current or short circuit , much quicker than a fuse will blow.. The idea is to protect the FET, which while cheap would be painful to replace.. The current limit point is set by the low value resistor ( in this schematic it is .1ohm for 4A ).. .2ohm would limit at 2A, etc..
(f) its four channels ( if thats not obvious )
(g) code for the pic, is almost identical to JEC's pixels. ( thanks for that John )

I've ordered enough parts from mouser to build quite a few of these. ( the parts for this are remarkably cheap

The FET is an automotive part, its $US0.37 at 100off, the lm339 is about 0.24c.. pic about $1.50, and the max485 about 0.80c.. a hand full of smt resistors and caps.. I figure all up cost of components at about $6.50.. PCB at about $10 a board.. Maybe a bit less. It will fit in a hunk of pvc pipe, will will all work fine for me..


This design will drive LEDS or Lamps.. If your driving LEDS you'll need to current limit of course.. Many of the MR16 bulbs that people are using are already current limited internally.


I'll keep you posted as things progress, with the pcb production. Gotta have this finished for this year.. Am sorta dependant on it now..

FYI. I won't be running a co-op or selling boards somewhere else.. However if anyone wants to, then all the info, files, code etc will be available for you.. PCB and schematics are in Eagle.

NB.. You'll note ( if you are observant ) that this is almost all SMT.. I've deliberately chosen components that woudl all be easy for a newbie to SMT to work with.. And since nne of the components are too expensive the risk factors are pretty low.

[mrpackethead]

Parts from mouser arrived.. Just four days from order to delivery, half way around the world.. I'm impressed.

[Wombat]

Dude I dont think that output protection circuit is going to work.

You have the output of the 339 normally high which will turn the transistor on all of the time shorting the gate to ground and mosfet will not turn on ever.

You also are shorting out an output of the PIC so that may not be good for the PIC. why didnt you just connect the output of the 339 directly to the gate of the Mosfet and have a 1k resistor to the output of the PIC. Also having a cap between the gate and Source will cause it to potentially oscillate as it will pulse the gate low every time the mosfet turns on not to mention the slower rise time of the gate the cap will cause.

I think you need to redo the output circuit
Wombat

[Matt_Edwards]

Guys,
Love this site.
I too have an interest in low voltage dimming.
Have you thought of using a series element like Polyswitch and then limit any surges with the Varistor?
Cheers
Matt