Dimmer to end all dimmers

[djacobow]

I've recently completed a small project (well, mostly) to create a simple incandescent lamp dimmer that follows a short program and dims the light from a starting set point to an ending set point over a longish period.

The application is putting the kids to bed with a light that almost imperceptibly slowly turns off.

Did it with a triac, opto-triac driver H11AA1 opto ZC detector, and an Arduino, and a board I sent to Seed studio.

Happy with it, but incandescents are on their way out and I am really keen learning more on power control, Internet of Things, etc, and would like to do this better.

The next stage in "better" seems to be reverse phase control. The topology I see most is an IGBT or FET shorting the DC side of a bridge rectifier, where the AC side is connected in series with the load to the AC power source.

The next next better is "true sine wave dimming, which is basically just PWM. I'm wondering, though, can this same topology work for PWM dimmingl? I would think something like 40kHz (above the range of human hearing). This should not be too much less efficient than reverse phase dimming (FET is switching more, but still mostly on or off, assuming its fast enough) and filtering back to a clean sine wave should be easy since 40kHz is so many octaves away from 60 Hz. And PWM allows me to get rid of the ZC detection entirely. This also allows me to get rid of some interrupt routines and just levergage the PWM hardware the uC. It sounds good if the circuit can be cheap and simple.

An alternate approach would perhaps dispense with the bridge rectifier and instead use complementary FETs, driven by inverted signals (I think, though, that this would require the ZC detection again, though.)

What are the gotchas here?
What is the gold standard circuit for high-frequency PWM of AC mains yielding a scaled "copy" of AC mains?
And can I buy this as a module already made?


Regards,
Dave J

[nickm]

First thing is figure out the compatability of the lights you are switching to.  Some LEDs won't work with adjusting the amplitude of the sine.

As far as problems with changing the pure sine scheme switching losses might get you.  40kHz is in between the transition from IGBT to MOSFET territory. 

Answering the last two questions with a device not intended with this application.  The easiest way to do what you want to do is buy a 70V or 100V audio amplifier and feed it with the signal you want to dim with.  Just make a 60Hz sine wave with whatever controller you are using and gradually increase the amplitude to mimic the pure sine you're talking about.  They might be fine with incandescents because they look like resistors but CFLs or LEDs might cause problems because they are non-linear loads.  Actually now that I think about it do what the cheap 70V/100V manufacturers do and take a 4/8ohm amp and put a step up transformer on the output.  1:2.5 turns ratio should do it.  It would be fun just to see how it performs.  Redneck dimming.

[johansen]

you can make an AC buck converter with 4 switches.

very painful to build and since you'll be limited to <50Khz, you will need a not very cheap inductor and capacitor on the output, as well as the input of the buck converter.

[djacobow]

Well, I understand that not all loads will be happy with all dimmer approaches. That's an unfortunate given. I would think, though that most loads would be happiest with the sine wave input, but I would not at all be surprised that some buck converter in an LED module is not.

Most of the information on dimming comes from the theater lighting industry. Triac dimmers are known for being flickery, hard on bulbs, and noisy, both electrically and audibly -- the filaments will often "sing." Reverse phase control helps relieve this since the voltage rises sinusoidally, and is only cut off abruptly. I'm not entirely sure why this is better than the other way around, but consensus is that it is. Most of the "for LED" or "for CFL" dimmers you see use this method.

The sine wave method obviously will have no high harmonics (if you filter them well) and that should be the easiest on incandescents for sure (and motors, etc) but yeah, who know what happens to power electronics downstream. However, I have tried dimming LED modules with the output of my variac, and the results were promising. So perhaps some of the LED drivers are happy with reduced voltage sine waves.

The theater industry appears to be moving to sine wave dimming using PWM at super-audible frequencies. They may be much more incandescent-oriented, though than the rest of the world is heading.

[johansen]

it would appear to me from a quick googling that sine wave dimming refers to the 4 switch ac buck converter i described above.. but without the output capacitor and inductor.. this is complete blasphemy lol. might work for lightbulbs and motors but that's it.

"igbt dimming" appears to refer to be what you describe, cutting off the tailing end of the sine wave.. this should work with led and CFL bulbs.
but by "work with" i mean: doesn't light on fire and or cut operating hours to less than 1000.
there's no guarantee the led will actually dim properly.. it may sit and flicker, as some of mine do.

[richard.cs]

Leading edge dimmers (where you get the first part of a smoothly rising sine wave) are better for feeding rectifier-capacitor input loads and normally specified for CFLs and LEDs that contain switchers. Trailing edge (a conventional triac dimmer where you get a sharp rise and then the remainder of the sine wave) are very bad for these as it puts huge spikes of current into the input capacitors but are better for inductive loads where you don't want to have to suddenly interrupt the current  (motors and iron core transformers).

It is possible to do leading edge dimming with SCRs/Triacs (and historically it was often done this way for theatre lighting) but a mosfet in a diode bridge or a pair of source-source mosfets is probably easier. To make it work with SCRs you have one SCR with a bit of series resistance that turns on at the start of a cycle and another that shorts it out through a capacitor robbing enough current from the first to turn it off but then ceasing conduction itself once the capacitor charges. Fets are definately better.

PWM on a since wave is much nicer and just gives you a lower amplitude sine wave after the filtering. It should work on any kind of load that can be dimmed. Note that some CFLs/LEDs will give constant output power for 90-260V in and go out entirely a little below that. They're just not dimmable by any technique. Also note that for rectifier-capacitor input loads all the dimming action happens in the first 90 degrees of the sine wave, it's largely immaterial how much of the downslope you get as the input capacitor is now charged.

[Kjelt]

Also note that a lot of intelligent dimmable led drivers do need a specific dimmertype for it to dim correctly.

[djacobow]

Thanks for the replies folks. This is making me thing that it actually isn't possible to make the dimmer to end all dimmers, because each target light source likes something better. And now that we're in the world of CFLs and LEDs, and who knows how many variations in ballasts and driver circuits, there are really many many types of light sources.

That said, it looks like the AC/AC converter concept is the most flexible. It can do sine wave dimmer (actually, this is more like wave-out-same-shape-as-wave-in dimming) but driven by the right software, could easily simulate reverse or forward phase angle dimming.

I think right now lighting is in the midst of a big transition and when things settle down, hopefully standards will arise. It's actually not cool at all that each luminaire comes with a specific and limited list of dimmers it will work with, etc.

[Codemonkey]

You could just get a Philips hue kit. The protocol they use (ZigBee Light Link) has commands such as "Move to level" and one of the parameters is the time that the bulb should take to do the transition. So, you can for example send it a command saying move to a brightness of zero and take an hour to do so. You can also do the reverse and have a light that comes on gradually to wake you. The technology seems to be going the way of putting the dimmer in the bulb.

[N2IXK]

Motorized Variac FTW!

[djacobow]

Unrelated to this question and project, I've been replacing a lot of lighting in my home with LED retrofits and discovering that the dimmers (simple triac phase control) don't work all that well. One one circuit, I bought a new top-of-the-line Lutron LED-compatible dimmer that _SPECIFICALLY_ listed my LED lights as compatible. Well, the dimmer works nicely, but the fixtures buzz.

Sorry, that's just unacceptable to me. And it makes me angry that the only way you can determine what works is through trial and error.

I honestly had not thought about household lighting very much until about a month ago and I can tell you for sure now, the situation is f*k'd up. There are no standards. There are no promises. Every fixture has a different driver circuit, and nobody will even tell you what "secret" circuit they have in their dimmer.

I have a feeling that the new buzzing from my lights is because the fancy new dimmer does reverse phase angle and as a result has a sharp turn off each cycle, resulting in a voltage spike. Should I need a snubber on a lighting circuit? No. I should not. It should have been in the dimmer.

[johansen]

I have a feeling that the new buzzing from my lights is because the fancy new dimmer does reverse phase angle and as a result has a sharp turn off each cycle, resulting in a voltage spike. Should I need a snubber on a lighting circuit? No. I should not. It should have been in the dimmer.

no, the buzzing is from the current flowing through the wire.. hook up a CT to your scope and make sure your beverage is well out of arms reach....

[Kjelt]

I've been replacing a lot of lighting in my home with LED retrofits and discovering that the dimmers (simple triac phase control) don't work all that well.

There are good and bad retrofit dimmable led lamps out there. That said I personally find that if you really change the lighting in your home to led you are much better off also changing the way you controll the lights. That means no more AC dimming but PWM DC dimming in the lamp or driver itself and controlling it with the appropriate dimming protocol (Dali/DMX/RS232/KNX/IP etc.). It wíll not be easy if you don't have a wooden house with wood and plaster walls like in the US. Here I have the same problem, all concrete and bricks and changing the layout is almost not done. Alternatives are RF controlled lamps (Zigbee) but they are expensive. The modern/new houses that are build with a nice budget are already future prove, each AC outlet, each lampfixture outlet, every and each electric outlet have their own seperate cable tube to the central power installation in the house. 

[djacobow]

Ah! This makes sense. The dimmer I have is a Lutron Maestro C*L unit, which I naively assumed used reverse phase angle because it's "for LEDs.  But it appears to be a normal triac dimmer with who knows what special magic. The Lutron dimmer with reverse phase control is called their "ELV" line. It's really expensive: $105. Going from a triac to an IGBT and bridge rectifier shouldn't cost that much, but...

Anyway. I may or may not spring for it.

I have a feeling that the new buzzing from my lights is because the fancy new dimmer does reverse phase angle and as a result has a sharp turn off each cycle, resulting in a voltage spike. Should I need a snubber on a lighting circuit? No. I should not. It should have been in the dimmer.

no, the buzzing is from the current flowing through the wire.. hook up a CT to your scope and make sure your beverage is well out of arms reach....

[djacobow]

I'm back with more dimming experiments and questions. :-)

First, I made a PCB for a simple triac dimmer that can be controlled by an AVR. Uses the H11AA1 and non-ZC detecting triac drivers and -- for incandescent lights -- it works great. It's cool because I can control it over a wireless network using NRF24L01 and similar. CFLs won't dim (even the kind that claim to dim with conventional dimmers) and I don't quite understand it. But looking at the current waveforms, it appears that the drivers for CFLs will pull a big spike of current every cycle the instant they can and so it's not surprising that they don't care how delayed the turn on. (Which makes me wonder why they work with "better" triac dimmers.

Another issue is that some lightbulbs "sing" and buzz. The noise can be annoying. I'm sure the electrical noise is even worse.

So... I'm revisiting the sine wave dimming concept. I've created the circuit below in LTspice and it seems to work, but I'm noticing that producing a filtered sine wave output requires some, umm, pretty "special" passive components.

If I want to run the PWM at 5 kHz and, say, want filter with a knee at ~1 kHz, I need 1uF capactors that can withstand 120 Vac (250V caps) are easy enough to come by, but 10mH inductors that can handle amps ... not seeing it.

Doing a Google image search on sine wave dimmers shows some theater equipment with truly massive filtering inductors.

I could up the frequency at which I drive the FET to make the filter easier, but then that gets harder to drive with my rather high impedance super janky "power supply."

Pic below attached. Sorry it's more vertical than horizontal. Bad planning while make the schematic.


Thoughts and ideas welcome. I don't like breadboarding AC powered projects, so when I have enough confidence that this will work well enough for experimenting, I'll do a PCB.

[johansen]

Inductors are great until you calculate Joules per dollar...
 ...I mean dollars per mJ

some IGBTs are up to the task at 30Khz.
most large motor inverters run at 2,4, and 8Khz, but they use the inductance of the motor as the filter.

One last thing: the issue with CFLs is even worse when you include the inductance between the cfl and the triac.
that's one of the reasons the cfl doesn't dim, resonant charging beyond the peak ac line votlage..

[ahakman]

Oooh, you had my hopes up with the topic title, and then when I read the thread, you just got stuck on the same things I did.

I also tired to simulate the output filter in spice for a "sinewave dimmer" and found it difficult to come up with anything that was very independent of the load. Was going to try some different filter topologies, but never got around to it. Shame too, because I'd really like to replace my X10 stuff with something more reliable, but like you, realized building a universal dimmer seems to be quite the feat not knowing what kind of bulb with what kind of driver you might end up with in the near future. I already have the control / network side of things done, in the form of something like the JeeLabs Jeenodes. I use almost the same hardware, but completely different scratch built firmware that's not arduino dependent for better performance, and because I enjoy low level coding. I also did completely custom PCB layouts that contains everything I need for each node, not in the modular arduino type approach, so a USB attached 'base node', and so far RGB controller, and eventually AC mains powered wall switch, and "lamp module" type nodes.

http://imgur.com/a/MP8Wn

Really would be nice if there was a standard to dim new tech lights like CFLs and LED fixtures the proper way (on the DC side of the fixture in the case of LEDs) without having to hack it into each one yourself. Oh well, for now I can build and control as many channels of RGB led controllers as I care for, at least after I get back to that project and do a new PCB layout for the different radio I switched to now that the RF12 is EOL, and to fix some other "rev 1" type issues!

http://youtu.be/qjKS70O8DNk

[NiHaoMike]

Best bet, if you're using LEDs, is to put a switching PSU and microcontroller in the fixture and use wireless or PLC for the control signals. Then put a transmitter in the control.

Dimmable CFLs (and even the regular ones, though to a much lesser extent) will dim by reducing the supply voltage, but many LED bulbs will not. They specifically look for the phase angle control. They don't make much noise when dimmed so no point using some other form of control.

[djacobow]

Dimming LEDs, if I control the fixture, is not that difficult, though it may be the most feasible way to dim stuff "in the future."

But we are living in a transitional time, and so I am still interesting in dimming existing plug loads. It occurs to me that with the FET (or IGBT) shorting a bridge rectifier topology, along with a ZC detecting circuit, one can actually delivery any of the three dimming techniques discussed here (forward phase control, reverse phase control, and PWM/sine wave). I think the passives would be slightly different, certainly for the sine wave dimming there is a need to do some serious filtering, but basically the bridge and the right FET can deliver all three.

So I think I'm going to either dead bug or maybe do a simple board to experiment with this and I'l report back with any learning.

The one thing that is stopping me from doing this right now is that I'm not sure how to drive the FET. I want to control it through an optoisolator, but those are open collector, so I can pull down from the opto, but I'll need a positive voltage supply to pull it up.  I'm not sure the best way to do this. It occurs to me that I can use the drain voltage itself, but that will go away when the FET is on.

[gxti]

Lutron costs so much you think they'd just bribe the lamp manufacturers to agree on a protocol where they could control the bulbs through the wiring, and let them sort it out. Can't be much more complicated than making a good "dimmable" bulb in the first place.

[SimpCirc]

It's a shame you can't just use a circuit that slowly increases resistance to the LED's. Simple, no flicker, and cheap. But that would require altering the bulbs yourself and using something similar to the X10 equipment or a signal piggy back through the mains supply. Shame a variac or variac simulator won't work cause more than likely the LED Bulbs have a Voltage Regulator of some kind in them that would output that voltage until it cutout. The singing and buzzing is common on frequency controllers, simple voltage and amperage control does away with that, or has in my experience. Heres hoping you can find a solution to this issue.

[ahakman]

It's a shame you can't just use a circuit that slowly increases resistance to the LED's.

You don't want to do that at all - resistor => burning lots of power in heat that you have to get rid of somewhere. Might as well just stick to incandescents if you're going to do that.

[ahakman]

But we are living in a transitional time, and so I am still interesting in dimming existing plug loads. It occurs to me that with the FET (or IGBT) shorting a bridge rectifier topology, along with a ZC detecting circuit, one can actually delivery any of the three dimming techniques discussed here (forward phase control, reverse phase control, and PWM/sine wave). I think the passives would be slightly different, certainly for the sine wave dimming there is a need to do some serious filtering, but basically the bridge and the right FET can deliver all three.

Yeah, that would be the best to be able to use any of the 3 methods. I think the IGBT might be the best bet, as I was reading an application guide before about using IGBTs for dimming, and you can do lots of good stuff with a few passives, like controlling the dv/dt of the gate to slow down the di/dt of the load current if doing reverse phase dimming, so less RF filtering is required, short circuit detection and shutdown, so no fuses are required, etc etc. Let me see if i can find that app note again... http://www.datasheetcatalog.com/datasheets_pdf/A/N/5/1/AN518.shtml weird that it doesn't seem to show up on ST's website anymore.

[johansen]

Yeah, that would be the best to be able to use any of the 3 methods. I think the IGBT might be the best bet, as I was reading an application guide before about using IGBTs for dimming, and you can do lots of good stuff with a few passives, like controlling the dv/dt of the gate to slow down the di/dt of the load current if doing reverse phase dimming, so less RF filtering is required, short circuit detection and shutdown, so no fuses are required, etc etc. Let me see if i can find that app note again... http://www.datasheetcatalog.com/datasheets_pdf/A/N/5/1/AN518.shtml weird that it doesn't seem to show up on ST's website anymore.

slowing dv/dt of the gate down isn't exactly an option because that's heat burned up in the igbt.

also, lagging edge phase control might work for led lighting and resistive loads, but that's it.
once you throw a lagging load on it you're going to blow up the igbt's snubber circuitry with a *LOT* of power.
an ac buck converter doesn't have that issue, but then you need an expensive inductor on the output.

[tom66]

Instead of the opto and associated driver circuit, consider an optoisolated gate driver, like FOD3182.

You could get 250kHz with that. 50kHz is probably a good limit as most high voltage FETs have high gate charge so your power dissipation will increase too much.