AVR Dimmer - Lead edge (reverse phase) control

[metal]

Found some bugs, fixed them and removed the triac simulation files, I no longer need them. I also modified some lines to ensure both ends 0 and 100 work correctly.

BTW. read about the type of loads that can be connected to this dimmer!

[metal]

Read the text file - you will like this project.

[DreamCat]

good project, but this kind of dimmer has big EMI, so the sine wave dimming maybe is only method for new standard.
of cource, it can not use Triac, the MOSFET and IGBT is good choice.

[metal]

This is what I thought as well. I suddenly forgot, that the triac can never be switched on at the zero-crossing and turned off after, say 300us, and "how silly", I tried to switch it on at zero-crossing and turn it off, no kidding :- )) then to say, damn it, this will never happen, sometimes I forget the basics while working!! So, the solution is to use reverse phase control, as you suggested, and with a MOSFET or IGBT. But I don't have experience working with mains and IGBTs or MOSFETs. I read some articles, and I was looking for something that will use two MOSFETs or IGBTs to be able to switch both halves, but I could not find examples doing it the way I was thinking. I don't know if I can connect two transistors to pass the bot wave potentials, some use a single MOSFET with a bridge, but this is not what I want.

Finally, it is really funny, I think it is meaningless to use a MCU while the triac still switches on for the rest of each half cycle, the worst case is at 90 and 270 degrees of the full wave, real bad indeed. A diac will do the job, so?

[metal]

Here you go, with MOSFET, reverse phase control, or the so called lead edge control. It took me quite some time to make it work, proteus is tricky. The funny thing is that I did not change a single code line. Very nice indeed, I am impressed that semitone implemented the MOSFET part, I stole it from there. I will see if I can get myself an IGBT to play with as well.

[metal]

Found some bugs, fixed them and removed the triac simulation files, I no longer need them. I also modified some lines to ensure both ends 0 and 100 work correctly.

BTW. read about the type of loads that can be connected to this dimmer!

[solutions]

You have the phase cut dimmer backwards, Metal.  

Your waveforms are IDENTICAL right now to that of a $3 TRIAC dimmer. You'd get the same action if you pulse the gate of a TRIAC at your TURN ON angle
:-)

You want to TURN OFF the trailing, not leading, edge.

[pickit2]

You have the phase cut dimmer backwards, Metal.  

Your waveforms are IDENTICAL right now to that of a $3 TRIAC dimmer.
:-)

You want to TURN OFF the trailing, not leading edge.

That is the C++wizzcoder MeTaL is using. text reads right to left.

[solutions]

That is the C++wizzcoder MeTaL is using. text reads right to left.

Aha! Thank you.

I need to get one of those epocsollicsO latigiD instruments for my bench.
:-)

[metal]

I knew it will hunt me back :- )

In posts #1 and #4, the part of the half wave you see is when the Triac or MOSFET are OFF, respectively.

In the Triac example, it stays OFF after the zero crossing and ON after a certain delay, for example when the phase is at 90 and 270 degrees which what you see in post #1, now compare reply #1 and #4, you will get the difference, extreme opposites! What I wanted is that the MOSFET is ON after each zero-crossing, and OFF after a certain delay, rather than the Triac OFF after the zero crossing and ON after the delay.

The oscilloscope is not mistaken, the point from which I took the signal is a bit deceiving because you will see that part of the half wave when the switching device (whether MOSFET or Triac) is OFF, and believe it or not, I spent many hours wondering why the Triac worked that way till I... you know... paid attention that this signal I saw on the oscilloscope was the inverse of the what actually happens, ON is OFF and vice versa :- ))

The problem is that if the device is ON when the wave is in its extreme ends i.e. 90 and 270 degrees, serious EMI develops, and I am not sure if a snubber can completely solve the EMI problem in the mains, because if I use a commercial dimmer on one of the outlets in my room, and turn the amp on without music, I can hear the buzz While the Triac is switching ON, and if you connect a soldering iron, you will start hearing that buzz, especially when the wiper is in its middles!

I still don't know and I am not sure if the MOSFET part will work correctly in the real world. I am mostly used to Triacs on the mains and MOSFETs on the amplifiers, but MOSFETs on the mains, I am really terrified... I have FQA24N50, I will start with it, and if things are just fine, I will go with locally available MOSFETs :- )

[solutions]

You have a few hundred nanosecond switching from 170V to zero.  They should know your lights are on a few stars over in the galaxy, but still won't confirm to them that the earth harbors intelligent life, especially with that kind of EMI coming off it.  

You might try one of those clampon ferrite beads on the power line feeding the MOSFET - that should slow the di/dt down to where it's less of a nuisance and will also act to block high frequency components from going out on the building wiring and trying to use that as an antenna.

MOSFETs aren't that scary on mains as long as you don't forget that they have that pesky intrinsic diode in them (S-D bias SHORT CIRCUIT)

[metal]

your lights are on a few stars from the galaxy

You should start writing manuscripts for IT world :- )

[metal]

Hi PICkit,

Look at this project: http://www.freebus.org/index.php/hardware-mainmenu-36/fb-applikationsboards-mainmenu-73/dimmer

There is a link to a forum, where the author and people who built that project discuss the problems of this method: http://www.freebus.org/distribution/viewtopic.php?p=2851#p2851 really interesting, flickering, MOSFETs burning, sudden strange behavior that can't be explained and not related to a software bug, etc... And someone says that these components in red should be added, why? I know not! http://www.freebus.org/distribution/viewtopic.php?f=12&t=1106&start=45#p4378 then on the last page someone comes up with a strange problem related to MOSFETs placement on the board!

Tell me, after you read the 5 pages, what do you think, I started to get really confused on which way to go? 2 MOSFETs, or a bridge and a MOSFET as in semitone?

Still I think this 2 MOSFETs method should work, but I not know how to make it work in proteus, at least I want to see it working in simulation. Look at this page, http://www.see-solutions.de/projekte/projekte.htm the project is called 2007_02: SPS-LightControl (Butzenbus), it is already working, but I don't know how to make it work without silly GMIN problems in proteus...

I attached a quick test.dsn file for this method, it works but GMIN problem can't solve it I.

Myself, I noticed that Semitone http://www.engbedded.com/semitone did not use the two MOSFETs method. Look in the downloads sections, and click on "Diamond_Schematics", download "sd1.0.1" file and look inside "sd1.0.1.tar.bz2\sd1.0.1\Schematics\PDF\Output\output_stage_fet_inverted.pdf". This is what I used in the current working simulation file in the sticky post.

[DreamCat]

here is a document for sinewave dimmer
http://www.iawco.com/parlights/Downloads_files/strand/Dimming/Sinewave%20Dimmer%20Technology.pdf
ST has a app note, but I forgot its link.

and there is also a project on http://www.digispeaker.com/blog/ .

[metal]

@ Dreamcat
Here is ST document attached, I found it on datasheetcatalog.org I think it is ok to upload it here.
Look at sticky post.

What really was new to me is driving the MOSFET correctly. Thanks to semitone which correctly did it. Seems that IGBT option is more complicated and expensive.

[DreamCat]

yeah, IGBT is expensive than MOSFET, but  it suite big power.

about MOSFET,  I remembered ST also have the special tpye for this.

oh, sorry, the load should be at front the bridge, other else will cause trouble when setup (install?) it.

I also has a question, for the wire to connect the load, it can also produce  EMI.
next day, I will use MicroCap simulate this..

[pickit2]

See Attachment for part solution, the problem this works in real life, Proteus only shows 1/2 wave. anyone able to sim a dc & ac supply in proteus?

as to 1 or 2 Mosfets is filement migration, one of the reasons lamps controlled by most dimmers have a short life.
with 2 mosfets fillement don't migrate as in full wave supply.


If you have a local microwave repair shop, that has a scrap bin, look for Panasonic Inverter supply.
they have two 900V Igbt when they fail one is blown oc, and other is ok. you maybe lucky and unit was scraped for other fault.

[metal]

PICKit, the current path will be the same for both methods I think. Both of them will pass through the diode bridge, so what advantage is the two MOSFETs introducing?

[pickit2]

with one mosfet its half wave and two = fullwave.
the problem proteus with two mosfets it displays contol as for one mosfet, I think there is a problem in proteus.

in the real world you have contol over fullwave, dimming is done without choping the wave, both cycles of the ac load are set by the amplitude of the supply.

[metal]

Kindly draw the path for the full wave in the two MOSFETs case?

[pickit2]

This is the AC & DC supply paths Note: AC is isolated from ground.

[metal]

Hi PICkit,

I think, these are the paths thru the MOSFETs, look at the attachment. Still I can't see any difference between single and two MOSFETs method. You will have the same result when using a single MOSFET and a bridge. At the same time, if you look at the patent attached, the actual implementation of GND doesn't refer to earth, rather to the point where D1 and D3 meet, as demonstrated from my arrows.

Edit: I added the single MOSFET with bridge.

[pickit2]

if you look at the patent attached it follows my original layout and drive for the two fets.
The patent has a load of extra stuff, you don't need to pulse or have zero crossing as your not chopping the ciruit, you are changing the Amplitude.
you can build a dimmer with less than 10 conponents.

I have drawn it out in proteus and it works, and there is no short from AC to the Neg of supply.
you have full wave across the load and also across the fets.

[max]

Hi friends,

Here is my collection of articles for mosfet based ac load control.
Specially see the article "mosfet-ac-relay".

Regards