AC Rectified - DC Mosfet dimmer - Hints please

[bingo600]

I was thinking about using this AC rectified DC-dimmer

http://www.instructables.com/id/AC-PWM-Dimmer-for-Arduino/



BrianHG writes here

Now, I have a method using a power mosfet and AC-DC bridge rectifier, but, it's something like a 10 component solution and you need to switch the optocoupler on only during the 0 crossing phase of the AC & you also need both access to the hot and neutral.

Would that apply to this construction (switching of the opto during zero cross) ?

I mean it's not AC anymore , but "unfiltered DC"

/Bingo

[bingo600]

I'll be dimming 2 x 40w - lightbulbs (The big ballon ones with golden glow)


I have gotten both some

Mosfets : STB10NK60Z - http://www.st.com/resource/en/datasheet/stb10nk60z.pdf

And some

IGBT's : SGP10N60  - https://www.digchip.com/datasheets/parts/datasheet/216/SGP10N60-pdf.php

100uF will be - NIPPON CHEMI CON series KMX 100uF 400V super LONG LIFE capacitors 105'C

And i have some 4N25 Optos in the drawer.


The author says the IGBT's are probably the best for this , but Mosfets are ok.

I don't have much experience with Mosfets or IGBT's , but i do with mains (no worries there).

MCU would be a ESP8266 , where i'll adapt the opto resistor to 3.3v


Can anyone see problems with the circuit , besides missing a fuse ?

/Bingo

[bingo600]

The ESP8266 is going to need 3.3v anyways , so if i get a 230->12v PSy , and a DC-DC switcher for the 12 -> 3.3v

Could i get rid of the Capacitive PSU , i mean the 100uF cap & diode.

And put the 12v psu in use for driving the Fet/IGBT ?

I'll loose the galvanic isolation , but the IF is going to be wireless via the ESP , so i'm not too worried.
Unless the "shared ground "Unsmooted 400v DC Ground" , and 12v DC-DC converter ground would intefere.


Any hint for what to modify ?
I mean i suppose i still need to have a pullup in the FET gate , as the opto is the "pulldown"

Do i need some 400v cap on the "mains (load) DC" ? - For smoothing ?
As i see it on the existing schmatic , there isn't any on the one in the picture, due to the diode.

A few hints would be appreciated.

/Bingo

[Benta]

That circuit has so many dumb things, it's difficult to know where to start (is "instructables" really like that? I've never visited them).

First: if you switch off your Arduino, the lamps will be lit always.

If you decide that the "always on" functionality is OK, then:

There's absolutely NO reason to use a 400 V cap. The cap will work just a well connected to the junction of R4/R5, so a 25 V type is enough.

D5 is unnecessary.

Reduce R5 to 5k6

Place a resistor between R4/R5 and the 4N35

Use a high-voltage resistor for R4, eg, VR68 from Vishay.

That's just a start...

[bingo600]

That circuit has so many dumb things, it's difficult to know where to start (is "instructables" really like that? I've never visited them).

Thanx for the comments
I knew about the PWM source must be always on


Author mentions this circuit (other creator) , and says it has some bad design faults , explains on Instructables page


He modified it , and says this is better



Unfortunately i'm an "Analog noob" , and "have to trust" the designers , that's why i asked for some hints (review)

It will be in my summerhous , and i don't want a fire 

/Bingo

[Benta]

Yes, the Giesberts circuit has a major problem (if PWM is 100%, there is no power for the control section).
The modification, placing the bulb on the DC side fixes this, and should work.

But the suggested "new" circuit is way off.

[james_s]

Is there a reason for using such an unconventional circuit rather than the classic triac phase control dimmer? If you use a zero crossing signal to trigger an interrupt in your microcontroller it's fairly easy to adjust the trigger point.

[Benta]

Is there a reason for using such an unconventional circuit rather than the classic triac phase control dimmer? If you use a zero crossing signal to trigger an interrupt in your microcontroller it's fairly easy to adjust the trigger point.

I asked myself the same question. The only rational reason is that the PWM controller (Arduino) can operate independently and does not have to sample the mains voltage (thus avoiding mains voltage connection).
That's OK for me.

[james_s]

Well if anyone builds this I'm curious to hear how well it works. Zero crossing detection on the mains is fairly simple, I used a comparator in the DMX dimmer pack I did years back, sampled the AC off the secondary of the power transformer before the rectifier. It's no problem for the AVR to sync to the zero cross and control dimming on 8 channels.

[bingo600]

I'd like to use it because apparently running zerodetection on the ESP8266 , would make ESP webpage responses laggy.

/Bingo

[BrianHG]

That circuit has so many dumb things, it's difficult to know where to start (is "instructables" really like that? I've never visited them).

Thanx for the comments
I knew about the PWM source must be always on


Author mentions this circuit (other creator) , and says it has some bad design faults , explains on Instructables page


He modified it , and says this is better



Unfortunately i'm an "Analog noob" , and "have to trust" the designers , that's why i asked for some hints (review)

It will be in my summerhous , and i don't want a fire 

/Bingo

Yes, this is the schematic is what I was thinking of...

[BradC]

Is there a reason for using such an unconventional circuit rather than the classic triac phase control dimmer? If you use a zero crossing signal to trigger an interrupt in your microcontroller it's fairly easy to adjust the trigger point.

The above circuit with the bridge rectifier in *series* with the load is a really neat way to allow you to use a DC device to control AC, so you can use a MOSFET rather than an IGBT. Additionally, it can be operated outside the parameters of being locked to a zero cross, so you can use either faster PWM *or* (as I've actually seen done) use the MOSFET as a resistor to give a clean but reduced voltage output (with the consequential power dissapation in the MOSFET). It'd even be a neat way of effectively implementing a trailing egde dimmer.

I have a real issue with supplying high voltage DC to anything a finger can get into (such as light socket).

[BrianHG]

You can also get rid of all the optocoupler and gate drive circuitry, replace it with a 500mw green led and a few photodiodes and a logic level mosfet.  Good for switch on and off, isolated, but it wont work at high speed, but, you never need worry about the gate supply for that second circuit.

[BrianHG]

Also, with the above AC circuit, a dumb 20Khz PWM allows for dimming of lights without the annoying 100/120hz buzz created by phase dimmers.  Also, with a 20khz tuned inductor & cap filter, you can make a nice AC full sine wave of the source good for driving AC motors.

[Zero999]

I'll be dimming 2 x 40w - lightbulbs (The big ballon ones with golden glow)

What sort of light bulbs? LED, compact fluorescent, incandescent?

is "instructables" really like that? I've never visited them

Unfortunately yes. Most instructables are very poor quality. There are one or two very good ones but most are bad. There's no quality control on that site. Anyone can post crap and it doesn't get removed because it generates advertising revenue. I like to call them destructables.

[bingo600]

I'll be dimming 2 x 40w - lightbulbs (The big ballon ones with golden glow)

What sort of light bulbs? LED, compact fluorescent, incandescent?

Standard light bulbs w. clear glass , and those wires that glows golden inside.

/Bingo

[bingo600]

Is there a reason for using such an unconventional circuit rather than the classic triac phase control dimmer? If you use a zero crossing signal to trigger an interrupt in your microcontroller it's fairly easy to adjust the trigger point.

The above circuit with the bridge rectifier in *series* with the load is a really neat way to allow you to use a DC device to control AC, so you can use a MOSFET rather than an IGBT. Additionally, it can be operated outside the parameters of being locked to a zero cross, so you can use either faster PWM *or* (as I've actually seen done) use the MOSFET as a resistor to give a clean but reduced voltage output (with the consequential power dissapation in the MOSFET). It'd even be a neat way of effectively implementing a trailing egde dimmer.

I have a real issue with supplying high voltage DC to anything a finger can get into (such as light socket).

Would you care to point out with one you mean ?

The top post (it think you mean that one) , or the one wo. the red cross

And i'd be using an ESP8266 for control , no fingers involved 
/Bingo

[BradC]

I meant the one with the Red Cross. Not that circuit specifically, but the same theory. Bridge I series with the the load.

[Seekonk]

I use opto isolators to switch FET often.  FET gate capacitance will make a very slow turn off with a 22K resistor. I wouldn't use any more than 1K to reduce transition heating. Electronic wall warts are a convenient way to power these circuits.  Surprised no one mentioned capacitive dropping to eliminate resistor heat.

[Kees van Gelder]

Actually there is more sense in using a 400 Volt capacitor in the position that it is in than you give credit.
When using Mosfets or IGBT you want them to work as a switch that switches fast between OFF and ON. You dont want the Gate voltage to  linger somewhere halfway, making the MOSFET or IGBT to dissipate a lot of heat. Tou want it to be either OFF or  ON on its RDSon.

With the capacitor as is in the given circuit, the only RCtime you have to worry about is that of the 100k resistor and the  capacity of the  tarnsistor.
For the average MOSFET that is in the range of 10pF. Hence your RC time being practically zero. Which means that the FET or IGBT will switch very fast.
Now lets see how that is with a 25uF capacitor where you say it should be. Then suddenly the picture is different. The RC time will be 2.5 seconds. Which means that it takes 2.5 seconds to get to 63% of its gate voltage. All that time the Transistor is not at RDSon but at a higher resistance, which translates in dissipating power -> heating up.

So I presume a bit more thought is gone in to that than you say.
Having said that, the 100uF probably could go down a bit, say 50, but I'd certainly keep it in the position it is,

The zener, as the author says is there just in case. Sure, you could leave it out, but it isnt a 'flaw'. I even see a circuit there with just the zener.

Coming to the 'resistor between the connection of R4/R5 and the  optocoupler', I presume also here fast switching was the goal, you want that Voltage to go down fast.
Now if you put a capacitor there as you propose, yes chances are you will kill not only the Transistor but also the optocoupler so yes then you need a resistor, but then not only you are slow in charging but also in discharging -> more time spent outside RDSon -> heating up your MOSFET

5k6 instead of 6k8 ? Cosmetic

Stays on when Arduino is off? yeah I guess that is the design, but I guess you could put the optocoupler in series with the 100k.

regarding the discussion on using a TRIAC and do zerocross detection. Sure that works as well, but you will not be able to use PWM which seems to be the point of this circuit

[Seekonk]

For the average MOSFET that is in the range of 10pF. Hence your RC time being practically zero. Which means that the FET or IGBT will switch very fast.

I'd sure like to get some of those FET!

[Benta]

I'd sure like to get some of those!

Me too!!

@Kees van Gelder:
I assume you are commenting on my first reply (#3)?

Concerning the cap: the original idea is a 310 VDC supply using a huge 100 uF, 400 V cap (do you realize the physical size of such a thing?)
My suggestion is to make a 16 V DC supply by placing the cap at the R4/R5 junction instead. 100 uF, 25 V is absolutely sufficient and can be smaller with no problems. And at 16 V, no Zener is needed.

The series resistor between cap and optocoupler is to prevent too much discharge of the cap when the coupler is on.

But quite frankly, an even better solution is to place the coupler transistor in series with the gate, plus a pull-down resistor on the gate. This would provide "positive logic" to the circuit and eliminate the "always on" issue.

As to switching losses: a circuit as simple as this is not laid out for 50 kHz switching. I'd switch it at perhaps 500 Hz. And the relatively slow transistions (due to the gate resistor) are actually a blessing if you think about EMC radiation.

[Seekonk]

This is the FET gate driven by one of those ebay 4 channel opto boards with a 5.1K pullup, just opposite of the pull down.  So, a 22K would be quite problematic.

[Benta]

This is the FET gate driven by one of those ebay 4 channel opto boards with a 5.1K pullup, just opposite of the pull down.  So, a 22K would be quite problematic.

I've no idea what eBay boards you are talking about, and I've no idea what your trace shows.

But let's put a few facts on the table:
IRF730 input capacitance: 700 pF typ.

This gives a time constant (at 22 kohm gate resistor) of: 22^3 x 700^-12 = 15.4 us

Now tell me that PWM'ing this circuit at 500 Hz is a problem. Not!

[Seekonk]

I never directly answer any one persons issue directly.  That FET is pretty dinky.  I'm doing 480Hz quite often.  My gates are at least six times that capacitance.  I also don't allow short pulses at either end of the duty cycle.  20KHz was mentioned. There doesn't seem to be any practical reason to have a resistance that high.  As long as you are aware.  This is for those that might consider copying parts of your circuit.