Using TRIAC to switch on a mains transformer

[metal]

I was thinking about using a TRIAC to switch on a mains transformer that is paralleled by a 47R resistor which is bypassed after 100ms. I am planning to design a softstart circuit to be used with such transformers to prevent the initial surge at power on. Now my questions are, 1. is it OK to use TRIACs, if yes, 2. what is the correct angle at which I should fire the TRIAC to ensure the minimum current surge? 0, 45, 60, 90, etc? 3. How long should I wait to bypass the resistor?

Edit: I found an interesting document that changes thinking a bit. I also am thinking about something like this:

[Sideshow Bob]

This may share some more lights on the problem
http://www.emeko.de/uploads/media/01-speech-from-berlin-at-cwieme-2004_01.pdf

[metal]

interesting, it took me quite some time thinking how he ensures the relay is switching correctly. Seems at the 4th magnetizing cycle he switches on both thyristor and relay to ensure there is no delay in relay switching for the following +ve half, is that true!!?? I don't know, is he using a relay after all?

Edit: in p. 16 it is mentioned that relay is activated at the peak of the last magnetization pulse.

[solutions]

K.I.S.S.

http://www.cantherm.com/products/thermistors/choosing_ntc.html

[Gallymimu]

+1 for thermistor.  we also used to use a relay bypass around the thermistors after the initial charge up of bus caps to improve Max power delivery and allow small thermistors. useful method if you have big power.

[metal]

frankly, I don't want to use thermistors. Yesterday I was investigating the possibility of using an existing code of a leading edge dimmer that I wrote few months ago to see if I can implement the magnetization pulsing method and then supply full wave. I added few lines of code and ended up with a random magnetization pulse direction (half-wave, no dimming), nevertheless I was still able to see that after three pulses, the third was a complete sine wave, happy ending. I am using a MOSFET in the leading edge dimmer, and I started thinking about reversing some lines in the code to implement a dimmer and see what results I get, any way, I will bypass the semiconductors by a relay, I will post more info later.

[zab]

The attached doc. require some careful study. For triac based soft start requires switching on the device near 180. then slowly reduce this angle to 120 90 60 so on. But one thing keep in mind that trigger pulse should remain on during the time of conduction. Switching near 0 and 180 is not possible.For voltage  across device is not sufficient to keep it on.The device in this region goes on and off rapidly creating high voltage pulses. a smart snubber is required across the device.

[metal]

Which document are you referring? if mine, where did you read I should switch on at 180 degrees and why do I have to change the angle this way, it is not mentioned in the document.

There is one thing I could not answer: when is it OK to turn off the transformer?? We were talking about when to turn it on and how, but what about turning it off?

[Atmelfan]

Why not use a MOC3041 (Zero crossing optoisolator with triac driver output) which already switches on at the zero crossing point and therefore no problems with surges?

[Sideshow Bob]

K.I.S.S.

http://www.cantherm.com/products/thermistors/choosing_ntc.html

There are several disadvantageous conditions then using both resistors and NTCs.

Posted on: November 16, 2012, 02:20:51 14:20 - Automerged

Why not use a MOC3041 (Zero crossing optoisolator with triac driver output) which already switches on at the zero crossing point and therefore no problems with surges?

No switching at zero crossing point give huge surges in transformers. That is shown in the document in post 1

[metal]

Here is a snapshot of what I was talking about yesterday, I mentioned it was an experimental change in an existing code, hence, the pulsing direction is random, at least, it is possible to test the idea in no time. But I am still thinking when I should turn it off, also what to use, if I use a TRIAC and then bridge it by a relay, it will be the best choice, provided that even with a TRIAC trigger should always stay ON during each half-wave time so that the current can flow correctly. At turn off, I still can turn the TRIAC on and take the relay out to control the turn off, but if the turn off requirements don't match how a TRIAC worked, this will be a problem and I will have to change the semiconductor device, I don't know yet.

[metal]

Hi,

I thought that I grabbed the idea, but apparently no I did not, there are many other issues to take care of in order to make this technique successful. I was looking at the documents on emeko site, and found this and this.

The guy doesn't give the transformer a complete half-wave, rather he gives it a part of it at 130 degrees which is equal to 237V, it seems that he measure the current according to this document to find out when is the right moment to turn the full power on. Moreover, the pulse width is variable depending on the measured current, and also the device he makes has a short-circuit protection and according to this document he solved the voltage dipping issues!

There are many pieces missing on how things are done with the minimal components count and the procedure to do them correctly to make this technique completely work. I don't know how to measure the current as he does, and I also don't know how he detects when it is the right moment to switch power on, add to it I don't know how he detects a missing half-wave and then waits the next half-wave to pass and then turns power back on. Really interesting....!

[Gallymimu]

We used to have issues with soft start of transformers when we custom designed large AC toroid stepdowns (5KW 208 to 10V).  Because the peak flux of the transformers was designed to be low, and because we were using iron powder core the risidual flux was pretty high so if we started to deliver power on the same polarity half wave as the thing turned off we'd get a core saturation event which blew breakers.

It was hard to fix.  We used slow ramp up phase angle firing which looks similar to your documents.  One of the biggest problems we had was any phase lead/lag would wreak havoc as the controllers we used didn't monitor both current and voltage so you could easily miss fire if for instance the transformer happened to be unloaded.

It's one of the reasons that Solutions and I were keen on the NTC approach.  It's not elegant (and engineers love cool and elegant) but it does the job.

We used controllers from HB Controls.  They had to build some custom firmware for us with a lot of back and forth before we got our system working right (and by right I mean well enough to make the customer happy, far from perfect though).

Posted on: November 16, 2012, 11:44:30 23:44 - Automerged

There is one thing I could not answer: when is it OK to turn off the transformer?? We were talking about when to turn it on and how, but what about turning it off?

Turn off only matters if you care where the residual flux is left on the transformer, or if you have a leaky transformer and you need to shunt the primary magnetizing current.  I'm assuming you have a tightly coupled transformer.  Now, as I mentioned if you turn off at the top of a half wave and turn on during the same polarity cycle you can drive your core flux to 2x.  This might be fine if you have a lot of flux density margin (and you can get transformers designed for phase angle firing for this issue with higher flux density).

So, the turn off depends.  If you can turn off at zero, then your turn on phase doesn't matter as much.  It's a little bit of an application specific question.

[metal]

Thanks for providing this info, it seems that large companies are working on such issues, this makes me think twice about working on such an issue as an individual. May be if I combine the solution in the document attached in the first post with NTC, or even resistors, things would be acceptable for a home appliance, I think this is better than creating an extremely complicated system that needs adjusting.

The thing that makes me revert is that I must monitor the wave and in case of a missing halfwave I will have to correctly recover, otherwise my system is useless. I will think more about it and read more to see what can be done. What I was thinking is completely wrong about providing complete halfwaves, they should be 237V instead of 310V which occurs at 130 degrees. If I figure some way to monitor the halfwave polarity and to track the last polarity before a power dip, then I will continue with emeko solution.

[metal]

Hi,

I found HCPL3700 opto-coupler to be a good candidate, this one has all necessary parts except the resistor. I am willing to use two zener diodes connected back to back in order to limit the voltage into the opto-coupler, also how should I calculate the resistor, based on RMS or peak voltage?

[metal]

I finally have got time to show something, the uploaded file includes a simulation video, proteus+hex files.

[metal]

Here is the zero-crossing detector part done yesterday. I will put this part inside a heatshrink, the rest will be isolated from mains.


still more to follow.


I better move this thread to projects section, may be later.

[Toxible]

"switch on a mains transformer that is paralleled by a 47R" ?
The resistor is in parallel?
BTW, I think that is not very useful to do a softstart for a transformer!
A transformer primary is a simple inductor, so current is (ideally) shifted 90° compared to voltage phase. Why use a softstart?

[sarah90]

The soft start is not for the transformer, it is for the capacitors in the power supply.

[Toxible]

Ok so we need to switch on the power on the zero crossing right?
The circuit posted has a little error:
Vin max = 230Vrms So the peak value is 230 * sqrt(2) = 325V
(325V - 1,4V) * 22k / (22k+220k+220k) = 15,4V... C1 must be 16V (25 is best)...
But i see another little error. C1 will charge through R1 + R2 and it discharge through R3 (of course Vbe of Q1). Is better to add a resistor on Q1 base.

[metal]

Don't bother, I meant the 47R in series with the transformer, not parallel, I don't know what was on my mind when I wrote "parallel".. I will draw a schematic to make things clearer..
I will try the ZCD circuit as is today, if problems occur, I will change it. But, I don't understand how it will discharge through Vbe junction, why would it do that any way while D5 should prevent that.

[Toxible]

Well well welll.... I've do a mistake... now i will make order on my ideas..
So, suppose C1 at 10V. now when Vin is approx zero, D1, D2, D3, D4 are off and so remove this form your mind. current can come out from C1 (+) and go through R3, go on Q1 base and reach C1 again (-).. so C1 will discharge. Q1 is a NPN bjt so you can see this as two diodes. But we are lucky twice:
1- C1 will not reach more of 4V, infact when Q1 goes on C1 can discharge throught R4. if you do some math you can see that this.
2- Discharge throught R3 is very short and will not make any problem.
Sorry for the mistake... my mind is a bit fused

[Gallymimu]

BTW, I think that is not very useful to do a softstart for a transformer!
A transformer primary is a simple inductor, so current is (ideally) shifted 90° compared to voltage phase. Why use a softstart?

You aren't really understanding how a real transformer works.  It isn't an inductor.  It does have parasitic magnetizing inductance but in the case of a "good" transformer that inductance is very small.

If the transformer has good coupling then to the line, the transformer looks like a reflection of the load multiplied or divided by the turns ratio (depending on how you reference things).

The inductance of the transformer has very little impact on whether you would want a soft start.  The load is the main determining factor.

The second factor you are missing is the saturation flux of the transformer which, when reached, will turn the transformer into a short circuit.  If you read the whole thread you will see some explanation and examples of this.

[sarah90]

Dear Toxible, it does look like to me that you are copying messages and posting not so useful information when looking at your first 6 posts. In your invite you said you wanted another chance to contribute, but I don't think this way it is helping the forum.

[metal]

Well well welll.... I've do a mistake... now i will make order on my ideas..
So, suppose C1 at 10V. now when Vin is approx zero, D1, D2, D3, D4 are off and so remove this form your mind. current can come out from C1 (+) and go through R3, go on Q1 base and reach C1 again (-).. so C1 will discharge. Q1 is a NPN bjt so you can see this as two diodes. But we are lucky twice:
1- C1 will not reach more of 4V, infact when Q1 goes on C1 can discharge throught R4. if you do some math you can see that this.
2- Discharge throught R3 is very short and will not make any problem.
Sorry for the mistake... my mind is a bit fused

Neither Q1 BE junction nor R3 will not be able to discharge C1 because of D5. Your assumptions are incorrect, rest your soul and think before you write something down, because I spent quite sometime analyzing the circuit, thu I did in the past, but you did confuse me.

Posted on: November 25, 2012, 11:55:56 23:55 - Automerged

Dear Toxible, it does look like to me that you are copying messages and posting not so useful information when looking at your first 6 posts. In your invite you said you wanted another chance to contribute, but I don't think this way it is helping the forum.

There are moderators on the forum for that job, and there is a report button in case you felt inconvenient with a post/member, moderators then can decide what to do.

[sarah90]

There are moderators on the forum for that job, and there is a report button in case you felt inconvenient with a post/member, moderators then can decide what to do.

Yeah, I know but it felt kind of bad to do things behind peoples back. Will use the button next time.

[Toxible]

@Gallymimu: You are correct! Thanks! I've not considered the load... big stupid error... sorry

@sarah90: I'm working hard as possible to give my contribution to the forum. I do not care to have a good score writing posts at random! But there are some topics that interest me, and some material that i can't download if i don't contribute (and it is correct!)
So I'm doing possible to help someone...

@metal: When vin=0 D5 are off. so the only path for the carge on C1 is R3 + BE junction.
If you don't belive, you can also simulatethe circuit. You can find a simple, useful, free simulator here: http://www.falstad.com/circuit/
If you don't want to spend time, i've done the simulation for you. (file attached).
In the image we can see the circuit and the o-scope plot of C1 voltage:
Sorry but i've placed D5 near a net so it seem that there isn't catode... it's an overlayer problem of sim interface....
1-your circuit (note: C1 is always <4V). the red circle is the LED
2- your circuit without load (LED). note C1 now reach 9.73V because the led don't discharge it. but if you see there is some ripple on signal! is the discharge due of current on Q1.
3- i've turned off Vin  (you can short but it's the same in this case). you can see that C1 discharge (throught Q1 base).

"There are moderators on the forum for that job..." Well said

[metal]

As long as you have simulated already, can you find out the amount of current passing through the BE junction when it discharges the capacitor? I am interested to see the actual ripple measurement in (fig 3) caused by this discharging as well as the current.

[metal]

Okay, I this will solve that issue you mentioned earlier. I moved R4 between the emitter and C1.

[Toxible]

Ok! Im here again Current throught BE is little (obvious) and it reaches 108uA..
So, I've attached the txt containing the schematic of java circuit simulator. if you want copy the text and use file>import to open the circuit.
in the graphics you can see Ibe and Vc1.
the file circuit 2 contain the modified schematic you've posted.
Ibe reaches 53uA and Vc1 3.38V
This solution is the same thing as before... and it work good as before
The only advantage is that you can drive more led in series with the same bjt (R1 on emitter fix the Ic current at about Vc1max / R1)

[metal]

Then, I don't have to change the  circuit, I can use it as is, when I moved the resistor down there, I noticed that the ZC pulse got worse, 108uA is not a big problem after all, I measured it yesterday and I thought a change is not required, this is why I asked you about the base current :- )

Great help Toxible.

BTW, the snubber is a must in such inductive circuits?

[Toxible]

Hi all folks!
@metal: lets talk about the snubber network...
If I understand correctly you use the triac to bypass the resistor. now there are some way.
In my opinion, a small snubber connected to triac is required to prevent damange when triac goes off during system turn off.
but to i want understand how you want use the resistor. I see 2 alternatives.
1- charge slowly the cap after the transformer
2- avoid the connection if the voltage is not zero.
this problem are similar but they are not the same thing. If you have a very big cap to charge, is not enought to switch on the triac on zero crossing. infact in this case the derivative of the voltage is high enough to turn the magnetothermic breaker off.
So we need to get a value for the capacitance "seen before primary" and then find the best solution.
If this value is quiet small, we can use triac only to solve problem 2 and the resistor is not essential.
If the value is high the resistor it's a method to solve problem 1 (and 2) but we need to calculate the power dissipated by resistor.

[metal]

Here is what I got until now. Simulation is truly nice for a lamp, there is a transformer in the component list, try it (without lamp) and see how things turn bells for the Christmas.. I don't like what I see with the transformer and TRIAC, may be I did not know how to setup transformer parameters, regardless of that fact, things started to become more and more worse so quickly. May be it is a bad idea to use a TRIAC? I believe it is really difficult to control inductive loads with a transformer, but how others use a TRIAC then?

I started with 4 pulses and the 5th continues connecting the mains. I tried many ways tbh, I don;t know, the more I dig in this topic the more I get lost :- ) damn power electronics...

@Toxible
Look at the schematic, I don't understand your last post.

[Toxible]

... there is a problem.. i have not proteus intalled.... can you post a screenshot?
In next days i will try to install proteus

[zab]

Hi,
The simulation can not give you a realistic picture. So my suggestion is try it practically.
The simulation is ok. the difference on inductive load is due to back emf of transformer . so donot worry about it as long as it is with in working range.
The difference after full conduction is because triac can not conduct at less voltage near after zero crossing.

The 5th pulse need to be delayed so that switching  can be at zero crossing. Short time triggering pulse create problems near zero crossing.
I use keep pulse on up next zero crossing for smooth fuctioning.
I am attaching pics and proteus file for ref,

[metal]

The 5th pulse need to be delayed so that switching  can be at zero crossing. Short time triggering pulse create problems near zero crossing.
I use keep pulse on up next zero crossing for smooth fuctioning.

I don't understand, can you please give more explanation.

[zab]

Just see the simulation, you will see that 5th pulse trigger the triac before voltage across triac reaches at zerocrossing.The first 4 pulses were triggering traic in positive half. while the 5th pulse trigger the triac in negative half pulse.


The width of trigger pulse is very important. Ii was my personal experience that when the trigger pulse is of short duration it some times does not
trigger the triac at all. the phenomena normally occurs when applied voltages on triac are not sufficient to hold it on in conduction mode.To overcome this problem I use to sense both positive and negative zero crossing.and trigger pulse remain on up to next coming zero crossing to ensure that triac conducts in all situations.
I think it would be clear now what I intend to tell.

I have been looking for some isolated way of measuring current with oscilloscope,What type of method you are using to monitor the current ,As the clamp meter does not show the peaks.

[metal]

This means I will have to keep the pulse even after zero-crossing, but how long after zero-crossing?

That's funny indeed, I was going to ask the same question because I am about to buy an oscilloscope because of this project, clamp meters can't catch this exact halfwave current, which is the current I am seeking.. I am thinking about tektronix tds2002c but I find its record length to be really bad.

[alexxx]

I am thinking about tektronix tds2002c but I find its record length to be really bad.

If you are to pay a respectable amount of money to buy a decent tool like tektronix, then in my humble opinion it should have 4 channels. Otherwise search for a cheaper oscilloscope. You will regret for paying lots of money on it when time comes and more than two channels are needed...

[sarah90]

I am thinking about tektronix tds2002c but I find its record length to be really bad.

I paid less than a third for my (now hacked to 200mhz) hantek scope and it has similar features to the tds2002c. If I would buy a scope today I would go for the newest rigols, or if money is no object the new agilents. Of course a second hand agilent/tek scope is an option too. I would not go for a recent tek, as they seem to lag behind.

[Gallymimu]

For high end brand I would only go Agilent at this point.  I have found the Tek scopes to be a bit behind in user interface quality and responsiveness.

For budget I use Rigol.  Their DS1000D series scopes are the same hardware and mostly the same firmware as the low end Agilent scopes (Rigol makes them for Agilent)

[sarah90]

For high end brand I would only go Agilent at this point.  I have found the Tek scopes to be a bit behind in user interface quality and responsiveness.

For budget I use Rigol.  Their DS1000D series scopes are the same hardware and mostly the same firmware as the low end Agilent scopes (Rigol makes them for Agilent)

Is that another way to say "I second that!"?

[Gallymimu]

Is that another way to say "I second that!"?

YES!

[metal]

if you want, I can split the posts where oscilloscope discussion starts and we can discuss it better over there, what do you think guys?

[Gallymimu]

if you want, I can split the posts where oscilloscope discussion starts and we can discuss it better over there, what do you think guys?

I don't know if there is a lot more to discuss but it does seem to be a useful topic if scopes haven't been covered so I'd split it as you suggest.  Maybe others will pipe in with opinions and we can have some good discussion!

[sarah90]

I agree. Oscilloscopes should not be part of this discussion. But I would like to see a topic where people can tell what equipment they have, their experiences and may be some tips for people looking to buy.

[zab]

This means I will have to keep the pulse even after zero-crossing, but how long after zero-crossing?

That's funny indeed, I was going to ask the same question because I am about to buy an oscilloscope because of this project, clamp meters can't catch this exact halfwave current, which is the current I am seeking.. I am thinking about tektronix tds2002c but I find its record length to be really bad.

As you are keeping the trigger pulse on permanently afterward so you need not for any wait for delay, you can switch on the trigger pulse as you get zero crossing pulse. At supply volt reaching to conduction level the triac will automatically on.

About the second point DSO will work. The simple solution is to use a resistor in series of load better at neutral point for measurement of current. But please insure that your scope should be supply thorough isolated transformer (not connected to any ground point).for safety reasons.As the probe share ground point. The other way is get some isolated current probe. But personally I too looking for some thing simple. and easily available.  I saw a topic where someone suggested to make it with hall sensor.But It performance is questionable.especially frequency response,
An other way is to use allegro ics. with some operational amplifiers.I saw any article in silicon chip about this way.but have not checked it.

[metal]

But what I really need is to switch triac on at specific timing for the first 5 pulses, you can see this in simulation, and then I switch the triac on and relay on then turn the triac off so that relay takes it from there. I don't think I will need to switch the triac at earlier time for the first 5 pulses.

Concerning the current, I will start with fast fuses to see if they burn first, there are many other cases to handle too, let's see, I don't have an oscilloscope at the moment, but will start testing to see if the concept works or not at least.