Inverse parallel SCR driver circuit with MOC3041/AC mosfets switch?

[eneuro]

Hello,
For the moment used two AC mosfets switches to triger inverse parallel SCRs in 230VAC circuit, but maybe it could be possible use only one AC mosfest switch, while it looks like those inverse parallel SCRs seams to work with optotriac in circuit below?

I'll use instead of one 360R rather 2 x 180R, but unsure if those diodes D1/D2 really will do the job and those SCRs gates are safe? 
They sugest 1N4001 diodes which are rated only 35Vrms 
I have 2 x 800V 25Arms SCRs thyristors (2N6509)  in this configuration and my concern is those 1N4001 diodes, while probably they will be fine for sinus shaped 230VAC mains, but if wanted use AC mosfets switch or phase control, so it can happen that there will be 325Vmax on SCR gate, so seams too much for 35Vrms 1N4001 diode.

Anyway, if one of SCRs will not switch on, than.... we have a toast because of during 10ms 230Vrms 77W (so ~38Wrms) disipated as heat in 360R +330R ?   
Probably I'll put there 1N4007 1000V diodes D1/D2, while do not like too much this 35Vrms of 1N4001 

[moffy]

I would go with the higher rated 1N4004 or higher, but the higher rating is not really necessary. It's the triggering that worries me, that single 360 ohm resistor, at peak voltage that will be a 1A pulse. But you need a lowish value to be able to trigger at lower phase angle/voltages. What do your data sheets for the SCRs say about gate current?

[matseng]

What is the advantage of using dual thyristors instead of a single triac? Easier heat dissipation since each device only handles half the average load?

[Circlotron]

so it can happen that there will be 325Vmax on SCR gate, so seams too much for 35Vrms 1N4001 diode.

Anyway, if one of SCRs will not switch on, than.... we have a toast because of during 10ms 230Vrms 77W (so ~38Wrms) disipated as heat in 360R +330R ?   

As soon as one of the SCRs trigger, the gate drive voltage is effectively removed. I have used in principle this same circuit but with an 8 amp triac between the gates of a pair of 1000 amps SCRs on a 415VAC line and 1N4004 diodes and a 15 ohm 1 watt resistor in series with the triac. We made about 500 three phase units and they worked just fine.

As to why not use a triac instead of two SCRs - an SCR can take about 20 times normal rated current for one mains half-cycle and also has much better dv/dt rating. Triacs are not nearly as good.

[matseng]

As to why not use a triac instead of two SCRs - an SCR can take about 20 times normal rated current for one mains half-cycle and also has much better dv/dt rating. Triacs are not nearly as good.

Thanks, this is a good titbit of info to file away in a corner of the brain, someday I might have use for it.

[eneuro]

I have used in principle this same circuit but with an 8 amp triac between the gates of a pair of 1000 amps SCRs on a 415VAC line and 1N4004 diodes and a 15 ohm 1 watt resistor in series with the triac.

If someone used similar inverse parallel SCRs driver circuit I'm much more confident now since it will control a few kW inductive transformer primary load (with additional snubber network), so
thanks for sharing this - anyway I'll use protective glasses for sure while turning on circuits like this just in the case "toast" happens

In this case using smaller series resistance than 360R with MOC3041 http://www.sos.sk/a_info/resource/d/moc304x.pdf   rated as 1A (ITSM) probably could work with nice sinus AC, but inductive transformer primary will be driven, so I decided to stay at safe 0.9Amax gate current, since I might want use instead of zero crossing also AC mosfets switch with phase control and in this case when turning at 230VAC voltage maximum 325Vmax 90 deg phase, than smaller series resistors could allow gate currents above 1A-my SCRs has 2A IGM, so not too much room and I have nice soldered SCRs with reinforced PCB tracks with additional copper wires, so I wouldn't like to kill them just by stressing their gate current in phase control mode.
Anyway, looking into my SCRs 2N6509 datasheet http://www.onsemi.com/pub_link/Collateral/2N6504-D.PDF it has 300A surge current capability which is nice and typical gate threshold current only IGT: 9mA, so probably it will triger at ~3.24V mains with 360R series resistance, which is 1/100 of 325Vmax at 90 deg phase angle, which means 0.571 DEG phase shift from zero crossing, which is at 50Hz mains: 20ms*0.571 DEG/360 DEG= 0.031ms= 31 us estimated typical delay from zero crossing at 230VAC 50Hz mains 

[Circlotron]

but inductive transformer primary will be driven, so I decided to stay at safe 0.9Amax gate current, since I might want use instead of zero crossing

Be careful turning on a large transformer at the AC voltage zero crossing because it will saturate and pull a lot of current for a fraction of a second. When the transformer is running normally, the voltage half cycle from 0-180 deg changes the magnetic flux from negative maximum to positive maximum. If you turn on at zero crossing the magnetic flux will start from zero, not negative maximum so when the voltage goes 0-90 deg the magnetic flux goes zero to maximum then from 90-180 deg it goes from maximum to very high and so current is limited only by DC resistance of windings, inductance is almost zero. Transformer makes a "BONGGGG" noise.

[eneuro]

When the transformer is running normally, the voltage half cycle from 0-180 deg changes the magnetic flux from negative maximum to positive maximum.

I've additional zero voltage detection circuit connected to MPU pin with 5kV optocoupler insulation from 230VAC mains, so I can detect zero voltage, but additionaly I have information about mains polarity while I've used two PC817 in anti parallel input diodes (with resostors of course to limit diodes current).
This means that I switch on ALWAYS with main polarity which is oposite to when switched off, since I ALWAYS try to turn thyristors for full AC mains period (360degs), so isn't transformer core be demagnetized and ready for another 20ms pulse?
In the case of phase control similar steps  are performed: deetcting the same polarity when zero crossing, than delay 0ms-10ms to achieve wanted phase and fire again in 2nd half period with the same delay from first 0-180deg.
Since I know when the same mains polarity period begins I always start new full wave pulse while transformer core is demagnetized from 2nd half period, so isn't such transformer control similar to its normal run? 
Additionally I have current mains monitoring, so can use also this feedback to control transformer-for the moment it is used as software fuse, which averages currents (Hall effect current sensor) within defined period of time and will force turn of switch at the end of current full wave period- still try to keep transformer core in well defined magnetized state for next turn on...

Anyway, resoldered prototype PCB and it looks like discussed inverse parallel SCR driver circuit survived simulated shorting MOC3041 optotriac output as 230VAC switch with 25W light bulb in series, so it is time for more serious tasks and probably AC mosfets switch will be fine to, since basicly it doesn't matter how we short those thyristors gates with those 2 x 180R 1W resistors in series with thyristors gates, so yeah-quite fersatile way to drive inverse parallel SCRs 

NOTE: I used 1N4007 1000V diodes D1/D2 @ 230VAC mains 50Hz instead of 1N4001 sugested in that MOC3041 datasheet from where this schematics comes 

[eneuro]

... additionaly I have information about mains polarity while I've used two PC817 in anti parallel input diodes (with resistors of course to limit diodes current).

In fact I don't need two PC817, since I'm only interested to detect lets say positive mains voltage period, so I've quick prototype and used 1N4007 anti parallel diode instead of another PC817 not needed no more, while I've MPU so can calculate when positive period begins and ends, so both AC mains period and small delay offset from real voltage zero crossing can be computed, so no need to even hardcode in software 20ms period, but for safety reasons I check if calculated mains period from this PC817 based VZC is within expected tolerance range (lets say +/- 20%), to ensure MPU runs with proper fuses programmed and is switched correctly to desired clock frequency, before send high power (full wavelength 20ms) pulses to transformer primary 
This is new prototype of my AC VZC detector which let me know when I have positive mains voltage period

It is designed (soldered) to be able hide it inside part of mains wire by adding thermal covers with hot glue inside and PC817 output open collector to drive MPU pin pulled up by 10k resistor or so

[Circlotron]

When the transformer is running normally, the voltage half cycle from 0-180 deg changes the magnetic flux from negative maximum to positive maximum.

This means that I switch on ALWAYS with main polarity which is oposite to when switched off, since I ALWAYS try to turn thyristors for full AC mains period (360degs), so isn't transformer core be demagnetized and ready for another 20ms pulse?

Yes the transformer core is demagnetised.
No it is not ready for another 20mS pulse. Not even a 10 mS pulse.

At startup, if the voltage is applied at zero crossing when the transformer is demagnetised, the transformer core will not support an entire half cycle of rated applied voltage. It will only get about halfway through that first half cycle and then saturate. The area under the curve of one half cycle (10mS) of rated voltage is enough to pull the transformer core magnetism from fully one direction to fully the other direction. If the magnetism starts from the middle, zero point as it would at switch-on it can only swing half as far and will hit the magnetic limit and amps will go to the moon.

[moffy]

If your load is at all inductive it can help the SCRs pull in if they have a nominal resistive load to enable the hold current of the SCR.

[eneuro]

At startup, if the voltage is applied at zero crossing when the transformer is demagnetised, the transformer core will not support an entire half cycle of rated applied voltage. It will only get about halfway through that first half cycle and then saturate.

Probably, I've used the word demagnetized, but I meant magnetized within negative second half 10ms pulse 
Probably you are right that when transformer core is not magnetized at all (to be right will have to read more how long transformer core can be in magnetized state, eg. when voltage is applied to it primary coil to the levels close to saturation, so next the same voltage (not reversed) saturates core...), than maybe it could be better idea after zero crossing detection, delay 90 degs and wait for voltage maximum when we start new a few 20ms pulses sequence and this is approach which is programmed in this "Beduino" (I prefere this name for those mainstream MPU boards  ) spot welder control software:
http://www.avdweb.nl/arduino/hardware-interfacing/spot-welder-controller.html#h11-programming

Code: [Select]

void weldCyclus(int weldTime_ms)
{ sinusMax();
  pulseWeld(preWeld_ms);
  delay(weldPause_ms);
  sinusMax();
  pulseWeld(weldTime_ms);
}
void sinusMax()
{ while(digitalRead(zeroCrossPin));
  while(!digitalRead(zeroCrossPin));
  delayMicroseconds(sinusMax_us); // to prevent inrush current, turn-on at the sinus max
}where they used their own zero voltage crossing on MPU pin (schematics can be found on the same page: http://www.avdweb.nl/Article_files/Arduino/Spotwelder-controller/Circuit.jpg ), so instead of MOC3041 they use classic optotriac without this feature (MOC3020).

It looks like they start each welding pulses with searching for mains voltage sinus maximum, which means 90 deg phase shift from zero crossing detection 

However, I've found also other DIY spot welder, where MOC3041 with zero voltage crossing support was used with decent results:
Semi-automatic spot welder 2,6V 1kA
Further, the latest NPN transistor switches zero-cross detection optotriac MOC3041, and the triac BT138 which is closing the primary circuit.   

Anyway, I have both MOC3041 and MOC3020 optotriack, but in this case in its datascheet http://www.ti.com/lit/ds/symlink/moc3020.pdf we can find rather small RMS values, 10 times lower than those for MOC3041? 

Output on-state current, total rms value (50-60 Hz, full sine wave): TA= 25 °C 100 mA
TA= 70° C 50 mA

Nonrepetitive are in the 1A range:

Output driver nonrepetitive peak on-state current (tw= 10 ms, duty cycle = 10%, see Figure 7) 1.2 A

In comparision for MOC3041 http://www.sos.sk/a_info/resource/d/moc304x.pdf we have

Peak Repetitive Surge Current ITSM 1A

which nicely fits into limited SCRs gate current with 360R total resistance giving us even at 325Vmax <1A , which in the case of MOC3020 with 0.1A rms current doesn't look great if we wanted to triger SCRs just at sinus maximum: @ 230VAC 325Vmax 

However, this is not such a big deal for me, since I've another way to triger those SCRs using... AC mosfets switch, where it is easy to find 1A N channel mosfets rated at 1000V, so maybe not worth messing with MOC3020.

More advanced resistove spot welders control (RSW) is described there Control and Power supply for resistance spot welding (RSW), but it is another 128 pages story 

Update: http://en.wikipedia.org/wiki/Inrush_current#Transformers

Worst case inrush happens when the primary winding is connected at an instant around the zero-crossing of the primary voltage, (which for a pure inductance would be the current maximum in the AC cycle) and if the polarity of the voltage half cycle has the same polarity as the remnance in the iron core has. (The magnetic remanence was left high from a preceding half cycle).

So, while I keep focus to turn on switch on the same polarity of the voltage half cycle, but will have usually reverse polarity of the remnance magnetization in the iron core, than it might not be the worst case scenario even if MOC3041 was used, and I saw also other more professional spot welders and they used full wave 20ms multiply pulses for sure, but unsure if they start each full wave pulse sequence at voltage maximum

When we read more than it looks like premagnetizing transformer core only makes sense and....

To avoid magnetic inrush, only for transformers with an air gap in the core, the inductive load needs to be synchronously connected near a supply voltage peak, in contrast with the zero voltage switching which is desirable to minimize sharp edged current transients with resistive loads such as high power heaters. But for toroidal transformers only a premagnetising procedure before switching on allows to start those transformers without any inrush current peak.

...I forgot mention that... I've in parallel to those SCRs switch... resizstive load .. old school light bulb 
I have also the same light bulb on transformer primary windings.

Now, question is, how to premagnetize transformer iron core... but maybe I've already did it with my light bulb in parallel with this switch?
I have no idea, how huge this premagnetization current should be-for the moment only 25W light bulbs are used as dummy loads, so this current in transformer primary is in the range of 0.1A or so, but maybe it helps? 

[eneuro]

If your load is at all inductive it can help the SCRs pull in if they have a nominal resistive load to enable the hold current of the SCR.

I've ~0.1A resistive load on transformer primary (in parallel), while another 25W light bulb is used as dummy load 

Another light bulb in parallel with SCRs, so there is small current bypassed via this SCRs switch, but output current on transformer secondary even if it is 100 x bigger it is as low as 10A, so no problem to touch two metal plates while prepering for (spot)welding.

Apropos this premagnetizing, I've found also this document:

researchgate.net: PDF: Controlled Energizing of Three-phase Transformer: Analysis of Test Results

Another way to reduce the transformer inrush current is using the controlled switching strategy, which
consist of closing each pole of a circuit breaker at defined instants of time at which the magnetizing
condition in the transformer iron-core coincides with that that would be produced by the voltage being
applied on the transformer windings.

This above is what I've already implemented in my ATTiny85 MPU to controll my big ass transformer in (spot)welder.
I've 10A slow fuse so input power is limited somehow to 2.3kW.

Lets test it and see if this inverse parallel SCR driver circuit with MOC3041 will work in (spot)welder, since I don't need lower power than one fullwave 50Hz 230VAC mains, so minimum 20ms pulse should be fine for this RSW for the moment   

[eneuro]

In this publication pdf: Residual magnetism we can find hints howto demagnetize core

Pre-magnetising might be opposite to this demagnetisation process shown above, I guess, so probabbly will use existing AC mosfets switch with upgraded Inverse parallel SCR driver circuit described there and before sending 20ms (spot)welder fullwave length pulses, pre-magnetization within a few seconds will be performed after turning transformer on througth mentioned light bulbs as samll dummy loads (without shorted secondary ), by apllying phase shifted full wave voltages to transformer primary.
I hope, it helps. I have no better idea for the moment. This is not a huge 3phase transformer-only a few kW... 

[Circlotron]

If you were starting from scratch, one low tech but very reliable solution would be to have a transformer with primary (and consequently secondary) voltage rating at 2X the intended operating voltage. That is to say, a 460V to 10V tranny running at 230V to 5V. That will never have a switch-on current surge because the area under the curve of the first half cycle of voltage will only bring the magnetic flux from zero to max rated. When operating continuously the flux would swing from -50% to +50%. Double size transformer probably not very economical though.

[eneuro]

Double size transformer probably not very economical though.

This is trafo I've upgraded SCRs switch for
Secondary is ~2.6VAC and copper pipes 15mm in diameter (>23mm2 cross section) provide spot welder current to electrodes.

Quite nice transformer-a lot of copper and electronis should disapear first in smoke, before this transformer primary will get even warm 

New mains voltage zero crossing sensor based on PC817 and 1N4007 diode in anti parallel assembled and ready for synchronizing pre-magnetization proces.


We'll see what happends, but I have Hall efect current sensor on transformer primary, so in my digital fuse I can change SCRs phase time, if averaged current will go above set limits-overwrite input power and make it dynamically lower if unexpected large currents will be detected.
For the moment current is sensed at ~10kHz frequency, so I have 10000 measurements per second, but analog comparator can be used to make continuous overcurrent monitoring and allow fast response to high currents present, but while SCRs are used it doesn't help too much since trigered SCR will stay on untill current falls below some level 

BTW: In next RSW upgrade 3 phase rectified mains or a few batteries in series  feed to full H-bridge mosfets driver to create nice sinus voltages should allow better controll this transformer primary current