Hi,please be patient and read the description of the situation:
A friend is having lot of problems trying to switch primary winding of a
400VA toroidal transformer by a triac driven by a zero crossing optotriac. Having a bit of measuring instruments i tried to help him. The net voltage is 220V,the secondary winding is left open,no load at all. Measuring the current peaks by means of a current clamp probe and a cheap digital oscilloscope i measured current spikes over 80A.
After a quick research on web i have plenty of explainations of that fact.This is a very simple one
formatting link
The last image should rapresent the situation i'm dealing with. It looks like at the zero voltage starting the current from zero,rather from the negative peak,this current is able to generate enough flux to saturate the core,with expected effects.
I 'm trying to implement a soft start using a random phase opto triac instead of a zero-crossing one. I arranged a zero crossing detector to trigger a microcontroller,and i can fire the optotriac and the triac in every moment during each semi-period.
The first soft start attempt consists in this
1)detect the zero
2)wait until 0.1 ms before next zero(end of semiperiod) and turn on the Triac
3)detect next zero,Triac turns off a bit later,i guess,when the current reaches zero
4)wait until 0.2 ms before next zero and turn on the Triac,current flows now in opposite direction than in 2)
5)same as 3) and so on
When i'm close to 90% of the semiperiods i turn off the Triac for a couple of seconds,i dont want to remain in ON state so to measure current spikes only during the "soft start"
Sadly ,big current spikes are still detected.
Could anyone please suggest me how a correct algorhithm should be done?
It's tricky using a triac in series with a transformer primary. It can be both hard to keep them on, and hard to turn them off. RC snubbing is critical, and gate drive may have to be sustained, not just pulsed.
NTC inrush limiters work well. They look like big black disk ceramic capacitors. Simple.
You can also switch in a series resistor with one triac or relay, to let things charge up/demagnetize, then use another for the direct connection. But that's hard on resistors.
Toroidal power transformers can be tricky that way. We had one amplifier that, when you switched it on, you could hear the wires in the wall jump. We were peaking around 1000 amps sometimes.
These transformers should have an auxiliary primary winding for startup. Maybe 20% extra primary turns but lots of copper resistance.
They are good but they won't work in quick-sequence brownout situations. Happens here on occasion. Power goes off, tries to come back on in rapid-fire bursts. Sometimes stays on, sometimes not.
But not all are that way. I've got one here in the office. 1000W, medical grade, the good stuff, I believe made in Sweden. It's on a 15A circuit and neither the lights flicker when turning it on nor did it ever trip the breaker.
The only weird thing is that it breaks into a faint growl once in a blue moon. Some sort of resonance with the chassis.
Then you might as well put a proper start-up circuit in front, it's smaller.
I did a CAMAC crate with a huge power transformer, and used inrush NTCs in the transformer primaries. They seemed to work fine, even for quick off/on cycles. Can't explain why.
Got any suggestions? A couple of triacs and a BIG resistor will work, but that's a hassle. And resistors don't like this longterm.
" A friend is having lot of problems trying to switch primary winding of a
400VA toroidal transformer by a triac driven by a zero crossing optotriac. "
** No real need exists for any soft start circuit with a 400 VA toroidal - the in-rush surge currents are manageable by use of slow acting fuses. The larger sizes may need one - starting at about 600 VA, particularly if the load is a big bank of filter caps.
The OP is barking mad trying to use uPs and opto-coupled triacs.
Toroids like to hold their magnetic state longer that EI cores. You might be turning on the core in the same direction that it was last magnetized. An inrush current limiter should help, long enough for the core to walk back to a balanced state. EI cores have a natural gap that help eliminate this effect. In larger equipment that is pulsed, such as X-ray, there is usually some sort of means to record the last magnetized state of the primary transformer. So that the core does not saturate and put stress on the SCR's.
Yep, that's why you shouldn't ever use zero crossing for the first cycle of turnon of an inductive load. Turn on at 90 degrees (voltage peak) instead, and KEEP it on by continuous gate drive on the triac,
If you must, for isolation, use a transformer to drive the triac, consider replacing it with a relay. It's more power-efficient, anyhow.
** Toroidal transformer cores are not left in a magnetised state by the simple act of switching off the AC supply.
A core may become magnetised if it is switched onto the AC supply for only a half or one cycle and so does not settle down - happens when the AC fuse blows or breaker trips at switch on.
** But not by any great amount.
E-cores of 800VA or more need soft start circuits too.
I designed several firing circuits for high power circuit breaker test sets, some of which use toroidal cores with a total of 10 kVA and output currents of 20 kA or more. We used dual SCRs, and the most recent firing circuit uses a PIC18F2420. We also had inrush current problems that were worse for the toroids. Here are some things we discovered and techniques we tried.
NTC thermistors were effective, but they caused waveform distortion at the initial firing, which was unacceptable for our purpose of high current primary injection testing of circuit breakers. They needed a clean starting waveform with minimal DC offset. Also, when the thermistors heated up, they were no longer effective until they cooled down, which could take several minutes.
We learned that the ideal initial firing angle was somewhat less than 90 degrees, and closer to 70 degrees, because the load was partially resistive. We used a variable initial phase angle control and adjusted it for equal peak amplitudes of all half-cycles.
We found that the transformer core would be magnetized if there were an unequal number of positive and negative half-cycles. So we designed our controller to produce even numbers of half-cycles. Actually we programmed it to drive the gates for about 4.7 cycles when we wanted 5, because the SCRs would remain in conduction until current reached zero. But if we used a time corresponding to exactly five cycles, sometimes the inductance of the load would carry into another half-cycle.
But sometimes we could not control how long current flowed, because the breaker under test would trip depending on the current and a net DC component would exist, which magnetized the core. We found that, under those circumstances, reversing polarity of the next pulse train did not produce the high instantaneous current we had otherwise. On a 480 VAC 200 amp service, this was sometimes enough to trip the main breaker for the entire building, and we estimated well over 2000 amps. We could hear the conductors slap against the conduit, and loose cables to the test set would jump.
We found that it is necessary to keep current on the gates at all times, even on the SCRs that were not normally conducting. This was because of the reactive load, where current and voltage are out of phase. Otherwise we often saw waveform distortion, especially at the crossover points.
A proposed modification to the design was to apply a series of diminishing phase-delayed pulses after the breaker tripped, to demagnetize the core. This was never implemented, but we proved its possible benefits by adjusting the test set to a lower output, initiating a pulse, and then returning to the higher setting where otherwise it would have pulled a huge current.
Another problem we had was unintentional half-wave triggering of the SCR when primary power was switched on and off at the source. We could not tolerate a large R-C snubber across the SCR because it caused leakage currents which were excessive and potentially unsafe. So we used a large snubber of perhaps 30 uF and 100 ohms across the load side of the circuit, which included a tap switch that was the main culprit. This reduced the random firing to some extent, and also improved the power factor of the load, but it also caused the output to remain on for a while after the SCRs were turned off because of the LC network. Ther series resistor was necessary to reduce the "Q" and minimize this effect, but the exact values of the components had to be adjusted for optimal overall performance, with some trade-offs.
I hope that helps answer your questions and provides guidance for a successful implementation. If you are interested in more details, or perhaps obtain one of our firing circuits for your own modification and use, please contact me.
Thanks,probably this will be an unavoidable option,but very expansive in terms of space If possible I would like to solve with "intelligent" start ,without adding bulky components.
"Martin Riddle" ha scritto nel messaggio news:hqdvlc$7n2$ snipped-for-privacy@news.eternal-september.org...
Thanks,Martin I hoped that feeding the primary with a growing "time slice" of current ,inverted in direction at every semiperiod was equivalent to alternatively start by a codition of alternatively growing induction,i.e. at semiperiod T1 B=1,then B=-2;then B=+3 and so on.
My impression is that rather i start every time by B=0;but i'm not shure
Yep, that's why you shouldn't ever use zero crossing for the first cycle of turnon of an inductive load. Turn on at 90 degrees (voltage peak) instead, and KEEP it on by continuous gate drive on the triac,
Thanks, Probably i'm wrong,but powering at 90 degree i still have peaks,but let me try again
If you must, for isolation, use a transformer to drive the triac, consider replacing it with a relay. It's more power-efficient, anyhow.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.