Sonsivri

Hello,

I am trying to understand how I can do a replica for this https://www.tauscher-transformatoren.de/assets/pdf/R_current_inrush_avoider.pdf
I can't understand how this work, how I should switch the traic, when, for how long? which quadrant, etc..

From what I understand TSR is using unipolar voltage impules to premagnetise the transformer. In the PDF you linked to the look to be in applied in the posetive half on the falling edge. Probably because this will most easy usning a solidstate switching element. Is also looks like the transfomer are switched fully on some time during the second pulse. The width of the pulse needed will depend on the transformer so TSR devices are have an adjusting pot.
Ps I have also seen seen pictures using the negative period. I do not think it matter as long as the premagnetise pulses are unipolar.

to minimize inrush current you should switch on the transformer near the positive or negative voltage maximum of the AC sinus.
Gate pulse length does not matter, once the triac is triggered it will stay ON until the load current drops to almost zero.
Just check the datasheet of your triac for minimum trigger puls length.

finally http://ww1.microchip.com/downloads/jp/AppNotes/jp587377.pdf

This is correct. The high inrush current is normally a result of core saturation during turn on, pushing the flux density into the saturation regions of the particular B-H curve. If you apply voltage at one of the peaks, you only accumulate 1/2 of the volt-seconds before the B-H trajectory moves through zero and in the reverse direction. Switching on at zero volts you accumulate a full half cycle of volt-seconds and  can push the core into saturation.

Do you mean that I have to switch either -ve or +ve - but not both - at 90 degrees for like 20 cycles before applying full power?

You just have to catch it on one of the two peaks and keep it on after this "trigger" point. Since you will be moving in the safe areas of the B-H curve after this, you should not experience a massive over-current.

I personally used this soft start with inrush current limiter

I see, then it is just a matter of switching at 90 degrees instead of zero crossing point and keep on, and that's it? I don't need to repeat switching at 90 degrees for many cycles? this is much easier than I thought, but the implementation of microchip and some papers I read is different, still in the papers there are missing information, unfortunately. Honestly I still don't get it.

more finding

https://www.youtube.com/watch?v=SVLGHB2IxxU

https://www.instructables.com/id/How-to-Make-a-Softstarter/

No you are wrong on this matter. Switching on toroidal transformer on peek voltage is NOT a good idea.
Also @Metal and @Vern take a look st this document https://www.noinrush.de/fileadmin/_migrated/content_uploads/01-speech-from-berlin-at-cwieme-2004_01.pdf

Sideshow Bob,

very good document indeed.
But it covers some very high power assemblies, where the extra effort for inrush protection does not matter.

However, for lower power and simpler designs it is still advisable to switch near the voltage peaks, because switching at zero crossing is far worse.

Yeh I can agree on it is kind of a lesser evil to switch near the peak. And to be pedantic the "correct" thing is to use a train of  premagnetise pulses with the correct with. The number of pulses and width may vary dependent of the transformer. But well as you said this will be needed only for transformers of some size. Kind of not common in the hobbyist world. I also found some more docs about TSR, they may only appeal to those who are interested in the TSR topic though. Both links are public.
https://www.emeko.de/fileadmin/_migrated/content_uploads/01-Doc-101-r4-en_TSR_Initial_Adjustment.pdf
https://www.noinrush.de/fileadmin/_migrated/content_uploads/01-Press-TSR-Speech-2009-en.pdf
https://www.noinrush.de/index.php?id=98&L=1

I think I am getting some where, I am premagnetizing at the +ve half, while TSR at the -ve half, doesn't really matter. But actual number of pulses that is needed for premagnetization is an unknown to me. Am I getting some where, or it is just an illusion :D

From what I have understod As long as the pulses are unipolar it does not matter. In the 01-Press-TSR-Speech-2009-en.pdf document I think I read that 40 pulses was needed. But that was not a torroid. They are somewhat iffy about how many pulses that are needed

The technique of switching at one of the peaks will only be very effective if you assume there is no residual magnetism left in the core and the B-H curve is close to zero crossing. If there is any residual magnetism left in the core, you should try and reset it back to the zero point on the curve.

This can be done by applying a lower amplitude AC voltage which decays towards zero over a period of time, effectively demagnetizing the core and moving the B-H point to near zero. After this period, you could switch safely on one of the peaks. This could be accomplished by using some PTC based current limiting circuit, but you should still monitor the current to help determine how long the decay must be to reset the core.

Using the pre-magnetization pulse method, you are also a bit in the dark on how many pulses will be required for a certain transformer without monitoring the current as well. There will be a steep rise in the current when one of the extremes of the B-H curve is reached.

I only see the inrush to be a real problem with large transformers above 1 kVA. With smaller transformers there will be enough copper losses to limit the inrush current.
  

here is the schematic of TSR

only schematic  :D :D

Here is a simple design concept for a TSR to play with. It works in simulation, but I have not tested the actual hardware.

No you don't have to repeat the switching, just make sure it turns on initially at the peak of the mains and keep the switch on. The inrush current will be the highest during the initial cycle and then will  gradually cease. That is the maximum you could do to limit the inrush current without adding hardware. I don't know what Microchip says but what I am saying is backed up by my experience. The only improvement you could do, if you are using digital control, is to store information at what point of the mains sinusoid the transformer switched off. That will determine at what point (-Br or +Br) the transformer core will remain. So if it is at -Br, you turn the switch on at +Vmax of the mains. Similarly, if the core is at +Br you turn the switch on at -Vmax. This will reduce the inrush current even further.

old school guru :)

were you able to simulate the inrush occurring? I shall share my Proteus simulation and PIC code, it is darn simple.

Posted on: September 06, 2020, 11:37:22 11:37 - Automerged

---

here is my proteus and PIC code. Unfortunately, I have no access to an oscilloscope to measure actual values concerning inrush current with/out this solution. If someone can check if this is an effective design, it will be really awesome. This is a test code which is very simple, sure I will change it to use an interrupt rather than just simple delays as in the simulation.

Yes, I added the simulation of the inrush current for hard switching and soft switching to my previous post. You can also see the comparison of the flux swing for both operations.

Attached is a Proteus file for the circuit I posted before., if someone want to play with it.

smart and simple.

@PM3295

If you tried my sim and been OK with how things work, let me know, so I will modify the code to use an interrupt and I will design a PCB for it as well, in case you are satisfied. The PIC solution is a lot simpler and less components to build.

metal, I have played around with your sim and it looks good to me. Do you have a way to vary the  switch pulse timing to compensate for the relay switching times? It will be nice if you could have an external preset pot that you can adjust to get it spot on.

Don't tell me you have not looked at the crappy code I wrote :] I will modify that, for sure. But, Relay switching times, frankly I don't know when the relay should switch ON, do you mean during the +ve half-wave, -ve half-wave, or should I wait few cyclyes after the SCR half-waves then switch the relay ON, guide me and I will translate that to the code using a POT and reup the simulation.

Yes PM3295 raised an important point because the relay switching time (from the moment of applying voltage to the coil till the relay contacts close) can be anywhere from 3ms to 20ms depending on the relay's type and contact's current. You could reduce the switching time by a factor of 2 or 3 by applying a higher (than the nominal) voltage to the coil for a short time. This requires unfortunately an additional voltage source. For example for a 12V relay  you should have a say 10V source to keep the relay's contacts closed without too much power dissipation in the coil and another 24V or 48V source that will be applied to the coil trough a series RC network at the moment of turning on. By playing with the RC values that control the duration (and the current) of the higher voltage pulse applied to the coil one can determine (experimentally) the shortest switching time achievable with the additional voltage source.

Honesntly, my first approach was using a TRIAC that switched at 17~18ms of each +ve half-wave for few cycles and then switched ON at 90 degrees of the +ve half-wave permanently for few seconds, then arelay switched ON. I am starting to see complications in what you have suggested.

No, I have not looked at the code itself.

As dikris said;
When you send the signal to switch on the relay to bypass the SCR, the relay will have a delay before the contacts actually close. This can vary from 5 ms to as much as 50 ms, depending on the size of the relay. So, if you send the signal to close the relay and the delay is too long, you may actually switch on the wrong polarity!

I see, then I better go with a TRIAC instead of a relay for the thyristor which can manager switching ON in a smoother manner than a relay. The relay might just short the TRIAC if necessary.

One way is to use dual SCR's for better control. Then you don't have to worry about relay switching times.

I modified my circuit to do this. So, you can still have the uni-polar pulses on the one SCR and at the switch point fire both while waiting for the relay to close.

In a practical circuit, one should use an MOC 3021/2 device for triggering.

Attached is the Proteus file for the modified dual SCR control. I added an RC time constant in the relay switch circuit as the current model doesn't have such a parameter.

I also modified mine to use a TRAIC - again indeed - and I no longer care about when the relay has to switch ON. Proteus and code attached. It is worth noting that I also modified the ZCD, and replaced the SCR.

Posted on: September 08, 2020, 12:30:40 00:30 - Automerged

---

BTW, if I wanted to add a "POT/DIP switch" this time, then I would do it to define how many pulses to do before fully switching ON :-D

I looked at your new sim, but I don't see switching on the negative half after the positive pre-mag pulses. After these pulses, the flux will be pushed up towards positive saturation, and you need to switch at the right moment, forcing it in the opposite direction. It appears from the waveform that you switch on the positive side, and that will push it up into hard saturation. The Proteus simple transformer model unfortunately can't display this saturation effect, so the peak current will be way more than shown.

I placed a cursor where the relay contacts should ideally be closed.

Your previous sim switched in the right sense.

Last attachment is from my modified Proteus cct. The relay switch time is set very long to be able to visually see it happening in the sim.

  

In my last sim, I am dependent on the TRIAC, not the relay, so I no longer care when the relay switches ON. The relay is switched ON after 1s from the time TRIAC is fully switched ON on both directions, when the relay stabilizes, I just turn the TRIAC off, I no longer need it, and this way also prolongs the relay life. Please look at the sequence in the code as well. Also if you look at the first attachment in this post (http://www.sonsivri.to/forum/index.php?topic=68555.msg197871#msg197871) and the sim in the current post here, I am imitating the exact thing.

Looking at your sim, seems I have to keep the rhythm of pre-mag pulses as is i.e. on the +ve half-wave, and on the last pulse I have to keep the TRIAC ON so that I don't have to care about the relay switch ON timing. This also ensures that the transformer has actually continued on the -ve half-wave as per your sim.

I don't trust Proteus sim that much, so I don't trust that the relay in the previous sim has actually closed at the right moment, don't believe this lie urself too.

Look at the attachment, is that what you are looking for?

That looks perfect. :)

Are you going to include a way to vary (shift pulse timing) the amount of pre-mag current to cater for various transformer sizes? Larger transformers will need a higher current.

See first document, Sideshow Bob reply #12

sure, I can do that, may be with a 3-bit DIP switch 16ms to 19ms in 0.5ms steps, sth like 1,2 ,3 ,4, and 5..

Can you post the latest file to test?

Here you go, 2-bit DIP switch, TSR3.rar, 6.PNG

and 3-dip switch, TSR4.rar, 7.png

Looking good!

Now I will re-write the code to become a real code, this code is a big lie! There will be an INT and a delay timer...

What about starting small and then go bigger in per-mag then switch ON, what's wrong with that?

The pre-mag level needs to be set to a fixed level (after experimentation) according to the type and size of the transformer you switching.
Naturally smaller transformers will have lower currents to get to +Bmax, which will be too low for a large transformer.

metal,

Have you made a working proto of this design yet?

I built a prototype, but I was not able to test it because my oscilloscope is broken.. I can't buy a new one because the business is completely destroyed these days..

I made a quick proto of my analog design to at least see if it works as per simulation. Below is a screen capture switching a 200W toroidal transformer. It seems to be working in principle. After the initial negative peak current, it settles down quickly to normal operating current.

Red CH    :  Is the AC half-cycle reference
Yellow CH : The gate firing pulses
Blue CH    : The transformer primary current
Green CH  : The bypass relay-ON signal, NOT the actual contact-closed point

Why are you switching at the zero-crossing? You are supposed to do pulses starting at nearly 80~90 degrees and start from there to see how things go, also you did not pulse the transformer for some cycles, look at my sims
you might want to proto mine, I will write the good the code for you

http://www.sonsivri.to/forum/index.php?topic=68555.msg197978#msg197978
http://www.sonsivri.to/forum/index.php?topic=68555.msg197983#msg197983

It worked in principle, but I still have to fine-tune the timing to switch exactly on the negative peak.

The previous waveform of the relay switch-on may be a bit misleading; The trace is where the relay gets the on-signal, but it actually switches sometime later, as expected. If you look at the current though the relay contacts alone below, it is switching very near the peak. The level of pre-magnetizing is also important, and you need to be able to monitor the current for the correct setting.

you are right, regardless, +ve or -ve pulses, still I am willing to see you trying my approach as it has lower components count and is much simpler to be honest with you.

I always make a point to at least test any design I came up with, regardless if it will be put to use or not.  ;)

No problem, I am reluctant to changing the topic, is this oscilloscope any good MP720011? if posts become more than 4 for MP720011 I will open a new thread. This one is available in stock, I don't have to wait for a long time to get it. I will use it for SMPS designs and this project, of course.

There are about 21 pre-mag pulses before switching, that should be enough I think.

Posted on: October 14, 2020, 07:40:37 19:40 - Automerged

---

You can't complain about the price for sure. I wonder if it is some re-branding of another popular name?

Yes, the price is good, I m not sure about rebranding thing, may be other members can shed light on this..

This is about the best I can set the balance between pre-mag level and inrush-current. It doesn't take much of a variation to flip it either way (too much or too little). I think smaller transformers may be more sensitive. When I get a chance, I will test it on the 1 kVA isolation transformer I have in the lab.

How much current at relay engage moment?

It appears that with no-load, a bit more pre-mag is required than when it is under some load.

can you try to see if more pulses at no load can make any difference? also do you know the actual inrush current for the tranfo you are working with?

If I catch it on the zero crossing it peaks at about 5A, so roughly twice the peak TSR gives me. With the smaller transformers, the copper resistance of the primary winding will help to limit the peak inrush. I expect the difference to be way more with a large transformer due to the low primary resistance.

I will experiment with more pulses to see what difference it makes.

PM3295, look at this (https://www.diyaudio.com/forums/power-supplies/217526-preventing-inrush-current-saturation-toroidal-ei-transformer.html), it is interesting!
I am wondering what your response will be concerning post #98 (https://www.diyaudio.com/forums/power-supplies/217526-preventing-inrush-current-saturation-toroidal-ei-transformer-10.html#post3137736) which is the actual scheme used. I still don't understand what he means by 500 times!

the scheme is kinda strange, the delay used on each half-wave is like this: 9ms, 8ms, 7ms, 6ms, then fully ON. May be repeating each delay for 125 times? I noted he doesn't use 5ms as he mentioned it caused inrush current to occur.

It appears that he implemented fundamentally a digital controlled soft-start scheme. Before full switch-on, he slowly increases the flux swing in the core and relies on the losses in the core to balance out any residual offset. This is basically the same method used in industry as I pointed out in one of my previous posts (#15), where they use a low-voltage  variable AC source to reset the residual component in the core.

He the applies the same technique (when he wants to switch off the transformer) by slowly ramping the voltage down, before switching off near zero. This will eliminate most of the residual  magnetism offset that can cause problems with the next initial turn-on cycle of the transformer.

He then tested this up/down ramping action 500 times. At least, this is how I understand this.