Power transformer snubber 2

[RyreInc]

Greetings,

I wish to switch a power transformer on the primary side using a reed relay, but am worried about voltage transients during switch-off. The XFMR will draw 15A RMS @ 60Hz when on. Is a snubber necessary to protect the relay's 300V maximum contact voltage? If so, is an RC or zener snubber (or other) best? Or, should I consider a zero-crossing SSR?

The relay is a Omron Electronics G2RL-1A-E DC12.

Thanks!
Ryan

 

[LionelHutz]

Also, if you are custom building a special PCB for this project then you should look at using a triac (~30A rated) and some heatsink and then add a driver circuit. Check these little drivers out.

http://www.fairchildsemi.com/pf/MO/MOC3021M.html

If you don't like that, then google search triac driver circuits and you'll find lots of possibilities.

 

[ScottyUK]

Gate firing becomes more demanding with highly inductive loads. This is one of the problems with many SSRs designed for resistive loads when applied to inductive loads. Keep away from triacs for anything inductive, it opens up a world of commutation problems: anti-parallel thyristors are a better solution.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[RyreInc]

I think I've ruled out 90 deg. turn-on/off since there are few/expensive options available (although this would be an ideal solution). Zero-crossing ruled out as well.

So, at this point I'm looking at either random-fire SSR or mechanical relay. SSR for no moving parts; mechanical for lower power dissipation and smaller formfactor, as well as less trouble with inductive loads. If I go with an SSR I will probably have to offload it from the PCB since I'm working with a fixed size and running out of room.

Right now I'm not seeing a way to limit in-rush current. Closest possibility is a thermistor, but this wouldn't perform predictably under repeated and variable cycles. With in-rush current a reality, how much should the relay be over-rated for a long life? Will the Panasonic ALFG2PF12 cut it?

Either way I'll add a MOV or RC snubber--no reason not to for 50 cents. I like the MOV better, any reason to go with RC instead?

Thanks again everyone, I'm learning quite a bit!

 

[Skogsgurra]

An RC snubber works two ways, it may reduce inrush also. The reason is that inrush not only depends on where on the sinewave you switch on. It also depends on what state you left the core in on last switch-off.

If you left the core with (for instance) +20 percent flux and switch on (randomly) at the positive zero crossing, you will get very high inrush.

You can use the snubber to demagnetize the core if you make the snubber so lossless (low R and quite large capacitor) that the output swings a few cycles as flux is taken down at switch-off. You will then switch on with near zero remanent flux and your inrush will never be extremely high - just high. And that is a lot better than extremely high.

I feel a measurement coming. Check in an hour or so. There might be a recording for you.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[RyreInc]

Ah, that makes perfect sense. I had even seen this oscillation in my SPICE simulation, but hadn't put two and two together. In this case I would want the snubber parallel to the transformer, not the relay, since I don't want leakage current, right?

 

[Skogsgurra]

Yes! That is exactly what you saw. Yes! Parallel to the transformer.

Did you check different R and C? Perhaps you put your results here. So I don't have to.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[RyreInc]

Well I won't discuss the boring details, but ultimately R and C are determined by a few simple equations.

First I measured open circuit RMS current (Im) on the primary. From there,

C=2*Im*V/(Vp^2*w), where Vp is desired maximum transient voltage and w=2*pi*f and V is operating RMS voltage.

To determine R, I played around with a few equations to compromise between power dissipation, relaxation time, and number of oscillations. Relaxation time is approximated by 4 times the time constant (alpha) and number of oscillations is approximated by the inverse of the damping ratio (delta),

alpha=2*R/Lm, Lm=V/(I*w)

delta=alpha/w0, where w0=1/sqrt(L*C)

The above determined Lm=0.6H, C=4.7uF, R=22ohm. This will limit transients to 270V, oscillate about 30 times in 0.25 seconds, and dissipate about 1W.

 

[Skogsgurra]

Does that minimize inrush, compared to no snubber at all? (I hope it does).

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[RyreInc]

In theory...

Everything's still on paper/computer at this point, and I don't have enough information about the transformer to model hysteretic effects in SPICE.

I'll post an update when I have that information.