Arrangement of X, Y Caps and Power Line Switch

[EHT]

Hi. I have a pretty basic question about how to arrange X & Y caps and a line switch. I want to add these two some equipment i have which has linear PSUs with fairly large transformers. These cause a large spike which interferes with other equipment when turned on/off, particularly audio amps and PC.

I'd like to know if it is best to add the X & Y caps and MOV to the incoming 240V AC line *before* the power switch, or between the power switch and transformer primary, or perhaps it doesn't  matter? Further, for the purpose of supressing this on/off switch spike, would it help to have a filter with a choke and/or to add a snubber cap across the switch?

I can see that often a line filter with X & Y caps is added before the power switch but this is often just because this filter is integral to an IEC socket. I'm adding filters to existing equipment where I can't fit an integral filter/socket.

Putting the caps before the switch obviously has the downside that the caps are powered up even when the equipment is off, but most designs I can see have it this way round.

FYI, the equipment I want to put this on is a TTi Linear PSU and my Hakko 951. Neither have any suppression. Thanks!

[Benta]

Hi. I have a pretty basic question about how to arrange X & Y caps and a line switch. I want to add these two some equipment i have which has linear PSUs with fairly large transformers. These cause a large spike which interferes with other equipment when turned on/off, particularly audio amps and PC.

Neither X nor Y caps will help you here. The correct solution is a soft-start for your power supples.

[EHT]

Thanks. In practice do you mean switching power to the transformer with a Solid State Relay that has a zero-crossing operation?

I can explain what I observed; I realised I have two pieces of equipment that both cause huge interference spikes when turned on/off and lots of other equipment also containing large linear PSUs that do not do this. I found the difference is these two items causing the interference have no mains filter components at all whereas the other equipment does. For example I have an HP PSU with a very large mains transformer; its also got a line filter and a big X cap. This doesn't cause interference when the mains is switched. I thought the big X-cap was the reason for this. I wondered if the X-cap needs to be on the live-side of the switch to have this effect.

[Benta]

Check out this thread: https://www.eevblog.com/forum/beginners/simple-amplifier-soft-start-for-inrush-protection/

It's a few days old, but op-topic.
You have two problems, which usually occur when your transformers are larger than ~300 VA.

Switching on the transformer at zero voltage results in a major current spike because the transformer core saturates. So ideally, you'd switch it on at peak mains voltage.

But this leads to a different problem: if you switch it on at peak mains voltage, the capacitors on the secondary side will draw an immense amount of current instead.

You're caught between a rock and a hard place.

That's why I suggested a soft-start circuit instead. The X/Y caps won't help one single little bit here.

[tggzzz]

You could consider the simple cheap old trick, which sometimes works: get a ferrite toroid a couple of inches in diameter, and wind the mains cable through it as many times as possible.

You could also consider inserting a "power line filter" with decent common mode and differential mode attenuation. Don't reuse an old one, since the RIFAs inside them will start to generate smoke

[EHT]

Thanks. I suppose I'll install a standard filter as per the other equipment and see. AFAIK equipment with linear PSU didn't _need_ to have EMI filtering but interesting to note that the high quality equipment like the HP PSUs do have it anyway.

[EHT]

You could also consider inserting a "power line filter" with decent common mode and differential mode attenuation. Don't reuse an old one, since the RIFAs inside them will start to generate smoke

So, FWIW, I added a new 6A rated line filter, a MOV and an extra 220n X-Cap in the TTi PSU. There are 3 E-I core transformers in this unit, I guess 100VA each or so. There was plenty of space to bolt the filter to the chassis. I've put these on the hot side of the switch since this seems to be the common thing. This made a lot of difference; I can't notice any interference on other equipment now!! I didn't bother trying to measure the pulse since this did the job so well.

Next, i'll do similar to the Hakko station. It's just annoying having these devices create big spikes!

[T3sl4co1l]

The first / most likely problem is the extreme risetime caused by the switch.

Turn-on isn't so easy to address, as you need a series R || L element to dampen that.  It's certainly doable, it's just not as often employed.  Note that simple ferrite beads won't do the trick; while they have the right characteristic (lossy inductance), they don't have enough impedance, or lose it quickly under load current (saturation).  Hm, maybe a one of those multi-hole wound beads, at least up to a few amperes -- they tend to perform better under bias, and can have fairly high impedances.  Otherwise, use a proper inductor, say 10uH, rated for maximum load RMS current; a type wound on a cheap powdered iron toroid will do.  The parallel resistor should be, say, 470 ohms or so (1/2W rated).

Again, not so commonly used, or needed, really, but if you find it is needed, that's the way to go about it.  If combining with a capacitor snubber, place the RL snubber in series with the switch alone, and wire the RC snubber outside of it.

Turn-off is usually the big offender, as the immediate load (power transformer) is very inductive at high frequencies (i.e., at fast edge rates), so can develop quite a high voltage, briefly; a voltage which causes the switch to spark, multiple times, within a few to tens of microseconds.  Each spark, discharges the stray wiring capacitance into the mains -- a miniature EMP, transmitted to anything connected or nearby.  This very likely either couples into audio equipment directly, or by way of RF rectification (for which, the equipment should have adequate input filtering to reject this -- but often doesn't).  This is easily dealt with by an R+C across the switch, say 100R + 10nF for a small inductive load (okay for a small transformer), up to 0.47uF or more for more inductive things.

Note that, because the switch is the fast-risetime element, the filtering needs to be done with respect to it.  Inductive kick is often blamed on the load (e.g., a motor, solenoid, transformer, etc.), but it's just doing its thing until interrupted by the switch; the inductance is an enabler, but the switch is ultimately the culprit.

If the filter might pass too much off-state leakage current (a very average snubber consists of 100R + 0.47uF, which leaks a whole 35mA at 240V, 50Hz), it can be rearranged to include neutral.  In this case, wire the snubber across the load side of the switch (hot to neutral), and bypass the source side of the switch to neutral with an X1/X2 type capacitor (of similar value to the C in the snubber).  These components should be placed near the switch; the lead length of the neutral wire isn't very important.

As for inrush, an NTC thermistor and optional bypass relay is an option, or more involved active circuitry.  Such as shown in the other thread.

Ed: and as for the line filter -- it's not quite the right tool for the job, but it does provide filtering to the mains side, so even if the switch is making noise, less is released into the surroundings.  And it provides some line-to-line capacitance, which acts to bypass (shunt) the noise near the source; the switch and load might still be making the noise, but it'll be less to begin with, plus the filtering value.  Win-win.  It may be that the noise must be dealt with directly, still requiring the snubber -- but it may also be fine like this.

Finally -- if it's good enough for the job, I mean, there you go.   A snubber is the most efficient (direct) plan of attack, but not the only possibility, and whatever it is, if it works, it works.  For example, it might still be emitting the noise, but below the threshold your other equipment picks it up at -- often it takes more than a few volts of noise to be detected in this way, so it only needs to be that much.  (Or it might only take some ~mV, because audio can be very sensitive; depends.)  Keep this in mind if you ever change your configuration (adding/removing equipment in the system) -- you never know when you're going to get something more or less sensitive, and the problem may seem to recur (when actually it never really went away, it just wasn't noticeable before).

Oh, one more thing about snubbers: since they suppress sparking, they also help to extend contact life -- hardly a concern with a power switch you might throw thousands of times in its lifetime, but good to know when working with relays, like for, Idunno, temp control or something like that -- anything that's switching every few minutes, or seconds, times many years, can rack up millions of contact closures!

Tim

[EHT]

Thanks very much for all the info Tim!

With the post, I was interested in understanding the function of the filtering components as well as fixing this problem. In the end, what I've done here is just copy the arrangement from the HP (linear) PSUs and other equipment with similar filter components. For clarity this is as per the first diagram attached.

My understanding is the MOV is to protect the equipment from external power spikes, I suppose not often operating but important for protection. It sounds like good practice in all equipment to also have the filter and X-caps to attenuate any high frequency noise from the equipment back to the line, and vice-versa. As you say, the side-effect is some filtering to dampen the pulse from the line switch. Maybe this is why the filters seem to always be placed on the line side?

I can see the snubber is the best way to directly tackle the problem though, so as per the 2nd pic. Maybe I'll try this on the soldering station. Much smaller transformer of course.

If the filter might pass too much off-state leakage current (a very average snubber consists of 100R + 0.47uF, which leaks a whole 35mA at 240V, 50Hz), it can be rearranged to include neutral.  In this case, wire the snubber across the load side of the switch (hot to neutral), and bypass the source side of the switch to neutral with an X1/X2 type capacitor (of similar value to the C in the snubber).

I wasn't quite sure what you meant - as per the 3rd pic?

[T3sl4co1l]

You got it!

Also yes, you need a snubber for each contact, if they're ganged, as you don't know in general which one opens first.  The second one opening, some ~ms later say, sees no current, so doesn't matter -- if you could know, that is.

Tim

[SeanB]

Most common method used to reduce inrush current is to use a NTC thermistor in the power line. Generally sized for the running current, but I have found that a common value for up to 2A input current is a 4R7 unit, in series with the power line. Also add in a MOV after the NTC, so that turn off spikes are clamped on the transformer, and the series resistance keeps the MOV from being damaged by mains transients, so it protects the transformer from spikes.

Incoming mains, fuse in live line, then double pole power switch. Then after switch your class X and Y capacitors, then the NTC, then the MOV, and finally the transformer primary. Provides good protection, and isolates the incoming power totally, so there is almost no risk of a faulty component causing issue, as the fuse is first to blow. Only issue is in some EU countries, where they want fusing on both line and neutral, while others forbid it, so a single fuse on line is about the best compromise. You can add the mains common mode filter instead of the X and Y capacitors, as all you are doing is putting in one component to replace 3, and getting a CM choke in the can, but for linear supplies not needed, though the X and Y capacitors plus a common mode choke probably work out cheaper.

[EHT]

Thanks for all the advice & sorry the delay giving feedback! I looked into this further; I hadn't heard of NTC thermistor inrush protectors before.

I experimented looking at the mains using a scopemeter (with insulated jack adaptor lead to the 13A plug). The screen is not great resolution of course, but I could see the interference from the Hakko, only when turning off. I can't see anything when switching the PSU that I've already added the mains filter to on/off. What I observe with the Hakko is not one big spike, but a burst lasting a few ms, so I expect this is arcing over the switch contacts as Tim explained. So a snubber should sort that out.

On the linear PSU, although there appears to be no line noise from it, I can just about hear some arcing in the main switch contact, so I'll add a snubber to increase its life. I suppose this alone would have sorted it out and been cheaper & easier to add than the filter.

Also, I can hear a thump/buzz when its turned on. The thump seems to be magnetic impact to the steel outer case, but it leads me to think adding an NTC would be "good". There is a downside with these that they dissipate heat. I can see what I think is one in an SMPS i have and that has been enclosed with some clear heatshrink. Maybe to protect in case it burns up? (might be a MOV). I've looked at schematics for other PSUs and I can see NTCs in the switch-mode ones (as you said Sean) but not in the Linear ones. Maybe that is just because of the age of the linear ones.

I'll add these components next time I put an order in and see how it goes.

[T3sl4co1l]

Thump is usually a transformer charging up due to random line switching.  Namely, as an inductive component, at idle it has a phase-shifted current, zero when the voltage is peak and max when voltage is zero.  Or instantaneously speaking, current rate-of-rise (dI/dt) is proportional to voltage.  So as long as voltage is positive, current is rising, and then it's falling when it flips negative.  Divide the voltage waveform into four quadrants: rising positive, falling positive, falling negative, rising negative.  Effectively, steady-state current ramps up during the middle pair (falling and falling), then up during the rest (rising and rising); the peak current corresponds to the area under only half a hump, as it cycles around zero.

When you switch on near a zero crossing, current starts at zero, but then goes up during both the rising and falling quarters of a full voltage hump. This would draw merely twice the peak magnetizing current, in a linear system, but we can't afford linear transformers; we use big iron cores to handle mains frequency, and avoiding this is literally double the cost for hardly any advantage*.  So we run out of capacity, and the core saturates (inductance drops sharply at high currents).  So, somewhere past midway through the voltage hump, current ramps up much faster, and ultimately causes more voltage drop across the primary's winding resistance (and anything else supplying it, if applicable).  Note that inductance drops faster than dI/dt rises, so winding voltage (the ideal magnetic EMF part) is forced towards zero.  The inrush current persists until voltage reverses somewhat; note this means there's now an initial current after zero crossing, so normal operation resumes, more or less.  (In practice, primary resistance and inductance distribute inrush over up to several cycles, and the waveform looks more like a sine wave, riding on top of an initial DC offset that decays at some L/R time constant, with some extra spiking at startup.  Most "shell" type or split-bobbin transformers behave this way, while toroidal transformers have lower stray inductance and tend to do one big gulp.)

*Core loss and noise are lower too, so there is some advantage, but still, only in applications that are highly insensitive to price.

This is all well and good, but something still needs to move, to be able to hear it -- there are two effects here.  One, current draw means magnetic fields around wires, and those wires can shake.  We're not talking huge currents here, so the power cable itself isn't going to move, but the windings on the transformer perhaps can.  (They're usually varnished in place to minimize movement, but maybe there's enough flex to hear it?)  Two, the core itself physically stretches, due to an effect called magnetostriction: under applied magnetic field, the core expands or contracts parallel to the field, and vice-versa perpendicular to it.  This is worse at high fields, and especially worse at saturation.  So you might hear merely a quiet humming in steady operation, but you can definitely get an audible bump or kick from inrush.  (Or you get an angry humming all the time, if the transformer is designed marginally in the first place: for example, microwave oven transformers are notoriously cost-reduced and peek into saturation even in steady state!)

Can also be some external fields around the transformer, which vibrate steel panels, etc.; it's usually very little field, so you need quite an unlucky combination of loose panels, plus a microscopic gap for them to rattle around in, so that you get not just a humming sound, but every cycle or a few, the panels slap against each other for an irritating buzz.

As for mitigation, yeah, an inrush NTC works well here.  If it's a small transformer, I wouldn't bother, its own resistance will dominate.  Bigger ones, or medium toroids and up, are a good application of it.  Value doesn't really matter; higher initial (cold) resistance means lower inrush current, but higher idle losses.  Obey the maximum RMS load current rating.  A 10mm dia. part is probably fine for low 100s VA loads.

Tim

[EHT]

Thanks very much Tim!
Hopefully others trying to reduce electrical and mechanical noise will also find this useful.

The thump is indeed from the steel case. With this removed you hear a loud buzz at power-on, then a fairly quiet hum. You can even feel the vibrations on the case. I was really just thinking that the thump would indicate a high inrush that would be best to dampen down.

There are 3 E-I core transformers, about 150VA each. One has a full frame, the other two only had bolts at the bottom (see pics). I added those at the top. The one showing some rust has the laminations coming apart a bit. I guess it would be quieter if i could find a frame piece to bolt on to the top, but not a big deal. Not sure how much residual hum is the core, windings, or the chassis.

The transformers do also cause a bit of bother to nearby CRT scope. I might experiment with a bit of mu-metal there.. or just move it away!

In one of the pics you can also see the metal-cased filter bolted to the chassis and the MOV & X cap on flying leads.

[SeanB]

Do not bother with shielding for magnetic fields, simply move it further away.