I'm trying to better understand the noise that a switch-mode power supply backfeeds into its AC supply. Our company currently has a machine being tested for EMI compliance and it looks like our power supply might be creating too much noise. We've gotten graphs of the noise pattern along with the limits we need to stay below (they didn't provide a number, but the graph matches EN55022 Class B requirements for conducted emissions from 150kHz to 30MHz.) Because of distance and language barriers it's VERY difficult to get any help from the test lab, so I'm trying to learn more about this on my own.

I've looked at the noise pattern of the power supply output (5VDC) and used our oscilloscope's FFT analysis to visualize it. What I'm seeing appears to match up fairly well with the graphs, provided by the test lab, showing our test failure. I believe our power supplies are the source of the noise, and I'm interested in trying other power supplies and/or looking into filtering methods to reduce the amount of this noise that feeds back into the AC mains. Towards that end, I want to do FFT analysis on the mains side of things so that I can see what effect different power supplies or different filters have on our noise issues. This brings up several questions:

1. I'm assuming that the EN55022 Class B conducted emissions test is looking at noise conducted back into the mains (as opposed to looking at other voltages, like our internal DC bus.) Is this correct?
2. I don't have suitable probes for safely measuring mains power directly. I planned to use a transformer that steps voltage down to ~17VAC. Will this transformer limit frequency response and prevent me from seeing the noise spectrum up to 5Mhz or beyond?
3. Seeing this low level noise is easy on an otherwise DC line, because anything that's not a flat line signal is noise. I can zoom the scope in to 100mV/division or tighter and there's plenty of noise signal to see and perform FFT analysis on. However, if I scope the AC side, I'd have to use a much higher scale in order to fit the full wave amplitude on screen, at which point the tiny ripple of the noise will be impossible to see (and presumably hard to run meaningful FFT calcs on.) Is there some trick that allows you to ignore that AC mains voltage (or rather, its transformer reduced equivalent) and only visualize the noise on it? I considered running a separate transformer on another outlet, scoping it, and viewing Probe A - Probe B to subtract out the baseline 60Hz AC signal, but if the noise is feeding back into the mains, won't it also be on the second transformer, in which case I'm visualizing nothing?

I realize I may going about this all wrong. What I ultimately need is a way to see if the various experiments we try in order to reduce noise are actually working. I feel like I need to see an FFT of the noise on the AC side in order to do this, but if there's a better way to go about this, I'm open to suggestions.

Thanks!

 

You can try to get one of these: http://www.gryphon-inc.com/Spec Sheets/Power Monitoring/917010A - ONEView.pdf It also has a name of ONEAC LV103. It's just the box and cables. Everything is potted, but schematic is provided in the manual.

So, ONEAC, POWERVAR and who knows what is the new name, if any. I remember a couple of you tube videos where they were comparing their UPS's against others and even the Trip-Lite ISOBAR.

In the 80's I was convinced at a show to get one of their power conditioners for about 1000W. I ended up with three for work. In one system during an upgrade, the conditioner went to serve another system for 17 years with no hard drive failure. Just a floppy drive and dust, I bough two more at that point. It was one of these and an ISOBAR suppressor from Tripp-lite.

There's a relative recent (last few months) article in the magazine www.nutsvolts.com about compliance and some easy ways to PRE-TEST.

 

You can view the AC noise on the mains with a standard oscilloscope, by using a capacitor and a resistor in a highpass filter configuration (capacitor in series with resistor to ground).
The corner frequency should be below the lowest frequency of EMI you want to see.
If it's 150Khz then the 60Hz will be suppressed by over 60dB for a single-pole filter.
If you need more suppression of the mains frequency you can go to a higher order filter.

Warning: You must have an isolation transformer between the mains and the unit under test so that the scope common and chassis won't be at a lethal voltage, and you won't be vaporizing a ground lead.

 

It is an error on your part to imagine that there is a single mechanism for noise to show up on the AC line. Emissions can be radiated and long power wires make pretty good antennas in the low frequency ranges. As far as conducted emissions, any conductor at all is a possible path for those emissions. Normally, your transformer will attenuate noise above a certain frequency. You can also include a common mode choke on the line input to your transformer. Common mode chokes are bi-directional and prevent line transients from making into your system as well.

 

As Pb said, a common-mode choke can help reduce conducted EMI.
You also can add typical EMI LC line filters.

 

As Pb said, a common-mode choke can help reduce conducted EMI.
You also can add typical EMI LC line filters.

I guess my new nickname could be "Lead"

 

Thanks for all the help! I'm going to attempt to use a transformer and RC filter tomorrow and see if it looks like I'm still seeing the same noise pattern. Hopefully the transformer doesn't prevent me from seeing the high frequencies I'm trying to see. I've simulated a filter setup in LTSpice which seems to make sense to me. I don't remember for sure what transformers we have at work, but I know they're 240VAC mains rated on the input, and I think they're 20:1. If not, something in that general range. It looks like this filter setup will knock out enough of the main voltage to see the noise reasonably well.

 

Ha! Just noticed that my AC voltage in the sim is double what it should be. I was using 350p-p as a ballpark equivalent for 250rms, but accidentally ran the sim with +/-350 relative to ground (700p-p) instead. Shouldn't change the filter behavior or the basic concept, but a silly mistake nonetheless.

Anyway, if you guys see any major problems with this approach, let me know. Otherwise I'll give it a try tomorrow morning. Once I find a way to get some baseline noise measurements, I can start trying filter and choke ideas, or alternate power supplies, to see how much noise reduction we can achieve.

 

I'm afraid the transformer will filter out a lot of the noise, but you can give it a shot.

See if you can find a 1:1 mains isolation transformer for doing the test.

 

I'm afraid the transformer will filter out a lot of the noise, but you can give it a shot.

See if you can find a 1:1 mains isolation transformer for doing the test.

The LTSpice perfect transformer passed the highs just fine, so I'm sure there won't be any difference in real life!

In all seriousness, I was afraid of that, especially after Pb (Lead?) mentioned the possibility. Still, I might as well try it, cause I don't have anything else to work with yet and it should be pretty quick and easy to try. I'm sure we don't have any 1:1 isos, unless they're built into something I can hack apart. Might try to order one in if I have no luck with the 20:1.

Perhaps this question should get its own thread, but here goes: Given that we have a reasonably limited, predictable range of input voltages (200-250VAC) and that efficiency doesn't matter at all (on a 10kW machine, who cares if the 10W control system is burning a few extra Watts,) is there any reason not to use a step down transformer, rectifier, and linear regulator instead of the current switch mode supply? Are there other advantages to switching supplies? I understand why they're chosen when efficiency or wide input voltages are required, but I don't see the advantage in our case.

 

Are there other advantages to switching supplies?

The main reasons for using switching supplies is to improve efficiency and eliminate any heat sink that a linear supply may require.
I know of no other reason.

 

The only other reason I can think of is cost. With the escalating price of copper a conventional tranny may well be more expensive than a complete SMPS. But I doubt the price difference would be significant for a 10kW machine.

 

Well, you guys were right about that transformer filtering the high frequency noise. Nothing useful to see on the scope that way. If only the EMI of our machine was as clean as the output of that transformer, we'd be golden! Would a 1:1 isolation transformer be likely to have more suitable frequency response? If so, anything special I should be looking for? Is toroidal better or worse than other designs for passing high frequencies? There's a fairly inexpensive toroidal I'd be tempted to try if it's suitable.

The other question would be CATIII probes - if I get a probe that's CATIII rated for the voltage in question, is it ok to directly probe the mains AC then? (I would assume yes, otherwise what else does that rating mean?!) It seems everywhere I look, people are quite adamant about not scoping mains voltage directly - is that just because very few people have the appropriate probes, or are there other reasons it's a really bad idea?

And thanks for the responses on linear supplies. It's pretty tempting to try one, but if I can't find a good way to scope the noise, I won't know whether or not we're doing any better than before! I guess if we can't figure out how to see what we're doing on this end, we may just be better off with an existing noise filter solution, maybe something like this:
http://www.mouser.com/ProductDetail...=sGAEpiMZZMt5zLV74dRd3W38HTEVH9xrt3aL3J/V0lk=

 

Would a 1:1 isolation transformer be likely to have more suitable frequency response? If so, anything special I should be looking for? Is toroidal better or worse than other designs for passing high frequencies? There's a fairly inexpensive toroidal I'd be tempted to try if it's suitable.

You measure the voltage between the isolation output and the equipment input, so the transformer is not in the signal path.

That transformer is only 25VA maximum.
Is that sufficient to power your equipment?

If you have two identical power transformers of sufficient VA rating you can connect them back-to-back to make an isolation transformer.

people are quite adamant about not scoping mains voltage directly

That's because of safety concerns since the mains are connected to earth ground and measuring the mains directly exposes you to lethal voltages to ground.
And if you don't do it properly it can also cause a short between the mains and the oscilloscope, since the oscilloscope common is connected to earth ground through the third plug safety ground.
That's the purpose of the isolation transformer, to isolate the circuit from ground.
The voltages can still be dangerous, but now there's no voltage between the isolated output and earth ground.

 

You measure the voltage between the isolation output and the equipment input, so the transformer is not in the signal path.

That transformer is only 25VA maximum.
Is that sufficient to power your equipment?

If you have two identical power transformers of sufficient VA rating you can connect them back-to-back to make an isolation transformer.

That's because of safety concerns since the mains are connected to earth ground and measuring the mains directly exposes you to lethal voltages to ground.
And if you don't do it properly it can also cause a short between the mains and the oscilloscope, since the oscilloscope common is connected to earth ground through the third plug safety ground.
That's the purpose of the isolation transformer, to isolate the circuit from ground.
The voltages can still be dangerous, but now there's no voltage between the isolated output and earth ground.

Doh! I'm embarrassed now. I was completely misunderstanding the function of the isolation transformer. I was picturing it between the machine and the scope. I'm all caught up now. Thanks for your clear explanation and your patience!

 

No, I spoke too soon. I thought I had it, but I'm still missing something.

If I've understood the last post correctly, the point of the isolation transformer is to make it such that the voltage in the device under test is no longer ground-referenced, and therefore make it *less* dangerous, albeit not totally safe.

Since the voltage in the DUT is now floating, I can't clip my probe's ground clip to earth ground and get meaningful voltage readings off of the DUT, right? I've read several descriptions online of clipping the ground clip of the probe to either the neutral or one of the hot legs (assuming split-phase or multi-phase system) in the DUT as a reference point. This makes sense in terms of getting meaningful readings from the scope, but that would then provide a ground reference for the voltage in the DUT, and it would no longer be floating, right? So the DUT is potentially safer before you hook up your probes to it, but once you do that, you're in the same ground-referenced, dangerous boat as before. It seems the only real advantage of this is that you run very little risk of frying your scope, since you get to define what the ground reference is - you no longer run the risk of clipping the ground clip to the wrong thing and burning your scope up, but you still have non-isolated voltage in the DUT once you make that ground connection, right?

I sincerely hope I'm missing something, because I was pretty sure I understood the intended hookup scheme, but it doesn't really seem safer to me (at least not for the humans involved - maybe it's a little safer for the scope!)

 

It seems the only real advantage of this is that you run very little risk of frying your scope, since you get to define what the ground reference is - you no longer run the risk of clipping the ground clip to the wrong thing and burning your scope up, but you still have non-isolated voltage in the DUT once you make that ground connection, right?

Right.
I've heard several times of someone connecting the probe ground lead to the wrong point on a main's connected circuit and vaporizing the probe ground lead.
With isolation, you can connect the ground lead to any point in the circuit and not generate any current to earth ground.
Of course lethal voltages are still there, so you have to be cautious of that.
So you want to connect the probe ground only when doing a measurement, and then immediately remove the ground lead after you are done.

 

Right.
I've heard several times of someone connecting the probe ground lead to the wrong point on a main's connected circuit and vaporizing the probe ground lead.
With isolation, you can connect the ground lead to any point in the circuit and not generate any current to earth ground.
Of course lethal voltages are still there, so you have to be cautious of that.
So you want to connect the probe ground only when doing a measurement, and then immediately remove the ground lead after you are done.

Got it! Thanks again for all your help. I still don't really know what we're going to do, but at least I understand our options now. This forum really is wonderful! I always leave just a little more knowledgeable than when I showed up.