Large Noise from Bridge Rectifier; Conducted Emission Measurement

...

...

Hi Klaus

That is a common problem:

How to Eliminate Diode Noise:

Bypass capacitors across rectifier diodes:

Noise from solid state rectifier in tube amp:

Quote: "... Diodes can be noisy when used as rectifiers due to reverse recovery induced spikes. Certain types of diodes perform better (schottky and so-called 'soft-recovery' diodes) or are essentially immune (silicon carbide) to this effect.

A well-designed circuit can address reverse recovery via the use of appropriate diodes (i.e schottkys and/or silicon carbide where appropriate) and by snubber networks, which absorb and dissipate the spike energy. ..."

Quote: "... They are there to shunt switching noise. If the capacitors are of the wrong type, they can create more noise than they attempt to remove. The size of the cap vs the expected 'load' is key to their effectiveness. ..."

Glenn

Might be the turn-on overshoot. Can you measure the voltage across one of the rectifiers?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Resonance due to under-loaded/damped EMC filter, high current lossy ferrite bead/beed/bede needed.

HI Glenn

Thanks for the links, but as I wrote that was the first thing I tried. Both purely capacitive snubbers and RC snubbers. There was no change in emissions

One suspicion I have is interaction of the diode bridge with the EMC filter, causing undamped ringing. But it's hard to measure, when we are talking 550V DC, to catch a small signal which the differential probe cannot resolve.

One idea I had was to drive the bridge rectifier with the same current, at low voltage, so it's possible to measure

Cheers

Klaus

Yes, I can measure the voltage. SO you think the diodes are slow to turn on and that the glitch is what we see?

I have measured the diode voltage, and it looks fine, except for a small un dershoot during turn-off (reverse recovery, but I could not see any change when adding snubbers)

Bytheway, the distance between the quasi peak and average is normally a fix ed ratio, say 10dB.

But in the event of periodic larger noise, the quasi peak get higher. Like we see in the plot. Sometimes that is from saturation of the common mode in ductor, but we have confirmed this is not the case, since we are running ve ry low currents and we have no CM noise (no CM saturation of the CM inducto r)

Cheers

Klaus

y" looking emission graph

er->DC Inductor->DC Link capacitor->6 switch Inverter->Motor

and bridge rectifier.

h a 2-stage EMC filter and a bridge rectifier with a capacitor on the outpu t and a resistive load)

noise also comes from a standard single phase rectifier

ry of the diodes, but the noise increases a less load (< 200W), and I would think the reverse recovery current glitch would be worse at higher current s running in the bridge rectifier

te bead/beed/bede needed.

Yes, we have thought about that. We then tried with a single stage EMC filt er, and the problem was almost exactly the same. I could try to add resista nce in the X and Y capacitors, just to rule it out. Thanks for the reminder :-)

Cheers

Klaus

I tried with 1nF and 500ohm for a 1kW drive. Any hints to the calculation of the size, just from critical damping of the reverse recovery pulse, or should the X capacitors in front of the 6 phase rectifier be included in the calculation?

Cheers

Klaus

The conduction angle shrinks as the load is reduced, so the peak current will occur over a smaller time at lower load. I've never seen where the peak current is actually higher at light load than heavy, but there might be conditions where this could actually happen, when the conduction angle shrinks to near zero width.

Jon

Make a pickup coil and hang it near the area, and also display current from a current probe, and sync to that. Even syncing to a voltage probe would probably be good enough. Then, correlate the pickup coil's signal to the waveform to understand what part of the rectifier's cycle is producing the output. it will almost certainly be pulsed at mains frequency.

Jon

Could still be reverse recovery spikes. Can you poke around with an oscilloscope?

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Aside from using current transformers or a current probe, I'd set a communications receiver to 150kHz, don the headphones, connect an EMCO I-probe (a smaller one) and sniff around. The human ear is much more sensitive and faster than even a fancy scope or analyzer.

Of course, every time I do that at a new client some people think that I am off my rocker ...

--
Regards, Joerg 

http://www.analogconsultants.com/

I all ready did that.

I used the Rohde und Schwarz test receiver in manual mode at 150kHz, listen ed to the speaker, even took out the signal to a scope. That's how we found out it was related to the bridge in the first place, nice 300Hz signal the re, but difficult to see anything else that is was mains related

Another nice trick is to extend the normal range (150kHz and up), by runnin g at say 10kHz and up and at 200Hz bandwidth, that reveals if the noise is broadband and the individual spikes comes out clearly

Cheers

Klaus

Interesting gets worse with lighter loads.

Where do I see what's inside your emi filter?

What's the physical location of all these parts?

What LISN are you using?

Who manufactured your 1N4007's?

You seem like you've been through a lot of these battles and really are caught off guard here. Sympathies. At least you're not at the Test Lab at

2am trying to sort this out.

Most SA's have a line sync option that let's you find a bit more information about the spikes and AC mains without having to delve into narrow bandwidth's, which also works.

The loop and finding out EXACTLY where the spike occurs sounds useful.

What I often do is pick one of the nastier spikes in the conducted emissions plot, dial the receiver onto it and then turn on a 300Hz crystal filter. By 600Hz it's rolled off 60dB or so, so that allows me to really zero in on stuff and sniff with smaller magnetic field probes. Otherwise the din of many other sources can muffle the signal you are hunting.

My hunch is that you'll probably need a better EMC filter, assuming the grounding at the EMC filter is ok.

--
Regards, Joerg 

http://www.analogconsultants.com/

But there's nothing like an oscilloscope to tell you WHY things are happening. Zoom in on the falling edge of the transformer secondary, just past the AC peak when the diodes are coming out of conduction; that can be horrific.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Sure, in this case it probably works. Maybe. But there were cases where folks had the whole arsenal on the bench. Scopes, analyzers, all five-digit Dollar gear, and nothing could be seen. My first case was in the late 80's when they were sure the noise was coming out of one of the signal processing boards. It had to. It just had to. Out came the receiver and head phones. Turned out it leaked from board at the other end of the system, a RAM bank on a video board. Everybody sat there in disbelief but the fix to prove it took only minutes.

--
Regards, Joerg 

http://www.analogconsultants.com/

looking emission graph

->DC Inductor->DC Link capacitor->6 switch Inverter->Motor

d bridge rectifier.

a 2-stage EMC filter and a bridge rectifier with a capacitor on the output and a resistive load)

ise also comes from a standard single phase rectifier

of the diodes, but the noise increases a less load (< 200W), and I would t hink the reverse recovery current glitch would be worse at higher currents running in the bridge rectifier

No insight into the cause, but the remedy usually turns out to be working o ut the path around which the high-frequency current circulates, and finding a way of making the area included within that loop smaller, and the path i tself shorter. It can help if you can twist the path, so that adjacent patc hes of the included area are radiating in opposite senses.

Ralph Morrison's book is helpful in getting you to think about the relevant physics

p/B000WCVEN8

I've got a copy of the 4th edition (which is what Amazon is selling), which has been up-dated to "the digital age" but the good stuff dates from the f irst edition, which I first read back in about 1967.

--
Bill Sloman, Sydney

are

That's a good tip. I will remember it.

?-)

That's because the conduction angle get smaller

I cannot disclose that, sorry

On a big 4 layer PCB board

Standard 3 phase LISN as per CISPR something something

We are not using the 1N4007, that's just to show what we are talking about. The diodes are integrated into a custom Fairchild power module

Tonight it was 10 PM :-)

We have full fledges internal lab, really cool equipment

I found the cause, it was the reverse recovery of the diodes. Putting a monster 1uF cap over each diode solved the problem. Now I will se how to pull back to smaller size and with resistors

I also solved a resonance in the filter that aggregated the problem

Cheers

Klaus