Large Noise from Bridge Rectifier; Conducted Emission Measurement

[Klaus Kragelund]

Hi

I am working on a high power frequency inverter (>1kW) and have a "funny" looking emission graph


The converter is the typical inverter type, EMC filter->3 Phase Rectifier->DC Inductor->DC Link capacitor->6 switch Inverter->Motor

Like this:

www.electronicsdesign.dk/tmp/Inverter.jpeg

We stripped down the circuits and isolated the noise to the EMC filter and bridge rectifier.

The problem is the very high conducted emission noise at 150kHz:

www.electronicsdesign.dk/tmp/NoisefromRectifier.pdf

(Conducted emission plot narrowed in from 150kHz to 2MHz, conducted with a 2-stage EMC filter and a bridge rectifier with a capacitor on the output and a resistive load)

www.electronicsdesign.dk/tmp/PowerModel_V0.pdf

The bridge rectifier is a part of an intregrated power module, but that noise also comes from a standard single phase rectifier

First thought would be that the noise is coming from the reverse recovery 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 currents running in the bridge rectifier

I have tried with snubbers and that did not help.

Any insights into the cause and the remedy?

Thanks

Klaus

[Glenn]

On 02/07/14 18.22, Klaus Kragelund wrote:
> Hi
...

>
> We stripped down the circuits and isolated the noise to the EMC filter and bridge rectifier.
>
> The problem is the very high conducted emission noise at 150kHz:
>
> www.electronicsdesign.dk/tmp/NoisefromRectifier.pdf
>
> (Conducted emission plot narrowed in from 150kHz to 2MHz, conducted with a 2-stage EMC filter and a bridge rectifier with a capacitor on the output and a resistive load)

...

> I have tried with snubbers and that did not help.
>
> Any insights into the cause and the remedy?
>
> Thanks
>
> Klaus

Hi Klaus

That is a common problem:

How to Eliminate Diode Noise:
http://www.ehow.com/how_7970048_eliminate-diode-noise.html

Bypass capacitors across rectifier diodes:
http://www.audiokarma.org/forums/archive/index.php/t-469495.html

Noise from solid state rectifier in tube amp:
http://electronics.stackexchange.com/questions/78813/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.
..."

http://www.valvewizard.co.uk/bridge.html

http://www.diyaudio.com/forums/solid-state/134243-what-does-01-cap-bridge-rectifiers-do.html
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.
..."

http://www.electro-tech-online.com/threads/capacitor-bypass-of-diodes-in-bridge-rectifiers.40733/

Glenn

[Phil Hobbs]

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

[bloggs.fred...@gmail.com]

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

>
>
>
> Thanks
>
>
>
> Klaus

[Klaus Kragelund]

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

[Klaus Kragelund]

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 undershoot 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 fixed 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 inductor, but we have confirmed this is not the case, since we are running very low currents and we have no CM noise (no CM saturation of the CM inductor)


Cheers

Klaus

[Klaus Kragelund]

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

Cheers

Klaus

[Klaus Kragelund]

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

[Jon Elson]

Klaus Kragelund wrote:


> First thought would be that the noise is coming from the reverse recovery
> 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
> currents running in the bridge rectifier

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

[Jon Elson]

Klaus Kragelund wrote:


> 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.

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

[John Larkin]

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

[Joerg]

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/

[Klaus Kragelund]

I all ready did that.

I used the Rohde und Schwarz test receiver in manual mode at 150kHz, listened 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 there, but difficult to see anything else that is was mains related

Another nice trick is to extend the normal range (150kHz and up), by running 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

[RobertMacy]

On Wed, 02 Jul 2014 09:22:20 -0700, Klaus Kragelund
<klau...@hotmail.com> wrote:

>> ...snip....

> Any insights into the cause and the remedy?
>
> Thanks
>
> 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.

[Joerg]

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.

[John Larkin]

On Wed, 02 Jul 2014 14:25:10 -0700, Joerg <inv...@invalid.invalid>
wrote:

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.

[Joerg]

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.

[Bill Sloman]

On Thursday, 3 July 2014 02:22:20 UTC+10, Klaus Kragelund wrote:
> Hi
>
> I am working on a high power frequency inverter (>1kW) and have a "funny" looking emission graph
>
> The converter is the typical inverter type, EMC filter->3 Phase Rectifier->DC Inductor->DC Link capacitor->6 switch Inverter->Motor
>
> Like this:
>
> www.electronicsdesign.dk/tmp/Inverter.jpeg
>
> We stripped down the circuits and isolated the noise to the EMC filter and bridge rectifier.
>
> The problem is the very high conducted emission noise at 150kHz:
>
> www.electronicsdesign.dk/tmp/NoisefromRectifier.pdf
>
> (Conducted emission plot narrowed in from 150kHz to 2MHz, conducted with a 2-stage EMC filter and a bridge rectifier with a capacitor on the output and a resistive load)
>
> www.electronicsdesign.dk/tmp/PowerModel_V0.pdf
>

> The bridge rectifier is a part of an integrated power module, but that noise also comes from a standard single phase rectifier

>
> First thought would be that the noise is coming from the reverse recovery 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 currents running in the bridge rectifier
>
> I have tried with snubbers and that did not help.
>
> Any insights into the cause and the remedy?

No insight into the cause, but the remedy usually turns out to be working out 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 itself shorter. It can help if you can twist the path, so that adjacent patches of the included area are radiating in opposite senses.

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

http://www.amazon.com/Grounding-Shielding-Techniques-Ralph-Morrison-ebook/dp/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 first edition, which I first read back in about 1967.

--
Bill Sloman, Sydney

[josephkk]

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

?-)

[Klaus Kragelund]

On Thursday, July 3, 2014 12:25:06 AM UTC+2, Robert Macy wrote:
> On Wed, 02 Jul 2014 09:22:20 -0700, Klaus Kragelund
>
> <klau...@hotmail.com> wrote:
>
>
>
> >> ...snip....
>
> > Any insights into the cause and the remedy?
>
> >
>
> > Thanks
>
> >
>
> > Klaus
>
>
>
> Interesting gets worse with lighter loads.
>
>

That's because the conduction angle get smaller

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

I cannot disclose that, sorry

>
>
> What's the physical location of all these parts?
>
>

On a big 4 layer PCB board

>
> What LISN are you using?
>
>

Standard 3 phase LISN as per CISPR something something

>
> Who manufactured your 1N4007's?
>
>

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

>
> 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.
>

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

[RobertMacy]

On Thu, 03 Jul 2014 15:33:21 -0700, Klaus Kragelund
<klau...@hotmail.com> wrote:

>> ...snip....
>

> 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

THANKS for coming back and telling the progress.

I asked about the EMI Filter for the reason that poor EMI Filter once bit
me! Some previous engineer had selected an absolutely WRONG filter. One of
those that looks good on paper, but turned out to be the WRONG impedance
match at EACH end!

I asked about the 'placement' of parts, because I've seen switching
currents bypass ALL the filtering, like windings on a transformer, with
the signals simply going around any attempt to EMI filter out stuff. [a
little copper tape shielding showed THAT weakness]

And, lastly, I asked about manufacturer BECAUSE FAIRCHILD has historically
made extremely LOUD rectifiers [having worked there once, I have no idea
why or how that happened but perhaps they wanted EFFICIENT snap OFF with
no regard for that reverse recovery time shutoff being so abrupt that it
almost produces noise out to MHz.] In our 'linear' supplies we simply
removed Fairchild from our approved vendors list and bought Motorolas,
General Radio(??) and such.

PS: it is really worth it to model your Test Setup, AC mains cabling,
LISN's and whole power supply and let LTspice show you where the tendency
towards ills are. Matches VERY closely out to around 10MHz and roughly on
up to 30MHz.

[Robert Baer]

Try adding a small resistor in series with each diode; value TBD.