Radiated emissions failure

Hi everyone

I recently submitted a new design for emc testing. Its passed all but the radiated emissions testing.

The design incorporates a camera connected to a controller. The camera is powered by 24vdc down a 15 m power cable.

When tested for emissions there are some pretty significant peaks in the results.



The top chart in the above image shows the emissions from just the camera powered via 15m cable.
The bottom chart shows the emissions after placing a ferrite clamp on the DC cable.

From the charts its clear that the noise is at harmonics of the 36MHz clock frequency of the digital video signal. After checking the 24v power rail with an oscilloscope with an FFT function it seems evident that there is a lot of 36 / 72 / 108 ... Mhz noise peaks on the 24v power rail. So i'm assuming that the emission failure is due to this 36Mhz fundamental and harmonics being conducted onto the 15m power cable and then radiated out as RF. I'm assuming that the power cable is the radiator because placing the ferrite on it reduces the emissions.

Am I on the right track with my assumption about the noise radiating from the cable ?

Any suggestions on what type of filter I need to kill these emissions?

Beetlejuice, this is a subject i have been looking at. I think to get a meaningful response you have to show your schematic, information on cable used, pictures of construction, pcb layout.

When it comes to EMI everything is important, components, lead lengths, layout construction.

You might get buy with a ferrite bead in both power lines or you may need a common mode choke. An RC filter may be enough, or even just a ceramic cap to ground on each power lead. Or a ceramic cap between the two leads.

Sometimes none of this will help because the inside of the camera box is filled with RF and it will couple to the power wires just because they are there. If this case you would use bulkhead filters.

A DC cable should be fairly easy to be made a non-radiator. Ferrite beads, yes, but also capacitors to ground on each line on each end. Is DC all that is on this cable? What OTHER parts of the design could be radiating?

As FlapJack said, we need more information to be able to help you.

Thanks both for the reply.

I am unable to show pictures and in depth details of the product. I can however expand on the information i have already given.

Below is the 24v input section of the camera ...




The supply feeds different parts of the camera through separate ferrites. So, you will see it is filtered to some extent, but quite obviously not enough.

Below is the 3.3v supply. It is fed 24v via one of the ferrites in the input curcuitry.



and similarly we have an internal 12v supply ...



As previously mentioned, the noise that is being radiated is at harmonics of the 36MHz pixel clock of the camera. Internally, an FPGA receives a 16bit image data and processes it at this clock frequency. So there is quite a lot going on at 36MHz. Its all in an earthed steel enclosure, however there is a gaping big hole for the camera to look through. Even though the camera enclosure does not provide great shielding, and will add to the emissions, I think the biggest issue is the radiation of noise on the 24v DC power cable.

The DC supply is derived from a 120W 24v SMPS from a world reputable brand with CE, FCC approvals. The SMPS connecte to a 15m cable which then connects to the camera. The 15m cable is bog standard 1mm 2 core power cable, (just like you would run your table lamp from !). I can see even as I am writing this that changing the power cable to a shielded cable would be a step in the right direction.

Any other suggestions would be greatly appreciated

New plan. You should always attack the EMI problem at the source. Is this 36MHz external on an open trace that you could put a ferrite bead at the source right out of a chip.

If not you may have to shield the 36MHz area with a small metal box. The shiny tin ones you see soldered to the PCB.

I think you should start there and try ceramic caps to ground on power lines second. Just guessing 0402 .01or .001uF. the capacitor must have very short connection to the power terminals and the ground plane of the noise signal as in directly to both. You are trying to give the noise a return path. If you do not have a good ground plane this will probably not work.

Hi FlapJack

The PCB is 4 layer with a very solid groundplane - ie no splits etc.
The 36MHz pixel clock is just one single at this frequency. There are also data lines changing every clock cycle. The tracks have been kept short as possible and all over good ground.
There are 22R resistors in the data and clock lines to reduce slew.

There is however about 80mm of ribbon cable from one PCB to another over which these signals pass. The ribbon connections have been alternated signal, gnd, signal, gnd ... etc to create the shortest possible return path.

I would normally agree - attack the route of the problem. However its a bit late in the day for that.

I need to just get a few more DB attenuation on the radiation from the DC cable and I'm through.
I've asked the test lab to put a ferrite ring onto the 24v input wires between the input connector and the PCB, and loop the wires through a few times.
Also asked them to add a 10nf cap across the 24v input.

Any other suggestions are appreciated.

I don't think your 24V input section is filtered much at all. You have ferrite beads, yes, but those are only part of the equation. Try putting 1000nF caps on the camera side of the beads.

beetlejuice said:

However its a bit late in the day for that.

Click to expand...

Not really... if you can't find a workable solution, then your product "ain't goin' nowheres" until it is fixed.

Hi SLK001

Thanks for your input.

I agree that the filtering on the input is very sparce. Although the bits of schematic are misleading of the layout. The ferrite beads shown are actually at the inputs to different sections of the camera board. For example the v_logic_filter ferrite feeds the 3.3v power supply, where there is a 10nF, and 330nF cap in parallel. Also v_camera_filter feeds the 12v supply where there are similar capacitor.

I think that this gives each separate section of the board a degree of filtering, but there is basically no filtering for noise leaving the camera via the 24v power input (con3 in the first diagram)

Also what i meant by "its a bit late in the day for that", was major redesign of the PCB. I am hoping I can find a filter combination that I can fix onto the 24v input, that will reduce the emissions enough to pass. This may well be an addition to the PCB, but a relatively easy change.

It sounds like you just want to get the product out the door and do not want to do another board rev. In that case it sounds like you have the simplest plan. It is a manual operation and will cost more but if it works you can move on.

Also were you have ferrite beads now on the 24v lines can you get more aggressive ones to try in the same package so no mods are needed.

If the EMI is getting past the ferrites then it is traveling up the fan wire as well. You may need ferrite rings on the fan wires also.

Yes absolutely, I'm looking to get it out of the door ASAP. Its a low volume industrial product, and adding a quick fix front end filter is basically what I'm after.

We have already seen that putting a ferrite Clamp on the DC cable reduces the emissions, but just not enough. From my understanding then, the ferrite on the cable is acting as a common mode choke, which suggests that the emissions are common mode.

The simplified block diagram of what we have at the moment is below...



The power supply is connected to earth internally so the 24v ground if referenced to earth there.
The camera metal enclosure is also earthed, but all the electronics inside it are isolated from the metalwork, so there is no connection between 0v and earth inside the camera.
Its my understanding that common mode currents are created because of differences in earth potential? There is no earth connection to the circuits within the camera so, but I assume there will be parasitic capacitances to earth. However, in a lab, the two earth points are very local.

I'm just not getting my head around this being common mode noise.

How about a small cap across the power supply at the camera end? Leave the ferrite beads as they are.

c_mitra said:

How about a small cap across the power supply at the camera end? Leave the ferrite beads as they are.

Click to expand...

Thanks for your input c_mitra

I've already asked the test house to try this - with a 10nf ceramic cap.
I'll let you know how it goes when I get the results.

Somehow you need to take a noise measurement to see if you are improving things. Just guessing can only take you so far.

If you have a spectrum analyzer try putting a few turns of wire on the end of the input lead and see if you can pick up any noise off of the power cable. Then try your fixes one at a time to see what is working.

- - - Updated - - -

SLK001 said:

A DC cable should be fairly easy to be made a non-radiator. Ferrite beads, yes, but also capacitors to ground on each line on each end.

Click to expand...

Putting a capacitor on each end does make a lot of sense, to return the noise. Are you supplying the power supply?

- - - Updated - - -

beetlejuice said:

There is however about 80mm of ribbon cable from one PCB to another over which these signals pass. The ribbon connections have been alternated signal, gnd, signal, gnd ... etc to create the shortest possible return path.

Click to expand...

I forgot about this. You may very well pick up several db with a ferrite clamp on this cable near the transmitting end.

Try a different cable length as well, a small change in cable length can sometimes have an affect especially when the cable length is near the wavelength of the clock or its harmonics (or a division 1/2 1/4 1/10 1/20 etc.)

Here's what I use to sniff out errant emissions:

SLK001 said:

Here's what I use to sniff out errant emissions:

Click to expand...

Very useful to know. Thanks

beetlejuice said:

There is however about 80mm of ribbon cable from one PCB to another over which these signals pass.

Click to expand...

Can you just try a Cu or Al tape on the ribbon cable? Please stick on both sides. Al tapes are common but may be too thin and you may need to put two layers. It may increase the cable capacitance and this is purely a debugging aid...

c_mitra said:

Can you just try a Cu or Al tape on the ribbon cable?

Click to expand...

I think if you do this, the tape just acts like another radiating element unless it is grounded.

FlapJack said:

I think if you do this, the tape just acts like another radiating element unless it is grounded.

Click to expand...

That is possible but unlikely; he has already taken care to put signal and ground wires side by side in the ribbon cable. I do not expect 8cm (3 inches) of ribbon cable to radiate much in any case and, if the cable just goes round it will simply act like a Faraday cage.

Did you ever try to use a regular common & differential mode filter for the power supply line? Although we are talking about radiated emissions, they'll hardly ever exit a small camera enclosure other than through the supply and signal cables. Next point would be to look at the enclosure ground and cable ground coordination.