3
\$\begingroup\$

I recently had one of my new rechargeable Lithium Ion products tested (RE pre-compliance scan) and I had a few frequencies (around 150 MHz) pop (7-10 dBuV/m) above the FCC and CISPR limits.

I'm not an expert at dealing with EMC/EMI problems by any means, but I am familiar with proper PCB layout/routing practices to minimize the chances of problems. I don't think the PCB is the problem. But lets see what you all think. I can provide images of the layout if needed.

I have done some probing using H and E near field probes and a StiCo 18" VHF antenna to try to pick up some of these harmonics. The fields do not seem very strong in the problem frequencies (+17-20 dBuV/m with the inductor inside the loop of my BeeHive 100C H field probe between 50-150 Mhz). This doesn't seem bad to me (but again, not an EMC expert). The field does have a 500 kHz harmonic, which confirms it's coming from the switcher which runs at that frequency. There's nothing else on the board that generates noise. The measurement drops to zero when moving more than about an inch from the emissions source. I don't pickup anything with the smaller loop probes. I see a (+5 dBuV) E field when probing the switcher and inductor directly.

I know it's very difficult to causally link near field measurements into far field results, but would it be safe to deduce from all of this, that i'm probably not radiating out from the PCB? Is there some other way to confirm this? It seems to me like the input wire is my problem. I do not have any other EMI filtering on the board other than ceramic bypass caps.

My next step is going to be to do a few more scans using a clip-on ferrite. If that works and gives me good margin, then i'll probably spec out a new AC/DC converter with the ferrite built in. I'm also looking at adding a proper EMI filter to the board, but I just don't think I have enough room on the PCB for those components, so the ferrite seems like my only option.

Some notes about the PCB design that I think may be pertinent;

  • PCB is small, 25mmx28mm, 6 layers with very short traces and power/ground planes
  • PCB is contained within a grounded metal (aluminum) housing that is open on only one end
  • 12V input, 3V output, 2.5A output
  • Maxim integrated 500kHz switcher capable of 4A continuous output current
  • The regulator is very stable and running at 90% efficiency with clean/no ringing switch node
  • Using a shielded 5x5mm Coilcraft inductor
  • The system uses a charging cradle base which has a long (1m) wire delivering input power

Any input or suggestions would be appreciated. Thanks

\$\endgroup\$
  • 1
    \$\begingroup\$ Any non-linearity (and any junction is a non-linearity) can mix multiple energies, and upconvert to 150MHz as the sum of input tones. \$\endgroup\$ – analogsystemsrf Oct 4 '18 at 3:26
  • 1
    \$\begingroup\$ I think it is save to say that it’s not a patch antenna type of radiation, because your board is too small for that and the wavelength at 150MHz is about 2 Meters. So it could be an inductive loop on your board or your cable that is radiating. This frequency 150 MHz is very likely to originate from the switcher design or layout. \$\endgroup\$ – Stefan Wyss Oct 4 '18 at 4:14
  • \$\begingroup\$ Is the aluminium casing connected electrically to the electroncis ground? Is it connected from multiple points? \$\endgroup\$ – user94729 Oct 4 '18 at 4:23
  • \$\begingroup\$ "I had a few frequencies (around 150 MHz)" ´Could you share the exact frequencies? Are the 1st and 6th layers of PCB mainly ground pour, and traces travel in the inner layers? Could you share any schematics / layouts? \$\endgroup\$ – user94729 Oct 4 '18 at 4:24
  • \$\begingroup\$ @analogsystemsrf Interesting point. \$\endgroup\$ – Ben Oct 4 '18 at 14:43
2
\$\begingroup\$

As far as I understand you have switching power supply that has unwanted emission. Your switching power supply is locked on the 500KHz and is generating 3V/2.5A output.

One of the confirmed methods in reducing the radiated EMI is using spread spectrum clock instead of the fix one (that in this case is 500KHz). If you simply could add some jitter to the clock (e.g. 500KHz +/- 20KHz) the concentration of the radiated energy will distributed over a wider range and the overall radiated energy of the switching device will be reduced. There is an IEEE article that may could be helpfull https://ieeexplore.ieee.org/document/1031552 Spread spectrum clock generation is very effective in reducing conductive radiation in micro controller based circuits too and some Freescale uCs have built-in SSCG engine.

There are some other ways. If your PCB had enough space, you could implement twisted pair on the PCB for line input and the same for line output.(some us patent exists for implementing twisted pair on PCBs) You may have a big inductor on your circuit some are shielded and some are not. In your case using shielded inductor is much better. If you had any control on reducing the rise time of the 500KHz signal, It could be also effective on the radiation.

\$\endgroup\$
  • \$\begingroup\$ I did consider spread spectrum. Unfortunately this is a fully integrated regulator. While some regulators have sync pins where you can vary the clock frequency, this one doesn't. I also don't have any control over the rise/fall time of the internal FET drivers. Space is definitely very tight on this design. I wouldn't want to infringe on any patents with the twisted pair designs, but I am looking at improving my layers and power path routing as much as possible. Thanks for those suggestions. \$\endgroup\$ – Ben Oct 4 '18 at 22:50
  • \$\begingroup\$ @Ben. Thank you in advance for your attention. Fortunately your design has violated the standard just for (7-10 dBuV/m). maybe it is helpful If on your 6 layers PCB you change the bottom and top layers to ground and make routing at the inner layers and put a grounded CAN over the regulator chip and its inductor. If you could reduce the drive strength (from 2.5A to 2A) It could help you too (If it has not conflict with your requirements) \$\endgroup\$ – BD_CE Oct 5 '18 at 6:35
  • 1
    \$\begingroup\$ Yes, i'm going to try to optimize my PCB some more. But I did [unfortunately] discover that my AC/DC wall adapter is contributing most of the noise in my test scans. The adapter's test report looks nearly identical in shape and noise levels (with a purely resistive load) to my test report with the complete product connected. \$\endgroup\$ – Ben Oct 8 '18 at 15:50
  • \$\begingroup\$ @Ben. So it is enough for you to change the AC/DC adapter with the one that has real CE/FCC standard or add Ferrite core line filters (kgs-ind.com/wp-content/uploads/2017/08/…) for both AC input lines and DC output lines. You also add internal I/O line filters (LC) to the adapter if it has enough space inside its case. \$\endgroup\$ – BD_CE Oct 9 '18 at 5:21
1
\$\begingroup\$

At 150Mhz it is almost certainly your cable which is radiating common mode emissions. Its very difficult to make an effective 2 meter differential antenna on a PCB (your PCB current loops). You will see PCB radiation if they are present at much higher frequencies.

100-300Mhz is a common "trouble range" for cables. You will likely solve this problem with a filter that reduces the noise conducted from the input node of your switch mode power supply to the cable. The switching action results in current spikes on the cable during switching. For low power electronics a ferrite will work well, for high power (main chargers for example) chokes and standard inductors are frequently used.

A few tens of mikroamperes of common mode high frequency current on a cable is enough to fail radiated emisisons. You can identify if the cable is a problem during testing by clamping a ferrite core around it. You can also estimate the magnitude and frequency of the current with a good current probe and a spectrum analyser.

There are two main sources of generated frequencies from switcher. The obvious one is the main switching waveform whose components are easy to identify as some multiple of the switching itself. The second source is the ringing generated by the parasitics between inductance and capacitances in the switcher design. This ringing can appear at a frequency not related to the regulator switching frequency! So if you have an unknown new frequency it is likely ringing somewhere. Usually the fundamental components of these will appear as "lobes" of considerable bandwidth compared to digital signals.

\$\endgroup\$
0
\$\begingroup\$

In my last EMC tests in the chamber, we had the same trouble: The DC Power Supply was a big noise source. To reduce this noise you can use a ferrite clamp of course.

A cheaper and more efficient way is to implement a common mode choke at the power input of your PCB. Additional add some MLCCs in the range of a few pF close to the input. In the size of 0402 or 0603 they have a high resonance frequency and should be capable to reduce the noise coming in from the power supply.

I had the same issue a few days ago and I´m now working on a board level fix.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.