This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMZ1420xH High Frequency Radiated Emissions

Other Parts Discussed in Thread: LMZ14201H

Hello,

I use an LMZ14201H in a certain circuit with the following parameters:

Vin: 16-32V

Vout: 11V

Iout: 600mA (max)

Frequency: 330KHz 

For some reason we have a peak of Radiated EMI at 70MHz approximately.

Are there any known issues or solutions?

We had in the past serious problems with conducted noises in low load conditions, due to the DCM/CCM transitions, we solved this problem, but the 70MHz radiated is hard to explain.

Thanks

  • Hi Emanuel,

    We're looking into this right now and will have a reply for you soon.

    Thanks,

    Anston

  • Thanks, I can send the schematics if needed and the plots
  • Yes please send schematics, layout and plots.
  • Here is the circuit and PCB layout

  • A few explanations skipped:

    C15 is a tantalum capacitor.

    This is the second revision of the circuit, we use the LMZ series in many applications with similar setup and layout, never had such issues.

    This circuit requires an extremely low standby current and the load can go down to 10mA.

    As mentioned, we had conducted emissions issues in the low load and took certain measures to avoid this, among other things the addition of the common mode choke, low frequency ferrite beads and increased increased capacitors.

    The old setup used the LC filter recommended but if compared to the one above used only C6, C7, C9,C10,C11 and I1.

    The evident 70MHz peak close to the allowed limit showed also there, barely passing the test but the CE wad bad.

    The CE is the reason we keep the frequency around 330KHz.

    These were the results of the old version:

    The 68MHz is there and so another peak at about 330MHz. There is a raise also at 150MHz, quite small but needs attention, see below.

    The new version passes quite well the CE test, but still gives the peaks now at 150MHz and 300MHz.

    The customer requires testing the operation of 2 units (each in its own housing, totally separated)

    When testing 2 units everything is amplified and fails. Here we can see the 70MHz back and big, at 150MHz completely fails and at 300MHz avery big peak.

    Quite strange behavior, we never expected noise in this range and in fact based on many projects with the LMZ series we newer observed such high frequency noise. In fact we were very happy with the success we had filtering the clock from the CE tests almost completely.

    I will really appreciate your assistance.

  • Can you set the frequency of one of the two converters to a slightly different frequency?  If the two frequencies are set at least 100kHz apart, they will not add together in quasi-peak mode in the spectrum analyzer.

  • Hi Joe,

    The converters are in individual devices, it is not that they are on the same board or even in the same box.
    As this is a product manufactured in quantities, there is no way to control which pair will be used in the same setup.
    The other thing is that the peak at 300MHz is there at the same amplitude with one or two devices, the one that accumulates when two are connected is the 150MHz.
    Same with the 70MHz peak which I think is the source of everything.
    What I don't understand is how such high frequencies noises result from a 300KHz clock.... black magic indeed.
    I can say for sure that there is nothing else in the circuit that might be the potential source.

    Emanuel
  • Understood. I think what your seeing is the amplification of very high-order harmonics of your switching frequency. This can be due to a resonant circuit formed by the parasitic components of the low-frequency filters. Spice modeling could be useful but time-consuming. I noticed the return path for the capacitors is on the top layer, this forms a loop that can make a good antenna. Is there a ground plane below the top layer?
  • The PCB is a simple double sided type, the bottom is almost in full a ground plane, can see the part under the filter and LMZ in my layout drawing above, the blue drawing in drawing.
    I use this setup in a lot of designs with the same input filter (simplified) and same layout, the difference in this circuit is the inductor, I used a smaller package type and more capacitors than usual. The radiated emissions are going out through the cables, I am almost sure about his, the plots shown are with the antenna in horizontal polarization, in vertical they are below the limits.
    So you think this is a layout problem? Not some resonant frequency of one of the filters?
  • If you are in shielded enclosure then radiation from the PCB may not be the issue but ,in general, the return circuits should be on the ground plane not coplanar with the current carrying trace. Also, the components on the "clean" side of T2 ( C4,C5,FB1,FB2) should be away from the "dirty" ground; no ground copper should be adjacent or below these components. The purpose of T2 is to reduce common mode noise so you don't want to re couple it afterwards.

    As far as the filters go, I think there is a potential for a resonant circuit to be formed at these high frequencies. It might be useful to rearrange the caps C9,C10,C11 and C12 to potentially change the high-frequency behavior. This won't affect your low-frequency performance.

    EMC testing is quite frustrating at times, sometimes the very littlest things can make a big difference.
  • I got your point about the grounding plane and I agree. And also I think maybe to reduce the number of capacitors, in similar designs they work and behave much better. By the way, speaking about ground plane sensitivity, I have several circuits assembled on single sided MCPCB and also work with no EMI noise at all.
    The problem I had here is a combination of two requirements:
    1- To work in standby mode (no load) with a very low current consumption
    2- The load changes from 10mA minimum to 600mA maximum.
    In standby and minimum load I had serious problems with the conducted emissions, as the LMZ is switching between DCM/CCM modes and creates a lot of noise at low frequencies, 10-150KHz.
    I solved part of the problem by shutting down the LMZ completely at standby, and filtering at minimum load.
    But this created the present problems.
  • Emanuel,

    The easiest way to tackle radiated EMI noise is to place very small capacitances, very close to the input and output of the IC. Start with values that are roughly between 470pF and 4.7uF. One or two small capacitors should be sufficient and will provide a high frequency bypass route.

    The other option is to employ a LISN device for each DUT. Very frequently the cabling to and from the board carry large chopping currents that radiate out and interfere with the readouts. I had this happen to me once and solved it by looping the input and output wires from the DUT around a ferrite bead or torroid.

    Let me know if this helps,
    Thanks,
    Anston
  • Hi Anston,

    Thanks for the answer, I will try this of course and also the ground plane change as suggested by Joe.

    Let me add that the situation we are facing is quite complicated. I had a lot to say about the test procedure which was made by the customer, which I think it was very wrong. The 461F allows reproducing the real installation conditions, but what they did is checked 2 units, first each one independently and somehow they passed, not brilliant results but below the limits. When connecting both of them the plots were as shown, and the over the limit peak at about 150MHz was only in horizontal antenna polarization, vertical passed. This clearly shows the noise is going out through the cables.

    They connected the cables separately for each DUT, giving a total of 4-5m non shielded cable. Of course, only 1 pair of LISNs was used.

    The problem is they won't agree to change anything, adding ferrite or using shielded cables. 

    They also claim the reputable lab insists the procedure is correct and above all they tested a competitor's product which passed the test with the same setup.

    I wish the solution will be as simple as adding some capacitors or changing the ground plane, given the expected quantities I think both us and TI will be happy.

    I will test the suggested solutions and keep you posted.

    Thanks,

    Emanuel