R-78 switching regulator noise

Discussion in 'Homebrew and Kit Projects' started by N5HXR, Aug 12, 2021.

ad: L-HROutlet
ad: l-rl
ad: abrind-2
ad: MessiPaoloni-1
ad: L-MFJ
ad: Left-2
ad: Left-3
ad: L-Geochron
ad: HRDLLC-2
  1. N5HXR

    N5HXR Ham Member QRZ Page

    I am experimenting with some switching regulators (the RECOM R-78 series), and trying to figure out what makes for acceptable switching noise levels. Internally, they use a 560kHz oscillator, and without any input or output filtering, I can see about -47dBV (RMS) at that frequency, with a bunch of harmonics following that up.


    In the datasheet, RECOM suggests a filter at the input. They recommend a 4.7uH inductor flanked by two 10uF capacitors. As luck would have it, I have a bunch of those on the bench. So, after playing around a bit, I found with that filter on both the input and the output, I observe -80dBV at the regulator's input and output.


    It was much worse with electrolytics. I soldered some 10uF MLCCs to some pin header so I could use those, and saw 10dB reduction in the switching noise :).

    Still not quite as good as LTSpice suggests this filter should do, though this is on a breadboard. Things wiggle around with different jumpers, etc., so I suppose this is kind of a "worst case". Maybe I should make up a little PCB with a decent ground plane and better caps to be sure.

    Is this a level of switching noise of any concern? Seeing levels at -80dBV makes me think it should be OK, but you guys are the real experts. I am somewhat attracted to the efficiency gains.
    Last edited: Aug 12, 2021
  2. SM0GLD

    SM0GLD Ham Member QRZ Page

    Yes a proper ground is very important!
    I would also use a similar filter on the output.
  3. N5HXR

    N5HXR Ham Member QRZ Page

    Aaaaand... on a PCB with traces and ground, etc., the result is actually worse. And this is with a filter on both input and output from the regulator. It's very stable, but worse, at -64dBV. That's somewhat dissatisfying -- makes me think the breadboard vagaries actually made things look better than they were.
  4. SM0GLD

    SM0GLD Ham Member QRZ Page

    Strange! I dont understand:(
  5. N5HXR

    N5HXR Ham Member QRZ Page

    Me either, though if I make it a higher order filter, I can get it down under -80dBV. I checked the datasheets for these little inductors I'm using, and they are supposed to be self-resonant at 38MHz, and have a series resistance at DC of 350 mΩ. The capacitors I'm using aren't the classiest ever, but they're 0805 MLCC 10uF caps.

    In case I might be getting spurious measurements, I coupled it into my TinySA (I know, not the best equipment, but it's what I have), and it seemed to jive pretty well with what the oscilloscope was saying.
  6. N5HXR

    N5HXR Ham Member QRZ Page

    OK, schematic below.


    I found that if I remove C2 and C3, the noise on the power rails drops from -64dBV to -79dBV. I read this interesting article from Analog Devices, which suggests that the regulator topology may influence where the filters should go. After experimenting with a bunch of permutations, it does seem like (A) need filtering on input and output, and (B) putting a shunt capacitor adjacent to the regulator seems undesirable.

    Right now, I have it humming along happily at -79dBV with just the two inductors + C1 and C4. I still don't know if that's a good level to attenuate to, or if I should be shooting for "invisible on the scope" or something :).
  7. SM0GLD

    SM0GLD Ham Member QRZ Page

    If you run without C2 make sure you dont have ringing on pin 1.
    Sometimes buck regulators becomes unstable without C2 if L1 is to big.
    The noise on the power rail drops down without C3 because C3 has a very low ESR witch creates high current spikes on the input. This is how a buck converter works.
    Low ESR caps usally results in higher efficiency but perhaps sometimes higher noice.
    As long as it runs stable you can use any filter you like.
    I like to take a look inside a Recom R-78. Is there any caps there at all?
  8. N7EKU

    N7EKU Ham Member QRZ Page


    I don't see the filter you are using on the datasheet. For EMC Class B they are specifying 10uF, 4.7uF, and a 12uH choke.

    Note on the spec sheet they say a 2% load is necessary to meet specs and their testing uses a 100nF chip cap on the output.

    N5HXR likes this.
  9. K3EO

    K3EO Ham Member QRZ Page

    -80 dBV is still undesirable if you intend to use the regulator in or near an HF receiver. -80 dBV is 100 uVolts which is way above the MDS of most HF receivers. You need another 40 to 60 dB of attenuation to get that ripple down to the MDS level.

    If you really like the efficiency of switchers, I would suggest you choose one or design one that has fundamental frequency that will not produce harmonics in your band of interest. Once again I am assuming that your regulator will go into or near an HF or VHF receiver.

    For example if you design/buy a switcher with a fundamental frequency of 4 MHz, none of the harmonics will fall within the HF/6m amateur bands with the exception of the 7th and 13th harmonic at 28 MHz and 52 MHz respectively. This will really simplify your filtering requirements.

    Years ago, I remember reading about Class E switching regulators. The design gave you high efficiency along with high frequency operation.
    N5HXR likes this.
  10. N5HXR

    N5HXR Ham Member QRZ Page

    Thanks -- that's confirming my suspicions. I guess these modules were attractive because they seem to be intended as drop-in replacements for an LM7805, etc. But the filtering situation is too problematic. I'll end up with as many parts surrounding them to filter them as it takes to use some of the other ones out there. I'll hang onto these for non-radio projects, though...

    Yep, exactly. Are there any regulators you'd recommend? I'm targeting a portable / hand-held form factor, and I'd like to run as power lean as I can (at least on receive). It bugs me that if I regulate 12V all the way down to 3.3V with a linear regulator, there is so much wasted power.

Share This Page