ad: expert-3

R-78 switching regulator noise

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

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

    N5HXR Ham Member QRZ Page

    Oh wow, you're right -- I must have crossed that up in my head when taking it to the breadboard. Though, I've experimented with different values, and gotten pretty close to that (I don't have a 12uH inductor, but got close!). With lots of permutations, I can't get it consistently below about -60dBV with a single filter on the input like that.

    I am confident with more parts I can get it down -- I set up a 5th order filter at one point (cutting some traces and flying inductors FTW!), and saw it drop down to something like -90dBV. But at that part count, it's going to take a lot of real estate, and I think I'd rather find a part that just might be a better choice :).
     
  2. N7EKU

    N7EKU Ham Member QRZ Page

    Hi,

    But did you test it at 2% load and with the 100nF cap? A small value cap like that can be very important for reducing noise at certain frequencies.

    73
     
  3. KA9JLM

    KA9JLM Ham Member QRZ Page

    I have never played with that regulator.

    Is it 560kHz for real or 56 k like others ?

    Good luck on your project.
     
  4. N5HXR

    N5HXR Ham Member QRZ Page

    Yes, I've tested at 2% load, and up to 40%, and with output capacitors at a few values from 10nF to 10uF (including 100nF, everything I have under 1uF is C0G). Tried inductances at 4.7uH, 10uH, 12uH (made one with a ferrite toroid), and 100uH (values I have around), on input and output, with some variation in capacitor chemistry and values. I'm sure there are some permutations I haven't tried, but there hasn't been anything yet that suggests a significant step in a better direction.

    Basically, I see the 560KHz ripple somewhere between -70dBV and -60dBV no matter what I do, and higher harmonics closer to my frequencies of interest (7MHz) in the -100dBV to -90dBV range (flirting with the noise floor, so likely at the limits of the scope), depending on configuration. If I go to a full 5th order filter (e.g., two inductors with shunt capacitors), I can make a lot more disappear, but the 560KHz line is still there, all prominent.

    But I don't have intuition for this stuff -- just have read scary stories about switching noise, and trying to do some diligence to understand what to do. The ARRL handbook doesn't have anything useful on this topic (like most topics, it stops tantalizingly close to what you really want to know, alas!), EMRFD doesn't touch on it (not to mention SSDRA), and I haven't been able to find blogs or other discussions about how to know what acceptable switching noise should look like. I'm sure there must be some out there, but it's not like I'm not searching :).

    It seems kind of irritating that I'd have a 12V supply, use a linear regulator to drop that to 3.3V for the uC and Si5351, and burn all that power. But then, I don't want to have 10+ parts for filtering either...
     
  5. N5HXR

    N5HXR Ham Member QRZ Page

    Thanks! I'm at least getting the hang of what happens when I change stuff :). It is 560KHz -- same as the TRACO DC/DC converters, and some others out there. I even swapped in a CUI part I found in a box, rummaging around, and it looks basically identical on the scope.

    I just rigged up a test with an LM2575, which is a 52KHz switching frequency, and holy cow, what a difference on the FFT! Though, to be fair, that regulator tails off pretty quickly and doesn't really show anything in the HF bands. But it takes a big inductor to tame the ripple (330uH), and big caps too (100uF).

    I am now in the middle of a switching regulator filtering presentation from TI, and learning a lot. Some of it explains what I'm seeing on my experiment board right now. I'm really starting to think I might be able to ignore the 560KHz ripple, and just make sure the 40m band is filtered down to the noise floor... maybe I'll build the thing with this part and just see what happens...
     
    KA9JLM likes this.
  6. SM0GLD

    SM0GLD Ham Member QRZ Page

    But you have presented so many nice RF projects, this is also RF.
    Put copper tape all around the R-78 and solder it to gnd. In other words shield it.
    To measure the residual ripple solder a coax directly across the input cap (or output cap)
    Do not use any probe with long gnd lead.
     
  7. K3EO

    K3EO Ham Member QRZ Page

    I took a look at the Linear Technology web page and they have the LTC3601 (see attached). It may be overkill for your load current requirements, but the switching frequency is programmable from 800 kHz up to ~4 MHz so you can move those nasty harmonics out of your band of interest. The regulator also can be synced up to an external oscillator which is rather neat.
     

    Attached Files:

  8. N5HXR

    N5HXR Ham Member QRZ Page

    OH. MY. GOODNESS, that was it. Wow. So it's bad measurement methodology... I don't have any copper tape or foil, but I thought I'd just start with the coax. I soldered an SMA connector to the board and connected it without using a probe (never mind the solder blobs; this thing's had a lot of solder, desolder, resolder today...):

    probe.jpg

    And the scope is so much prettier... so, the fundamental is at -92dBV, and harmonics drop across HF quite nicely:

    SDS00124.png

    Wow, I didn't realize the probes were such a bad idea for measuring something like this. Lesson learned, thanks for sticking it out on the thread to get me where I need to be.

    Maybe I'll pick up some copper tape or foil or something over the weekend and play some more.

    I guess these things aren't that bad. Color me sheepish :oops:
     
    KA9JLM and SM0GLD like this.
  9. N5HXR

    N5HXR Ham Member QRZ Page

    I wanted to post an update on this, now that I've built a whole transceiver with the R-78E in the power supply section, and tested it "for real." The result is unfortunately not good. I put it all together, and with some minor board layout issues I could bodge around, it seemed to be going quite well:

    build.jpg

    In testing with my signal generator, I was expecting the minimum detectable signal to be S0, based on my intended gain distribution. But no matter what I did, there was an internal noise floor I couldn't get away from. The smallest signal I could detect by ear (or see on my scope's FFT) was about -105dBm at the antenna input. Not what I'd hoped!

    I changed the gain distribution... adjusted some filters here and there... added some ferrites... added an optional LPF going into the RF preamplifier... nothing made a difference.

    Finally, I went on a trip and mulled over it for a few days, and realized I had to go back to the power supply question. Even though the major spikes in the R-78E's switching harmonics passed over the 40m band OK, and they were very heavily attenuated (more than shown in the last post!), I thought maybe swapping out for an LM7805 would prove if it was at fault...

    And... now my minimal detectable signal at maximum gain is -130dBm. So... I guess the R-78E has harmonic spikes, but is also throwing out a general increase of the noise floor internal to the receiver? Maybe there's a power rail trace too close to something, or maybe just having the enclosed box of the R-78E too close to input filters exacerbates it... but anyway, I guess I have to work out something else.

    psu-change.jpg

    Annoying too, because with the LM7805 in there, the RX current shoots from 42mA to 91mA (ugh!). Looks ugly too.
     
  10. AI5DH

    AI5DH Ham Member QRZ Page

    Using any wire for a signal is going to be a problem at higher frequencies. Additionally, get rid of all the components using wire leads. Go SMD using planes for power and return.
     

Share This Page

ad: M2Ant-1