Taming a switching supply?

I've been comparing an external 5V USB battery versus assorted wall warts, to get a sense of the extra noise introduced by assorted 5V switching supplies. There are definite differences, mainly at the low end below1 MHz. I wonder if something as basic as a large capacitor (such as a 1F super capacitor) placed across the 5V supply outlet would suppress the extra RF noise. Or is it still necessary to use a(n) inductor(s) in combination with a capacitor to truly "tame" a switching supply? And yes, I realize that I shouldn't be using a switching supply to power a KiwiSDR.

Comments

  • There is normally a fair amount of mixed mode noise around switchers so a lot depends on the supply and how it is getting in to the radio. I won't pretend to be any kind of expert but in my tests it's important to stop radiated noise and doing that is easier before you tame generated noise from the supply.

    Get some large clamp on ferrite beads (multi turn on the lead) and use Kiwi channels* to do simple comparative measurements, have a couple of RX channels on the go, one real slow and others zoomed in to problem area(s), add ferrite to the lead from the supply and to the kiwi, work on the principle that both are antennas and you are looking to make them really bad antennas more like simple resistors turning the unwanted energy into heat

    Once you have made the leads more immune to pickup by suppression and separation/orientation, look at adding common mode filtering to the input and output. Many cheap PSU's skip fitting the input filter componenets so feed RF back to the mains (which can be a decent antenna). Dealing with the noise conducted through the DC lead is only a part of the issue and perhaps the simpler area to work on, skip the rest and gains may be hard to find as it's rarely just a bit of riple on the DC getting past the Kiwi filtering.


    *One slow channel will show changes that are marginal but can be detected over time (make notes of change against time), more zoomed in channels can show more comb detail lost in the wider sampling.


    If Martin G8JNJ drops by he can give a more learned response.

  • edited April 12

    Sorry to say all this yet again...

    As one who successfully uses SMPS on multiple Kiwi's and has systems with noise floor below propagated noise levels at least through mid-HF, I would suggest that ferrites are almost never the solution. The problem they attempt to solve is worse than the ability of any inductor of any kind to solve. The impedances involved are often several hundred ohms and against the requirement for perhaps 30 dB reduction, no inductor can provide the extremely high Z over the several decade bandwidth. Even tuned systems can't reliably do this is in a narrowband case.

    For a Kiwisdr, at least, 'ground isn't ground' and coaxial cable only shields TEM and differential signals. Common mode signals and noise is injected even when every conductor of a cable is at equipotential with respect to the others. These that have zero differential noise can support current that then passes through the so-called "ground" of a kiwi and develops IZ (it's not just IR) drop that appears between the preamp input and the reference to the ADC and appears as unwanted response.

    The kiwi can detect signals that are only -157 dBm in 1 Hz bw. The amount of current required to achieve this with the Z of these common paths through the Kiwi is so small that many orders of magnitude of reduction can be required in typical situations in order to push it down far enough to not be detectable.

    By injecting common mode noise one can easily verify that only 80-100 dB of isolation is present where maybe 30-60 dB more than this is necessary to avoid impairing system performance. MIcroamperes of unwanted signal and noise flowing through the Kiwi common, the so-called 'ground' is enough to destroy potential performance as can be seen by examining the bulk of publicly available Kiwis. Most systems do not do this well. The few good ones are markedly different.

    Rather than fussing with inductors, C's and L's at the PS I suggest first being sure that the balance/isolation against the common mode path out the SMA connector be addressed. Generally a 1:1 transformer like a MiniCIrcuits T1-1 located right at the SMA is a good starting place. These transformers have no more than about 1 pFof inter-winding capacitance - the path which CM noise must traverse - and may be able to provide 30 dB of additional rejection at 30 MHz, more at lower frequencies. Paths other than those involving the antenna SMA are possible but this is usually the worst.

    As I've written on this forum way too many times before, the biggest issue is common mode paths. The Kiwisdr doesn't have a significant differential noise ingress problem. Capacitors and series L's don't solve the fundamental problem. Please see my Notes at wsprdaemon.org.

  • edited April 12

    I approached the problem more "pragmatically" than "scientifically". 😉 My Kiwi is solar powered, so 12..14 V are supplied (which I also use directly for my homebrew antenna switch). For the Kiwi I use a small 3A step-down converter, which I built completely shielded into a tinplate case, with all lines to the outside going through ferrite chokes and feedthrough capacitors. (It is even placed right next to the kiwi.) Subjectively, I am very satisfied with the result, and Kiwi's own SNR calculation often reports values of more than 40 dB, so it can't be all that bad. However, a comparison test with a linear power supply is still pending.

    I wonder if something as basic as a large capacitor (such as a 1F super capacitor) placed across the 5V supply outlet would suppress the extra RF noise.

    This could be problematic as it will slow down voltage rise while powering on. Digital circuits, especially processors, often don't like that and won't properly start then, so it could lead to strange misbehavior. However, don't know how critical the Kiwi is in that regard.

  • You can generally 'tame' switched mode supplies on frequencies > 1MHz, by the use of suitable screening, capacitors, inductors and ferrites.

    However as Glenn has already indicated at length, the main problem is common mode noise, and the 'ground' path through the KiWi which is difficult to fix, especially on the lower frequencies, due to the size of the componnets required to be effective.

    All sorts of common mode paths can exist, and some experimentation is required to minimise the effects. Although it's usual to find that as soon as you fix a problem on one set of frequencies a new set pop up in another part of the spectrum.

    It's a never ending task, like playing 'whack a mole' or peeling an onion, to reveal yet another layer.

    In my experience, most switched mode supplies benefit from a 0.1 (SMD) and a 1000uF conencted directly across the DC output at the PCB, then several turns (8 to 12) on a FT240-31 ferrite ring fitted to the Mains and DC cables.

    Building a set of E and H field 'sniffer' probes to use in conjunction with your KiWi, other SDR or Spectrum analyser, is the quickest and best way to locate, identify and tame noise sources.


    Images from https://www.emcstandards.co.uk/updates



    Regards,

    Martin

  • I really appreciate the detailed and thoughtful replies to my original query. Setting up a second Kiwi at a new location (EM83 grid) prompted me to explore this question around switching power supplies. The one I am currently using only causes "mischief" below about 0.6 MHz, with a series of noise blobs that raise the floor 10-15 dB - enough to take away the fun of trolling for the weaker NDBs. And given that NDBs are on their way out, this seemed like a good time to put a few in the log. It definitely seems like the common mode pathway is the issue, since I can re-introduce the same noise blobs simply by connecting the SMPS negative connector to the (battery powered) Kiwi ground. Overall not a big deal, but will be fun to work on mitigating it via the assorted suggestions here.

  • This could be problematic as it will slow down voltage rise while powering on. Digital circuits, especially processors, often don't like that and won't properly start then, so it could lead to strange misbehavior. However, don't know how critical the Kiwi is in that regard.

    Absolutely correct. The power management IC (PMIC) on the Beagle enforces a 50 msec rise time on the 5V input. It simply won't power on if it is slower than that.

  • The inexpensive ~960 kHz buck converter mentioned elsewhere on the forum will likely get you low end performance uncompromised. I think they cost about US$1 each...

  • @n6gn can you give a little more detail on your Minicircuits T1-1 setup? Is it mounted to a small circuit board with SMA connectors? Or just free hanging in a small project box? The center tap is just left disconnected?

  • edited May 15

    Just 1:1 on small PCB, SMA male to Kiwi Completely isolated from SMA female to antenna feed line. This leaves only about 1 pF of common mode coupling between Kiwi and antenna line.


    top side of board shown above, Center tap, if present T1-1T, is unused.

    T1-1 from Mini-Circuits,

    SMA edge connectors from eBay,

    unzip & drop .brd file onto OSHPark.com. 3 boards postpaid in US in about 10 days, probably < US$5.


  • edited May 14

    The minicircuit isolation transformers are also available in a metal can with BNC connectors, sometimes there are cheap used ones available on ebay.

    FTB-1-1 with the T1-1 transformer (0.2-500 MHz at the -3 dB insertion loss points)

    I use FTB-1-6 with T1-6 transformer, they go a bit lower in frequency (0.05-150 MHz at -3 dB) at the cost of slightly higher coupling capacity. I measure 2 pF (with the metal can).

  • As Glenn has mentioned "Generally a 1:1 transformer like a Mini-CIrcuits T1-1 located right at the SMA is a good starting place. These transformers have no more than about 1pF of inter-winding capacitance - the path which CM noise must traverse - and may be able to provide 30 dB of additional rejection at 30 MHz, more at lower frequencies"

    It is possible to improve the common mode rejection of the transformer on the HF bands by also including a ferrite common mode choke on the transformer connecting cable.

    The hybrid combination of a flux coupled transformer and transmission line choke can work well, if the high frequency transmission line choke is designed to complement the lower frequency common mode rejection of the transformer.

    In practice I have found that a further 20dB of common mode rejection (S21 measured in a 50 Ohm system) can be achieved at 30MHz.

    The eight KiWi sdr's at Weston, use a dipole and special balun I built, consisting of a 12:1 flux coupled transformer and transmission line choke, the combination of which works very well as a broadband receive antenna. There is a diagram of the receive system on the KiWi header.

    http://websdr.uk:8078/?wifi=0

    Coil-Craft also have a number of miniature wide band transformers, that may be easier to obtain than Mini-Circuits parts in some areas of the world.

    https://www.coilcraft.com/en-us/products/transformers/wideband-rf-transformers/smt/wb_smt/

    I have used both of the following 1:1 types with good success, but other ratios are available.

    WB1-6TSL_

    WBT1-6SL_

    Regards,

    Martin

  • see edited file previously posted for .brd file to have fab'd.

  • Thanks @n6gn for the OshPark files, I have ordered some. And thanks @HB9TMC and @G8JNJ for your insight.

    Does anybody know if the Minicircuits ADT1-1+ will also work? Looking at the specs, it seems to be comparable, but I'm not sure. 1 pF inter-winding capacitance is not on the spec sheet, is more capacitance better?

    The reason I'm asking is that shipping costs from Minicircuits for the T1-1-KK81+ is more than 5x the cost of the transformer. Mouser has the ADT1-1+, 1.6mm end SMA connectors, and also the BBAI in stock, so maybe I'll do a bigger upgrade 😀

    I realize that the ADT1-1 is much narrower than the pads on Glenn's board, but I'm confident I can wire mod it up. Pin spacing is still the same at 0.100" as far as I can tell.

  • I don't know about the ADT1-1+ but I do know that the T1-6T+, spec'd to go lower in frequency is NOT a good idea. Rather than being a simple flux coupled transformer with low inter-turn capacitance it appears to be a transmission line type, even though they don't say.

    Without verifying low capacitance I would stay away from alternatives, though there may be some.

    It's handy to order the CT versions since these can be useful for CAT5 to 50 ohm use as well.

    Mini-Circuits will sample a small number of parts if you have a commercial/academic connection. Otherwise order 10-20 of them, pay the shipping and share them with your friends😀

Sign In or Register to comment.