Quiet switch mode power supply (SMPS) for KiwiSDR

2

Comments

  • I think you may have a fake.

    I have tested several genuine Apple 5w compact plug top power supplies, they hav a switching frequency of around 100KHz (dependent upon load), and the noise spurs tend to be relatively low on frequencies greater than 500kHz.


    This plot shows the RF noise measured on the DC output of the supply from 0Hz to 1MHz.

    The yellow trace is an Apple 5w charger running with a 1A load.

    The purple trace is a linear power supply using a LT1083 5A LDO regulator.

    The cyan trace is the ambient noise level with the power supply turned off.

    Cheap Chinese chargers can easily produce 15 to 20dB more noise at around 1MHz.

    Obviously, this is not the whole story. Earth loops and radiated emissions are other mechanisms by which power supply noise can cause interference. But this at lest give some indication of typical performance.

    Regards,

    Martin

  • Hi, Martin. How did you take the measurements? I am most interested in LT1083. My linear power supply has 3 LT1083 pieces. There are also L7812 and L7805 and the like. But I used good LC filters for the output voltage. Should I be worried that this noise is spreading into the airwaves?

  • An easy test is to terminate the KiWi RF input with a 50 ohm load, and see if the noise floor rises with the power supply you wish to test.

    If it does, then the supply is really noisy, and it will be very evident if the setup shown below is used.

    The diodes protect the Spectrum Analyser from excessive voltage spikes, and the transformer helps to eliminate possible ground loops.


    In practice, earth loops and radiated emissions are still likely to be the most significant factors,

    So if it only seems noisy when the antenna or coax is connected, then the test setup that I detailed earlier in this thread, is more suitable to help identify the common mode noise contribution.

    Regards,

    Martin


  • I performed some measurements on the cheap Chinese filter boards mentioned earlier in this thread. I tried the 50V 4A version which had 22uH inductors and 470uF capacitors fitted.

    The differential mode rejection is pretty good, but the common mode rejection is poor, and somewhat disappointing.

    One of the boards had a damaged inductor, so I swapped both for 470uH 1A inductors of the same size and style.

    This improved the common mode rejection on the LF bands, but made the overall performance worse on the HF bands.


    Regards,

    Martin

  • I have one of those and it is very quiet, BUT, transistor on back get WAY too hot. Too lazy to take apart and put a big Heatsink on it !

    For what its worth, I'm now using a 12 V 1 amp supply from Jameco that was under $20. It is part number 1953612. It been on for several weeks now and cool to the touch.

  • edited July 6

    Clint KA7OEI have a series of articles in his blog that may be helpful:

    As stated in his articles and earlier in the thread be careful when working on the mains !

    https://ka7oei.blogspot.com/2012/12/reducing-switching-supply-racket-rf.html

    All that "iron" we removed to make the psu "efficient and lightweight" we have to put back as filter on the outside 😉

    73:s de Kenneth / SM0OHC

    FrankinAu

  • In order to prevent noise on PE one can fit in a choke, withstand a shorting current.

  • Sometimes, when used inside a Switched Mode Power supply, the additional choke inductor in the ground (L3 in the circuit) can sometimes be more problematic than useful.

    I tend to wire the DC 0v / -ve directly to the mains earth, but depending upon the local configuration, this can also make things worse.

    Unfortunately, it's one of those cases where, although it may be good practice to do certain things, you generally have to experiment, in order to find the best solution for your own setup.

    Regards,

    Martin

    Radiofan_01
  • edited July 10

    The way to see that problem is important. Schematic above shows the configuration for an electric assy, which could catch up interferences on "polluted" mains PE lead, not in opposit way connecting leads marked with "load" to a switching PSU. Last situation prevents - as fully correctly stated above by @G8JNJ - conduct RF towards PE. Choke works there as an unwanted resistor then, and great deal of RF remains in assy. However a dimensioning of 100µH seems to be a compromise for both situations.

  • The choke and other grounding configurations depend on if you are trying to prevent mains interference from reaching the receiver, or if you are trying to stop noise from a Switched Mode Power supply being radiated.

    Sometimes you can do both, but other times it's a trade between them, and you have to consider what works best for you.

    Having separate Antenna and Mains supply grounds, further complicate the problem, and introduce safety issues.

    Regards,

    Martin

  • edited July 11

    Lets having an example here. The isulation resistor beween protection earth of chassis of my linear PSU here in use and the KiWi minus DC lead should read as stated in picture. Even better as local regulations say. In my opinion, thus, no problems in earthing of Antenna circuitry, here. However testing voltage approx. 1000V, better to test with 2kV, I think.

  • Hi Martin,

    Your test page is the best discussion of receive site power supply issues I have ever found. Could you post it somewhere that I could refer to on my wsprdaemon.org site?

    As you have shown and Glen N6GN has taught me, many receive site RFI problems come from common mode currents flowing through the milli-ohms of Kiwi PCB ground plane resistence. The PSU is frequently the major, but not the only, source of those currents.

    Many additional currents flow when there is an antenna lead connected to the Kiwi's RF input SMA. It is of course important to minimize the PSU as a source of ground loop currents, but in a working receive site the antenna + LNA (with its own PSU) + feed line almost always introduces one or more ground loops which are additional sources of ground loop RFI currents.

    At my KPH, KFS and AI6VN/KH6 sites where the PSU ground currents are well suppressed, I have still needed the 30+ dB of common mode noise suppression provided by the MiniCircuits T1-1 isolation transformer: https://www.minicircuits.com/WebStore/dashboard.html?model=T1-1-KK81%2B


    Every site differs, so the T1-1 may be of no help but it has been an essential component at my sites. Adding a LNA at or near the antenna feed point can with care also improve the signal levels while not adding to the ground currents.

    My sites are powered by 12VDC battery backup systems fed by a 20A linear regulator and I have found these $1.25 3A switching supplies to be very useful to convert that 12V to the 5V needed by the Kiwis. These supplies ramp up quickly enough to match the Kiwi's power-up requirement, and they switch at 925 KHz so the fundamental RFI and the first few harmonics are outside HAM bands, spaced by almost 1 MHz, and the common mode currents can be well suppressed by small torroids on the input and output leads



    To see the spectrum of a Kiwi system using those components, open: http://www.kphsdr.com:8073/


    73,


    Rob

    studentkra
  • jksjks
    edited July 15

    It's a real shame our efforts to place a T1-1 on the Kiwi-2 failed (it made matters worse for reasons that were never fully understood, but no doubt due to the cramped conditions on the board). But you can bet a T1-1, and more, will be present on Kiwi-3 where there will not be the space constraints.

    Nate_Rnitroengine
  • Oh man, I missed this board from @n6gn, had not seen that thread where it was posted! I'll have to keep that in mind if some of my CM interference is found to be coming through the antenna line.


    -Nate

    N8BTR

  • edited July 16

    Tried out on "Local" connected to notebook via 1m CAT7e cable using the same PE as the power supply of the notebook nearby, causing that shown interference. You may recognize the little bit better SNR, at the time,when outer screen of the CAT7 cable was connected to PE. The interference comes from the notebook's switching power supply, indeed. Nevertheless most of records were made in battery operation, to make prevention of interferences of that kind as successfully as possible. Conclusion: Notebook's PSU may cause interference, too. Using CAT 7 cable seems to be best way to protect LAN connection of the KiWi itself, in an almost inexpensive way.


  • Hi Rob,

    I'm not sure which test page of mine you are referring to ?

    My website hasn't been updated for some time due to various issues, so I don't really have a suitable repository for various bits of information at the moment.

    Regards,

    Martin

  • OK, little bit off topic. But I tried find out the difference in noise produced by the internet router itself. Surprisingly the delivery contained switching PSU causes a noise between approx. 7 and 14 MHz. Just for fun connected a conventional transformer power supply. And you can see the difference, at once. Will say, one ought prefer using some better noise suppressing SNTs. But did'nt find any suitable on store. I am looking forward to the results of tests here.


  • All the broadband routers I have had provided, have had noisy Switched Mode "Wall Wart" type power supplies as standard.

    They produced far more interference than the routers, which generally tended to be of good construction and had been EMC tested as part of the design process.

    The power supplies always seemed to be an afterthought, and may have been sourced differently from the actual router, depending on local electrical regulations and requirements.

    It is likely that the power supplies have been chosen on the basis of cost, and we all know what happens then.

    Because of their small size, it is usually not possible to add extra filtering, and such internal modifications can also result in safety and insurance issues.

    I tend to look in Charity (Thift) stores or at Car Boot sales for older, cheap electronics, with separate linear power supplies. As it's difficult and expensive to buy new linear power supplies, especially since the audiophiles are willing to pay silly prices for them. The rest of the electronics are useful for parts too.

    On an incidental, but related subject

    A friend of mine, who allows me to host a KiWi at the historic Goonhilly satellite earth station in Cornwall UK, was good enough to resite the KiWi to a new location on the site, in an attempt to resolve some long standing interference problems. However, he couldn't get it to connect properly to the network. He could see it acquired a DHCP address, but he couldn't ping it. He then transferred the KiWi back to the original location, and it still didn't work properly.

    Thinking that somehow the build had become corrupted, we arranged for him to return the KiWi to me, so that I could fix it, upgrade the OS and take the opportunity to check things like the fan (which as expected was stuck solid).

    The actual fault was the Switched Mode Power supply, which seemed to be working correctly, as the output voltage was correct when checked with a multimeter, it would also still happily deliver 5A output.

    However, I noticed that the output voltage dipped just a bit more than I would have expected when on load, but only by about 0.2v.

    I took the power supply apart and discovered that the main high voltage DC storage capacitor (150uF at 450v) had dried out over time, and was completely open circuit. This is why the DC regulation was so poor, and also why the KiWi was having difficulties.

    Replacing the capacitor fixed the problem, and the KiWi was back to working as normal.

    Most of my past KiWi problems have been power supply related in one way or another, so my rule of thumb now is:-

    If your KiWi is misbehaving in any way, check the power supply first, and pay close attention whilst you do it.

    Regards,

    Martin

    studentkraWA2ZKDRadiofan_01
  • The main reasons for using switching power supplies in my country are the EU regulations. For example ErP Directive (2009/125/EC) A Kiwi with a linear power supply would run the risk of not getting a CE mark and would therefore probably not go through European countries customs. That's why it's better if a power supply is not included in the scope of delivery. And is purchased separately according to country-specific local standby and energy saving regulations. However, no one can prohibit me from building and using a linear power supply that meets my expectations. I compared the total power consumption. Using SNT for the mentioned router can only save around 50 euros per year. I owe you a beer for the savings.

    B. regs. Radiofan01

  • If anyone is interested, I use a linear transformer power supply. I have three SDR receivers. One KiwiSDR, one FlydogSDR and one RSP1A running Orange Pi4 lts. I also use a Microtik LTE router, an ethernet switch, and active loop antenna and a downconverter. Depending on user activity, the total electricity consumption is 60 watts per hour.

    Nate_R
  • @studentkra I'm interested, do you have a wiring diagram or PCB? Thank you.

  • @IK8SUT , I don't have a diagram. This is a classic power supply consisting of a step-down transformer, a diode bridge and a 1084 linear stabilizer.

    I used a toroidal transformer from a BBK household amplifier. The secondary windings of this transformer had too much voltage, so I simply wound two windings on top with a thick stranded wire of 12 and 5 volts. I used four diode bridges to power three 5 volt channels and one 12 volt channel. After the diode bridges, I installed electrolytic capacitors and linear stabilizers. One 7812 and three 1084. (I bought some MP1584 high frequency converters. I'll test them for noise later). After linear converters, I used filters from inductors and capacitors. I'll attach some photos.


    F5AFY
  • edited July 25

    No, it is not a mistake, the T3 is a PNP and correctly wired. That sophisticated circuit allowes shorting on output, and after two seconds shorting output's voltage will be totally cut off, in order to avoid damage due to overheating and so on. You must then disconnect mains power for at least 10 seconds, waiting until C1 will be discharged. And power on again -

    Works in a rough environment, particular developped for CB-radio purposes. It's a pity no more available. This one could be used for the internet router, adjusted to appropriate voltage.


  • Getting rid of differential noise is fine, though the power management IC regulators in the Kiwi's BB is so good that this is seldom to never a problem, in my experience. (Try deliberately introducing differential noise and check the Kiwi's response. I haven't done this for a while but I think even 100 mV tends to be pretty invisible)

    But just looking the schematic of this typical linear regulator and wondering ...

    What would happen if we built a PS, either linear or SMPS and then added another series pass device in the 'GND' leg of the output to raise the impedance and reduce CM current? Against typical CM source/load impedances, would it significantly improve the situation in real circumstances?

    It likely might not since typical total CM impedance may be in the 100's of ohms territory and well above the effective resistance of the series pass devices, but it wouldn't be difficult to try.

    Radiofan_01
  • Using center tap secondaries saves 2 diodes ( and condensors) and lowers ground impedance. That kind of PSU worked fine since a couple of years. I even tested it with CB-handheld-radio pushing the transmit button in order to see, what happens. Only a short krack, nothing else. Other parts do react on such a provocational RF level with varying speed of cassette recorder, hum on loudspeaker and something like that.


  • Yes, that was sort of the general direction of what I was wondering about but I suspect it might best be done at a high Z location, perhaps on the mains side. There total current might be 25-50 times lower than at 5V. Two buck converter SMPS that convert down to near 5+VDC from 160/320 VDC would be necessary to keep dissipation low but could that higher Z 'transformation' be helpful in reducing unwanted CM that is sourced/sunk at 100's of ohms? The goal is to raise the CM path Z while still providing low voltage high current (low Z) power to run the HW.

    I don't know, I still haven't thought about it carefully, this may be a silly idea.

    No doubt a better way is to manage CM current paths in the SDR radios to keep from adding in I^2Z drop between the ADC reference and the preamp common. Ground path management goes a long way and even simply going to differential preamplifier input from single ended probably adds enough balance/symmetry to reduce the problem by 20 dB.

  • edited July 26

    My experience is that there is seldom an issue on the secondary side of the power supply. Short cables and rather good filtering in PS and in the Kiwi. However on the primary side it's an disaster sometimes. Bad filtering and since it's connected to the power grid it radiates a lot. I Have built a shielded PS with a (good) switched PS and a really good EMI filter (no photo inside though). All five outputs have a filter each (just simple low voltage filters). I have also build some boxes with EMI filter and placed them at "critical" places in the house where I have router, switches etc. The result is remarkable. 10-15 dB lower noise on my Beverage antennas. No need to replace all SMPS to linear PS. But choose good SMPS even if you have the filter. My opinion is that most people think the issue is "bad" voltage from the PS to the kiwi which causes noise spurs etc. But it's not, in most cases it's radiation from the power grid that causes the problems...




  • This is my point as well. Differential QRN really isn't a problem in any of the many Kiwi's I've seen. The problems seem to always be related to common mode currents in/out and through the Kiwi's.

    At low frequencies it's certainly not "radiation", if by that one means inverse square propagation of the type we generally want to receive. The extremely low radiation resistance of even long lines is simply to low to generate that. Rather, It's near-field and/or currents going where they shouldn't and getting turned into QRN that the Kiwi has to deal with.

    This is paired with the problem that has so many thinking that coax is a shield. It is, but only for TEM mode. For CM or TM current it doesn't help at all.

    Maybe people get tired of my talking about CM but it really is the problem rather than PS noise of the type that seems to get so much attention, IMO.

  • edited September 1

    I had to build my own linear supply, to get the quietest possible power for my Kiwi. It was definitely MUCH better than ANY switching supply I tried. Unfortunately, as others have mentioned, there is a LOT of RFI being generated by now innumerable SMPS devices ALL AROUND US! I live in a neighborhood of mostly single-family houses. My two neighbors are only about 15' away. My Kwi picks up a LOT of noise from the neighborhood! :( My LAN Router doesn't help, although I've put Ferrite RF chokes on the wires going to it. I just have to live with this. :( I've tried unplugging various devices in the house, with minimal improvement. (Sensitivity min is set to -120dB)


Sign In or Register to comment.