Severe IMD but no OV indication

rrobinetrrobinet
edited September 2018 in Problems and Issues
The Kiwi at KPH (http://kphsdr.com:8072/) fed by a Marconi T through a Nooelec broadcast band blocking filter. Nevertheless, there is severe IMD across the whole spectrum even though no signal is greater than -30 dB and the OV never lights up. There are no active components in the transmission chain ahead of the Kiwi so I'm at a loss as how to debug it.

Comments

  • G8JNJG8JNJ
    Hi,

    I think it may not be a typical ADC overload but it's simply a case of a lot of regularly spaced BC signals that are close to the ADC threshold causing intermodulation products.

    SDR's don't behave in quite the same way as 'analogue' receivers WRT to IMD products and you do get some weird effects as you get near the clipping theshold.

    Have you got a sensitive broadband analogue power meter (50 Ohm resistor with a signal diode feeding an analogue voltmeter) that you can use to measure the total RF input power present at the antenna. You can use an un-calibrated detector and measure the level by connecting a calibrated signal generator and turning up the output level to get the same reading.

    The usual trick with 'analogue' receivers is to insert something like a 3dB attenuator at various points in the receive chain and see if the IMD falls by 3dB or more. If it's only 3dB the IMD source is before that point, if it's more than 3dB it's after that point (as IMD products fall away at a faster rate than 'genuine' signals'.

    As you have got such a huge MF antenna connected to the KiWi I think you probably need more BC band attenuation than the Nooelec filter is currently providing. Try an attenuator to see how much of a further reduction you need to cure the problem.

    You could try stacking Nooelec filters or something like this equaliser circuit (change the attenuator resistor values if required) which provides more attenuation at the LF end (where you don't need as much gain because the natural noise floor defines the overall S/N ratio)



    Other comments are in this thread http://forum.kiwisdr.com/discussion/comment/4491

    Incidentally one other possibility is that electrostatic charge on the antenna may be causing the KiWi input protection diodes to start to conduct (which could be another source of IMD) if you have such a huge antenna connected I'd also think about some additional RF / ESD protection ahead of the KiWi in order to protect it from lightning / static surges.

    Be sure to let us all know what you find.

    Regards,

    Martin - G8JNJ
  • jksjks
    You are seeing IMD at 10 kHz intervals because there are simply too many stations at -30 dBm (~ S9+40) at the low end of the AM BCB. Might be IMD in the preamp even if the ADC isn't seeing overflows. That Nooelec BPF is not as good as you might think. Looking at their published response curve it is only down about -12 dB at 580 (http://www.nooelec.com/store/downloads/dl/file/id/78/product/283/distill_am_dc_to_5mhz_sweep.png). Have a look at the screenshots below.

    Also, who the heck is on 580? The only listing I found was for KMJ Fresno, 50 kW. At 10 PM local they were fading up above -20 dBm. Almost S9+60! Yes, just now I heard a KMJ ident at the top of the hour -- incredible!

    Do this experiment: Note the general average values of the lower AM stations on the Kiwi spectrum. Get yourself an ordinary lowish-value pot and put it across the antenna input. Zoom the Kiwi into an IMD product e.g. 460 kHz. Slowly turn the pot down until there is an exponential drop in the IMD and it disappears into the noise floor. Go back to the BCB and note how much the average values have dropped (5 - 15 dB I would guess). Get a better filter that can attenuate by this additional amount.

    And it didn't help that a Navtex popped up on 518 kHz at S9+60. I assume that's coming from NMC in Bolinas. Look how much additional IMD in the NDB band that causes! But even after it leaves the 10 kHz IMD remains (see screenshot).

    image

    image

    Attachments:
    https://forum.kiwisdr.com/uploads/Uploader/67/4320439cf12dd37709a382f55881fb.png
    https://forum.kiwisdr.com/uploads/Uploader/a2/8b1fee477059807a62e09ace26cede.png
  • G8JNJG8JNJ
    Another thought is that although unlikely, it could possibly be the very high levels of BC station RF present at the antenna that may be saturating the small SMD inductors in the Nooelec filter causing IMD.

    Trying the attenuator trick before and after the filter to see if the IMD levels drop to the same level regardless of the attenuator position, will test for this possibility.

    Regards,

    Martin - G8JNJ
  • jksjks
    edited September 2018
    Martin brings up an excellent point about the TVS diodes. Do you know if there is a DC path from the Marconi to the Kiwi input? If there is then a DC offset from the antenna could be forward biasing the diodes prematurely. It would be interesting to add a couple of parallel series caps of different values ahead of the Kiwi input and see what happens.
  • jksjks
    edited September 2018
    I just looked at the VLF reception of this Kiwi for the first time. Completely unbelievable for a wire antenna. There's DHO38 Germany 23.4 kHz plain as day. NWC Australia 19.8 of course, TFK Iceland on 37.5 and NLK WA booming in at S9+30!
  • ChristophChristoph
    The Kiwi at KPH (http://kphsdr.com:8072/) is excellent for receiving NDBs from the Pacific region, even with the IMD products. Many NDBs from Australia/NZL can be regularly heard just before local sunrise. I will attach some logs later.
  • rrobinetrrobinet
    Thanks for the suggestions.

    In the week up to last Saturday I had a custom band reject filter installed which was so effective that I was getting complaints about the lost AM reception. I need to confirm it, but I think the IM was still there with the custom filter, so that suggests that the source precedes the filter. A colleague at KPH has suggested that the input section of an active distribution amp on the other leg of the Minicircuits 2-way may be overloaded and the base-emitter junctions are acting as rectifiers.
    To diagnose and correct this I will go out to KPH next Saturday and experiment with attenuators as you suggest, the custom filter, and disconnecting the unused active distribution amp.

    This is the custom filter response


    I've attached the design documents by Roget KK6EEW which was built for KPH by Tom KJ6MKI.
  • jksjks
    edited September 2018
    So there are active devices upstream of the Kiwi but just not on the Kiwi's path? (i.e. on the other leg of the 2:1 splitter?)
  • rrobinetrrobinet
    edited September 2018
    I seem to have double-entered my post and can't figure out how to delete the second. [fixed, -jks]
    That LF/MF Kiwi was donated by JKS
    There are 4 more Kiwis at that site connected to a TCI-530 3-30 Mhz omni, 3 of which you can access at kphsdr.com:8073
    The 4th Kiwi is in 8 channel mode and my kiwiwspr.sh is spotting all HF bands. It is the #2 spotter in North America after K9AN.
    I have a 40 dB GPS antenna and all Kiwis are locked on 10-12 satellites; we have verified that the Kiwis are (most of the time) within 1 Hz of accurate at 14 Mhz. Perhaps further improvements are possible, but instead we will be upgrading to an external Bodnar GPSDO so we will be accurate to 10e-9 and thus ready for John's millihertz FFT feature.
  • rrobinetrrobinet
    yes, I just learned of those devices this evening
  • ChristophChristoph
    Below are some outstanding NDB logs made @KPH (:8072).

    Although there is IMD on some channels which makes it hard to copy 260-NF Norfolk Island the main limitation for weak signal work on LF is thunderstorm noise.

    Christoph

    Format:
    date, and the following lines UTC frequency, call, (location, ), state, country.


    8/29/2018 13:28 272 LHI Lord Howe Island, NW
    8/29/2018 13:44 260 NF, NFK
    8/29/2018 13:19 305 GTH Griffith, NW
    8/29/2018 13:37 316 MAJ Majuro Atoll (Dalap Island), MHL
    8/29/2018 13:26 343 ML MTS
    8/29/2018 13:42 348 HN Honiara, SLM
    8/29/2018 13:25 360 OX VOI
    8/29/2018 13:32 366 PNI FSM
    8/29/2018 13:24 395 PMQ Port MacQuarie, NW
    8/29/2018 13:30 400 MDY 'Midway' Gooneyville (Sand Island), MDW
    8/29/2018 13:40 403 TUT, SMA

    8/30/2018 13:28 403 TUT, SMA
    8/30/2018 13:29 360 OX, VOI
    8/30/2018 13:30 343 ML, MTS
    8/30/2018 13:32 238 KT Kaitaia, NZL
    8/30/2018 13:34 377.5 MO, OCE
    8/30/2018 13:35 366 PNI, FSM
    8/30/2018 13:40 316 MAJ, MHL
    8/30/2018 13:41 327 VYI HI
    8/30/2018 13:41 356 TG Tauranga, NZL
    8/30/2018 13:42 364 CS Cairns, QD, AUS
    8/30/2018 13:45 395 PMQ, Port MacQuarie, NW, AUS
    8/30/2018 13:52 320 AI, CKS
    8/30/2018 13:55 400 MDY, MDW (still there @14:02)

    9/6/2018
    13:23 320 AI CKS
    13:23 318 BR BRU good
    13:25 302 WYD Wynyard, TA AUS
    13:25 299 CWR Cowra, NW AUS
    13:27 278 CG Coolangatta, QD AUS
    13:28 293 COM Cooma, NW AUS
    13:34 505 PI Yuzhno-Sakhalinsk/Khomutovo, RSE
    13:36 355 NI Naurau, NRU
    13:37 359 AMB Amberley, QD
    13:39 395 PMQ Port MacQuarie, NW AUS
    13:39 395 MER Merimbula, NW AUS
    13:43 348 HN SLM
    13:48 238 KT NZL
    13:52 403 TUT SMA
    13:55 380 SU Sunshine Coast, QD AUS
    14:02 393 UKS Kosrae (Kosrae Island), FSM
    14:08 272 LHI Lord Howe Island, NW AUS

    9/9/2018
    13:25 366 SF Springfield, NZL
    13:26 346 TG Tauranga, NZL
    13:28 326 WR Whangarei, NZL
    13:32 278 CG Coolangatta, QD
    13:39 318 BR Brunei, BRU
    13:43 395 PMQ Port MacQuarie, NW
    13:46 376 NP Napuka, TUA
    13:47 377 ROM Roma, QD
    13:53 400 MDY MDW
    14:06 343 ML MTS

    9/10/2018
    12:47 437 OG Okha, RSE
    13:14 505 SL Yuzhno-Sakhalinsk/Khomutovo, RSE
    12:51 377 ROM Roma, QD
    12:53 343 ML MTS
    12:55 320 AI CKS
    13:03 360 OX VOI
    13:18 376 NP Napuka, TUA
    13:20 393 UKS FSM
    13:20 395 PMQ Port MacQuarie, NW AUS
    13:23 364 CS Cairns, QD AUS
    13:24 355 NI Nauru, NRU
    13:29 346 TG Tauranga, NZL
    13:36 400 MDY MDW
    13:36 403 TUT SMA
    13:37 278 CG Coolangatta, QD AUS
    13:39 305 GTH Griffith, NW AUS
    14:02 348 HN SLM
    14:05 327 VYI HI
    14:07 238 KT Kaitaia, NZL
    14:09 366 PNI FSM

    9/12/2018
    11:03 403 TUT SMA
    11:05 278 CG QD AUS

    9/13/2018
    13:35 299 CWR Cowra, NW AUS
    13:36 302 BUD Bundaberg, QD AUS
    13:44 403 TUT SMA
    13:47 293 COM Cooma, NW AUS
    13:50 395 PMQ
    13:50 376 NP
    13:51 366 PNI
    13:51 360 OX
    13:52 359 AMB Amberley, QD AUS
    13:53 348 SN
    13:54 338 MA Mount Isa, QD AUS
    13:54 324 EML Emerald, QD AUS
    13:55 316 MAJ MHL
    13:56 305 GTH
    13:57 308 MK
    13:58 293 CDU Ceduna, SA AUS
    13:59 269 CV Charleville, QD AUS
    14:03 355 NI NRU
    14:04 377 ROM Roma, QD AUS
    14:05 392 LHR Lockhart River, QD AUS
  • WA2ZKDWA2ZKD
    I solved a lot of IMD issues here on my public kiwi with the AMBC notch filters I posted a long time ago. The Q is high enough to allow reception of most of the rest of the AMBC band and of course there's no out of band IMD from combinations of the notched heavy hitters. http://rx.jimlill.com:8073
  • rrobinetrrobinet
    Did you ever see IMD but no OV indicator? That is my problem at KPH. On Saturday I hope to verify my suspicion that the IMD is being produced ahead of the blocking filter already in place. Only yesterday did I learn there is an active distribution amp in the other leg of a 2-way splitter, so my theory is that amp is horribly overloaded and the base-emiiters are acting as diodes and the resulting IMD is leaking back through the 2 way splitter.
  • WA2ZKDWA2ZKD
    I have never seen evidence of my active loop LNA being pickled. I mention the notches as that is a way to get protection for any active device in the path without wiping out a whole range,
  • VK3KHZVK3KHZ
    edited January 2021
    I also think that individual notches ahead of any splitting regime is the way to go. As a keen MW DX'er, I want to preserve the MW sensitivity, but knock out some big hitters in my neighbourhood...
    I use simple single series L+C tuned circuits on each big MW hitter.
    Each Series LC has enough Q and will take out 10-15dB of carrier and side-bands, for more attenuation, just duplicate circuit values and tap them together onto the main RX buss. Use a T-Piece on RX input....
    As far as component choices and values are concerned, for each notch I use fixed inductor, in series with a fixed cap, with parallel trimmer adjusted to resonance. See below for EXCEL calculator
    73' Paul VK3KHZ

    http://sdr-amradioantennas.com:8071/?f=7125.00amz8

    Excel calculator:- http://www.amradioantennas.com/Temp/MW_Notch_Calculator_VK3KHZ.xlsx
    image

    Attachments:
    https://forum.kiwisdr.com/uploads/Uploader/84/d63dbfbf9d20880d2f56b14f7ef2fe.jpg
  • G8JNJG8JNJ
    edited September 2018
    Hi Paul,

    You can also trade the overall notch depth against the notch bandwidth by varying the L/C ratio.

    In most cases you don't need to have more than 10 to 20dB notch depth in order to bring the worst offenders down to an acceptable level.

    Here's a couple of filters I made using SMD parts. If you remove the 5.6R resistors and are using components with a good value of Q (L >40)
    you can obtain >20dB notch depth.




    This is the frequency response with the resistors fitted.



    Regards,

    Martin -G8JNJ
    Powernumpty
  • WA2ZKDWA2ZKD
    I use notches similar to what Martin shows. I also have 1:4 and 4:1 xfrmrs in/out of the filter assembly which results in a somewhat higher Q
  • rrobinetrrobinet
    Thanks for your suggestions, but KPH may be an outlier in that the IMD appears to be generated ahead of the the filters I have put before the Kiwi. Even when I inserted a 5th order elliptic filter of this design with >65 dB of predicted and measured BCB rejection, the IMD was still present on signals in the passbands. This filter was so effective that I received complaints that the local AM stations which previously had severely overloaded the Kiwi could no longer be heard. So my current theory is that an active distribution amplified upstream of the Kiwi but on the other leg of a passive splitter is generating the IMD which is leaking through the splitter into the feed to the Kiwi. I will test that theory on Saturday afternoon when I return to KPH. You can listen for yourselves at http://kphsdr:8072/ if you want to see/hear the LF/MF spectrum before and after my experiments.

    The measured filter response:



    The predicted filter response:
  • WA2ZKDWA2ZKD
    Link to my filter info http://forum.kiwisdr.com/discussion/593/am-overload-solved#latest
  • G8JNJG8JNJ
    "I received complaints that the local AM stations which previously had severely overloaded the Kiwi could no longer be heard."

    You could always put a variable resistor between the filter input and output connections so that you could vary the depth of the notch.

    The match would still be OK if you did this.

    e.g approximate resistor values

    240R = 10dB
    1K = 20dB
    3.3K = 30dB
    9.6K = 40dB

    Any chance of showing the Nooelec filter schematic ?

    Regards,

    Martin - G8JNJ
  • G8JNJG8JNJ
    "I also have 1:4 and 4:1 xfrmrs in/out of the filter assembly which results in a somewhat higher Q"

    Hi Jim,

    I tried this method, but I found that if I got the L/C ratio correct and used moderately good inductors, I didn't need to use any transformers as I could achieve very similar amplitude / frequency response curves.

    I used RFSIM99 (free)

    https://www.electroschematics.com/835/rfsim99-download/

    to compare component values using an inductor Q of 40 and capacitor Q of 5,000.

    The modeled curves matched the measured ones very closely.

    Regards,

    Martin - G8JNJ
  • rrobinetrrobinet
    Currently the Nooelec is inserted before the Kiwi and the max AM signal is KNBR 680 at -30 dBm, which correlates well with the -20 dBm KNBR I get with no filter and 20 dB of attenuation in front of the Kiwi.
    So with the Nooelec in circuit I am well below OV. I can't find a schematic, but from the sweep and seeing only two inductors inside the filter, it is much simpler than the above design.
    Since the Nooelec has a lot more 630M attenuation than I would like, I may follow your suggestion and put an attenuator around the custom filter.
    But I don't think filters are the primary solution to the current IMD problem.

    This is the Nooelec filter sweep from their web site:

  • WA2ZKDWA2ZKD
    Martin.... the xfmrs were needed when I had six notches going, not so much with 3 or less
  • VK3KHZVK3KHZ
    These wide filters are no good for AM DX, so I still stick with my individual notches as per my calculator...

    Amplifiers are notorious for 2nd / 3rd order IMD ( Distortions ), so I don't run any.

    Splitting devices with transformers are a No-No, that is; if you want to maintain a flat frequency response from 1KHz -> 30MHz as I do.

    I use WYE splitters, with Zo = 50 Ohms, this simple formulae to calculate the resistors required - > R=Zo x (N-1)/(N+1), where N is the number of ports required...

    Sure this type of splitter results in splitting loss, but it maintains a flat frequency response right across the full frequency range and the gain deficit can be made up ( calibrated ) by adjustment of the Kiwi S-meter calibration (dB) & Waterfall calibration (dB) on the ADMIN Config Page.

    The WYE Splitter can be used on the SDR inputs, because it is completely un-necessary to have receiver to receiver isolation ( as there is no local oscillator to cause
    inter-receiver hetrodynes / beats & birdies )

    73'
    Paul
    VK3KHZ
    www.amradioantennas.com
  • WA2ZKDWA2ZKD
    Wye spiltters can be subject to filters that are applied post splitter affect the other leg
  • G8JNJG8JNJ
    Hi Paul,

    "These wide filters are no good for AM DX, so I still stick with my individual notches as per my calculator..."

    The point I was trying to make is that by varying the L/C ratio you can make the notches as wide or as narrow as you wish (within limits).

    I gave an example that would cover the whole AM BC range (like the Nooelec).

    To try and further clarify the point, here are two MF notches with different L/C ratios.



    Note the difference in bandwidths and maximum depth.

    Generally speaking the larger the value of C the wider the notch bandwidth and the deeper the notch depth. However there are limits to the maximum value of L and the minimum value of C that can be used in order to provide a reasonable notch. As Jim has also indicated, if you want to use a lot of narrow notches with close frequency spacing's, then you may have to use transformers in order to provide a different characteristic impedance for the through path.

    Maybe your calculator could be modified to also give an indication of the -3dB points (BW) and maximum notch depth at the centre frequency ?

    "Amplifiers are notorious for 2nd / 3rd order IMD ( Distortions ), so I don't run any"

    agreed they can be problematic and some are worse than others. I use this design as a distribution amplifier, which has quite good characteristics.



    Measured performance

    Gain = 13dB
    Frequency range 19KHz to 40MHz @ -1dB
    Sat O/P power = +28dBm
    OIP2 = +76dBm
    OIP3 = +45dBm


    Regards,

    Martin - G8JNJ
Sign In or Register to comment.