Radio W4KAZ

Thanks for stopping by the virtual KazShack. Feel free to comment - I often approve them.

W4KAZ Softrock Based CW Skimmer Station – 02) Hardware, Technical Notes


The skimmer station is built around the SoftRock SDR RX kits by Tony Parks, KB9YIG.  His basic concept has been refined and has evolved over the 2007-2011 period.  The simplicity of the hardware and the circuit is fascinating.  The current kit is the Softrock Lite II combined receiver.

In the W4KAZ skimmer implementation, several earlier versions of the SoftRock kit are also used.  Performance of these kits RX sensitivity using ‘regular’ antennas seems on par with the primary rigs(Elecraft K2, FT-920,TS-590s) on 160m thru 20m.  They are somewhat less sensitive on 15m and 10m.  Their selectivity can be compromised by overload from strong signals, which is seen as a “mirror image” on the opposite side of the center frequency.  The best of the skimmers here have shown themselves to begin showing mirror images whenever the SNR of the skimmer approches 45db.

A  ‘smoke test’ over 2012 ARRL Field day resulted in 14000 station spots on two bands(80m&40m), with no system crash, running two skimmers per dual core CPU.   Quad core cpu’s will be much more able to handle multiple CW skimmer instances.

Preamps and bandpass filtering:

The W3LPL RX band pass filters are used on all bands, and have eliminated any decoding of out-of-band signals.  These were constructed here with common NP0 ceramic caps[rather than recommended silver-mica’s], hand wound inductors, and tuned by adjusting for minimum SWR on the passband using an MFJ model 259 antenna analyzer.  If there is a fall off in performance due to the ceramic caps, it is not an issue in this usage.  The filters as built are sufficient to eliminate any detectable out-of-band signals on the softrocks.

The preamps are all of the W7IUV design. [20150503 Note: a kit version W7IUV amp is available from] The power supply for the system is a bit lower voltage than recommended for these, using 12.8v where 13.5v is recommended.  No issues from that.  The pre-amp may be too much in this usage – overload of the softrocks and mirror image generation is an issue/trade-off(see Note#2).  Heat sink the 2N5109,  it WILL run hot!!  One of these has been running 24/7/365 since late 2012, no problems.  Interested now in the IK4AUY design.


The small active verticals seem to be a popular solution for a lot of skimmer installations, in particular the short vertical/preamp from Clifton Labs.  A good product,  a very good value, and certainly easy to site-locate  –  but that $olution, de$pite its bargain price, retail$ for more ca$h than is inve$ted in everything el$e combined.

Not. Happening. Here.

A small vertical is not my personal first choice RX solution anyway, so little heartache(i.e., “none”) is generated over that.  (ed, 20150406, w4kaz, but in hindsight it may well be my ‘best’ choice, as a tradeoff .  Too much gain is counterproductive for softrock based CW skimmers.)

Instead, the K9AY array was be pressed into regular service for the 160/80/40 bands.  The antenna input from the K9AY array is routed:

  1. through an ICE antenna limiter
  2. into W7IUV Pre-amp
  3. through a three way splitter(a cascaded magic-T splitter with 40m on -3db port, 80m and 160m on -6db ports)
  4. W3LPL design RX bandpass filter
  5. to the SoftRock box for 160m/80m/40m
  6. to each respective sound card/computer.

For 20m/15m/10m the initial antenna will be start off with a coaxial dipole mounted as a sloper.  A W7IUV pre-amp will be used if it improves performance.  A magic-t will probably be used as a combiner at the base of the antennas when(if) 15m and 10m are added.  There will then be a single coax run return to the shack.  Based on the results with the 20m skimmer, that may revise itself to some other solution.[Revision, 20150406,w4kaz- The antenna for 20m/15m/10m may wind up being a simple vertical for each band, or  a multi-band vertical fed with a single coax line.  The magic-T type splitter will be used only if the W3LPL RX band pass filters on the front end of the softrocks prove insufficient for signal separation.  Or if there is too much interference generated back up the line by the softrocks themselves.  Hopefully the W7IUV preamps will block some of that.]

CPU’s – Being Revised 20150502

The SoftRocks currently feed their I/Q signals into a set of two Dell cpu’s.   40m is into a win XP box running a single skimmer instance.   160m/80m/20m/15m/10m are all being run on the second win8.1 computer.  The major difference being the second cpu is a core 2 quad  box.  If it were a more modern quad core, or had an AMD 8 core processor a single unit would be sufficient.  For six skimmer instances I suspect the 8 core processor would be ideal.  Into the current core 2 quad(2 cores with 4 threads) with five skimmers at 96khz bandwidth the CPU load is about 50% on dead bands.  More speculation below

skimmer03 – dell optiplex 360 minitower(pentium dual core e2200, Win XP Pro Sp3+)

Soundcard/band pairs

  1. onboard soundmax       48/96/192khz(@96khz)…….[Currently unused, may be used to run 40m at 192khz]
    MAudio delta 66            48/96(@96khz)………………..[40m, 7.055015 center, occasional use at 192khz bandwidth]
skimmer04 – Dell OptiPlex 960 SMT, 2.66 GHz Core 2 Quad, Win 8.1

Soundcard/band pairs

  1. Asus Xonar DS(PCI) @96khz …………………………[160m, 1.842950 center]
  2. Diamond Xtreme XD71HDU (USB)………………..[80m,3.533950 center]
  3. onboard soundmax @96khz …………………………..[20m, 14.046950 center, occasional use at 192khz bandwidth] [?????????????]
  4. Soundblaster Audigy 2     @96khz ……………………[15m, 21.044xxx] [?????????????]
  5. Diamond Xtreme XD71HD(PCI-E) 48/96/192khz(@96khz) …..[10m,28.090000 center]
General Discussion/Lessons Learned/”fiddley bits”  Being Revised 20150502

Issue 1] Load Balancing.   The CPU with the 4 threads was an experiment in a couple of areas.  First was to see if were able to handle more skimmer instances than the ‘skimmer03’ cpu that could run 3@96Khz.  Answer: Yes, and quite handily-but….192Khz might be pushing it past its capabilities.  It will probably handle one 192Khz instance and four others at 96Khz.  If it can, I’d rather have it run at least two of the skimmer instances at 192Khz.  Probably set it up for 2015 WPX CW. Load balancing issue.

Issue 2] The sound card issue #1.  To run multiple skimmers requires multiple sound card installs(or on the number of stereo inputs for cards with multiple inputs like the M-Audio 44 or 66).  This had previously been a problem with the Xonar DG/DS/DX line of cards.  Previously had been unable to get any of the Xonar cards to co-exist with any of the other outboard cards, though they did allow the integrated motherboard sound devices to function.  With a copy of windows 8 available, decided to test if if the Xonars would be better behaved under a current OS(All of the other skimmer cpu’s are/were win XP).  Answer: Yes, but…..The Asus cards seem to require being declared the default device at the time the skimmer session using it is started, otherwise no joy.  Since the other devices were not conflicting, I just set the ASUS card as the default.  Note: Changing the default sound record device changes the order of the devices in the windows sound controls.  Each skimmer’s sound card selection setting needs to be checked after devices are added to the system or the default sound device is changed.  This includes plugging in a USB card.  The USB device is not always being detected on boot.

Issue 3] Multiple skimmer instances, Win8.1, and “Quad” core.  The 960 box is really an early dual core system that imitates being a quad core via hyperthreading.  With windows 8.1, it idles at about 50% CPU usage with five skimmer sessions running.  Also, each skimmer is set to scan only the bottom 90khz of each cw band via skimmers band plan assignment(bandplan.ini).  Each skimmer is also further limited by setting the number of skimmer decoders to either 200 or 250.   With 40m on its own seperate box, the natural flow of activity from high bands to low help balance the load.  It should also be noted that the bandwidth being scanned does effect the amount of CPU load, with CPU loads for 192Khz being abut twice the load of a 96Khz bandwidth.  A modern i5 or i7 Intel processor, a xeon box,  or one of the newer 6 or 8 core AMD CPU chips would be ideal.  Still experimenting with the CPU loading combinations on this box as of 20150502.

Issue 4]  The sound card issue #2.  Serendipitously discovered(circa 2015-03) a ‘new’ sound card that showed promise based on manufacturer specs, the Diamond xTreme XD71HD series.  Released in 2013(after the initial skimmer builds went ‘live’ at W4KAZ), this newer HD version(s) of the Diamond XD71 card brought with it several useful features.  It is available in PCI or USB form.  The PCI version is half-height, meaning it could be used in any of the available skimmer CPU’s already in the KAZshack(i.e., minitower or small form factor).    They come with ADC chips capable of 24bit 192Khz input, and are highly rated by audio reviewers.  And the price is very reasonable for a 192Khz capable device, either version available for under $50USD(sometimes).  BAM!  Both versions work very well at any bandwidth, and both can coexist with each other as well as a Soundblaster and Xonar.

Issue 5]  The apparent memory leaks.  The skimmers being run under win XP on any of the CPU’s require restart/reboot of the host platforms on about a weekly basis.  The skimmers occasionally lock-up/barf/croak if these systems are not restarted regularly.  I suspect some sort of memory leak in Win XP rather than cw skimmer itself.  So far the only system crash on the Win8.1 box was caused by the crash of the 40m skimmer running on the other system..  That cascaded to the crash of aggregator on the Win8.1 box.  So…regular restarts required.  check.

Issue 6]  The Si570 Oscillator for 10m.  The si570 oscillator project for the 10m skimmer worked out really well.  The images on 10m seem to occur at the 40-to-45db snr ratio I generally see on the lower bands, so currently the 10m skimmer is performing better than either 20m or 15m in that regard.  Selecting the oscillator frequency for 10m initially to center 10m on 28.045 brought attention to the interaction of the softrocks.  That choice caused quite a bit of interference with 15m, and a moderate amount of interference on 20m.  These three are installed in the same box with inadequate shielding.  so the oscillators were interacting and causing spurs.  Moving the 10m center up to 28.090 eliminated all of the interaction spurs from 15m and 20m, and reduced the spurs on 10m to a few on either side of center.  So 10m is using 28.090 as its center, the skimmer and sound card at 24bit 192khz scan rates, and decoding only 28.00 through 28.090.   Also a good way to eliminate a lot of the mirror images.  😉

Issue 7]  The sound card issue, #3.  The USB sound card is not always(ever??) being detected when the machine is booted(cold boot only??)  Temp resolution….unplug and replug the USB sound card after reboot.

Issue 8] Mirror Images Summary.  Mitigating the mirror images is a matter of deciding what level of false spots you are willing to live with.  Since the skimmer station is primarily intended to push spots up to the RBN, I’d like to minimize the amount of bad spots sent.  Judgement calls required.  !20150503, need separate post on mirror images!A subject worthy of its own seperate novella…..

Note1] 20150502 Using the W3LPL band pass filters as splitters for antenna sharing is also working for 10m/15m.

Note2]  20150502 The use of band pass filters is necessary to keep the softrocks from decoding signals from harmonically related bands.  The pre-amps follow the filters for the same reason.  Depending on the signal strength from the antennas used, it may also be necessary to add attenuation before the preamp or be able to switch the preamp out of the circuit.  Signals over about 45db SNR seem to cause skimmer mirror image problems with the Softrock Lite’s.  So it becomes a trade-off between amplification for weak signal copy or suppression of the mirror image spots.


[Revised 20150502, This info below is Obsolete.  It remains only for reference, as it was the original set-up circa 2012.]


The SoftRocks feed their I/Q signals into a set of two Dell Optiplex 360’s, and one Dell Optiplex 745.  The onboard Soundmax cards in the 360 boxes are capable of 192khz bandwidth, but will normally run at 96khz.  The onboard Soundmax in the 745 box is capable of 48khz.

The CPU platforms will initially be paired up SoftRocks as follows:

skimmer01 – dell optiplex 745sff (Pentium D Dual Core, Win XP Pro SP3+)
  1. onboard soundmax     @48khz ………………………….[80m,3.533950 center] [?????????????]
  2. Asus Xonar DX(PCI-E) 48/96/192khz(@96khz) …..[10m,28.087780 center]
skimmer02 – dell optiplex 360SDT (pentium dual core e2200, Win Vista Bidness)
  1. onboard soundmax    48/96/192khz(@96khz)…….[15m,  21.044xxx, occasional use at 192khz bandwidth]
    Asus Xonar DG(PCI)  @96khz   …………………………[20m,  14.046950 center]
skimmer03 – dell optiplex 360 minitower(pentium dual core e2200, Win XP Pro Sp3+)
  1. onboard soundmax       48/96/192khz(@96khz)…….[40m, 7.055015 center, occasional use at 192khz bandwidth]
    MAudio delta 66            48/96(@96khz)………………..[160m, 1.842950 center]

The Optiplex 360 minitower will likely be the 24/7/365 CPU platform because its larger case keeps both its power supply and its CPU running cooler, it is running Win XP, and its onboard sound does an acceptable job at the 192khz bandwidth.  The Xonar DS has higher fidelity, so I’m hoping that will help with the normally higher noise on 160m.

The other combinations are subject to change depending on band conditions and system performance.  If the 10m SoftRock solution actually develops into real hardware, it will like remain the band relegated to the 48Khz bandwidth until a better CPU makes itself available.



Leave a Reply

You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>