Loading ...
Sorry, an error occurred while loading the content.

Re: Decoding performance

Expand Messages
  • rt_tclay
    ... with Similar results here on the CQWW file. Running with graphics window shrunk down. XP Pro SP2. 1) old 1 GHz AMD Athlon. 1.25 GB RAM maxed out 100% CPU,
    Message 1 of 19 , Apr 25, 2008
    • 0 Attachment
      --- In dxatlas_group@yahoogroups.com, Pete Smith <n4zr@...> wrote:
      >
      > At 05:19 PM 4/23/2008, bill_w4zv wrote:
      > >I then tried the CQWW file below and got the following results,
      with


      Similar results here on the CQWW file. Running with graphics window
      shrunk down. XP Pro SP2.

      1) old 1 GHz AMD Athlon. 1.25 GB RAM

      maxed out 100% CPU, decoded only 69 calls

      2) 1.8 GHz AMD Sempron ("3100+" model I think), 1 GB RAM

      max CPU 70-80%, 155 calls


      Given that CPU never hit 100% on the second system, I would say that
      throwing more CPU at it won't give more calls decoded. Of course if
      it ran for longer, the number of decoders might increase enough to
      hit 100%. I think to test performance on faster CPU's you need a
      longer sample.

      Tor
      N4OGW
    • bill_w4zv
      Thanks Pete and Tor! Yes Tor, I simply had not discovered how to increase the window size on the display. How do you feel about Skimmer s ability to decode
      Message 2 of 19 , Apr 26, 2008
      • 0 Attachment
        Thanks Pete and Tor! Yes Tor, I simply had not discovered how to
        increase the window size on the display.

        How do you feel about Skimmer's ability to decode weaker signals in
        QSB and QRN? I'm somewhat unimpressed so far but it could be I'm not
        doing something right. Does the DSP filter width affect all decoders
        or just the one displayed with the green bar? Based on what I've seen
        so far, Skimmer would be great for identifying strong stations CQ-ing
        or in a pileup but it doesn't work very well for weaker signals, which
        would likely be the very ones you want to know about. But this could
        be due to my not setting Skimmer up correctly or due to my weak
        computer. BTW, I noticed 3GHz P4's on eBay go for <$150 so I may pick
        up one of these to solve the latter issue.

        73, Bill W4ZV


        --- In dxatlas_group@yahoogroups.com, Pete Smith <n4zr@...> wrote:
        >
        > At 05:19 PM 4/23/2008, bill_w4zv wrote:
        > >I then tried the CQWW file below and got the following results, with
        > >CPU utilization hitting 100% at times in all cases:
        > >
        >
        >http://cid-e39b59612b664f63.skydrive.live.com/self.aspx/Public/CQWWCW_2005_1_of_42.wav
        > >
        > >33 calls, 7009.0>7020.10, everything running on computer.
        > >33 calls, same QRG range, Internet connection and browser off.
        > >40 calls, 7007.0>7021.9, AVG anti-virus off.
        > >58 calls, 7002.0>7024.6, Skimmer display minimized.
        > >
        > >I could probably do a little better by shutting down more background
        > >processes but I'm curious what someone with a 2-3 GHz CPU gets when
        > >running the same CQWW file above?
        >
        >
        > Your wish is my command, sir ;^)
        >
        > With a 2.2 GHz Celeron, 768 MB DDR RAM (a VERY basic Dell 2400, four
        years
        > old, With XP home SP2) I got the following results off the same file.
        >
        > Note that when playing back recordings the sound card is not involved;
        > Skimmer is processing the I/Q file directly, so settings like playback
        > sampling rate have no effect:
        >
        > Decoders Other variables Calls Decoded CPU
        peak
        > utilization
        >
        > Adaptive Not minimized 156 96%
        > 250 "" 131 75%
        > 300 "" 152 81%
        > 350 "" 154 86%
        > 400 "" 153 100%
        > Adaptive Telnet server off 154 100%
        > Adaptive Skimmer minimized 154 ?
        >
        > Interesting, no? I would have to other programs minimized or in the
        > background do not affect decoding performance; one of the critical
        > questions I hope to answer in coming months is whether the reverse
        is true,
        > since we're talking about using Skimmer to feed spots to a contest
        logging
        > program, which has to be able to respond quickly, send CW, etc. My
        > personal tests with N1MM logger, using a Winkey for CW, have been very
        > encouraging; I suspect that MM simply takes precedence when it needs
        to,
        > and that the Winkey avoids any issues with making good CW.
        >
        > 73, Pete N4ZR
        >
      • rt_tclay
        ... not ... decoders ... seen I don t know for sure, but I think the program just uses the green bar setting for the audio output filter width. I can copy
        Message 3 of 19 , Apr 26, 2008
        • 0 Attachment
          --- In dxatlas_group@yahoogroups.com, "bill_w4zv" <w0zv@...> wrote:
          >
          > Thanks Pete and Tor! Yes Tor, I simply had not discovered how to
          > increase the window size on the display.
          >
          > How do you feel about Skimmer's ability to decode weaker signals in
          > QSB and QRN? I'm somewhat unimpressed so far but it could be I'm
          not
          > doing something right. Does the DSP filter width affect all
          decoders
          > or just the one displayed with the green bar? Based on what I've
          seen

          I don't know for sure, but I think the program just uses the green
          bar setting for the audio output filter width.

          I can copy better than Skimmer on weak signals. If you watch it
          trying to copy a weak signal, it usually gets a few letters correct
          each time, but has a hard time putting together the full correct
          callsign. That's where the brain obviously has an advantage, when you
          are copying you can say definitely if you copied a particular letter
          and mark that mentally as "good copy".

          If on your computer Skimmer is reaching 100% cpu, then yes, it is
          having to reduce the number of decoders. I don't know the algorithm
          used to do that, but yes, it might do that based signal strength. Try
          reducing the bandwidth. My old 1GHz Athlon can't keep up with 96 KHz,
          but is ok at 48 KHz if the band isn't too busy.

          Tor
          N4OGW
        • Pete Smith
          ... Hi Bill - I think you may need to do some experimenting to maximize Skimmer s weak-signal performance. I may have mentioned this already, but a great deal
          Message 4 of 19 , Apr 26, 2008
          • 0 Attachment
            At 09:20 AM 4/26/2008, bill_w4zv wrote:
            >How do you feel about Skimmer's ability to decode weaker signals in
            >QSB and QRN? I'm somewhat unimpressed so far but it could be I'm not
            >doing something right. Does the DSP filter width affect all decoders
            >or just the one displayed with the green bar? Based on what I've seen
            >so far, Skimmer would be great for identifying strong stations CQ-ing
            >or in a pileup but it doesn't work very well for weaker signals, which
            >would likely be the very ones you want to know about. But this could
            >be due to my not setting Skimmer up correctly or due to my weak
            >computer.

            Hi Bill - I think you may need to do some experimenting to maximize
            Skimmer's weak-signal performance. I may have mentioned this already, but
            a great deal depends on the dynamic range of the line input of your
            soundcard. Whichever card you use, you need to make sure that the input
            gain is set just short of clipping, in order to get the closest possible
            approximation of the sensitivity of the receiver alone. and of course, you
            are comparing it with one of the best receivers ever made - no fair hi.

            73, Pete N4ZR
          • Pete Smith
            ... One of the most effective tricks is to uncheck the Adaptive decoder box, and set a specified maximum number of decoders. Hitting 100% isn t so
            Message 5 of 19 , Apr 26, 2008
            • 0 Attachment
              At 02:01 PM 4/26/2008, rt_tclay wrote:
              >If on your computer Skimmer is reaching 100% cpu, then yes, it is
              >having to reduce the number of decoders. I don't know the algorithm
              >used to do that, but yes, it might do that based signal strength. Try
              >reducing the bandwidth. My old 1GHz Athlon can't keep up with 96 KHz,
              >but is ok at 48 KHz if the band isn't too busy.


              One of the most effective tricks is to uncheck the "Adaptive" decoder box,
              and set a specified maximum number of decoders. Hitting 100% isn't so
              important, in my view, as eliminating most of the junk you get from
              Skimmer's trying to decode on RTTY signals. If you set a maximum number of
              decoders, Alex says that Skimmer accommodates that by narrowing the band of
              frequencies that it attempts to decode, centered around the center
              frequency of the waterfall display. Experimentally, I discovered that 150
              decoders seemed pretty much to eliminate anything above about 14060 on a
              normally busy weekend band, when I set the waterfall on 14030. Today,
              running 300 decoders, I'm getting a lot of "AA4EE" type "callsigns" in the
              14070 and above range. During a major contest, you might have to set it to
              a higher number to get the same approximate frequency coverage.

              73, Pete N4ZR
            Your message has been successfully submitted and would be delivered to recipients shortly.