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

Re: [softrock40] Audio Card Sampling Rate Errors

Expand Messages
  • Chris Wilson
    ... 20/07/2013 08:46 Excellent info, thanks for sharing this Warren. I ll have a bash at seeing what my sound card sampling rate error is like. Does HDSDR have
    Message 1 of 8 , Jul 20, 2013
      >
      > For anyone else in search of precision frequency performance from
      > the Softrocks radios I would like to share some work I have been
      > doing on sampling rate errors in the audio cards.

      > There is a very easy way to check your audio card for sampling rate
      > errors. Using HDSDR tune WWV or your calibrations signal with the LO
      > set as close as possible to the tune frequency and still avoid the
      > "hump". I use 9.999 for the LO. Calibrate your Si5170 here.

      > Then move the LO as far as possible away from WWV so you can just
      > see the signal at the edge of the spectrum. I did about 9.978 which
      > put the 10 MHz signal just inside the edge of the display. Now check
      > your 10 MHz calibration again. Any difference is the result of PC/Soundcard rate errors.

      > Once you know you have an error it is easily compensated for. You
      > just need to be sure you calibrate with the offset you normally use.
      > The error amount increases as you move away from the center frequency.

      > My beloved iMic showed a total error of more than 3 Hz between the
      > center and edge points. My E-MU showed about 200 millihertz which would not be resolvable normally.

      > The point of this is, for any real precision frequency
      > measurements, you must account for this sampling error. If you are
      > using the RXTX for WSPR you should calibrate the Si5170 with a 13500
      > Hz offset as this is where it would normally operate on this mode.
      > Simply assuring your radio is calibrated and on frequency is not
      > sufficient and can easily result in you being 1-2 Hz off, even when
      > you think you are precisely on frequency.

      > I have been using an RXTX referenced to GPS for 20 meter precision
      > WSPR work and finding I still had 20 millihertz of unexplainable error.

      > Thanks to Light Coder from the HDSDR development team in helping me
      > to understand this came from the sound card and guiding me through offsetting to correct for it.

      > Warren Allgyer - 9V1TD



      20/07/2013 08:46


      Excellent info, thanks for sharing this Warren. I'll have a bash at
      seeing what my sound card sampling rate error is like.

      Does HDSDR have any sort of user / developer forum?

      Have a good weekend!


      --
      Best Regards,
      Chris Wilson. 2E0ILY
    • Alan
      ... Subject: Re: [softrock40] Audio Card Sampling Rate Errors ... Chris, No, I have thought of offering to do this for the developers but I d probably make a
      Message 2 of 8 , Jul 20, 2013
        ----- Original Message -----
        Subject: Re: [softrock40] Audio Card Sampling Rate Errors


        >
        > Does HDSDR have any sort of user / developer forum?
        >

        Chris,

        No, I have thought of offering to do this for the developers but I'd probably make a mess of it. LC and Mario read several groups
        from time to time, including this one.

        73 Alan G4ZFQ
      • Alan
        ... Subject: [softrock40] Audio Card Sampling Rate Errors ... Yes Warren, because you can! But how far was WWV away from it s exact frequency? Many times
        Message 3 of 8 , Jul 20, 2013
          ----- Original Message -----
          Subject: [softrock40] Audio Card Sampling Rate Errors


          > For anyone else in search of precision frequency performance from the Softrocks radios I would like to share some work I have been
          > doing on sampling rate errors in the audio cards.

          Yes Warren, because you can!

          But how far was WWV away from it's exact frequency?
          Many times signals can be seen to vary and be a few Hz from their actual transmitted frequency after only one refraction from the
          ionosphere. There are many screenshots like this <https://sites.google.com/site/g4zfqradio/ok0eu-beacon-screenshot-3-4-august-2010>
          showing the OK beacons or other signals.

          I'd suggest directly measuring the frequency from a frequency derived from GPS. My Rockwells provide a 10KHz output.
          But for off-air signals mHz accuracy seems very difficult to confirm.
          With good strong signals there may not be much shift but multipath can show two, or more signals separated by a relatively
          significant frequency.

          I have never examined frequency standard stations with anything like mHz precision, maybe something interesting to do?

          73 Alan G4ZFQ


          >
          > There is a very easy way to check your audio card for sampling rate errors. Using HDSDR tune WWV or your calibrations signal with
          > the LO set as close as possible to the tune frequency and still avoid the "hump". I use 9.999 for the LO. Calibrate your Si5170
          > here.
          >
          > Then move the LO as far as possible away from WWV so you can just see the signal at the edge of the spectrum. I did about 9.978
          > which put the 10 MHz signal just inside the edge of the display. Now check your 10 MHz calibration again. Any difference is the
          > result of PC/Soundcard rate errors.
          >
          > Once you know you have an error it is easily compensated for. You just need to be sure you calibrate with the offset you normally
          > use. The error amount increases as you move away from the center frequency.
          >
          > My beloved iMic showed a total error of more than 3 Hz between the center and edge points. My E-MU showed about 200 millihertz
          > which would not be resolvable normally.
          >
          > The point of this is, for any real precision frequency measurements, you must account for this sampling error. If you are using
          > the RXTX for WSPR you should calibrate the Si5170 with a 13500 Hz offset as this is where it would normally operate on this mode.
          > Simply assuring your radio is calibrated and on frequency is not sufficient and can easily result in you being 1-2 Hz off, even
          > when you think you are precisely on frequency.
          >
          > I have been using an RXTX referenced to GPS for 20 meter precision WSPR work and finding I still had 20 millihertz of
          > unexplainable error.
          >
          > Thanks to Light Coder from the HDSDR development team in helping me to understand this came from the sound card and guiding me
          > through offsetting to correct for it.
          >
        • Chris Wilson
          ... 20/07/2013 11:37 In some ways *NOT* having a user forum must be a Godsend to code writers, they often seem to descend into personality differences and long
          Message 4 of 8 , Jul 20, 2013
            > Chris,

            > No, I have thought of offering to do this for the developers but
            > I'd probably make a mess of it. LC and Mario read several groups
            > from time to time, including this one.

            > 73 Alan G4ZFQ



            20/07/2013 11:37


            In some ways *NOT* having a user forum must be a Godsend to code
            writers, they often seem to descend into personality differences and
            long drawn out argument over trivial items. I guess the time input
            from the developers, should they themselves participate, would take a
            significant amount of time away from doing any "proper" work. And a
            group with no input at all from the developers would be pretty
            pointless. Must be a hard call. They are doing something right, for me
            it's a tight bit of software that usually just works, and isn't too
            bloated or PC demanding.

            The few times I have felt cause to contact
            Mario he's been absolutely superb in helping me out.


            --
            Best Regards,
            Chris Wilson. 2E0ILY
          • warrenallgyer
            Yes Alan, you are absolutely correct. In fact, measuring the Doppler shift on HF signals is one of the applications that I am building this mess for! :-) For
            Message 5 of 8 , Jul 20, 2013
              Yes Alan, you are absolutely correct. In fact, measuring the Doppler shift on HF signals is one of the applications that I am building this mess for! :-)

              For the purposes of measuring sound card sampling errors it would be better to use a stable, local frequency source. We are only interested in how much it changes when measured at the center of the SDR spectrum as compared to the edge. The absolute frequency of the source is not particularly important in this case.

              One interesting (to me anyway) observation made since I posted this: My beloved iMic which showed close to 3 Hz variance across 22 KHz on a Win 7 machine, when measured on a slower Win XP laptop, only showed about 210 milliHertz change across the same range. It seems that the drivers and perhaps even the PC itself come into play and Win 7 seems to make the problem worse in the case of the iMic.

              One additional thought: This concept does impact those who transmit from a different sound card than the one they use on receive. Calibration of the receiver would not necessarily extend to the transmit side in such cases.

              But before everyone starts throwing their sound cards in the river remember: for the most part here the variance we are talking about is considerably less than one Hertz. This is not a practical issue for anyone but we nerds who are trying to make cheap equipment jump through extraordinary hoops.



              Warren Allgyer - 9V1TD

              --- In softrock40@yahoogroups.com, "Alan" <alan4alan@...> wrote:
              > Yes Warren, because you can!
              >
              > But how far was WWV away from it's exact frequency?
              > Many times signals can be seen to vary and be a few Hz from their actual transmitted frequency after only one refraction from the
            • Bill Cromwell
              ... Hi Warren, Thanks for mentioning the practical aspects of all this at the end of your post. I have read a lot of posts and online documents regarding
              Message 6 of 8 , Jul 20, 2013
                On 07/20/2013 07:17 AM, warrenallgyer wrote:
                >
                > Yes Alan, you are absolutely correct. In fact, measuring the Doppler
                > shift on HF signals is one of the applications that I am building this
                > mess for! :-)
                >
                > For the purposes of measuring sound card sampling errors it would be
                > better to use a stable, local frequency source. We are only interested
                > in how much it changes when measured at the center of the SDR spectrum
                > as compared to the edge. The absolute frequency of the source is not
                > particularly important in this case.
                >
                > One interesting (to me anyway) observation made since I posted this:
                > My beloved iMic which showed close to 3 Hz variance across 22 KHz on a
                > Win 7 machine, when measured on a slower Win XP laptop, only showed
                > about 210 milliHertz change across the same range. It seems that the
                > drivers and perhaps even the PC itself come into play and Win 7 seems
                > to make the problem worse in the case of the iMic.
                >
                > One additional thought: This concept does impact those who transmit
                > from a different sound card than the one they use on receive.
                > Calibration of the receiver would not necessarily extend to the
                > transmit side in such cases.
                >
                > But before everyone starts throwing their sound cards in the river
                > remember: for the most part here the variance we are talking about is
                > considerably less than one Hertz. This is not a practical issue for
                > anyone but we nerds who are trying to make cheap equipment jump
                > through extraordinary hoops.
                >
                > Warren Allgyer - 9V1TD
                >
                Hi Warren,

                Thanks for mentioning the practical aspects of all this at the end of
                your post. I have read a lot of posts and online documents regarding
                frequency measurement (as related to digital sound card radio software)
                and the tone of much of the information would have us believe we are out
                of bounds if we are a fraction of a millicycle from a "specified"
                frequency. It is worthwhile to make your measurement gear as accurate as
                you can (accuracy and resolution are not the same) but one order of
                magnitude greater than the rules (FCC?) is sufficient. More is better.
                Ham radio *IS* "just" a hobby. Frequency measurement is a part of that
                hobby and more critical measurements are a valid interest. So long as we
                keep in mind that level of accuracy is not really needed for day to day
                operation. Even for various other radios services the channel
                specifications have non zero error tolerances. In amateur radio we only
                have to stay between two relatively very wide limits. In the main my
                approach and the tone of my posts reflects my amateur radio operation
                point of view.

                I have used the info about sound card calibration to be aware of those
                frequency errors and take them into account for my own frequency
                measurements. I use Spectran and Spectrum Lab in my shack/shop and most
                of the time WWV. I know that errors can be introduced by WWV directly
                and from transmission anomolies. But to my knowledge the adored GPS
                signals are also transmitted so why would those not be subject to
                similar transmission errors (even if they are smaller errors)? I think
                we can spend a lot of money for "standards" and still have errors. If
                frequency measurement is a big part of *your* hobby - go for it. For the
                rest of us good enough is good enough and the present batch of sound
                card software and SDR is well more than good enough.
                73,

                Bill KU8H
              • pe0fko
                ... beloved iMic which showed close to 3 Hz variance across 22 KHz on a Win 7 machine, when measured on a slower Win XP laptop, only showed about 210
                Message 7 of 8 , Jul 20, 2013
                  >> One interesting (to me anyway) observation made since I posted this: My beloved iMic which showed close to 3 Hz variance across 22 KHz on a Win 7 machine, when measured on a slower Win XP laptop, only showed about 210 milliHertz change across the same range. It seems that the drivers and perhaps even the PC itself come into play and Win 7 seems to make the problem worse in the case of the iMic.

                  The USB connection need a stable (with in 500ppm I think) frequency from the device, so many devices used a crystal for it. When the USB chip is working on a RC (internal) oscillator most of the time the oscillator is locked (software pll) to the USB Start Of Frame signal from the PC site (1ms). Maybe that is the reason why the USB sound-card freq changed with the change of PC?

                                    
                  73, Fred
                  PE0FKO



                  2013/7/20 warrenallgyer <allgyer@...>
                   

                  Yes Alan, you are absolutely correct. In fact, measuring the Doppler shift on HF signals is one of the applications that I am building this mess for! :-)

                  For the purposes of measuring sound card sampling errors it would be better to use a stable, local frequency source. We are only interested in how much it changes when measured at the center of the SDR spectrum as compared to the edge. The absolute frequency of the source is not particularly important in this case.

                  One interesting (to me anyway) observation made since I posted this: My beloved iMic which showed close to 3 Hz variance across 22 KHz on a Win 7 machine, when measured on a slower Win XP laptop, only showed about 210 milliHertz change across the same range. It seems that the drivers and perhaps even the PC itself come into play and Win 7 seems to make the problem worse in the case of the iMic.

                  One additional thought: This concept does impact those who transmit from a different sound card than the one they use on receive. Calibration of the receiver would not necessarily extend to the transmit side in such cases.

                  But before everyone starts throwing their sound cards in the river remember: for the most part here the variance we are talking about is considerably less than one Hertz. This is not a practical issue for anyone but we nerds who are trying to make cheap equipment jump through extraordinary hoops.

                  Warren Allgyer - 9V1TD



                  --- In softrock40@yahoogroups.com, "Alan" <alan4alan@...> wrote:
                  > Yes Warren, because you can!
                  >
                  > But how far was WWV away from it's exact frequency?
                  > Many times signals can be seen to vary and be a few Hz from their actual transmitted frequency after only one refraction from the


                Your message has been successfully submitted and would be delivered to recipients shortly.