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

Re: [softrock40] RX Lowpass filter

Expand Messages
  • Alan
    ... Subject: [softrock40] RX Lowpass filter ... Rob, How many dB down do you calculate at your critical frequency? One thing to consider is that we need a flat
    Message 1 of 8 , Sep 20, 2013
    • 0 Attachment
      ----- Original Message -----
      Subject: [softrock40] RX Lowpass filter


      > Hi all,
      >
      > I had a theoretical question about the design of the RXTX ensemble. I notice that the feedback on the op-amp has a 5k resistor
      > and a 220pf cap in parallel. As far as I can tell this is the only real low pass filter before the signal is passed off to the
      > soundcard. 220pf with a 5k resistor works out to a critical frequency of 909 khz. I'm no expert at this stuff, but that seems
      > really high. Shouldn't that make that RXTX susceptible to overloading from strong out of band signals? Why not just up the value
      > of C44/C45? 1 nanofarad would work out to 200khz which seems like a more reasonable value.
      >

      Rob,

      How many dB down do you calculate at your critical frequency?
      One thing to consider is that we need a flat response up to 100KHz if a 192KHz card is used.
      Agreed the effect of high frequencies into the input is unknown but I'd guess there is not much coming from the Softrock and that
      what is is just lost.
      I'm not aware of anyone suggesting of spurious emissions occurring.

      73 Alan G4ZFQ
    • Anthony Casorso
      Sound cards contain a low pass filter at the input to limit the spectrum based on the sample rate. I doubt that you would gain anything by playing with the op
      Message 2 of 8 , Sep 20, 2013
      • 0 Attachment
        Sound cards contain a low pass filter at the input to limit the spectrum based on the sample rate.  I doubt that you would gain anything by playing with the op amp rolloff.
         
        Tony

         

        To: softrock40@yahoogroups.com
        From: alan4alan@...
        Date: Fri, 20 Sep 2013 10:17:35 +0000
        Subject: Re: [softrock40] RX Lowpass filter

         

        ----- Original Message -----
        Subject: [softrock40] RX Lowpass filter

        > Hi all,
        >
        > I had a theoretical question about the design of the RXTX ensemble. I notice that the feedback on the op-amp has a 5k resistor
        > and a 220pf cap in parallel. As far as I can tell this is the only real low pass filter before the signal is passed off to the
        > soundcard. 220pf with a 5k resistor works out to a critical frequency of 909 khz. I'm no expert at this stuff, but that seems
        > really high. Shouldn't that make that RXTX susceptible to overloading from strong out of band signals? Why not just up the value
        > of C44/C45? 1 nanofarad would work out to 200khz which seems like a more reasonable value.
        >

        Rob,

        How many dB down do you calculate at your critical frequency?
        One thing to consider is that we need a flat response up to 100KHz if a 192KHz card is used.
        Agreed the effect of high frequencies into the input is unknown but I'd guess there is not much coming from the Softrock and that
        what is is just lost.
        I'm not aware of anyone suggesting of spurious emissions occurring.

        73 Alan G4ZFQ


      • warrenallgyer
        I think your idea does hold water Rob and it should be fairly easy to test using HDSDR and recording the level as you tune through a steady carrier every 10
        Message 3 of 8 , Sep 20, 2013
        • 0 Attachment

          I think your idea does hold water Rob and it should be fairly easy to test using HDSDR and recording the level as you tune through a steady carrier every 10 KHz or so.


          In my experience overload of the op amp is indeed a problem. Unmodified RXTX will begin clipping in this stage with input signals at about -20 dBm or about 50 over S9. It has also been my experience that strong signals 300-500 KHz away from the tune frequency can overload the op amps and cause intermodulation problems. So this may be a worthwhile experiment to improve the RXTX.


          Phase issues may arise and we need to watch out for them. That said, the IQ balance is only really valid at a single point in the audio passband anyway. I would think that most software IQ balance controls would take care of any component mismatch.


          When I can get into the lab later today I will run some tests...........


          Warren Allgyer

          9V1TD



          --- In softrock40@yahoogroups.com, <physicsrob@...> wrote:

          Hi all,

          I had a theoretical question about the design of the RXTX ensemble. I notice that the feedback on the op-amp has a 5k resistor and a 220pf cap in parallel. As far as I can tell this is the only real low pass filter before the signal is passed off to the soundcard. 220pf with a 5k resistor works out to a critical frequency of 909 khz. I'm no expert at this stuff, but that seems really high. Shouldn't that make that RXTX susceptible to overloading from strong out of band signals? Why not just up the value of C44/C45? 1 nanofarad would work out to 200khz which seems like a more reasonable value.

          Does my argument hold water? Is this theoretical concern just a non-issue in the real world? Am I missing something? I haven't finished assembling my RXTX so experimenting is a bit premature, otherwise I might just give it a try.

          73,

          Rob
          NE6U
        • Robert Porter
          Hi all, Thanks for the feedback. Tony Parks messaged me off list mentioning that I had forgotten a factor of 2pi. The actual cutoff frequency is 192khz -- a
          Message 4 of 8 , Sep 20, 2013
          • 0 Attachment
            Hi all,

            Thanks for the feedback.  Tony Parks messaged me off list mentioning that I had forgotten a factor of 2pi.  The actual cutoff frequency is 192khz -- a much more reasonable choice than I was thinking.  Still lowering this frequency a bit might make sense for some (for instance if you are only sampling at 44khz, or if you prefer dynamic range over bandwidth).

            Cheers,
            Rob
            NE6U

            On Sep 20, 2013, at 3:59 PM, <allgyer@...> wrote:

             

            I think your idea does hold water Rob and it should be fairly easy to test using HDSDR and recording the level as you tune through a steady carrier every 10 KHz or so.


            In my experience overload of the op amp is indeed a problem. Unmodified RXTX will begin clipping in this stage with input signals at about -20 dBm or about 50 over S9. It has also been my experience that strong signals 300-500 KHz away from the tune frequency can overload the op amps and cause intermodulation problems. So this may be a worthwhile experiment to improve the RXTX.


            Phase issues may arise and we need to watch out for them. That said, the IQ balance is only really valid at a single point in the audio passband anyway. I would think that most software IQ balance controls would take care of any component mismatch.


            When I can get into the lab later today I will run some tests...........


            Warren Allgyer

            9V1TD



            --- In softrock40@yahoogroups.com, <physicsrob@...> wrote:

            Hi all,

            I had a theoretical question about the design of the RXTX ensemble. I notice that the feedback on the op-amp has a 5k resistor and a 220pf cap in parallel. As far as I can tell this is the only real low pass filter before the signal is passed off to the soundcard. 220pf with a 5k resistor works out to a critical frequency of 909 khz. I'm no expert at this stuff, but that seems really high. Shouldn't that make that RXTX susceptible to overloading from strong out of band signals? Why not just up the value of C44/C45? 1 nanofarad would work out to 200khz which seems like a more reasonable value.

            Does my argument hold water? Is this theoretical concern just a non-issue in the real world? Am I missing something? I haven't finished assembling my RXTX so experimenting is a bit premature, otherwise I might just give it a try.

            73,

            Rob
            NE6U


          • warrenallgyer
            I ran some tests on three different RXTX this afternoon. The test was done using a signal generator at about -25 dBm into HDSDR. I put a scope on the Audio
            Message 5 of 8 , Sep 21, 2013
            • 0 Attachment
              I ran some tests on three different RXTX this afternoon. The test was done using a signal generator at about -25 dBm into HDSDR. I put a scope on the "Audio In" directly out of the board and tuned the HDSDR LO to 10 KHz below the test frequency. This put a nice 10 KHz sine wave on the audio out for a reference. I then lowered the HDSDR LO 10KHz at a time, which increases the audio out frequency by the corresponding amount, until I reached output voltages that equaled -3, -10, and -20 dB.

              On two of the rigs the -3dB point was at about 150KHz, 10 dB was about 400 KHz, and 20 dB at about 1.5 MHz.

              The third rig will bear some investigation as it was 3 dB down at 75 KHz, 10 dB at 185 KHz, and 20 dB at 420 KHz. Clearly something inconsistent here.

              If you are using 192 KHz displays then the numbers for the first two rigs make some sense.... it should be relatively flat out to 100KHz. If you use 96 or 48 KHz displays then a response curve like the third rig would work fine and give a little bit of dynamic range relief for strong signals 200 KHz or so removed from the operating frequency.


              Warren Allgyer - 9V1TD

               



              --- In softrock40@yahoogroups.com, <softrock40@yahoogroups.com> wrote:

              Hi all,

              Thanks for the feedback.  Tony Parks messaged me off list mentioning that I had forgotten a factor of 2pi.  The actual cutoff frequency is 192khz -- a much more reasonable choice than I was thinking.  Still lowering this frequency a bit might make sense for some (for instance if you are only sampling at 44khz, or if you prefer dynamic range over bandwidth).

              Cheers,
              Rob
              NE6U

              On Sep 20, 2013, at 3:59 PM, <allgyer@...> wrote:

               

              I think your idea does hold water Rob and it should be fairly easy to test using HDSDR and recording the level as you tune through a steady carrier every 10 KHz or so.


              In my experience overload of the op amp is indeed a problem. Unmodified RXTX will begin clipping in this stage with input signals at about -20 dBm or about 50 over S9. It has also been my experience that strong signals 300-500 KHz away from the tune frequency can overload the op amps and cause intermodulation problems. So this may be a worthwhile experiment to improve the RXTX.


              Phase issues may arise and we need to watch out for them. That said, the IQ balance is only really valid at a single point in the audio passband anyway. I would think that most software IQ balance controls would take care of any component mismatch.


              When I can get into the lab later today I will run some tests...........


              Warren Allgyer

              9V1TD



              --- In softrock40@yahoogroups.com, <physicsrob@...> wrote:

              Hi all,

              I had a theoretical question about the design of the RXTX ensemble. I notice that the feedback on the op-amp has a 5k resistor and a 220pf cap in parallel. As far as I can tell this is the only real low pass filter before the signal is passed off to the soundcard. 220pf with a 5k resistor works out to a critical frequency of 909 khz. I'm no expert at this stuff, but that seems really high. Shouldn't that make that RXTX susceptible to overloading from strong out of band signals? Why not just up the value of C44/C45? 1 nanofarad would work out to 200khz which seems like a more reasonable value.

              Does my argument hold water? Is this theoretical concern just a non-issue in the real world? Am I missing something? I haven't finished assembling my RXTX so experimenting is a bit premature, otherwise I might just give it a try.

              73,

              Rob
              NE6U


            • warrenallgyer
              On retest with a higher signal generator level and an attenuator to set the reference levels: Across 4 RXTX and one RXII I am measuring the following audio
              Message 6 of 8 , Sep 21, 2013
              • 0 Attachment
                On retest with a higher signal generator level and an attenuator to set the reference levels: Across 4 RXTX and one RXII I am measuring the following audio responses:

                -3dB - From 80-100 KHz
                -10dB - From 200 to 290 KHz
                -20dB - From 400 to 460 KHz

                It seems, like most aspects of the RXTX design, once you dig into it you find a very well thought out execution given the variables of component variances and construction techniques.

                The system was clearly engineered with the 192KHz display in mind. 48K and 96K users could get slightly better out of band rejection by tightening the filter but I think it is a lot of trouble for very little benefit.

                Warren Allgyer - 9V1TD

                 



                --- In softrock40@yahoogroups.com, <allgyer@...> wrote:

                I ran some tests on three different RXTX this afternoon. The test was done using a signal generator at about -25 dBm into HDSDR. I put a scope on the "Audio In" directly out of the board and tuned the HDSDR LO to 10 KHz below the test frequency. This put a nice 10 KHz sine wave on the audio out for a reference. I then lowered the HDSDR LO 10KHz at a time, which increases the audio out frequency by the corresponding amount, until I reached output voltages that equaled -3, -10, and -20 dB.

                On two of the rigs the -3dB point was at about 150KHz, 10 dB was about 400 KHz, and 20 dB at about 1.5 MHz.

                The third rig will bear some investigation as it was 3 dB down at 75 KHz, 10 dB at 185 KHz, and 20 dB at 420 KHz. Clearly something inconsistent here.

                If you are using 192 KHz displays then the numbers for the first two rigs make some sense.... it should be relatively flat out to 100KHz. If you use 96 or 48 KHz displays then a response curve like the third rig would work fine and give a little bit of dynamic range relief for strong signals 200 KHz or so removed from the operating frequency.


                Warren Allgyer - 9V1TD

                 



                --- In softrock40@yahoogroups.com, <softrock40@yahoogroups.com> wrote:

                Hi all,

                Thanks for the feedback.  Tony Parks messaged me off list mentioning that I had forgotten a factor of 2pi.  The actual cutoff frequency is 192khz -- a much more reasonable choice than I was thinking.  Still lowering this frequency a bit might make sense for some (for instance if you are only sampling at 44khz, or if you prefer dynamic range over bandwidth).

                Cheers,
                Rob
                NE6U

                On Sep 20, 2013, at 3:59 PM, <allgyer@...> wrote:

                 

                I think your idea does hold water Rob and it should be fairly easy to test using HDSDR and recording the level as you tune through a steady carrier every 10 KHz or so.


                In my experience overload of the op amp is indeed a problem. Unmodified RXTX will begin clipping in this stage with input signals at about -20 dBm or about 50 over S9. It has also been my experience that strong signals 300-500 KHz away from the tune frequency can overload the op amps and cause intermodulation problems. So this may be a worthwhile experiment to improve the RXTX.


                Phase issues may arise and we need to watch out for them. That said, the IQ balance is only really valid at a single point in the audio passband anyway. I would think that most software IQ balance controls would take care of any component mismatch.


                When I can get into the lab later today I will run some tests...........


                Warren Allgyer

                9V1TD



                --- In softrock40@yahoogroups.com, <physicsrob@...> wrote:

                Hi all,

                I had a theoretical question about the design of the RXTX ensemble. I notice that the feedback on the op-amp has a 5k resistor and a 220pf cap in parallel. As far as I can tell this is the only real low pass filter before the signal is passed off to the soundcard. 220pf with a 5k resistor works out to a critical frequency of 909 khz. I'm no expert at this stuff, but that seems really high. Shouldn't that make that RXTX susceptible to overloading from strong out of band signals? Why not just up the value of C44/C45? 1 nanofarad would work out to 200khz which seems like a more reasonable value.

                Does my argument hold water? Is this theoretical concern just a non-issue in the real world? Am I missing something? I haven't finished assembling my RXTX so experimenting is a bit premature, otherwise I might just give it a try.

                73,

                Rob
                NE6U


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