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

Re: poor performance of WBIR in powersdr1.19.3.15 with non-flex radios

Expand Messages
  • sv1eia
    Hi Andrew, As I said, it works ok here so I do not have anything to say for your setup more than what I explained already. You didnt mention though if you have
    Message 1 of 24 , Jun 1, 2010
    • 0 Attachment
      Hi Andrew,

      As I said, it works ok here so I do not have anything to say for your setup more than what I explained already.
      You didnt mention though if you have any transformers or filters on your setup.

      73,
      Christos SV1EIA



      --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@...> wrote:
      >
      > Hello Christos,
      > I have now done some measurements on the iq balance of various setups.
      >
      > Measurements for baseband from +/-10kHz to +/-40kHz using a gain/phase meter
      > Softrock 6.1 at 7.055MHz: <0.4dB/<4 degs
      > 6.3 TxRx plus mobo at 7MHz: <0.7dB/<5 degs
      > 6.3 TxRx plus mobo at 14MHz: <0.5dB/<7.5 degs
      > 6.3 TxRx plus mobo at 28.2MHz: <0.2dB/9.2 degs
      >
      > All measurements look good on an oscilloscope but then its not so easy to see 5 degrees error, hence using the meter.
      >
      > Generally the lower the baseband frequency the better the balance. So overall one would expect 20-35dB image suppression un-tweaked. Some offsets were within 1 degree.
      >
      > However I get the same result with any kit and any offset and that is that until signals get above about -50dBm input you get the basic performance of the hardware and sometimes worse. When you exceed -50dBm the image starts to fall eventially reaching the noise floor in some cases.
      >
      > Thought it might be my pc / sound card so I repeated the tests with another soundcard and pc with the same result.
      >
      > Another test was to feed the line inputs with a quadrature baseband signal generated using the soundcard lineout and some software. With an older version of powersdr I looked at the panadapter and adjusted the iqgenerator program for at least 30dB image suppression. I then switched to powersdr 1.19.3.15 and the image rejection was about 30dB, so far so good. Then I increased the signal to the magic -50dBm and the image started to drop to the noise floor. Finally I changed band and back again to an input below -50dBm and the image rejection had decreased to 20dB, a figure I see a lot.
      >
      > This is all very confusing, it really looks like the WBIR does nothing until this threshold is reached and in some instances makes the image worse? I wonder if anyone else is seeing this???
      >
      > I would be good if we could find out more about how the WBIR works, for example is there a maximum amplitude and phase error above which it can't correct, or an snr needed for the algorithm to work?
      > Is there any reference material anywhere?
      >
      > Another test I did was to tune to either 7 or 14MHz dead with an antenna connected. With good image rejection all the cw signals to the right should stop at the band edge. With poor rejection you can easily see many cw signals to the left of the band edge.(all signals were -70dBm or lower).
      >
      > Any thoughts Christos? It all causing a lot of head scratching this end!!
      > Thanks,
      > Andrew
      >
      >
      > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
      > >
      > > It works all the time for the complete incoming stream.
      > >
      > >
      > >
      > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
      > > >
      > > > Hi Christos,
      > > > Thanks for the further explanations, is there a threshold for the WBIR or does it start working right from the noise floor?
      > > > Thanks,
      > > > Andrew
      > > >
      > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
      > > > >
      > > > > Andrew,
      > > > >
      > > > > One more note.
      > > > >
      > > > > All these algos need 'linear' behaviour to work best.
      > > > > When non-linear distortion in either phase or gain of the I/Q signals happen, then the algos do not operate as they suppose to.
      > > > > (non-linear according to frequency axis)
      > > > > Avoid long cables, chokes, coils, and ANY isolation transformer after the opamp output, these are causes of non-linear phase distortion.
      > > > > At some discussion there was a voice that said, "ok so what, its only about 10 degrees distortion, s/w should keep up with it..", well I guess he did not knew that when we try do TRUE WBIR on real frequency agile rigs (and NOT based on center frequency), then we use floating point arithmetic having 20 something decimal digits on phase and amplitude counting on even the most tiny fraction (Complex/imaginery numbers involved too) so 10 whole degrees of non-linear deviation will throw us somewhere between the Moon and the Sun without a way back..
      > > > > even a fraction of a degree makes a large difference in signals around noise floor.
      > > > >
      > > > > There is already such a distortion effect in the rig itself, so adding more distortion outside of it will always make things worse than better.
      > > > >
      > > > > I hope you do get some hints now for your homemade rigs to perform better :-)
      > > > >
      > > > > 73,
      > > > > Christos
      > > > >
      > > > >
      > > > >
      > > > >
      > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
      > > > > >
      > > > > > Hello Christos,
      > > > > > Thanks for the reply and the further words about how the cancellation works. I was wondering what the coefficient was for!
      > > > > > This evening I have tried a fixed softrock lite receiver at 7MHz and got much the same results, perhaps a little better. I suspect that the better the i/q balance is in the first place the better the WBIR works. There is definately a degredation though when you add a second signal as you can see both images pop up 10-20dB which is interesting.
      > > > > > I have checked quadrature balance on the 6.3rxtx but not taken any results (I think it was within 0.5dB and about 5 degrees typically). I have a gain/phase meter so I shall report back with some hard data hopefully tomorrow.
      > > > > > Thanks for your help,
      > > > > > Andrew
      > > > > >
      > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
      > > > > > >
      > > > > > >
      > > > > > > Andrew,
      > > > > > >
      > > > > > > Nobody told you that immediately you would not see any image.
      > > > > > > Initially you would see images, but in an small amount of time it will be nulled out automagically.
      > > > > > >
      > > > > > > In 1.12.20 there was a coeficient put by me there that was giving to the user some 'adjustment', faster<->better for larger signals, slower<->better for smaller signal.
      > > > > > >
      > > > > > > In 1.19 WBIR, Flex has done some kind of assesment and put there similar internal coefficients but for all-cases moderate scenario and the user cannot adjust them.
      > > > > > >
      > > > > > > I hope that clear it out.
      > > > > > >
      > > > > > > Yet again, if you do see some problem there, I honestly think that you better look at the IQ outs of your rig. They need to be equal in signal level and quadrature.
      > > > > > > You can do a test with a o-scope attaching each x-channel to I and y-channel to Q, then goto PSDR to a frequency, eg VFO = 10MHz and inject a signal equal to VFO - IF + 1 (9992KHz) then theoretically you should have two equal x and y signals of 1KHz in 90 degrees phase diff on your o-scope.
      > > > > > > Do the same test at the end of the cables you use before they go into the the line-in (Rx IQ) input of your audio card.
      > > > > > >
      > > > > > > If you do not have these conditions in your signals (1KHz, equal level, quadrature), the your rig or cables needs some checkout.
      > > > > > >
      > > > > > > And do not use ANY isolation transformers, they do only damage.
      > > > > > >
      > > > > > > 73,
      > > > > > > Christos SV1EIa
      > > > > > >
      > > > > > >
      > > > > > >
      > > > > > >
      > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
      > > > > > > >
      > > > > > > > Hello Christos,
      > > > > > > > I am interested to see that you don't have these problems. So if you inject a very small signal that is below the noise floor and gradually increase its level so that it ends up at say -30dBm, you don't see any image on the panadapter at any time?
      > > > > > > > Any chance you could capture a wav iq file so that I could run it on my system?
      > > > > > > > Thanks,
      > > > > > > > Andrew
      > > > > > > >
      > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
      > > > > > > > >
      > > > > > > > > Hi Andrew,
      > > > > > > > >
      > > > > > > > > I do not share your experience.
      > > > > > > > > Here it works absolutely marvelous and I have three SR 6.2 rigs.
      > > > > > > > > The same has been reported by anyone else I know it has it working.
      > > > > > > > > One common though problem is rig's I/Q out.
      > > > > > > > > Check your I/Q output, it is probably having a problem either cabling or opamps are damaged.
      > > > > > > > >
      > > > > > > > > Also in 1.12.20, the EIA is actually the same algo as WBIR..
      > > > > > > > > so if you have the exact same cabling for both applications and it works ok in one but not ok when you switch to the other, then ..I do not know what it is!! :-)
      > > > > > > > >
      > > > > > > > > 73,
      > > > > > > > > Christos
      > > > > > > > >
      > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
      > > > > > > > > >
      > > > > > > > > > Hi Christos,
      > > > > > > > > > I am finding that the WBIR feature doesn't work very well at all with my softrock + mobo radio.
      > > > > > > > > >
      > > > > > > > > > I find that the automatic image cancellation only seems to start working with input signals above -50dBm (you can see it kick in and reduce the image). Once it has done its magic it holds about 40dB or better until you tune somewhere else. If you add a second signal above -50dBm then the image increases to about -40dB. If you tune to another frequency you are back to what the softrock can do without tweaking the amplitude and phase setup which you can't do in this version. So the result is that with any signals below -50dBm the auto cancellation does nothing and you are back to the basic softrock performance which might be as bad as 20-30dB on 28MHz. By the way this test was done with signal generators not off air signals.
      > > > > > > > > > I suspect with real flex hardware the radio is firmware calibrated over its whole tuning range to get say 40-50dB and then the auto canceller takes care of the bigger signals.
      > > > > > > > > >
      > > > > > > > > > Do you see these issues?
      > > > > > > > > >
      > > > > > > > > > Is there any way in the code to reduce the threshold at which the WBIR cuts in? Or any way to re-enable the gain and phase sliders so at least the image can be optimised with weaker signals?
      > > > > > > > > >
      > > > > > > > > > I would actually prefer to switch back to your powerSDR-iq v1.12.20 as it has manual gain/phase adjust plus no echo on tx, but it doesn't have the panafall display (how difficult to add panafall???).
      > > > > > > > > >
      > > > > > > > > > Anyway I thought I'd highlight the shortcomings I'm seeing in case anybody else has seen similar.
      > > > > > > > > > Regards,
      > > > > > > > > > Andrew
      > > > > > > > > >
      > > > > > > > >
      > > > > > > >
      > > > > > >
      > > > > >
      > > > >
      > > >
      > >
      >
    • irbsurfing
      Hello Christos, No transformers, 6.3RxTx straight into soundcard. Also when I did the tests using the lineout fed back into the line in that was a direct
      Message 2 of 24 , Jun 2, 2010
      • 0 Attachment
        Hello Christos,
        No transformers, 6.3RxTx straight into soundcard. Also when I did the tests using the lineout fed back into the line in that was a direct connection.
        This is puzzling, I can only assume your rigs have much better iq balance than mine.
        Perhaps you could test this for me, if you were to run PowerSDR-IQ version 1.12.20 and set the software i/q trim to 0, what image rejection would you get in the 7/14/28MHz bands?
        I get 20-35dB depending on baseband frequency and RF frequency.
        Andrew

        --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@...> wrote:
        >
        > Hi Andrew,
        >
        > As I said, it works ok here so I do not have anything to say for your setup more than what I explained already.
        > You didnt mention though if you have any transformers or filters on your setup.
        >
        > 73,
        > Christos SV1EIA
        >
        >
        >
        > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
        > >
        > > Hello Christos,
        > > I have now done some measurements on the iq balance of various setups.
        > >
        > > Measurements for baseband from +/-10kHz to +/-40kHz using a gain/phase meter
        > > Softrock 6.1 at 7.055MHz: <0.4dB/<4 degs
        > > 6.3 TxRx plus mobo at 7MHz: <0.7dB/<5 degs
        > > 6.3 TxRx plus mobo at 14MHz: <0.5dB/<7.5 degs
        > > 6.3 TxRx plus mobo at 28.2MHz: <0.2dB/9.2 degs
        > >
        > > All measurements look good on an oscilloscope but then its not so easy to see 5 degrees error, hence using the meter.
        > >
        > > Generally the lower the baseband frequency the better the balance. So overall one would expect 20-35dB image suppression un-tweaked. Some offsets were within 1 degree.
        > >
        > > However I get the same result with any kit and any offset and that is that until signals get above about -50dBm input you get the basic performance of the hardware and sometimes worse. When you exceed -50dBm the image starts to fall eventially reaching the noise floor in some cases.
        > >
        > > Thought it might be my pc / sound card so I repeated the tests with another soundcard and pc with the same result.
        > >
        > > Another test was to feed the line inputs with a quadrature baseband signal generated using the soundcard lineout and some software. With an older version of powersdr I looked at the panadapter and adjusted the iqgenerator program for at least 30dB image suppression. I then switched to powersdr 1.19.3.15 and the image rejection was about 30dB, so far so good. Then I increased the signal to the magic -50dBm and the image started to drop to the noise floor. Finally I changed band and back again to an input below -50dBm and the image rejection had decreased to 20dB, a figure I see a lot.
        > >
        > > This is all very confusing, it really looks like the WBIR does nothing until this threshold is reached and in some instances makes the image worse? I wonder if anyone else is seeing this???
        > >
        > > I would be good if we could find out more about how the WBIR works, for example is there a maximum amplitude and phase error above which it can't correct, or an snr needed for the algorithm to work?
        > > Is there any reference material anywhere?
        > >
        > > Another test I did was to tune to either 7 or 14MHz dead with an antenna connected. With good image rejection all the cw signals to the right should stop at the band edge. With poor rejection you can easily see many cw signals to the left of the band edge.(all signals were -70dBm or lower).
        > >
        > > Any thoughts Christos? It all causing a lot of head scratching this end!!
        > > Thanks,
        > > Andrew
        > >
        > >
        > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
        > > >
        > > > It works all the time for the complete incoming stream.
        > > >
        > > >
        > > >
        > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
        > > > >
        > > > > Hi Christos,
        > > > > Thanks for the further explanations, is there a threshold for the WBIR or does it start working right from the noise floor?
        > > > > Thanks,
        > > > > Andrew
        > > > >
        > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
        > > > > >
        > > > > > Andrew,
        > > > > >
        > > > > > One more note.
        > > > > >
        > > > > > All these algos need 'linear' behaviour to work best.
        > > > > > When non-linear distortion in either phase or gain of the I/Q signals happen, then the algos do not operate as they suppose to.
        > > > > > (non-linear according to frequency axis)
        > > > > > Avoid long cables, chokes, coils, and ANY isolation transformer after the opamp output, these are causes of non-linear phase distortion.
        > > > > > At some discussion there was a voice that said, "ok so what, its only about 10 degrees distortion, s/w should keep up with it..", well I guess he did not knew that when we try do TRUE WBIR on real frequency agile rigs (and NOT based on center frequency), then we use floating point arithmetic having 20 something decimal digits on phase and amplitude counting on even the most tiny fraction (Complex/imaginery numbers involved too) so 10 whole degrees of non-linear deviation will throw us somewhere between the Moon and the Sun without a way back..
        > > > > > even a fraction of a degree makes a large difference in signals around noise floor.
        > > > > >
        > > > > > There is already such a distortion effect in the rig itself, so adding more distortion outside of it will always make things worse than better.
        > > > > >
        > > > > > I hope you do get some hints now for your homemade rigs to perform better :-)
        > > > > >
        > > > > > 73,
        > > > > > Christos
        > > > > >
        > > > > >
        > > > > >
        > > > > >
        > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
        > > > > > >
        > > > > > > Hello Christos,
        > > > > > > Thanks for the reply and the further words about how the cancellation works. I was wondering what the coefficient was for!
        > > > > > > This evening I have tried a fixed softrock lite receiver at 7MHz and got much the same results, perhaps a little better. I suspect that the better the i/q balance is in the first place the better the WBIR works. There is definately a degredation though when you add a second signal as you can see both images pop up 10-20dB which is interesting.
        > > > > > > I have checked quadrature balance on the 6.3rxtx but not taken any results (I think it was within 0.5dB and about 5 degrees typically). I have a gain/phase meter so I shall report back with some hard data hopefully tomorrow.
        > > > > > > Thanks for your help,
        > > > > > > Andrew
        > > > > > >
        > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
        > > > > > > >
        > > > > > > >
        > > > > > > > Andrew,
        > > > > > > >
        > > > > > > > Nobody told you that immediately you would not see any image.
        > > > > > > > Initially you would see images, but in an small amount of time it will be nulled out automagically.
        > > > > > > >
        > > > > > > > In 1.12.20 there was a coeficient put by me there that was giving to the user some 'adjustment', faster<->better for larger signals, slower<->better for smaller signal.
        > > > > > > >
        > > > > > > > In 1.19 WBIR, Flex has done some kind of assesment and put there similar internal coefficients but for all-cases moderate scenario and the user cannot adjust them.
        > > > > > > >
        > > > > > > > I hope that clear it out.
        > > > > > > >
        > > > > > > > Yet again, if you do see some problem there, I honestly think that you better look at the IQ outs of your rig. They need to be equal in signal level and quadrature.
        > > > > > > > You can do a test with a o-scope attaching each x-channel to I and y-channel to Q, then goto PSDR to a frequency, eg VFO = 10MHz and inject a signal equal to VFO - IF + 1 (9992KHz) then theoretically you should have two equal x and y signals of 1KHz in 90 degrees phase diff on your o-scope.
        > > > > > > > Do the same test at the end of the cables you use before they go into the the line-in (Rx IQ) input of your audio card.
        > > > > > > >
        > > > > > > > If you do not have these conditions in your signals (1KHz, equal level, quadrature), the your rig or cables needs some checkout.
        > > > > > > >
        > > > > > > > And do not use ANY isolation transformers, they do only damage.
        > > > > > > >
        > > > > > > > 73,
        > > > > > > > Christos SV1EIa
        > > > > > > >
        > > > > > > >
        > > > > > > >
        > > > > > > >
        > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
        > > > > > > > >
        > > > > > > > > Hello Christos,
        > > > > > > > > I am interested to see that you don't have these problems. So if you inject a very small signal that is below the noise floor and gradually increase its level so that it ends up at say -30dBm, you don't see any image on the panadapter at any time?
        > > > > > > > > Any chance you could capture a wav iq file so that I could run it on my system?
        > > > > > > > > Thanks,
        > > > > > > > > Andrew
        > > > > > > > >
        > > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
        > > > > > > > > >
        > > > > > > > > > Hi Andrew,
        > > > > > > > > >
        > > > > > > > > > I do not share your experience.
        > > > > > > > > > Here it works absolutely marvelous and I have three SR 6.2 rigs.
        > > > > > > > > > The same has been reported by anyone else I know it has it working.
        > > > > > > > > > One common though problem is rig's I/Q out.
        > > > > > > > > > Check your I/Q output, it is probably having a problem either cabling or opamps are damaged.
        > > > > > > > > >
        > > > > > > > > > Also in 1.12.20, the EIA is actually the same algo as WBIR..
        > > > > > > > > > so if you have the exact same cabling for both applications and it works ok in one but not ok when you switch to the other, then ..I do not know what it is!! :-)
        > > > > > > > > >
        > > > > > > > > > 73,
        > > > > > > > > > Christos
        > > > > > > > > >
        > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
        > > > > > > > > > >
        > > > > > > > > > > Hi Christos,
        > > > > > > > > > > I am finding that the WBIR feature doesn't work very well at all with my softrock + mobo radio.
        > > > > > > > > > >
        > > > > > > > > > > I find that the automatic image cancellation only seems to start working with input signals above -50dBm (you can see it kick in and reduce the image). Once it has done its magic it holds about 40dB or better until you tune somewhere else. If you add a second signal above -50dBm then the image increases to about -40dB. If you tune to another frequency you are back to what the softrock can do without tweaking the amplitude and phase setup which you can't do in this version. So the result is that with any signals below -50dBm the auto cancellation does nothing and you are back to the basic softrock performance which might be as bad as 20-30dB on 28MHz. By the way this test was done with signal generators not off air signals.
        > > > > > > > > > > I suspect with real flex hardware the radio is firmware calibrated over its whole tuning range to get say 40-50dB and then the auto canceller takes care of the bigger signals.
        > > > > > > > > > >
        > > > > > > > > > > Do you see these issues?
        > > > > > > > > > >
        > > > > > > > > > > Is there any way in the code to reduce the threshold at which the WBIR cuts in? Or any way to re-enable the gain and phase sliders so at least the image can be optimised with weaker signals?
        > > > > > > > > > >
        > > > > > > > > > > I would actually prefer to switch back to your powerSDR-iq v1.12.20 as it has manual gain/phase adjust plus no echo on tx, but it doesn't have the panafall display (how difficult to add panafall???).
        > > > > > > > > > >
        > > > > > > > > > > Anyway I thought I'd highlight the shortcomings I'm seeing in case anybody else has seen similar.
        > > > > > > > > > > Regards,
        > > > > > > > > > > Andrew
        > > > > > > > > > >
        > > > > > > > > >
        > > > > > > > >
        > > > > > > >
        > > > > > >
        > > > > >
        > > > >
        > > >
        > >
        >
      • Alan
        ... From: irbsurfing
        Message 3 of 24 , Jun 2, 2010
        • 0 Attachment
          ----- Original Message -----
          From: "irbsurfing" <
          Sent: Wednesday, June 02, 2010 12:20 PM
          Subject: [powersdr-iq] Re: poor performance of WBIR in powersdr1.19.3.15
          with non-flex radios



          >I get 20-35dB depending on baseband frequency and RF frequency.

          This is typical of the Softrock and presumably anything like it that has no
          hardware adjustment.

          73 Alan G4ZFQ
        • sv1eia
          Andew, Since you mention upper bands, I have to remind you this - http://groups.yahoo.com/group/powersdr-iq/message/1336 in general though, without any
          Message 4 of 24 , Jun 2, 2010
          • 0 Attachment
            Andew,

            Since you mention upper bands, I have to remind you this
            -> http://groups.yahoo.com/group/powersdr-iq/message/1336

            in general though, without any correction these are more or less also the numbers here 35db in 40m, down to 25db in 10m.

            73,
            Christos SV1EIA



            --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@...> wrote:
            >
            > Hello Christos,
            > No transformers, 6.3RxTx straight into soundcard. Also when I did the tests using the lineout fed back into the line in that was a direct connection.
            > This is puzzling, I can only assume your rigs have much better iq balance than mine.
            > Perhaps you could test this for me, if you were to run PowerSDR-IQ version 1.12.20 and set the software i/q trim to 0, what image rejection would you get in the 7/14/28MHz bands?
            > I get 20-35dB depending on baseband frequency and RF frequency.
            > Andrew
            >
            > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
            > >
            > > Hi Andrew,
            > >
            > > As I said, it works ok here so I do not have anything to say for your setup more than what I explained already.
            > > You didnt mention though if you have any transformers or filters on your setup.
            > >
            > > 73,
            > > Christos SV1EIA
            > >
            > >
            > >
            > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
            > > >
            > > > Hello Christos,
            > > > I have now done some measurements on the iq balance of various setups.
            > > >
            > > > Measurements for baseband from +/-10kHz to +/-40kHz using a gain/phase meter
            > > > Softrock 6.1 at 7.055MHz: <0.4dB/<4 degs
            > > > 6.3 TxRx plus mobo at 7MHz: <0.7dB/<5 degs
            > > > 6.3 TxRx plus mobo at 14MHz: <0.5dB/<7.5 degs
            > > > 6.3 TxRx plus mobo at 28.2MHz: <0.2dB/9.2 degs
            > > >
            > > > All measurements look good on an oscilloscope but then its not so easy to see 5 degrees error, hence using the meter.
            > > >
            > > > Generally the lower the baseband frequency the better the balance. So overall one would expect 20-35dB image suppression un-tweaked. Some offsets were within 1 degree.
            > > >
            > > > However I get the same result with any kit and any offset and that is that until signals get above about -50dBm input you get the basic performance of the hardware and sometimes worse. When you exceed -50dBm the image starts to fall eventially reaching the noise floor in some cases.
            > > >
            > > > Thought it might be my pc / sound card so I repeated the tests with another soundcard and pc with the same result.
            > > >
            > > > Another test was to feed the line inputs with a quadrature baseband signal generated using the soundcard lineout and some software. With an older version of powersdr I looked at the panadapter and adjusted the iqgenerator program for at least 30dB image suppression. I then switched to powersdr 1.19.3.15 and the image rejection was about 30dB, so far so good. Then I increased the signal to the magic -50dBm and the image started to drop to the noise floor. Finally I changed band and back again to an input below -50dBm and the image rejection had decreased to 20dB, a figure I see a lot.
            > > >
            > > > This is all very confusing, it really looks like the WBIR does nothing until this threshold is reached and in some instances makes the image worse? I wonder if anyone else is seeing this???
            > > >
            > > > I would be good if we could find out more about how the WBIR works, for example is there a maximum amplitude and phase error above which it can't correct, or an snr needed for the algorithm to work?
            > > > Is there any reference material anywhere?
            > > >
            > > > Another test I did was to tune to either 7 or 14MHz dead with an antenna connected. With good image rejection all the cw signals to the right should stop at the band edge. With poor rejection you can easily see many cw signals to the left of the band edge.(all signals were -70dBm or lower).
            > > >
            > > > Any thoughts Christos? It all causing a lot of head scratching this end!!
            > > > Thanks,
            > > > Andrew
            > > >
            > > >
            > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
            > > > >
            > > > > It works all the time for the complete incoming stream.
            > > > >
            > > > >
            > > > >
            > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
            > > > > >
            > > > > > Hi Christos,
            > > > > > Thanks for the further explanations, is there a threshold for the WBIR or does it start working right from the noise floor?
            > > > > > Thanks,
            > > > > > Andrew
            > > > > >
            > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
            > > > > > >
            > > > > > > Andrew,
            > > > > > >
            > > > > > > One more note.
            > > > > > >
            > > > > > > All these algos need 'linear' behaviour to work best.
            > > > > > > When non-linear distortion in either phase or gain of the I/Q signals happen, then the algos do not operate as they suppose to.
            > > > > > > (non-linear according to frequency axis)
            > > > > > > Avoid long cables, chokes, coils, and ANY isolation transformer after the opamp output, these are causes of non-linear phase distortion.
            > > > > > > At some discussion there was a voice that said, "ok so what, its only about 10 degrees distortion, s/w should keep up with it..", well I guess he did not knew that when we try do TRUE WBIR on real frequency agile rigs (and NOT based on center frequency), then we use floating point arithmetic having 20 something decimal digits on phase and amplitude counting on even the most tiny fraction (Complex/imaginery numbers involved too) so 10 whole degrees of non-linear deviation will throw us somewhere between the Moon and the Sun without a way back..
            > > > > > > even a fraction of a degree makes a large difference in signals around noise floor.
            > > > > > >
            > > > > > > There is already such a distortion effect in the rig itself, so adding more distortion outside of it will always make things worse than better.
            > > > > > >
            > > > > > > I hope you do get some hints now for your homemade rigs to perform better :-)
            > > > > > >
            > > > > > > 73,
            > > > > > > Christos
            > > > > > >
            > > > > > >
            > > > > > >
            > > > > > >
            > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
            > > > > > > >
            > > > > > > > Hello Christos,
            > > > > > > > Thanks for the reply and the further words about how the cancellation works. I was wondering what the coefficient was for!
            > > > > > > > This evening I have tried a fixed softrock lite receiver at 7MHz and got much the same results, perhaps a little better. I suspect that the better the i/q balance is in the first place the better the WBIR works. There is definately a degredation though when you add a second signal as you can see both images pop up 10-20dB which is interesting.
            > > > > > > > I have checked quadrature balance on the 6.3rxtx but not taken any results (I think it was within 0.5dB and about 5 degrees typically). I have a gain/phase meter so I shall report back with some hard data hopefully tomorrow.
            > > > > > > > Thanks for your help,
            > > > > > > > Andrew
            > > > > > > >
            > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
            > > > > > > > >
            > > > > > > > >
            > > > > > > > > Andrew,
            > > > > > > > >
            > > > > > > > > Nobody told you that immediately you would not see any image.
            > > > > > > > > Initially you would see images, but in an small amount of time it will be nulled out automagically.
            > > > > > > > >
            > > > > > > > > In 1.12.20 there was a coeficient put by me there that was giving to the user some 'adjustment', faster<->better for larger signals, slower<->better for smaller signal.
            > > > > > > > >
            > > > > > > > > In 1.19 WBIR, Flex has done some kind of assesment and put there similar internal coefficients but for all-cases moderate scenario and the user cannot adjust them.
            > > > > > > > >
            > > > > > > > > I hope that clear it out.
            > > > > > > > >
            > > > > > > > > Yet again, if you do see some problem there, I honestly think that you better look at the IQ outs of your rig. They need to be equal in signal level and quadrature.
            > > > > > > > > You can do a test with a o-scope attaching each x-channel to I and y-channel to Q, then goto PSDR to a frequency, eg VFO = 10MHz and inject a signal equal to VFO - IF + 1 (9992KHz) then theoretically you should have two equal x and y signals of 1KHz in 90 degrees phase diff on your o-scope.
            > > > > > > > > Do the same test at the end of the cables you use before they go into the the line-in (Rx IQ) input of your audio card.
            > > > > > > > >
            > > > > > > > > If you do not have these conditions in your signals (1KHz, equal level, quadrature), the your rig or cables needs some checkout.
            > > > > > > > >
            > > > > > > > > And do not use ANY isolation transformers, they do only damage.
            > > > > > > > >
            > > > > > > > > 73,
            > > > > > > > > Christos SV1EIa
            > > > > > > > >
            > > > > > > > >
            > > > > > > > >
            > > > > > > > >
            > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
            > > > > > > > > >
            > > > > > > > > > Hello Christos,
            > > > > > > > > > I am interested to see that you don't have these problems. So if you inject a very small signal that is below the noise floor and gradually increase its level so that it ends up at say -30dBm, you don't see any image on the panadapter at any time?
            > > > > > > > > > Any chance you could capture a wav iq file so that I could run it on my system?
            > > > > > > > > > Thanks,
            > > > > > > > > > Andrew
            > > > > > > > > >
            > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
            > > > > > > > > > >
            > > > > > > > > > > Hi Andrew,
            > > > > > > > > > >
            > > > > > > > > > > I do not share your experience.
            > > > > > > > > > > Here it works absolutely marvelous and I have three SR 6.2 rigs.
            > > > > > > > > > > The same has been reported by anyone else I know it has it working.
            > > > > > > > > > > One common though problem is rig's I/Q out.
            > > > > > > > > > > Check your I/Q output, it is probably having a problem either cabling or opamps are damaged.
            > > > > > > > > > >
            > > > > > > > > > > Also in 1.12.20, the EIA is actually the same algo as WBIR..
            > > > > > > > > > > so if you have the exact same cabling for both applications and it works ok in one but not ok when you switch to the other, then ..I do not know what it is!! :-)
            > > > > > > > > > >
            > > > > > > > > > > 73,
            > > > > > > > > > > Christos
            > > > > > > > > > >
            > > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
            > > > > > > > > > > >
            > > > > > > > > > > > Hi Christos,
            > > > > > > > > > > > I am finding that the WBIR feature doesn't work very well at all with my softrock + mobo radio.
            > > > > > > > > > > >
            > > > > > > > > > > > I find that the automatic image cancellation only seems to start working with input signals above -50dBm (you can see it kick in and reduce the image). Once it has done its magic it holds about 40dB or better until you tune somewhere else. If you add a second signal above -50dBm then the image increases to about -40dB. If you tune to another frequency you are back to what the softrock can do without tweaking the amplitude and phase setup which you can't do in this version. So the result is that with any signals below -50dBm the auto cancellation does nothing and you are back to the basic softrock performance which might be as bad as 20-30dB on 28MHz. By the way this test was done with signal generators not off air signals.
            > > > > > > > > > > > I suspect with real flex hardware the radio is firmware calibrated over its whole tuning range to get say 40-50dB and then the auto canceller takes care of the bigger signals.
            > > > > > > > > > > >
            > > > > > > > > > > > Do you see these issues?
            > > > > > > > > > > >
            > > > > > > > > > > > Is there any way in the code to reduce the threshold at which the WBIR cuts in? Or any way to re-enable the gain and phase sliders so at least the image can be optimised with weaker signals?
            > > > > > > > > > > >
            > > > > > > > > > > > I would actually prefer to switch back to your powerSDR-iq v1.12.20 as it has manual gain/phase adjust plus no echo on tx, but it doesn't have the panafall display (how difficult to add panafall???).
            > > > > > > > > > > >
            > > > > > > > > > > > Anyway I thought I'd highlight the shortcomings I'm seeing in case anybody else has seen similar.
            > > > > > > > > > > > Regards,
            > > > > > > > > > > > Andrew
            > > > > > > > > > > >
            > > > > > > > > > >
            > > > > > > > > >
            > > > > > > > >
            > > > > > > >
            > > > > > >
            > > > > >
            > > > >
            > > >
            > >
            >
          • irbsurfing
            Hi Christos, Your numbers are not that different to mine, so for some reason you don t see the same issues as me. What sound card do you use? Thanks, Andrew
            Message 5 of 24 , Jun 2, 2010
            • 0 Attachment
              Hi Christos,
              Your numbers are not that different to mine, so for some reason you don't see the same issues as me. What sound card do you use?
              Thanks,
              Andrew

              --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@...> wrote:
              >
              > Andew,
              >
              > Since you mention upper bands, I have to remind you this
              > -> http://groups.yahoo.com/group/powersdr-iq/message/1336
              >
              > in general though, without any correction these are more or less also the numbers here 35db in 40m, down to 25db in 10m.
              >
              > 73,
              > Christos SV1EIA
              >
              >
              >
              > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
              > >
              > > Hello Christos,
              > > No transformers, 6.3RxTx straight into soundcard. Also when I did the tests using the lineout fed back into the line in that was a direct connection.
              > > This is puzzling, I can only assume your rigs have much better iq balance than mine.
              > > Perhaps you could test this for me, if you were to run PowerSDR-IQ version 1.12.20 and set the software i/q trim to 0, what image rejection would you get in the 7/14/28MHz bands?
              > > I get 20-35dB depending on baseband frequency and RF frequency.
              > > Andrew
              > >
              > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
              > > >
              > > > Hi Andrew,
              > > >
              > > > As I said, it works ok here so I do not have anything to say for your setup more than what I explained already.
              > > > You didnt mention though if you have any transformers or filters on your setup.
              > > >
              > > > 73,
              > > > Christos SV1EIA
              > > >
              > > >
              > > >
              > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
              > > > >
              > > > > Hello Christos,
              > > > > I have now done some measurements on the iq balance of various setups.
              > > > >
              > > > > Measurements for baseband from +/-10kHz to +/-40kHz using a gain/phase meter
              > > > > Softrock 6.1 at 7.055MHz: <0.4dB/<4 degs
              > > > > 6.3 TxRx plus mobo at 7MHz: <0.7dB/<5 degs
              > > > > 6.3 TxRx plus mobo at 14MHz: <0.5dB/<7.5 degs
              > > > > 6.3 TxRx plus mobo at 28.2MHz: <0.2dB/9.2 degs
              > > > >
              > > > > All measurements look good on an oscilloscope but then its not so easy to see 5 degrees error, hence using the meter.
              > > > >
              > > > > Generally the lower the baseband frequency the better the balance. So overall one would expect 20-35dB image suppression un-tweaked. Some offsets were within 1 degree.
              > > > >
              > > > > However I get the same result with any kit and any offset and that is that until signals get above about -50dBm input you get the basic performance of the hardware and sometimes worse. When you exceed -50dBm the image starts to fall eventially reaching the noise floor in some cases.
              > > > >
              > > > > Thought it might be my pc / sound card so I repeated the tests with another soundcard and pc with the same result.
              > > > >
              > > > > Another test was to feed the line inputs with a quadrature baseband signal generated using the soundcard lineout and some software. With an older version of powersdr I looked at the panadapter and adjusted the iqgenerator program for at least 30dB image suppression. I then switched to powersdr 1.19.3.15 and the image rejection was about 30dB, so far so good. Then I increased the signal to the magic -50dBm and the image started to drop to the noise floor. Finally I changed band and back again to an input below -50dBm and the image rejection had decreased to 20dB, a figure I see a lot.
              > > > >
              > > > > This is all very confusing, it really looks like the WBIR does nothing until this threshold is reached and in some instances makes the image worse? I wonder if anyone else is seeing this???
              > > > >
              > > > > I would be good if we could find out more about how the WBIR works, for example is there a maximum amplitude and phase error above which it can't correct, or an snr needed for the algorithm to work?
              > > > > Is there any reference material anywhere?
              > > > >
              > > > > Another test I did was to tune to either 7 or 14MHz dead with an antenna connected. With good image rejection all the cw signals to the right should stop at the band edge. With poor rejection you can easily see many cw signals to the left of the band edge.(all signals were -70dBm or lower).
              > > > >
              > > > > Any thoughts Christos? It all causing a lot of head scratching this end!!
              > > > > Thanks,
              > > > > Andrew
              > > > >
              > > > >
              > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
              > > > > >
              > > > > > It works all the time for the complete incoming stream.
              > > > > >
              > > > > >
              > > > > >
              > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
              > > > > > >
              > > > > > > Hi Christos,
              > > > > > > Thanks for the further explanations, is there a threshold for the WBIR or does it start working right from the noise floor?
              > > > > > > Thanks,
              > > > > > > Andrew
              > > > > > >
              > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
              > > > > > > >
              > > > > > > > Andrew,
              > > > > > > >
              > > > > > > > One more note.
              > > > > > > >
              > > > > > > > All these algos need 'linear' behaviour to work best.
              > > > > > > > When non-linear distortion in either phase or gain of the I/Q signals happen, then the algos do not operate as they suppose to.
              > > > > > > > (non-linear according to frequency axis)
              > > > > > > > Avoid long cables, chokes, coils, and ANY isolation transformer after the opamp output, these are causes of non-linear phase distortion.
              > > > > > > > At some discussion there was a voice that said, "ok so what, its only about 10 degrees distortion, s/w should keep up with it..", well I guess he did not knew that when we try do TRUE WBIR on real frequency agile rigs (and NOT based on center frequency), then we use floating point arithmetic having 20 something decimal digits on phase and amplitude counting on even the most tiny fraction (Complex/imaginery numbers involved too) so 10 whole degrees of non-linear deviation will throw us somewhere between the Moon and the Sun without a way back..
              > > > > > > > even a fraction of a degree makes a large difference in signals around noise floor.
              > > > > > > >
              > > > > > > > There is already such a distortion effect in the rig itself, so adding more distortion outside of it will always make things worse than better.
              > > > > > > >
              > > > > > > > I hope you do get some hints now for your homemade rigs to perform better :-)
              > > > > > > >
              > > > > > > > 73,
              > > > > > > > Christos
              > > > > > > >
              > > > > > > >
              > > > > > > >
              > > > > > > >
              > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
              > > > > > > > >
              > > > > > > > > Hello Christos,
              > > > > > > > > Thanks for the reply and the further words about how the cancellation works. I was wondering what the coefficient was for!
              > > > > > > > > This evening I have tried a fixed softrock lite receiver at 7MHz and got much the same results, perhaps a little better. I suspect that the better the i/q balance is in the first place the better the WBIR works. There is definately a degredation though when you add a second signal as you can see both images pop up 10-20dB which is interesting.
              > > > > > > > > I have checked quadrature balance on the 6.3rxtx but not taken any results (I think it was within 0.5dB and about 5 degrees typically). I have a gain/phase meter so I shall report back with some hard data hopefully tomorrow.
              > > > > > > > > Thanks for your help,
              > > > > > > > > Andrew
              > > > > > > > >
              > > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
              > > > > > > > > >
              > > > > > > > > >
              > > > > > > > > > Andrew,
              > > > > > > > > >
              > > > > > > > > > Nobody told you that immediately you would not see any image.
              > > > > > > > > > Initially you would see images, but in an small amount of time it will be nulled out automagically.
              > > > > > > > > >
              > > > > > > > > > In 1.12.20 there was a coeficient put by me there that was giving to the user some 'adjustment', faster<->better for larger signals, slower<->better for smaller signal.
              > > > > > > > > >
              > > > > > > > > > In 1.19 WBIR, Flex has done some kind of assesment and put there similar internal coefficients but for all-cases moderate scenario and the user cannot adjust them.
              > > > > > > > > >
              > > > > > > > > > I hope that clear it out.
              > > > > > > > > >
              > > > > > > > > > Yet again, if you do see some problem there, I honestly think that you better look at the IQ outs of your rig. They need to be equal in signal level and quadrature.
              > > > > > > > > > You can do a test with a o-scope attaching each x-channel to I and y-channel to Q, then goto PSDR to a frequency, eg VFO = 10MHz and inject a signal equal to VFO - IF + 1 (9992KHz) then theoretically you should have two equal x and y signals of 1KHz in 90 degrees phase diff on your o-scope.
              > > > > > > > > > Do the same test at the end of the cables you use before they go into the the line-in (Rx IQ) input of your audio card.
              > > > > > > > > >
              > > > > > > > > > If you do not have these conditions in your signals (1KHz, equal level, quadrature), the your rig or cables needs some checkout.
              > > > > > > > > >
              > > > > > > > > > And do not use ANY isolation transformers, they do only damage.
              > > > > > > > > >
              > > > > > > > > > 73,
              > > > > > > > > > Christos SV1EIa
              > > > > > > > > >
              > > > > > > > > >
              > > > > > > > > >
              > > > > > > > > >
              > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
              > > > > > > > > > >
              > > > > > > > > > > Hello Christos,
              > > > > > > > > > > I am interested to see that you don't have these problems. So if you inject a very small signal that is below the noise floor and gradually increase its level so that it ends up at say -30dBm, you don't see any image on the panadapter at any time?
              > > > > > > > > > > Any chance you could capture a wav iq file so that I could run it on my system?
              > > > > > > > > > > Thanks,
              > > > > > > > > > > Andrew
              > > > > > > > > > >
              > > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
              > > > > > > > > > > >
              > > > > > > > > > > > Hi Andrew,
              > > > > > > > > > > >
              > > > > > > > > > > > I do not share your experience.
              > > > > > > > > > > > Here it works absolutely marvelous and I have three SR 6.2 rigs.
              > > > > > > > > > > > The same has been reported by anyone else I know it has it working.
              > > > > > > > > > > > One common though problem is rig's I/Q out.
              > > > > > > > > > > > Check your I/Q output, it is probably having a problem either cabling or opamps are damaged.
              > > > > > > > > > > >
              > > > > > > > > > > > Also in 1.12.20, the EIA is actually the same algo as WBIR..
              > > > > > > > > > > > so if you have the exact same cabling for both applications and it works ok in one but not ok when you switch to the other, then ..I do not know what it is!! :-)
              > > > > > > > > > > >
              > > > > > > > > > > > 73,
              > > > > > > > > > > > Christos
              > > > > > > > > > > >
              > > > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
              > > > > > > > > > > > >
              > > > > > > > > > > > > Hi Christos,
              > > > > > > > > > > > > I am finding that the WBIR feature doesn't work very well at all with my softrock + mobo radio.
              > > > > > > > > > > > >
              > > > > > > > > > > > > I find that the automatic image cancellation only seems to start working with input signals above -50dBm (you can see it kick in and reduce the image). Once it has done its magic it holds about 40dB or better until you tune somewhere else. If you add a second signal above -50dBm then the image increases to about -40dB. If you tune to another frequency you are back to what the softrock can do without tweaking the amplitude and phase setup which you can't do in this version. So the result is that with any signals below -50dBm the auto cancellation does nothing and you are back to the basic softrock performance which might be as bad as 20-30dB on 28MHz. By the way this test was done with signal generators not off air signals.
              > > > > > > > > > > > > I suspect with real flex hardware the radio is firmware calibrated over its whole tuning range to get say 40-50dB and then the auto canceller takes care of the bigger signals.
              > > > > > > > > > > > >
              > > > > > > > > > > > > Do you see these issues?
              > > > > > > > > > > > >
              > > > > > > > > > > > > Is there any way in the code to reduce the threshold at which the WBIR cuts in? Or any way to re-enable the gain and phase sliders so at least the image can be optimised with weaker signals?
              > > > > > > > > > > > >
              > > > > > > > > > > > > I would actually prefer to switch back to your powerSDR-iq v1.12.20 as it has manual gain/phase adjust plus no echo on tx, but it doesn't have the panafall display (how difficult to add panafall???).
              > > > > > > > > > > > >
              > > > > > > > > > > > > Anyway I thought I'd highlight the shortcomings I'm seeing in case anybody else has seen similar.
              > > > > > > > > > > > > Regards,
              > > > > > > > > > > > > Andrew
              > > > > > > > > > > > >
              > > > > > > > > > > >
              > > > > > > > > > >
              > > > > > > > > >
              > > > > > > > >
              > > > > > > >
              > > > > > >
              > > > > >
              > > > >
              > > >
              > >
              >
            • sv1eia
              Its a Terratec Phase 24 FW.
              Message 6 of 24 , Jun 2, 2010
              • 0 Attachment
                Its a Terratec Phase 24 FW.


                --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@...> wrote:
                >
                > Hi Christos,
                > Your numbers are not that different to mine, so for some reason you don't see the same issues as me. What sound card do you use?
                > Thanks,
                > Andrew
                >
                > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
                > >
                > > Andew,
                > >
                > > Since you mention upper bands, I have to remind you this
                > > -> http://groups.yahoo.com/group/powersdr-iq/message/1336
                > >
                > > in general though, without any correction these are more or less also the numbers here 35db in 40m, down to 25db in 10m.
                > >
                > > 73,
                > > Christos SV1EIA
                > >
                > >
                > >
                > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
                > > >
                > > > Hello Christos,
                > > > No transformers, 6.3RxTx straight into soundcard. Also when I did the tests using the lineout fed back into the line in that was a direct connection.
                > > > This is puzzling, I can only assume your rigs have much better iq balance than mine.
                > > > Perhaps you could test this for me, if you were to run PowerSDR-IQ version 1.12.20 and set the software i/q trim to 0, what image rejection would you get in the 7/14/28MHz bands?
                > > > I get 20-35dB depending on baseband frequency and RF frequency.
                > > > Andrew
                > > >
                > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
                > > > >
                > > > > Hi Andrew,
                > > > >
                > > > > As I said, it works ok here so I do not have anything to say for your setup more than what I explained already.
                > > > > You didnt mention though if you have any transformers or filters on your setup.
                > > > >
                > > > > 73,
                > > > > Christos SV1EIA
                > > > >
                > > > >
                > > > >
                > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
                > > > > >
                > > > > > Hello Christos,
                > > > > > I have now done some measurements on the iq balance of various setups.
                > > > > >
                > > > > > Measurements for baseband from +/-10kHz to +/-40kHz using a gain/phase meter
                > > > > > Softrock 6.1 at 7.055MHz: <0.4dB/<4 degs
                > > > > > 6.3 TxRx plus mobo at 7MHz: <0.7dB/<5 degs
                > > > > > 6.3 TxRx plus mobo at 14MHz: <0.5dB/<7.5 degs
                > > > > > 6.3 TxRx plus mobo at 28.2MHz: <0.2dB/9.2 degs
                > > > > >
                > > > > > All measurements look good on an oscilloscope but then its not so easy to see 5 degrees error, hence using the meter.
                > > > > >
                > > > > > Generally the lower the baseband frequency the better the balance. So overall one would expect 20-35dB image suppression un-tweaked. Some offsets were within 1 degree.
                > > > > >
                > > > > > However I get the same result with any kit and any offset and that is that until signals get above about -50dBm input you get the basic performance of the hardware and sometimes worse. When you exceed -50dBm the image starts to fall eventially reaching the noise floor in some cases.
                > > > > >
                > > > > > Thought it might be my pc / sound card so I repeated the tests with another soundcard and pc with the same result.
                > > > > >
                > > > > > Another test was to feed the line inputs with a quadrature baseband signal generated using the soundcard lineout and some software. With an older version of powersdr I looked at the panadapter and adjusted the iqgenerator program for at least 30dB image suppression. I then switched to powersdr 1.19.3.15 and the image rejection was about 30dB, so far so good. Then I increased the signal to the magic -50dBm and the image started to drop to the noise floor. Finally I changed band and back again to an input below -50dBm and the image rejection had decreased to 20dB, a figure I see a lot.
                > > > > >
                > > > > > This is all very confusing, it really looks like the WBIR does nothing until this threshold is reached and in some instances makes the image worse? I wonder if anyone else is seeing this???
                > > > > >
                > > > > > I would be good if we could find out more about how the WBIR works, for example is there a maximum amplitude and phase error above which it can't correct, or an snr needed for the algorithm to work?
                > > > > > Is there any reference material anywhere?
                > > > > >
                > > > > > Another test I did was to tune to either 7 or 14MHz dead with an antenna connected. With good image rejection all the cw signals to the right should stop at the band edge. With poor rejection you can easily see many cw signals to the left of the band edge.(all signals were -70dBm or lower).
                > > > > >
                > > > > > Any thoughts Christos? It all causing a lot of head scratching this end!!
                > > > > > Thanks,
                > > > > > Andrew
                > > > > >
                > > > > >
                > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
                > > > > > >
                > > > > > > It works all the time for the complete incoming stream.
                > > > > > >
                > > > > > >
                > > > > > >
                > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
                > > > > > > >
                > > > > > > > Hi Christos,
                > > > > > > > Thanks for the further explanations, is there a threshold for the WBIR or does it start working right from the noise floor?
                > > > > > > > Thanks,
                > > > > > > > Andrew
                > > > > > > >
                > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
                > > > > > > > >
                > > > > > > > > Andrew,
                > > > > > > > >
                > > > > > > > > One more note.
                > > > > > > > >
                > > > > > > > > All these algos need 'linear' behaviour to work best.
                > > > > > > > > When non-linear distortion in either phase or gain of the I/Q signals happen, then the algos do not operate as they suppose to.
                > > > > > > > > (non-linear according to frequency axis)
                > > > > > > > > Avoid long cables, chokes, coils, and ANY isolation transformer after the opamp output, these are causes of non-linear phase distortion.
                > > > > > > > > At some discussion there was a voice that said, "ok so what, its only about 10 degrees distortion, s/w should keep up with it..", well I guess he did not knew that when we try do TRUE WBIR on real frequency agile rigs (and NOT based on center frequency), then we use floating point arithmetic having 20 something decimal digits on phase and amplitude counting on even the most tiny fraction (Complex/imaginery numbers involved too) so 10 whole degrees of non-linear deviation will throw us somewhere between the Moon and the Sun without a way back..
                > > > > > > > > even a fraction of a degree makes a large difference in signals around noise floor.
                > > > > > > > >
                > > > > > > > > There is already such a distortion effect in the rig itself, so adding more distortion outside of it will always make things worse than better.
                > > > > > > > >
                > > > > > > > > I hope you do get some hints now for your homemade rigs to perform better :-)
                > > > > > > > >
                > > > > > > > > 73,
                > > > > > > > > Christos
                > > > > > > > >
                > > > > > > > >
                > > > > > > > >
                > > > > > > > >
                > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
                > > > > > > > > >
                > > > > > > > > > Hello Christos,
                > > > > > > > > > Thanks for the reply and the further words about how the cancellation works. I was wondering what the coefficient was for!
                > > > > > > > > > This evening I have tried a fixed softrock lite receiver at 7MHz and got much the same results, perhaps a little better. I suspect that the better the i/q balance is in the first place the better the WBIR works. There is definately a degredation though when you add a second signal as you can see both images pop up 10-20dB which is interesting.
                > > > > > > > > > I have checked quadrature balance on the 6.3rxtx but not taken any results (I think it was within 0.5dB and about 5 degrees typically). I have a gain/phase meter so I shall report back with some hard data hopefully tomorrow.
                > > > > > > > > > Thanks for your help,
                > > > > > > > > > Andrew
                > > > > > > > > >
                > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
                > > > > > > > > > >
                > > > > > > > > > >
                > > > > > > > > > > Andrew,
                > > > > > > > > > >
                > > > > > > > > > > Nobody told you that immediately you would not see any image.
                > > > > > > > > > > Initially you would see images, but in an small amount of time it will be nulled out automagically.
                > > > > > > > > > >
                > > > > > > > > > > In 1.12.20 there was a coeficient put by me there that was giving to the user some 'adjustment', faster<->better for larger signals, slower<->better for smaller signal.
                > > > > > > > > > >
                > > > > > > > > > > In 1.19 WBIR, Flex has done some kind of assesment and put there similar internal coefficients but for all-cases moderate scenario and the user cannot adjust them.
                > > > > > > > > > >
                > > > > > > > > > > I hope that clear it out.
                > > > > > > > > > >
                > > > > > > > > > > Yet again, if you do see some problem there, I honestly think that you better look at the IQ outs of your rig. They need to be equal in signal level and quadrature.
                > > > > > > > > > > You can do a test with a o-scope attaching each x-channel to I and y-channel to Q, then goto PSDR to a frequency, eg VFO = 10MHz and inject a signal equal to VFO - IF + 1 (9992KHz) then theoretically you should have two equal x and y signals of 1KHz in 90 degrees phase diff on your o-scope.
                > > > > > > > > > > Do the same test at the end of the cables you use before they go into the the line-in (Rx IQ) input of your audio card.
                > > > > > > > > > >
                > > > > > > > > > > If you do not have these conditions in your signals (1KHz, equal level, quadrature), the your rig or cables needs some checkout.
                > > > > > > > > > >
                > > > > > > > > > > And do not use ANY isolation transformers, they do only damage.
                > > > > > > > > > >
                > > > > > > > > > > 73,
                > > > > > > > > > > Christos SV1EIa
                > > > > > > > > > >
                > > > > > > > > > >
                > > > > > > > > > >
                > > > > > > > > > >
                > > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
                > > > > > > > > > > >
                > > > > > > > > > > > Hello Christos,
                > > > > > > > > > > > I am interested to see that you don't have these problems. So if you inject a very small signal that is below the noise floor and gradually increase its level so that it ends up at say -30dBm, you don't see any image on the panadapter at any time?
                > > > > > > > > > > > Any chance you could capture a wav iq file so that I could run it on my system?
                > > > > > > > > > > > Thanks,
                > > > > > > > > > > > Andrew
                > > > > > > > > > > >
                > > > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "sv1eia" <sv1eia@> wrote:
                > > > > > > > > > > > >
                > > > > > > > > > > > > Hi Andrew,
                > > > > > > > > > > > >
                > > > > > > > > > > > > I do not share your experience.
                > > > > > > > > > > > > Here it works absolutely marvelous and I have three SR 6.2 rigs.
                > > > > > > > > > > > > The same has been reported by anyone else I know it has it working.
                > > > > > > > > > > > > One common though problem is rig's I/Q out.
                > > > > > > > > > > > > Check your I/Q output, it is probably having a problem either cabling or opamps are damaged.
                > > > > > > > > > > > >
                > > > > > > > > > > > > Also in 1.12.20, the EIA is actually the same algo as WBIR..
                > > > > > > > > > > > > so if you have the exact same cabling for both applications and it works ok in one but not ok when you switch to the other, then ..I do not know what it is!! :-)
                > > > > > > > > > > > >
                > > > > > > > > > > > > 73,
                > > > > > > > > > > > > Christos
                > > > > > > > > > > > >
                > > > > > > > > > > > > --- In powersdr-iq@yahoogroups.com, "irbsurfing" <irbsurfing@> wrote:
                > > > > > > > > > > > > >
                > > > > > > > > > > > > > Hi Christos,
                > > > > > > > > > > > > > I am finding that the WBIR feature doesn't work very well at all with my softrock + mobo radio.
                > > > > > > > > > > > > >
                > > > > > > > > > > > > > I find that the automatic image cancellation only seems to start working with input signals above -50dBm (you can see it kick in and reduce the image). Once it has done its magic it holds about 40dB or better until you tune somewhere else. If you add a second signal above -50dBm then the image increases to about -40dB. If you tune to another frequency you are back to what the softrock can do without tweaking the amplitude and phase setup which you can't do in this version. So the result is that with any signals below -50dBm the auto cancellation does nothing and you are back to the basic softrock performance which might be as bad as 20-30dB on 28MHz. By the way this test was done with signal generators not off air signals.
                > > > > > > > > > > > > > I suspect with real flex hardware the radio is firmware calibrated over its whole tuning range to get say 40-50dB and then the auto canceller takes care of the bigger signals.
                > > > > > > > > > > > > >
                > > > > > > > > > > > > > Do you see these issues?
                > > > > > > > > > > > > >
                > > > > > > > > > > > > > Is there any way in the code to reduce the threshold at which the WBIR cuts in? Or any way to re-enable the gain and phase sliders so at least the image can be optimised with weaker signals?
                > > > > > > > > > > > > >
                > > > > > > > > > > > > > I would actually prefer to switch back to your powerSDR-iq v1.12.20 as it has manual gain/phase adjust plus no echo on tx, but it doesn't have the panafall display (how difficult to add panafall???).
                > > > > > > > > > > > > >
                > > > > > > > > > > > > > Anyway I thought I'd highlight the shortcomings I'm seeing in case anybody else has seen similar.
                > > > > > > > > > > > > > Regards,
                > > > > > > > > > > > > > Andrew
                > > > > > > > > > > > > >
                > > > > > > > > > > > >
                > > > > > > > > > > >
                > > > > > > > > > >
                > > > > > > > > >
                > > > > > > > >
                > > > > > > >
                > > > > > >
                > > > > >
                > > > >
                > > >
                > >
                >
              • Rob
                Hi Christos et. al. Sorry to make this thread rumble on. I did tests on image rejection at 27.5 and 28.5MHz using a signal generator to broadcast a (small)
                Message 7 of 24 , Jun 5, 2010
                • 0 Attachment
                  Hi Christos et. al.

                  Sorry to make this thread rumble on. I did tests on image rejection at 27.5 and 28.5MHz using a signal generator to broadcast a (small) signal.

                  My antenna is only a small modified CB vertical for 10m, so rarely receives signals above -60dBm. I used v1.12 and v1.19 beta for these tests.

                  With the sig gen turned down, and PSDR started up, I see an arbitary starting value of -90dBm on the S meter. The image reads -110dBm, only 20dB down. The system sits at these values indefinetly. In v1.12 I have EIA set to 0.300.

                  Increasing the sig gen output, the image rises with the real signal. I reach -60dBm (approximately) on the S meter. In a few seconds in v1.12, the image drops to the noise floor. On v1.19 beta, the image also drops to the noise floor, but more slowly. Perhaps the EIA setting will control the speed of the "nulling".

                  Obviously the S meter calibration affects the absolute level, and my calibration is not accurate - it was done against the voltage readings from a scope rather than a calibrated attenuator on a proper signal generator. I am using a basic PLL project I made years ago as a signal gen.

                  Anyway my conclusion is ALL versions of PSDR-IQ need a clean signal of 60dB or more above the noise floor on the higher frequencies before the image null kicks in. On lower bands this happens easily - those broadcasters on 7.2MHz are high like Mt. Everest!

                  Poor I/Q balance here may play a part, but I humbly suggest there is a software problem somewhere.

                  Can anyone repeat these tests? The rest of my system is:

                  Edirol FA-66
                  SR V6.3 with MoBo V4.3
                  All latest drivers installed AFAIK

                  And please don't be upset Christos that this thread has continued... we need your software expertise desperately for our Softrock rigs, and I know other people on the group will second me on that.

                  73s,
                  Rob
                  (M0RZF)
                • uli
                  Hi Rob, hi Christos, using here WinXP, EMU-0202 PowerSDR-V1.12.24-SV1EIA and SoftRock V9. Now my measurements on 10m/40m. LO=28.040, Signal=28.060 and
                  Message 8 of 24 , Jun 5, 2010
                  • 0 Attachment
                    Hi Rob, hi Christos,

                    using here WinXP, EMU-0202 PowerSDR-V1.12.24-SV1EIA and SoftRock V9.
                    Now my measurements on 10m/40m.

                    LO=28.040, Signal=28.060 and Image=28.020, Noisefloor abt. -135dBm
                    Signal Image(without EIA) Image(with EIA)
                    -88dBm -100dBm -107dBm
                    -83dBm -95dBm -106dBm
                    -73dBm -88dBm -106dBm
                    -63dBm -78dBm -114dBm
                    -53dBm -68dBm -123dBm
                    -45dBm -58dBm in the noisefloor !!!!!

                    LO=7.040, Signal=7.060 and Image=7.020, Noisefloor abt. -135dBm
                    Signal Image(without EIA) Image(with EIA)
                    -88dBm -103dBm -121dBm
                    -83dBm -98dBm -121dBm
                    -73dBm -88dBm -126dBm
                    -63dBm -78dBm in the noise
                    -53dBm -68dBm in the noise
                    -43dBm -58dBm in the noise

                    Hope this helps

                    73
                    Uli
                    DF5SF





                    > Hi Christos et. al.
                    >
                    > Sorry to make this thread rumble on. I did tests on image rejection at 27.5 and 28.5MHz using a signal generator to broadcast a (small) signal.
                    >
                    > My antenna is only a small modified CB vertical for 10m, so rarely receives signals above -60dBm. I used v1.12 and v1.19 beta for these tests.
                    >
                    > With the sig gen turned down, and PSDR started up, I see an arbitary starting value of -90dBm on the S meter. The image reads -110dBm, only 20dB down. The system sits at these values indefinetly. In v1.12 I have EIA set to 0.300.
                    >
                    > Increasing the sig gen output, the image rises with the real signal. I reach -60dBm (approximately) on the S meter. In a few seconds in v1.12, the image drops to the noise floor. On v1.19 beta, the image also drops to the noise floor, but more slowly. Perhaps the EIA setting will control the speed of the "nulling".
                    >
                    > Obviously the S meter calibration affects the absolute level, and my calibration is not accurate - it was done against the voltage readings from a scope rather than a calibrated attenuator on a proper signal generator. I am using a basic PLL project I made years ago as a signal gen.
                    >
                    > Anyway my conclusion is ALL versions of PSDR-IQ need a clean signal of 60dB or more above the noise floor on the higher frequencies before the image null kicks in. On lower bands this happens easily - those broadcasters on 7.2MHz are high like Mt. Everest!
                    >
                    > Poor I/Q balance here may play a part, but I humbly suggest there is a software problem somewhere.
                    >
                    > Can anyone repeat these tests? The rest of my system is:
                    >
                    > Edirol FA-66
                    > SR V6.3 with MoBo V4.3
                    > All latest drivers installed AFAIK
                    >
                    > And please don't be upset Christos that this thread has continued... we need your software expertise desperately for our Softrock rigs, and I know other people on the group will second me on that.
                    >
                    > 73s,
                    > Rob
                    > (M0RZF)
                    >
                    >
                    >
                    >
                    >
                    > ------------------------------------
                    >
                    > Yahoo! Groups Links
                    >
                    >
                    >
                  • Pete Ferrand
                    ... Yes, I m seeing exactly what Rob is seeing here, with slightly different hardware. The amplitude numbers I m seeing, with limited calibration gear, are in
                    Message 9 of 24 , Jun 5, 2010
                    • 0 Attachment
                      Rob <robbrown@...> writes:

                      >Anyway my conclusion is ALL versions of PSDR-IQ need a clean signal of 60dB or more above the noise floor on the higher frequencies before the image null kicks in. On lower bands this happens easily - those broadcasters on 7.2MHz are high like Mt. Everest!
                      >
                      >Poor I/Q balance here may play a part, but I humbly suggest there is a software problem somewhere.
                      >
                      >Can anyone repeat these tests? The rest of my system is:


                      Yes, I'm seeing exactly what Rob is seeing here, with slightly different hardware. The amplitude numbers I'm seeing, with limited calibration gear, are in line with Rob's statements. Whilst I have a reasonably sized vee beam, here in the upper midwest of the USA there are no powerful broadcasters around the ham bands, so I have never seen any station loud enough to trigger the automatic interference reduction. The use of a signal generator as Rob describes will do it.

                      As I am running Windows 7 I have only used the SoftRock with 1.19.3.15. The rest of the station is the same as I have used successfully with my SDR-1000, a Delta 44. I am not using any transformers in the audio lines although I have tried and found they make no difference with the above described effect. I am using a 6.3 in its stock configuration with the I2C USB adapter, no mobo.

                      The only possible quibble I have with Rob is the need for a clean signal. I have tried the same operation with a signal with about 30% 60Hz hum modulation and found the WBIR takes hold, but only at a level that no actual signal can matc. Most of my experiments have been conducted in the 20 meter band, but I note the same effect on the others.

                      I am very grateful for Christos for developing this package, and I'm sure this is a solvable problem for anyone who undestands the workings of the WBIR mechanism.

                      Thanks to all for any help.

                      -Pete
                      WB2QLL
                      Somers, WI
                    • Rob
                      Hi Folks, I carefully read all previous posts on this topic and thought about it overnight. We can see that the algorithm needs a good reference before it
                      Message 10 of 24 , Jun 6, 2010
                      • 0 Attachment
                        Hi Folks,

                        I carefully read all previous posts on this topic and thought about it overnight. We can see that the algorithm needs a good reference before it starts to null the image. To null on a noisy signal with poor balance may make things worse. So they don't do it.

                        >Anyway my conclusion is ALL versions of PSDR-IQ need a clean signal >of 60dB or more above the noise floor on the higher frequencies >before the image null kicks in.

                        So why not provide one? Switch on a signal gen and do an "image null" on the offending band before using the system. That's a workaround.

                        Apparently the software does not store the image rejection settings between restarts. It maybe expensive on disk space to do so, and also need manual intervention to tell it when the image is nulled OK.

                        > The only possible quibble I have with Rob is the need for a clean >signal. I have tried the same operation with a signal with about 30% >60Hz hum modulation and found the WBIR takes hold, but only at a >level that no actual signal can matc. Most of my experiments have >been conducted in the 20 meter band, but I note the same effect on >the others.

                        Agreed Pete. Anything with a visible "carrier" like an AM signal but not SSB will work.

                        OK, I will stop taking bandwidth now!

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