Removing Zero IF Image signals in FM Broadcast Band
Posted: Mon Feb 02, 2015 11:06 pm
When first using SDRPlay I noticed that an image of the strongest station - Heart 102.8 - was present. It turns out that this is an inherent feature of SDR receivers that use Zero IF Quadrature Switching Detector - of which SDRPlay is one. In order to get the best image (i.e. unwanted sideband) rejection the amplitude difference between the I and Q channels must be as close to zero as possible and the 90 degree phase difference must be very precise. An article in the RSGB Handbook mentions 0.01dB and 0.01 degrees respectively to achieve 60dB rejection (RSGB Handbook).
Wherever there was a strong signal to one side of the centre frequency I could find its corresponding image on the other side of the centre if the frequency wasn't already occupied by a genuine signal. For signals that were less than around 45dB above the noise floor the inmages were too weak to detect.
Ian directed me to the option in HDSDR to adjust the phase/amplitude balance manually. [Options/Input Channel Calibration Rx].
This picture shows the presence of Heart 102.8 at 103.8 on the oppsoite side of a centre frequency of 103.3. The amplitude and phase adjustments had been reset to zero. https://app.box.com/s/0z0u9gu5sjppskhxm8w4hr8dg9bkxzku
This picture shows what happens whens the amplitude and phase are corrected to minimise the image https://app.box.com/s/21xgc53grzp6f8lzeznquoav20fluppi I could now hear a much weaker station on 103.8
I have found that it is possible to have one setting which is effective across the whole of the FM Band at least when a wide IF bandwidth is used. When an IF bandwidth of 1.536 MHz was used I found the settings were appreciably different from the setting needed at the wider bandwidth and were noticeably different from one end of the band to the other. Also when using a 5MHz bandwidth I've found that the amplitude doesn't need any tweaking once it is set, but from one day to another a slight tweak of the phase might be needed to re-optimise suppression.
Now SDR Console and SDR Sharp both, I understand from the forums, have automatic correction of the I and Q balance. I've checked on MW and on SW and can find no image signals from poor suppression of the unwanted sideband with either of these software packages - but there the image of Heart 102.8 is very noticeable on both. My speculation is that the algorithms used have been developed for the HF bands when QSD SDR receivers were all relatively narrow bandwidth and HF only. I also gather that further work on improving the auto correction is in mind.
Using the HDSDR software - with no further adjustment from those arrived at for the FM Band I could find no images from poor sideband suppression in Medium Wave or Shortwave. But please take note that all my observations are visual/aural. I don't have test equipment or an RF generator to confirm this for any of the three software packages and this certainly is an area for analysis by someone with all the gear.
The bottom line is that with the image suppression optimised the combination of SDRPlay plus HDSDR plus RDSSpy software I reckon I've got a very good set up for FM-DX-ing.
As a footnote I've also compared the sensitivity of SDR Console and RSDR Sharp for RDS decoding against the above and the SDDR/HDSDR/RDSSpy comes out significantly better. It also outperforms my modified analogue tuners (most of the time) for how quickly it will grab a PI code from a marginal signal. Comparison based on the analogue tuners feeding RDSSpy software.
David
Sittinbourne
Wherever there was a strong signal to one side of the centre frequency I could find its corresponding image on the other side of the centre if the frequency wasn't already occupied by a genuine signal. For signals that were less than around 45dB above the noise floor the inmages were too weak to detect.
Ian directed me to the option in HDSDR to adjust the phase/amplitude balance manually. [Options/Input Channel Calibration Rx].
This picture shows the presence of Heart 102.8 at 103.8 on the oppsoite side of a centre frequency of 103.3. The amplitude and phase adjustments had been reset to zero. https://app.box.com/s/0z0u9gu5sjppskhxm8w4hr8dg9bkxzku
This picture shows what happens whens the amplitude and phase are corrected to minimise the image https://app.box.com/s/21xgc53grzp6f8lzeznquoav20fluppi I could now hear a much weaker station on 103.8
I have found that it is possible to have one setting which is effective across the whole of the FM Band at least when a wide IF bandwidth is used. When an IF bandwidth of 1.536 MHz was used I found the settings were appreciably different from the setting needed at the wider bandwidth and were noticeably different from one end of the band to the other. Also when using a 5MHz bandwidth I've found that the amplitude doesn't need any tweaking once it is set, but from one day to another a slight tweak of the phase might be needed to re-optimise suppression.
Now SDR Console and SDR Sharp both, I understand from the forums, have automatic correction of the I and Q balance. I've checked on MW and on SW and can find no image signals from poor suppression of the unwanted sideband with either of these software packages - but there the image of Heart 102.8 is very noticeable on both. My speculation is that the algorithms used have been developed for the HF bands when QSD SDR receivers were all relatively narrow bandwidth and HF only. I also gather that further work on improving the auto correction is in mind.
Using the HDSDR software - with no further adjustment from those arrived at for the FM Band I could find no images from poor sideband suppression in Medium Wave or Shortwave. But please take note that all my observations are visual/aural. I don't have test equipment or an RF generator to confirm this for any of the three software packages and this certainly is an area for analysis by someone with all the gear.
The bottom line is that with the image suppression optimised the combination of SDRPlay plus HDSDR plus RDSSpy software I reckon I've got a very good set up for FM-DX-ing.
As a footnote I've also compared the sensitivity of SDR Console and RSDR Sharp for RDS decoding against the above and the SDDR/HDSDR/RDSSpy comes out significantly better. It also outperforms my modified analogue tuners (most of the time) for how quickly it will grab a PI code from a marginal signal. Comparison based on the analogue tuners feeding RDSSpy software.
David
Sittinbourne