I am doing digital FM demodulation using the common ATAN2 and first difference method. When the carrier is offset from center, there is a non-linear region that appears and seems to be tied to the FM demodulation sample rate. Without be able to change the FM demodulation sample rate, are there any work arounds from this non-linear region? This non-linear region is just outside my bandwidth of interest, but causes false tone detections, as I don't want any tone detections outside my bandwidth of interest. I have played with increasing my I&Q filtering prior to demodulating(>120dB attenuation in my stopband) but since the ATAN2 system is so sensitive 1's and 0's can give you valid tone results. Implimenting AFC, would only shift this non-linear region by how much the AFC tracks. I have tried to use the signal strength as an indication of when to ignore valid tones (when signal strength is low), but this causes my receivers sensitivity to drop (with centered carrier) my receiver must be pretty sensitive (better than -104dBm). I am running out of ideas to try and have exausted the internet for stuff. Changing my I&Q data rate would fix my problem, but it's not an option currently.
FM carrier offset and aliasing effects
Started by ●September 27, 2010
Reply by ●September 27, 20102010-09-27
jacobfenton wrote:> I am doing digital FM demodulation using the common ATAN2 and first > difference method. When the carrier is offset from center, there is a > non-linear region that appears and seems to be tied to the FM demodulation > sample rate. > Without be able to change the FM demodulation sample rate, are > there any work arounds from this non-linear region? > This non-linear region is just outside my bandwidth of interest, but causes > false tone detections, as I don't want any tone detections outside my > bandwidth of interest.Why did you decide that the tone detection is affected by nonlinearity?> I have played with increasing my I&Q filtering prior > to demodulating(>120dB attenuation in my stopband)Band limiting in FM causes nonlinear distortion. The signal shifted from center gets shaved from one side. This is natural property of FM; choose filter parameters carefully. but since the ATAN2 system is so sensitive 1's and 0's can give you valid tone results. Use a threshold. If I^2 + Q^2 is less then valid threshold, then set the output to 0.> Implimenting AFC, would only shift this non-linear region by how much the > AFC tracks. I have tried to use the signal strength as an indication of > when to ignore valid tones (when signal strength is low), but this causes > my receivers sensitivity to drop (with centered carrier) my receiver must > be pretty sensitive (better than -104dBm).The criteria should be not the signal strength, but the signal to noise ratio.> I am running out of ideas to try > and have exausted the internet for stuff. Changing my I&Q data rate would > fix my problem, but it's not an option currently.I can fix your problem in one day. However, this is business. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●September 27, 20102010-09-27
> >Why did you decide that the tone detection is affected by nonlinearity? >Outside of my FIR filter stopband, I get false tone indications, that correlate to the region of the DC offset in audio which is not linear. As you watch the DC offset in the audio vs. carrier shift, it is not linear.>Band limiting in FM causes nonlinear distortion. The signal shifted from >center gets shaved from one side. This is natural property of FM; choose > filter parameters carefully. >Correct, I will investigate this further. I am currently just using a FIR filter on I and Q.>Use a threshold. If I^2 + Q^2 is less then valid threshold, then set >the output to 0. >I have done this. Distinguishing between -107dBm and offset carrier does not work. Perhaps it relates back to my I and Q filtering, which might be causing all the non-linearities.>The criteria should be not the signal strength, but the signal to noise >ratio. >I will also look into this.
Reply by ●September 28, 20102010-09-28
On Sep 27, 9:59�pm, "jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> wrote:> >Why did you decide that the tone detection is affected by nonlinearity? > > Outside of my FIR filter stopband, I get false tone indications, that > correlate to the region of the DC offset in audio which is not linear. As > you watch the DC offset in the audio vs. carrier shift, it is not linear. > > >Band limiting in FM causes nonlinear distortion. The signal shifted from > >center gets shaved from one side. This is natural property of FM; choose > > �filter parameters carefully. > > Correct, I will investigate this further. I am currently just using a FIR > filter on I and Q. > > >Use a threshold. If I^2 + Q^2 �is less then valid threshold, then set > >the output to 0. > > I have done this. Distinguishing between -107dBm and offset carrier does > not work. Perhaps it relates back to my I and Q filtering, which might be > causing all the non-linearities. > > >The criteria should be not the signal strength, but the signal to noise > >ratio. > > I will also look into this.What bandwidth are you working with? To put things in perspective, AMPS (old analog) cellphones using FM in a 30kHz wide channel had to be "ringable" at or below -117 dBm.
Reply by ●September 28, 20102010-09-28
>What bandwidth are you working with? To put things in perspective, >AMPS (old analog) cellphones using FM in a 30kHz wide channel had to >be "ringable" at or below -117 dBm. > >I am only working the digital portion of this project, I fear that the RF front end is not clean enough. 180kHz bandwidth is the minimum (30kHz deviation, multiple tones up to 30kHz, some margin for doppler)
Reply by ●September 28, 20102010-09-28
On Sep 28, 1:05�pm, "jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> wrote:> I am doing digital FM demodulation using the common ATAN2 and first > difference method. When the carrier is offset from center, there is a > non-linear region that appears and seems to be tied to the FM demodulation > sample rate. Without be able to change the FM demodulation sample rate, are > there any work arounds from this non-linear region? > > This non-linear region is just outside my bandwidth of interest, but causes > false tone detections, as I don't want any tone detections outside my > bandwidth of interest. I have played with increasing my I&Q filtering prior > to demodulating(>120dB attenuation in my stopband) but since the ATAN2 > system is so sensitive 1's and 0's can give you valid tone results. > Implimenting AFC, would only shift this non-linear region by how much the > AFC tracks. I have tried to use the signal strength as an indication of > when to ignore valid tones (when signal strength is low), but this causes > my receivers sensitivity to drop (with centered carrier) my receiver must > be pretty sensitive (better than -104dBm). I am running out of ideas to try > and have exausted the internet for stuff. Changing my I&Q data rate would > fix my problem, but it's not an option currently.Nevger mind all this digital crap - use an analogue PLL. You will only get a dc offset if there is an offset from the free running frequency. Hardy
