Hi everybody, I demodulate a FM signal with this : IdQ - QdI But in a paper I see that after this, we have to add an Integrate and Dump !! But for this operation, we need and integrate clock time over the symbole to demodulate.... If I don't put this Integrate and Dump block I think that in presence of noise the demodulator works bad !!! Do you have suggestions ? Thanks.
Digital FM demodulation
Started by ●September 26, 2006
Reply by ●September 26, 20062006-09-26
Patrick, Please more info for the group. How do you implement the dI and dQ? Is the signal bandlimited to a range appropriate for dI and dQ implementation? Do you account for any delay between I and dI, Q and dQ? Did you try to normalize by sqrt(I^2+Q^2)? Dirk Bell DSP Consultant patrick.me...@dmradiocom.fr wrote:> Hi everybody, > > I demodulate a FM signal with this : IdQ - QdI > > But in a paper I see that after this, we have to add an Integrate and > Dump !! But for this operation, we need and integrate clock time over > the symbole to demodulate.... > > If I don't put this Integrate and Dump block I think that in presence > of noise the demodulator works bad !!! > > Do you have suggestions ? > > Thanks.
Reply by ●September 26, 20062006-09-26
I compute de dI =3D I(n)-I(n-1), the same for dQ... I first normalize I and Q by dividing I and Q by SQRT(I=B2+Q=B2) before demoduling with the derivative. So after the derivative, what's the better : 1 - Integrate and Dump then symbol decision 2 - Compare the symbol with a treshold beetween Max and Min For 1 I need a decision clock, so a clock bit synchronizer For 2 I don't need a bit clock, so if the transmission is asynchronous, I just need to send the output of the comparator to an UART thanks
Reply by ●September 26, 20062006-09-26
patrick.melet@dmradiocom.fr wrote:> I compute de dI =3D I(n)-I(n-1), the same for dQ... > > I first normalize I and Q by dividing I and Q by SQRT(I=B2+Q=B2) before > demoduling with the derivative. > > So after the derivative, what's the better : > > 1 - Integrate and Dump then symbol decision > > 2 - Compare the symbol with a treshold beetween Max and Min > > For 1 I need a decision clock, so a clock bit synchronizer > > For 2 I don't need a bit clock, so if the transmission is asynchronous, > I just need to send the output of the comparator to an UART > > thanksYou should pass the discriminator through a one-bit long boxcar filter in either case. John
Reply by ●September 26, 20062006-09-26
So your I,Q are 1/2 sample out of alignment with your dI,dQ, and your differentiator only differentiates for that part of the spectrum that is grossly oversampled. Dirk Bell DSP Consultant patrick.melet@dmradiocom.fr wrote:> I compute de dI =3D I(n)-I(n-1), the same for dQ... > > I first normalize I and Q by dividing I and Q by SQRT(I=B2+Q=B2) before > demoduling with the derivative. > > So after the derivative, what's the better : > > 1 - Integrate and Dump then symbol decision > > 2 - Compare the symbol with a treshold beetween Max and Min > > For 1 I need a decision clock, so a clock bit synchronizer > > For 2 I don't need a bit clock, so if the transmission is asynchronous, > I just need to send the output of the comparator to an UART >=20 > thanks
Reply by ●September 28, 20062006-09-28
Yes I have a problem in my Matlab simulation... After the demod which computes the demodulated signal, I've got an amplitude offset and my signal is not normalized to 1 I have divided the output of the substractor by I=B2+Q=B2... What do you mean by the delay beetwen I and dI and Q and dQ ? you talk about a 1/2 sample ? The sampling frequency or the center frequency of I and Q ? My I and Q signals are center to 50 kHz with +/- 40 kHz deviation... and I'm sampling at 2 MHz
Reply by ●September 28, 20062006-09-28
hi I have simulated the FM demodulation under matlab and I have a problem At the output of the substractor, there's a DC offset I made the differentiator with I[n]-I[n-1] My I and Q signals are center to 50 kHz, with +/- 40 kHz deviation What do you mean by I and dI (Q and dQ) alignment, we put a one sample (frequency sampling at 2 MHz delay on the I and Q branch before multpliying by dI and dQ I don't understand why I have this DC offset thanks
Reply by ●September 28, 20062006-09-28
patrick.melet@dmradiocom.fr wrote:> hi > > I have simulated the FM demodulation under matlab and I have a problem > > At the output of the substractor, there's a DC offset> > My I and Q signals are center to 50 kHz, with +/- 40 kHz deviation > I don't understand why I have this DC offsetThe DC offset is there because your signal are centered at 50 kHz. Isn't it obvious? You should center your signal at zero frequency. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●September 28, 20062006-09-28
hi I have simulated the FM demodulation under matlab and I have a problem At the output of the substractor, there's a DC offset I made the differentiator with I[n]-I[n-1] My I and Q signals are center to 50 kHz, with +/- 40 kHz deviation What do you mean by I and dI (Q and dQ) alignment, we put a one sample (frequency sampling at 2 MHz delay on the I and Q branch before multpliying by dI and dQ I don't understand why I have this DC offset thanks
Reply by ●September 28, 20062006-09-28
hi I have simulated the FM demodulation under matlab and I have a problem At the output of the substractor, there's a DC offset I made the differentiator with I[n]-I[n-1] My I and Q signals are center to 50 kHz, with +/- 40 kHz deviation What do you mean by I and dI (Q and dQ) alignment, we put a one sample (frequency sampling at 2 MHz delay on the I and Q branch before multpliying by dI and dQ I don't understand why I have this DC offset thanks