Hi,

I found my problem on I=B2+Q=B2...

I fact, I don't have the same amplitude on I and Q... and so it distord
my I=B2+Q=B2

I had a problem with my filters in the downconverter

Now all works fine...

Best regards

In fact, I've simulated the FM demodulation under Simulink.

But know I'm on the real world, I've described the FM demodulation with
VHDL langage and implented the recevier under Altera Stratix chip. So I
see my new problems in the real world, not in VHDL simulations...

So I have compute I and Q with a downconverter.

I first subsample my 10.7 MHz FM at 5.136 MS/s and so I've got a 428
kHz FI, then I downconvert to banseband by multiplying this 428 kHz FI
with cosine to have I and sinus to have Q, these cosine and sinus have
a frequency of 428 kHz.

And so when I compute I=B2+Q=B2, I and Q are not "really" in phase
because they are produced with cosine and sinus. I is in-phase and Q is
in-quadrature of phase.

It seems that the oscillation that I have on I=B2+Q=B2 is due to a delay
of 13 samples beetween I and Q... I don't try ever to delay I by 13
samples and to see if I've already have the oscillation on I=B2+Q=B2...
because I don't undersatnd why I have to put this delay !!

I don't use a FIR to compute dI and dQ, but I think IdQ-QdI is right
because I have a pure sinus at 1 kHz at this step, and I sent a 1 kHz
into my FM modulation...

It's just when I do (IdQ-QdI)/(I=B2+Q=B2) that my sinus is distorted by
the oscillation that I can see on (I=B2+Q=B2).. I put IdQ-QdI in phase
with I=B2+Q=B2 before doing the division

But I think my problem is only on the I=B2+Q=B2 calculation...=20

Thanks...

patrick.melet@dmradiocom.fr wrote:

> Hi,
>=20
> I modulate a 10.7 MHz center frequency with 20 kHz deviation.
> When I send a 1 kHz sinus signal to modulate the 10.7MHz, I demodulate
> this FM signal with IdQ-QdI.
> So this result is a goos sinusoide signal at 1 kHz
>=20
> But when I calculate I=B2+Q=B2, there's a 20 kHz sinus on this signal a=
nd
> so when I do (IdQ-QdI)/(I=B2+Q=B2) I have this 20 kHz signal modulated =
on
> my 1 kHz sinus wanted..
>=20
> Is for calculate I=B2+Q=B2 I must add a delay on the I branch to have I=

> and Q in phase before compute I=B2+Q=B2.
>=20
> Or may I filter the I=B2+Q=B2 filter before doing division ? But If I
> filter it, the division process won't remove a AM signal added to my FM=

> signal. (AM is unwanted).
>=20

All members of the (I dQ/dt - Q dI/dt)/(I^2 + Q^2) should correspond to=20
the same instant of time, and your sample rate should be considerably=20
higher then the deviation. You may have to compute the derivatives with=20
FIR differentiators.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Patrick,

You have been on this problem for a while.

The I and Q must be in phase.
The dI, dQ must be in phase to match each other and the I and Q.

If you have a purely FM signal, smoothly modulated as you describe,
wouldn't you expect the AM signal to be constant?

If it is not too long, post your MATLAB script.

Dirk Bell
DSP Consultant

patrick.melet@dmradiocom.fr wrote:
> Hi,
>
> I modulate a 10.7 MHz center frequency with 20 kHz deviation.
> When I send a 1 kHz sinus signal to modulate the 10.7MHz, I demodulate
> this FM signal with IdQ-QdI.
> So this result is a goos sinusoide signal at 1 kHz
>
> But when I calculate I=B2+Q=B2, there's a 20 kHz sinus on this signal and
> so when I do (IdQ-QdI)/(I=B2+Q=B2) I have this 20 kHz signal modulated on
> my 1 kHz sinus wanted..
>
> Is for calculate I=B2+Q=B2 I must add a delay on the I branch to have I
> and Q in phase before compute I=B2+Q=B2.
>
> Or may I filter the I=B2+Q=B2 filter before doing division ? But If I
> filter it, the division process won't remove a AM signal added to my FM
> signal. (AM is unwanted).
>=20
> Thanks.

Hi,

I modulate a 10.7 MHz center frequency with 20 kHz deviation.
When I send a 1 kHz sinus signal to modulate the 10.7MHz, I demodulate
this FM signal with IdQ-QdI.
So this result is a goos sinusoide signal at 1 kHz

But when I calculate I=B2+Q=B2, there's a 20 kHz sinus on this signal and
so when I do (IdQ-QdI)/(I=B2+Q=B2) I have this 20 kHz signal modulated on
my 1 kHz sinus wanted..

Is for calculate I=B2+Q=B2 I must add a delay on the I branch to have I
and Q in phase before compute I=B2+Q=B2.

Or may I filter the I=B2+Q=B2 filter before doing division ? But If I
filter it, the division process won't remove a AM signal added to my FM
signal. (AM is unwanted).

Thanks.