Forums

OFDM demodulation question

Started by Harry April 3, 2006
Here's my question:

In doing the OFDM (like DVB-T) demodulation, why do people just talk
about
carrier frequency sync and carrier phase noise but not carrier phase
itself?
Does that mean OFDM is not coherently demodulated?

Thank you for your help!


-- Harry

Harry said the following on 04/04/2006 00:31:
> Here's my question: > > In doing the OFDM (like DVB-T) demodulation, why do people just talk > about > carrier frequency sync and carrier phase noise but not carrier phase > itself?
Can you give an example of when "people don't talk about carrier phase"? The only reason I can think of for your conclusion is that in a typical OFDM receiver architecture, phase recovery is obtained implicitly via channel estimation, and does not require a separate step.
> Does that mean OFDM is not coherently demodulated?
Depends. OFDM based on PSK or QAM is coherently demodulated (e.g. DVB), OFDM based on DPSK isn't (e.g. DAB). -- Oli
Hi Oli,

Suppose the OFDM signal received is

f(t) = I(t)*cos(wt) + Q(t)*sin(wt).

where w is the carrier frequency and I(t), Q(t) are IFFT outputs on the
transmitting side.

On the receiving side, the LO will generate two signals
cos(vt + p) and sin(vt + p) to downconvert f(t) to baseband.

I know that the receiver will try to make "v" as close as possible to
"w" (frequency offset tracking, using pilot signals) but what about "p"
the phase offset?

Will the receiver also try to track the phase offset?  Is this phase
offset "p" the same as the so-called "common phase error" -- CPE?   I
have the impression that the CPE is only caused by the phase noise of
the local oscillator but not the regular unknown phase that should be
tracked out in a normal PLL.

Is that possible that this "p" will be tracked out by the FFT that
follows it? (OFDM symbol/FFT block  sync)
Or is it a don't care stuff in a OFDM system like DVB-T.

Do I misunderstand anything?  Is there any on-line tutorial that
explains it?

Thank you very much!

-- Harry

Harry said the following on 04/04/2006 16:35:
> Hi Oli, > > Suppose the OFDM signal received is > > f(t) = I(t)*cos(wt) + Q(t)*sin(wt). > > where w is the carrier frequency and I(t), Q(t) are IFFT outputs on the > transmitting side. > > On the receiving side, the LO will generate two signals > cos(vt + p) and sin(vt + p) to downconvert f(t) to baseband. > > I know that the receiver will try to make "v" as close as possible to > "w" (frequency offset tracking, using pilot signals) but what about "p" > the phase offset? > > Will the receiver also try to track the phase offset? Is this phase > offset "p" the same as the so-called "common phase error" -- CPE? I > have the impression that the CPE is only caused by the phase noise of > the local oscillator but not the regular unknown phase that should be > tracked out in a normal PLL.
As I said, phase correction is usually an implicit part of channel estimation. If the received sub-carriers Y[k] can be expressed in terms of the transmitted sub-carriers X[k] as: Y[k] = H[k].X[k] + N[k] Then we need to estimate the H[k], so that we can recover X[k] (in a least-squares sense). This is usually done with pilot information. However, in the case of a phase offset, phi, the relationship will be: Y[k] = exp{j*phi}.H[k].X[k] + N[k] But there's no reason we can't introduce a new variable H'[k] = exp(j*phi}.H[k], which means we can express Y[k] as: Y[k] = H'[k].X[k] + N[k] This is now an identical estimation problem to the first equation, i.e. phase offset is recovered implicitly. -- Oli
>Here's my question: > >In doing the OFDM (like DVB-T) demodulation, why do people just talk >about >carrier frequency sync and carrier phase noise but not carrier phase >itself? >Does that mean OFDM is not coherently demodulated? > >Thank you for your help! > > >-- Harry > >
I dont think its the same case like Single carrier, in which your coherent modulation is done on the carrier,in OFDM the coherent modulation is done at baseband. On the receive side, once you downconvert, at which point you're at baseband, demodulation is done in digital domain. You use the pilots to determine phase offsets, and accordingly demodulate the coherent baseband modulated data.