[Discuss-gnuradio] Re: Regarding OFDM implementation

[Tom Rondeau]

Shravan,

I copied this to the listserv as I hope others find the discussion useful.

Shravan Rayanchu wrote:
> Hi Tom,
>
> I have a few questions regarding the OFDM implementation in gnuradio
> (I am a networking guy and I have a little knowledge of DSP, so some
> of my questions might be trivial):
>
> 1. Can you point me to the version with 16-QAM (or 64-QAM) running
> over OFDM on the air?
The code is checked in, but I can't guarantee that it will work 
(especially the receiver). We have currently separated the modulators 
into gr_ofdm_<mod>_mapper; there's one for QAM, which was meant to be a 
generalized QAM modulator, but we just hard-coded 16QAM for now.

Eventually, these will go away into a single gr_ofdm_mapper function 
that takes a constellation vector as an argument, allowing any type of 
modulation per subcarrier.

> 2. Does the argument 'occupied tones' refer to subcarriers? What is
> the frequency band occupied by each subcarrier? Can I vary the number
> of subcarriers used? What does "cp length" correspond to?
Yes. The fft_length is the total number of subcarriers, and the total 
number of subcarriers that carry actual data is the occupied_tones 
value. We currently allocate the middle number of subcarriers to carry 
data (instead of allowing random subcarries to be turned on and off as 
in non-contiguous OFDM).

The frequency bands of each subcarrier depends on the interpolation rate 
you use in the transmitter.

The cp_length value is the number of samples in the cyclic prefix. The 
total number of samples in time is the fft_length + cp_length per symbol.

> 3. What PHY data rate does each of the modulation correspond to ? (For
> eg. in OFDM, 64-QAM supports PHY data rates of 56Mbps and 48 Mbps)
We were able to get a maximum of 2.4 Mbps using the 16-QAM, so 600 kHz 
of occupied bandwidth. The limitation in speed is mostly in the receiver 
complexity, so as the code is optimized, the data rates will increase.

Just a side note, this is a problem I always run into. OFDM is _not_ 
802.11g. 802.11g uses OFDM as it's PHY layer modulation, so it's 
incorrect to say that OFDM at 64QAM gives you 54 Mbps; this is only true 
in the 802.11g implementation. In our system, at the maximum rate, we 
would get 3.6 Mbps with 64QAM.

> 4. Is there any other documentation available regarding the
> implementation (I looked at the presentation available on the trac)
Nope, not at this time. I think we've scattered a few references inside 
the code to papers we used when designing the system.

> Thanks much for the help :)
>
> Shrava

Tom