[Ns-developers] spectrum modeling and wifi device

[Pei, Guangyu]

Hi, Nicola 

> -----Original Message-----
> From: Nicola Baldo [mailto:nbaldo at cttc.es] 
> Sent: Wednesday, February 25, 2009 5:00 AM
> To: ns-developers
> Subject: Re: [Ns-developers] spectrum modeling and wifi device
> 
> Hi Ruben,
> 
> Ruben Merz wrote:
> > So, you want to do two things in the wireless PHY of ns3: 
> (1) support 
> > for multiple transmission channels at the PHY and (2) interference 
> > modeling at the receiver.
> 
> Our short-term goal is to simulate cognitive radio/dynamic 
> spectrum access systems using the wifi MAC and the baseband 
> part of the PHY. 
> These systems are expected to operate in bands other than 
> ISM, with different channel spacings. And we also want to 
> evaluate the interference of this modified Wifi onto 
> technologies other than Wifi (i.e., primary users). As a 
> longer-term objective, we are also looking at possible ways 
> to simulate OFDMA system, which need a rather complex 
> spectrum modeling. Ideally, we would like to write a single 
> piece of code which can be reused in all these cases. I think 
> such a spectrum modeling feature could also be used for other 
> scenarios, such as underwater communications.
> 
> Both the support for multiple channels in 802.11 and the 
> modeling of adjacent channel interference are kind of a 
> by-product of what we aim at.
> 

This is a very nice by-product.  Tom Henderson and I are working on
studying the interference and coexistence issues for 2.4 GHz ISM band.
Accidentally, we also planned to add off-channel model in ns-3 and
compare it to the testbed measurements described in figure 8 and 9 of
the paper by Glenn Judd and Peter Steenkiste
http://www.cs.cmu.edu/afs/cs/project/cmcl/archive/2008/Monet.pdf. It
seems that your effort is exact in line with our plan. When are you
planning to complete the basic spectrum modeling framework? Do you have
any validation thoughts on your spectrum model? 

> Talking about performance, a detailed spectrum modeling would 
> increase to some extent the computational load. That's why I 
> liked the idea of allowing to plug below YansWifiPhy both the 
> simple and fast YansWifiChannel, and a more detailed spectrum 
> model for those who want it.
> 
> 

I agree. I also like the idea of tagging packet with RF signatures.

> 
> Ruben Merz wrote:
>  >>> For interference modeling at the receiver, I think you 
> don't have to  >>> insert an object between YansWifiPhy and 
> YansWifiChannel. What you  >>> may want, is to get all the 
> packets (irrespective of their  >>> transmission channel) 
> back to YansWifiPhy, and then process those  >>> packets in 
> YansWifiPhy. Typically, you would do that at the same  >>> 
> time than you compute the SNR.
>  >> Yes, this would be one way of doing this. You could also 
> simply  >> modelize this as an extra loss for incoming 
> packets which is  >> calculated based on the center frequency 
> of the incoming packet and  >> the center frequency of the receiver.
>  >
>  > Actually, it is complementary to the SNR calculation, 
> right? But, yes,  > the point of doing this is that it would 
> allow you to properly model  > cases where, for instance, 
> transmission is deferred because of a  > strong packet on an 
> adjacent channel. You could also then properly  > model the 
> reception of a strong packet that is on an adjacent channel  
> > than the receiver. Although I don't have any reference at 
> hand, I seem  > to remember that this can happen.
> 
> I like the idea of keeping the interference calculation in 
> YansWifiPhy as-is, and just add the calculation of the loss 
> due to use of different channels somewhere.
> 
> As for transmission deferral due to strong adjacent channel 
> interference, wouldn't that be automatically handled by the 
> "maybeCcaBusy" feature?
> 
> 
> Regards,
> 
> Nicola
>