spectrum framework

[Gary Pei, 2010-04-02 21:25:25]

Hi,
Sorry for late reply. I just got chance to take a look at this patch. I think
this patch provides many useful and general features that strengthen overall
ns-3 wireless capabilities. 

The following are my initial comments. One general comment I have is that I have
not seen an easy way to use this for modeling an arbitrary signal source that is
not a network device. For example, a duty-cycled microwave oven model to
interfere with 802.11 devices. In particular, how does the spectrum channel
(either single or multi-model) handle in this case. An example that illustrates
this would be helpful.

Best,
Gary

http://codereview.appspot.com/230045/diff/1/2
File examples/spectrum/adhoc-ofdm-aloha.cc (right):

http://codereview.appspot.com/230045/diff/1/2#newcode129
examples/spectrum/adhoc-ofdm-aloha.cc:129: c.Create (2);
I suggest that a test program with > 2 nodes. This case does not test the
interference/collision resolutions.

http://codereview.appspot.com/230045/diff/1/5
File examples/spectrum/waveform-generator.cc (right):

http://codereview.appspot.com/230045/diff/1/5#newcode88
examples/spectrum/waveform-generator.cc:88: *txSpectrum2 *=  0.5;
It seems that user has to make sure the sum is 1. I am wondering whether it make
sense to enforce that at the model.

http://codereview.appspot.com/230045/diff/1/5#newcode107
examples/spectrum/waveform-generator.cc:107: Simulator::Schedule(Seconds(1),
&WaveformGenerator::Start, wg);
For some reason, the first spectrum analyzer output differs from txSpectrum
specified above. Would you please check it?

1.05 2.4e+09 0
1.05 2.401e+09 0.245
1.05 2.402e+09 0.49
1.05 2.403e+09 0.49
1.05 2.404e+09 0.245
1.05 2.405e+09 0
1.05 2.406e+09 0
1.05 2.407e+09 0

....

1.1 2.4e+09 0
1.1 2.401e+09 0.25
1.1 2.402e+09 0.5
1.1 2.403e+09 0.5
1.1 2.404e+09 0.25
1.1 2.405e+09 0
1.1 2.406e+09 0
1.1 2.407e+09 0
1.1 2.408e+09 0

http://codereview.appspot.com/230045/diff/1/28
File src/devices/spectrum/multi-model-spectrum-channel.cc (right):

http://codereview.appspot.com/230045/diff/1/28#newcode32
src/devices/spectrum/multi-model-spectrum-channel.cc:32: 
I don't believe the test codes cover this MultiModelSpectrumChannel. Since this
model also use SpectrumConverter. My suggestion is that a test case that
exercises these are helpful.

http://codereview.appspot.com/230045/diff/1/30
File src/devices/spectrum/single-model-spectrum-channel.cc (right):

http://codereview.appspot.com/230045/diff/1/30#newcode89
src/devices/spectrum/single-model-spectrum-channel.cc:89:
PhyList::const_iterator recvphyit = m_phyList.begin ();
might consider to rename for better reading: recvPhyIterator

http://codereview.appspot.com/230045/diff/1/34
File src/devices/spectrum/spectrum-error-model.cc (right):

http://codereview.appspot.com/230045/diff/1/34#newcode53
src/devices/spectrum/spectrum-error-model.cc:53: capacity += (bi->fh - bi->fl) *
(*vi);
I assume that (*vi) has value of Log2(1+sinr). This really is the upper bond,
not sure whether it is too optimistic here.

http://codereview.appspot.com/230045/diff/1/45
File src/devices/spectrum/wifi-spectrum-value.cc (right):

http://codereview.appspot.com/230045/diff/1/45#newcode77
src/devices/spectrum/wifi-spectrum-value.cc:77: 
I am not sure whether these match with 802.11. In particular, Figure 15-11 in
802.11-2007 is for 802.11b, Figure I.1 on page 1144.

[Nicola Baldo, 2010-05-21 15:08:43]

Hi Gary,

thank you for your review, and sorry for the very late reply. Below some
responses to your comments. Most of the issues are addressed in the new version
of the code, which I will publish for review very soon.

http://codereview.appspot.com/230045/diff/1/2
File examples/spectrum/adhoc-ofdm-aloha.cc (right):

http://codereview.appspot.com/230045/diff/1/2#newcode129
examples/spectrum/adhoc-ofdm-aloha.cc:129: c.Create (2);
On 2010/04/02 21:25:25, Gary Pei wrote:
> I suggest that a test program with > 2 nodes. This case does not test the
> interference/collision resolutions.

I added an example with a a microwave oven interferer in the revised version.

http://codereview.appspot.com/230045/diff/1/5
File examples/spectrum/waveform-generator.cc (right):

http://codereview.appspot.com/230045/diff/1/5#newcode88
examples/spectrum/waveform-generator.cc:88: *txSpectrum2 *=  0.5;
On 2010/04/02 21:25:25, Gary Pei wrote:
> It seems that user has to make sure the sum is 1. I am wondering whether it
make
> sense to enforce that at the model.

There is no such constraint. In this example, txSpectrum2 sums to 1 only by
chance. The values are intended to be arbitrary.

http://codereview.appspot.com/230045/diff/1/5#newcode107
examples/spectrum/waveform-generator.cc:107: Simulator::Schedule(Seconds(1),
&WaveformGenerator::Start, wg);
On 2010/04/02 21:25:25, Gary Pei wrote:
> For some reason, the first spectrum analyzer output differs from txSpectrum
> specified above. Would you please check it?

SpectrumAnalyzer calculates the average received spectrum over a given time
interval. The "first output" that you see is the average of the first interval.
What probably happens is that the first piece of signal is not active for the
whole duration of the interval, thus resulting in a slightly lower value than
txSpectrum.

http://codereview.appspot.com/230045/diff/1/28
File src/devices/spectrum/multi-model-spectrum-channel.cc (right):

http://codereview.appspot.com/230045/diff/1/28#newcode32
src/devices/spectrum/multi-model-spectrum-channel.cc:32: 
On 2010/04/02 21:25:25, Gary Pei wrote:
> I don't believe the test codes cover this MultiModelSpectrumChannel. Since
this
> model also use SpectrumConverter. My suggestion is that a test case that
> exercises these are helpful.

there are test cases for SpectrumConverter in spectrum-value-test.cc; they are
located in that file (rather than a separate file) because they share some code
with the test cases for SpectrumValue.



SpectrumModelConverterTestSuite

http://codereview.appspot.com/230045/diff/1/30
File src/devices/spectrum/single-model-spectrum-channel.cc (right):

http://codereview.appspot.com/230045/diff/1/30#newcode89
src/devices/spectrum/single-model-spectrum-channel.cc:89:
PhyList::const_iterator recvphyit = m_phyList.begin ();
On 2010/04/02 21:25:25, Gary Pei wrote:
> might consider to rename for better reading: recvPhyIterator

fixed in the new revision

http://codereview.appspot.com/230045/diff/1/34
File src/devices/spectrum/spectrum-error-model.cc (right):

http://codereview.appspot.com/230045/diff/1/34#newcode53
src/devices/spectrum/spectrum-error-model.cc:53: capacity += (bi->fh - bi->fl) *
(*vi);
On 2010/04/02 21:25:25, Gary Pei wrote:
> I assume that (*vi) has value of Log2(1+sinr). This really is the upper bond,
> not sure whether it is too optimistic here.

This error model is intended to represent exactly the theoretical upper bound.
Of course it would be nice to have more realistic models as well (for example
the Effective SNR model). However, I think they should be implemented as
alternative models, i.e., on separate classes.

http://codereview.appspot.com/230045/diff/1/45
File src/devices/spectrum/wifi-spectrum-value.cc (right):

http://codereview.appspot.com/230045/diff/1/45#newcode77
src/devices/spectrum/wifi-spectrum-value.cc:77: 
On 2010/04/02 21:25:25, Gary Pei wrote:
> I am not sure whether these match with 802.11. In particular, Figure 15-11 in
> 802.11-2007 is for 802.11b, Figure I.1 on page 1144.

You're right, this is fixed in the new version (note that the name of the file
changed to wifi-spectrum-value-helper).

[Gary Pei, 2011-02-28 23:44:44]

I believe it should be Pr = Pt / (4 PI d / lamda) ^2.

http://codereview.appspot.com/230045/diff/10001/src/common/friis-spectrum-pro...
File src/common/friis-spectrum-propagation-loss.cc (right):

http://codereview.appspot.com/230045/diff/10001/src/common/friis-spectrum-pro...
src/common/friis-spectrum-propagation-loss.cc:89: double loss = ( d * f * f) /
((3e8 * 3e8) / (4 * M_PI)) ;
This does not seem to be correct.

How about replacing the this line to the following:
double loss = ( d * f) / (3e8 / (4 * M_PI)) ; 
loss *= loss;