Classical SAR simulators can be used to simulate 
this process. Given two slightly different satellite 
trajectories, we simulate two SAR images that we 
coregister to compute the phase differences. In 
fact, this simulation approach simply imitates the 
experimental interferogram generation process 
without presenting the interest of real data. The 
techniques to set up for such a simulation are 
complicated and may not be applicable on real 
interferograms. 
Instead, we chose to use a simulation technique 
that simplifies the interferometric process, makes it 
easy to set up and interpret and allows further 
experiments. 
  
fig. 3: portion of a SPOT DEM 
From a given SPOT DEM (figure 3), considered as 
the reference terrain, and from two satellites 
trajectories, we compute directly the phase 
difference for each point of the terrain. Here, the 
interferogram generated keeps the SPOT DEM 
geometry making it easy to evaluate. However it is 
easy, as shown figure 4, to resample this 
interferogram into the slant range geometry of the 
reference satellite if we want to get close to what 
we would obtain in reality. With classical simulation 
the simulated interferogram would be in radar 
(slant range) geometry precluding direct 
comparison with geocoded SPOT DEM. 
This approach, whereas being easy to implement, 
allows us to simply visualize interferograms without 
dealing with the registration and geometrical 
problems. The interferograms generated differ from 
‘real interferograms” since they do not take into 
account the noise effects as well as the 
uncorrelated pixels. They can still illustrate some of 
the behavior of the interferometric process. 
2.1 Parameters influence 
The first use of this simulation has been to assess 
and demonstrate the potential interferometric 
accuracy and the influence of the different 
parameters. Figures 5, 6 and 7 describe different 
influence of the parameters: 
* Simulations with different wavelength (simulating 
different radar sensors) show that the smaller the 
wavelength the smaller the fringe patterns and thus 
the more accurate the terrain restitution for a given 
satellites geometry. 
For a given radar wavelength (in this case 
SEASAT parameter) simulations with different 
baselines (100, 500 and 1000 meters), show that 
the larger the baseline, the smaller the fringe 
patterns and the better the elevation accuracy. 
Since the interferograms are generated with 256 
quantification levels for each interfringe region 
(corresponding to 2x radians phase rotation), we 
88 
  
fig. 4: Simulated interferogram in Slant Range 
geometry for a given satellite trajectory 
can easily assess the potential elevation accuracy 
obtainable after unwrapping. In the SPOT DEM the 
elevation range is about 2048 meters in this area. It 
comes from figure 5 where we hardly have 1 
interfringe region for the whole scene, that the 
potential restituted elevation accuracy with such a 
geometry (100 meters baseline) is 8 meters. For 
the 1000 meters baseline geometry (figure 7), we 
can count 6 different fringe orders which leads to a 
potential accuracy of 1.6 meters. 
  
fig. 5: Simulated interferogram with a 100 meters 
baseline (same effect as for a large wavelength) 
fig. 6: Simulated interferogram with a 500 meters 
baseline (same effect as for a medium wavelength) 
  
  
fig. 7: Simulated interferogram with a 1000 meters 
baseline (same effect as for a small wavelength) 
  
  
  
Ste 
dif 
ele