order of 
retrieve 
ties. For 
level is 
s in the 
quality 
je kine- 
of coor- 
ery new 
the geo- 
1e ambi- 
and re- 
quality 
n time. 
mbigui- 
requires 
gies are 
nd have 
id kine- 
exploit 
? ambi- 
ibiguity 
| a con- 
Ig strat- 
ve two 
of the 
ve been 
he vali- 
[994]. 
the rms 
2) 
neces- 
' |s not 
he am- 
| of the 
1991]. 
ths are 
etrieve 
994 in 
aneous 
juality, 
1995 
  
341 
  
  
  
  
Ies # 
  
  
  
  
  
Discrim. Factor 
  
0 2 4 6 8 10 12 14 16 18 20 22 24 
Time in Hours 
Phase quality (single difference) Code quality (single difference) 
ame ms= 05m 
nnd e" mszcib5m 
ms=5 m 
rms = 0.005 m 
  
Figure 2. Theoretical instantaneous discrimination for dual frequency measurements 
Ionospheric sensitivity of the ambiguity search: In the theoretical investigation presented above, systematic errors 
have been neglected. In practice, however, they are often the limiting factor. The high discrimination potential of the 
ambiguity search in the dual frequency mode precisely lies in the fact that unbiased L1- and L2 phase measurements are 
available. The systematic errors introduced by the ionosphere, however, are frequency dependent and therefore leads to 
lack of discrimination. Fortunately for many applications over short distances the biases introduced by the ionosphere 
are small enough to be entirely neglected for the ambiguity search. This maximal distance is strongly depending on the 
overall ionospheric activity and on the latitude and may typically range between 3 and 15 km. In the case where the 
ionospheric bias can not be neglected, a stochastic modeling may be applied [Bock et al. 1984, Beutler et al. 1989, Co- 
card 1994]. This leads to a better, i.e. bias-free estimation, but weakens the discrimination of the ambiguity search and 
thus longer integration times become necessary in order to securely retrieve the correct integer values of the ambigui- 
ties. 
6. INFLUENCE OF NOT SOLVED AMBIGUITIES ON THE COORDINATE DETERMINATION 
If it is not possible to securely fix the ambiguities to integer, because the GPS-data is too poor (for example. in the case 
of single frequency data over short time spans, or distances of some hundreds of kilometers between reference station 
and rover), it is still possible to compute the coordinates by using the best real valued estimate of the ambiguities. Con- 
sequently the GPS-coordinates are affected by these wrong ambiguity values. In a first approximation the introduced 
error shows a linear behavior (offset and linear drift) over time. Such ‘floating point’ solutions have often been and are 
still used in the aerophotogrammetry, i.e., where the linear behavior of the GPS-solution is then modeled by additional 
drift parameters in a combined GPS/bundle block adjustment. In this case, however, additional perpendicular strips at 
the end of the blocks become necessary in order to stabilize the geometry. 
7. ON-LINE VERSUS OFF-LINE MODE 
The major problem in realizing an on-line solution is the data link between the reference station and the moving re- 
ceiver. In the case where only a differential code solution at the meter level is sought after, only range correction values 
have to be transmitted with an update rate of 5-10 sec. In many countries such services have been already or will be 
installed commercially in the near future. In the case of Switzerland a DGPS-service is going to be installed in 1995. Its 
aim is a country-wide coverage. Range correction values are broadcast at a baud rate of about 50 bits/sec. The user has 
to be equipped by a special receiver with integrated decoder connected to the GPS-receiver and providing for the range 
correction values. 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995