Asset Model tuning

[davidbrayan]

I tried tuning a model in Asset,
got this as an output,
so can you people verify if its ok or what?
guide me for a better and realistic prediction,also its not smooth.
TX antenna used is omini 11db gain 27m height.
clutter losses and offset values are set by the tool itself while auto tuninig.
i dont know if they are right or wrong.
just stuck into it,help me out

Please send me any good docs about model tuning.

BR
db

[justdream]

what about your STD and ME?

this is very important and tell me if you are near from actual or not.

Also I would like to know your colors legend and what's your operating frequency?

Good luck

[davidbrayan]

std 9.5
me 0
color legends are
0 to -65 dark green
-65 to -75 light green
-75 to -85 yellow
-85 to -95 orange
-95 to -102 red
excluded signal below -94 in cw data
operating freq is 1851 MHz

[davidbrayan]

std around 7.5 to 8 is better but im unable to drag it down to 8
its always above 9

[DCA]

It could be that your drivetest dat is not accurate enough (to less samples). Also the used clutter and height in Aircom could cause the error. A std dev 7dB with an accurancy of 80% for every clutterclass or more is possible.....

[davidbrayan]

here are otherdetails if you can look into

It could be that your drivetest dat is not accurate enough (to less samples). Also the used clutter and height in Aircom could cause the error. A std dev 7dB with an accurancy of 80% for every clutterclass or more is possible.....

[venom]

If your STD is above 9 then it means your model is not well tuned, verify the drive measurements, You should use the Scanned Drive Data with scanning done from an external antenna (preferable).

Model Tuning is always a seperate project and drive test is usually done by transmitting the clean unmodulated frequency. This GIMS Drive Test data is one of the examples for it.

STD should be below 8, the lesser value means your model is more accurate. Above 8 is always considered as the in-accurate model !

Clutter Definitions and weights are defined according to the area at which you have done the drive test or for which you are actually making a model, for example if you are making a model of URBAN AREA then the more/highest weightage should be given to it ! Please also note that the lesser resolution you use, the better model you can make !

Smoothing Effect changes with the rollout factor. But before using smoothing, you must make sure that your model is correct. If it is incorrect, then prediction with smoothing is also very un-reliable !

[justdream]

Guys,

what do you think about letting ASSET automatically tune the clutters through and offset losses?

is that reliable? what did you do in your past expereince?

[davidbrayan]

now the problem is,
its not going to tune in any case since the mean error is around -25 to -30.
the approach i used is .
1. converted CW data to signia TX and RX files, for that i exported CW drive data to text with latitude longitude and RSSI, renamed the text file to siteid.dat this is the RX file for signia format, now created a header TX file with details of the site used for CW transmission like latitude,longitude,
height 27m
power(EIRP) 52 i.e. 20W=43dbm,Ga=11dbi,Losses=2dbm, so EIRP=43+11-2=52
Antenna omini, i've the pattern for that.
ground height
tilt
azimuth
with few other details.
created this text file and renamed to .hd
2. now created a fresh project with all data like clutter,heights and vectors.
3. imported the omni antenna.
4. created a standard macrocell3 model with default parameters.
5. created a site with same information as in CW signia .hd file, like same latitude, longitude, height ,EIRP,tilt, model etc.
6. loaded the .hd file,generated pathloss,
now here is the problem ,
if i analyse the CW data , it gives mean error around -25 or -30 and the mmodel is far tuned whatever changes i do.
also if i remove the antenna pattern from signia.hd file then all is well and i can tune the model.
but whatever antenna i use its not going to tune.

Any way out?
somthing missing?
or somebody can help me with this?

BR
db

[MINTO]

Hi David..i have worked on atoll, in atoll first we will do automatic calibration..then we will run prediction & then we compare it through analysis tool..then check the error in the sample..delete below & above -20&+20 ..then save it & again calibrate the model..finally you have to play with k1 & k2