Use proper scaling for dBfs
This commit is contained in:
@@ -36,11 +36,10 @@ class modes_output_print(modes_parse.modes_parse):
|
|||||||
reference = float(reference)
|
reference = float(reference)
|
||||||
timestamp = float(timestamp)
|
timestamp = float(timestamp)
|
||||||
|
|
||||||
#TODO this is suspect
|
|
||||||
if reference == 0.0:
|
if reference == 0.0:
|
||||||
refdb = -150.0
|
refdb = -150.0
|
||||||
else:
|
else:
|
||||||
refdb = 10.0*math.log10(reference)
|
refdb = 20.0*math.log10(reference)
|
||||||
output = "(%.0f %.10f) " % (refdb, timestamp);
|
output = "(%.0f %.10f) " % (refdb, timestamp);
|
||||||
|
|
||||||
try:
|
try:
|
||||||
|
|||||||
Reference in New Issue
Block a user