Below is a sample of data I took recording popular news channels for 3mins exactly and then analysing the results. This was done between 0530g and 0614g 01/04/12 here was the test subject list of freely available news streams
Here was my process
All was done on my iMac in OSX (10.7.2)All testing never left the digital domain.
All audio file settings/encodes were 48000/16bit/Stereo. All playback locally used quicktime player 10.1 All web playback used safari 5.1.1
1) Create - 18dbFS file in adobe audition
2) Used http://www.audiofile-engineering.com/spectre/ to confirm that playback level was -18dbFS on a digital meter and 4PPM on a BBC meter when played back and all audio controls set to max (all checked out)
3) Created video file in iMovie of a -18dbFS tone and uploaded to youtube http://www.youtube.com/watch?v=tziQZGjIspM
4) With the youtube flash player volume set to max, played back and calibrated my flash player to the spectre meters (only needed about -1.5db adjust, not sure why yet)
5) Went through popular news channels with their flash player volumes set to max, auto recorded 3min files to a PCM wav file, also did the same with the calibrated tone from the youtube player.
6) Put the .wav files through a tektronix cerify file checker, obtained results and plotted on graph below.
So I think its a fair test, as I ensured a unity path at all stages and times and recorded all files with the same setup.
As you can see CNN, Bloomberg, Aljazeera and BBC all look to group well as you would suspect. The 2 other channels NDTV and Euronews I have no conclusion yet.I was thinking I would see more of a spread of results on the internet streams as opposed to TV baseband audio levels… but I guess and hope this is following baseband TV levels, and I guess with modern encoders there is no real difference between baseband TV levels and the internet stream levels.
a) Im assuming -18dbFS ref level….. they could be -20/24 or anything really.
b) Im assuming the flash players all have 0dbFS at their max setting.
c)) This was only a specific sample window I would need more samples to be a lot more accurate.