Re: [Ekiga-list] Bad frequency response with Ekiga
Le 05/05/2013 01:18, *Craig Southeren* wrote: The Ekiga output is definitely showing a filtered response. Today I have made two other tests using Pulseaudio for sending white noise without wire interconnecting soundcards. (The application playing the white noise audio file directly into the softphone capture device.) This time, Ekiga and Linphone give about the same result, as you can see on the attached image. It seems to me that Linphone makes some kind of preemphasis on the audio gotten from soundcard only. Good Idea, don't you think ? attachment: Spectrums.png___ ekiga-list mailing list ekiga-list@gnome.org https://mail.gnome.org/mailman/listinfo/ekiga-list
Re: [Ekiga-list] Bad frequency response with Ekiga
Sorry, my records were polluted by an unmuted microphone. To see cleaner spectrums, please look at http://pix.toile-libre.org/?img=1367856298.png for Ekiga and at http://pix.toile-libre.org/?img=1367856367.png for Linphone (Indeed, the difference is still here.) ___ ekiga-list mailing list ekiga-list@gnome.org https://mail.gnome.org/mailman/listinfo/ekiga-list
Re: [Ekiga-list] Bad frequency response with Ekiga
*fred k* wrote: Using Ekiga v.3.9.90 ... You can see it here: http://pix.toile-libre.org/?img=1367698438.png My observation: When you (or Ekiga, or the sound card) took samples at 8 kHz and transmit two-sample running average(*) off the stream, you'll get 0.7 dB down at 1 kHz 3.0 dB down at 2 kHz 8.3 dB down at 3 kHz 12.6 dB down at 3.4kHz Fairly close to 1367698438.png but not exact. Attractive hypothesis! Only developpers know if samples are averaged this way. (*) a+b, b+c, c+d, ... where samples are a,b,c,d,... Another possibility is down-sampling from 44.1kHz sound card but I have no idea how it works. The white noise whas generated at 8000 samples per second in the same way during the two conversations. ___ ekiga-list mailing list ekiga-list@gnome.org https://mail.gnome.org/mailman/listinfo/ekiga-list
Re: [Ekiga-list] Bad frequency response with Ekiga
Thanks for taking the time do this analysis and share the results! The Ekiga output is definitely showing a filtered response. The only filter in the audio chain is the Automatic Echo Cancellation - did you have that enabled when doing the measurements? This would certainly explain the output. Craig On 5/05/2013 6:49 AM, geo cherchetout wrote: Hello, Using Ekiga v.3.9.90 and the line input of his soundcard, and PCMA codec, Foo sends white noise to Bar who receives and records this sound with Linphone v.3.5.99.0. (Recent versions of Linphone make recording easy, the recorded sound being extracted from rtp streams, not from soundcard.) Then, using Audacity, Bar displays the spectrum of the recorded sound. You can see it here: http://pix.toile-libre.org/?img=1367698438.png Now, the same experiment is made with Foo and Bar using the same Linphone v.3.5.99.0 softphone, and the spectrum is much better as you can see there: http://pix.toile-libre.org/?img=1367698567.png Perhaps there is something like a wrong equation in Ekiga's or some library's code ? Or this is a feature and not a bug ? I am able to compile Ekiga but not to modify the code by myself... ___ ekiga-list mailing list ekiga-list@gnome.org https://mail.gnome.org/mailman/listinfo/ekiga-list -- --- Craig Southeren Post Increment – VoIP Consulting and Software cra...@postincrement.com.au www.postincrement.com.au Mobile: +61 417231046G+: craig.southe...@gmail.com US: +1 415 800 4201 MSN: craig_southe...@hotmail.com Science is the poetry of reality. Richard Dawkins ___ ekiga-list mailing list ekiga-list@gnome.org https://mail.gnome.org/mailman/listinfo/ekiga-list