[Montevecchi]
Hi Nanang,
I changed jbuf.c according your request.
I performed 650 calls for a total amount of 2600 PESQ evaluations on each
side (4 evaluations per call). Pjsua's configuration is the same as in the
previous tests.
Results are quite similar than the last time.
I am summing up:
Minimum of OT.PesqScoreMin = 1.10
Average of OT.PesqScoreAvg = 4.4
Minimum of TO.PesqScoreMin = 2.81
Average of TO.PesqScoreAvg = 4.4
TotalNumebr of PESQ evaluations executed on each side = 2600
Number of PESQ_OT < 4.20 Total 30
Number of PESQ_TO < 4.20 Total 17
I will send you directly logs relative to 3 tests with low scores.
Best Regards
Massimiliano
>From your SIP trace/pjsua log file, it seems the jitter buffer did
some shrinkings and surely this could degrade the audio quality (not
sure if this is the only suspect though). So, I think it would be
great if you could do another test to measure how far those shrinkings
affect the audio quality by disabling the jbuf's shrinking, e.g:
replace jbuf.c line 470:
diff = cur_size - jb->jb_max_hist_level;
with
diff = 0;
If the jbuf shrinking really could degrade the audio quality
significantly, perhaps we should find a way to smoothen or enhance the
jbuf shrinking process.
Thanks,
nanang
On Fri, Sep 26, 2008 at 7:15 AM, Benny Prijono <bennylp at pjsip.org> wrote:
>
> On Thu, Sep 25, 2008 at 2:20 PM, Massimiliano Montevecchi
> <massimiliano.montevecchi at elsagdatamat.com> wrote:
>>
>> Hi Benny
>>
>>
>>
>> This morning I have run about 1500 PESQ evaluations.
>>
>> As I told you, I have used pjsua 1.0-rc2 + changes coming from
enhancement
>> #638.
>>
>>
>>
>> In attach you can find the pjsua's configuration used:
>>
>> codec PCMU
>>
>> no vad
>>
>> no echo suppressor
>>
>>
>>
>> I am summing up the results:
>>
>>
>>
>> Result
>>
>> Minimum of OT.PesqScoreMin
>>
>> Average of OT.PesqScoreAvg
>>
>> Minimum of TO.PesqScoreMin
>>
>> Average of TO.PesqScoreAvg
>>
>> Total
>>
>> Executed
>>
>> Failed %
>>
>> TcRun #
>>
>> PASSED
>>
>> 3.56
>>
>> 4.38
>>
>> 3.26
>>
>> 4.38
>>
>> 150
>>
>> 150
>>
>> 0
>>
>> 150
>>
>>
>
> Cool!
>
>>
>> A single test consists of 4 PESQ evaluations on each side of the call:
>> OT.Pesq = PESQ performed on called side; TO.Pesq = Pesq performed on
caller
>> side.
>>
>> The speech sample is 7 seconds long and a call is about 1 minute long.
>>
>>
>>
>> You can see that the average of the PESQ score is good, but there is some
>> test with low values.
>>
>> In general the PESQ score with lower value is related to the first
>> evaluation of the call, but that it is not always true.
>
> I see.
>
>>
>>
>>
>> Now I am doing the same tests after the activation of the VAD.
>>
>> The first results highlight that there is a considerable worsening of the
>> PESQ measures stability.
>
> I think I've encountered this before. It's got something to do with the
> jitter buffer too, and what happen was there was a stutter during the
start
> of talksprut. This didn't happen when VAD is off of course, since the
packet
> will be continuous.
>
> But since then I couldn't reproduce it again, hence I'm not sure if the
> problem has been fixed or not. Probably not, reading your description. So
> this is definitely something that we need to look at.
>
> Thanks
> Benny
>
>>
>>
>> Let me now how to send you logs.
>>
>>
>>
>> Best Regards
>>
>> Massimiliano
>>
>> _______________________________________________
>> Visit our blog: http://blog.pjsip.org
>>
>> pjsip mailing list
>> pjsip at lists.pjsip.org
>> http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org
>>
>
>
> _______________________________________________
> Visit our blog: http://blog.pjsip.org
>
> pjsip mailing list
> pjsip at lists.pjsip.org
> http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org
>
>
