jitter buffer size?

[Rambobino]

Hi,

I just received my new OBi110 and I'm testing it. I already have a PAP2T-NA that I have been running for a year. I'm comparing the two devices to see some differences and I've noticed one in the voice delay. I'm using voip.ms as a service provider and I can test the echo by dialing 4443. Under the same conditions (Internet line and router QoS), there's a noticeable additional delay when using the obi110 (by ear I would say ~750ms with the obi and ~250ms with the PAP2T). I know that the jitter buffer setting can create delay when it is set too high, and by checking the call status information in the obi110, I see that the buffer is at 190ms which is a little bit high. I've searched for a setting but I didn't find anything so I'm asking here if it's possible to change the jitter buffer size like I can do in the PAP2T? If not, how can I reduce the delay?

Here's the call status information:

Call 1             
Terminal 1   Terminal 2
Terminal ID   PHONE1   SP1
State   connected   connected
Peer Name      
Peer Number   4443   4443
Start Time   22:58:41   22:58:41
Duration   00:12:09   00:12:09
Direction   Outbound   Outbound
Peer RTP Address      *removed*
Local RTP Address      *removed*
RTP Transport      UDP
Audio Codec      tx=G711U; rx=G711U
RTP Packetization (ms)      tx=20; rx=20
RTP Packet Count      tx=36483; rx=36430
RTP Byte Count      tx=6275076; rx=6265960
Peer Clock Differential Rate      -34 PPM
Packets In Jitter Buffer      9
Packets Out-Of-Order      0
Packets (10ms) Interpolated      30
Packets Late (Dropped)      0
Packets Lost      4
Packet Loss Rate      0 %
Packet Drop Rate      0 %
Jitter Buffer Length      190 ms
Received Interarrival Jitter      1 ms
DTMF Digits Received      0
Jitter Buffer Underruns      0
Jitter Buffer Overruns      0
Sequence number discontinuities      4
skew compensation      0 ms

Thanks!

[Rambobino]

Nobody has a clue?
Ok so forget about the PAP2T, how can I reduce the voice delay on the obi110? There's got to be a way to customize that??

Thanks

[OBiSupport]

OBi device has a fully adaptive jitter buffer integrated, and there is no manual setting for jitter buffer configuration, nor user customization required.

You may contact our support at support@obihai.com if you have concerns regarding voice latency in normal phone calls, or experience any voice quality issues, thank you,

[crazyk4952]

I am also using voip.ms with my OBi100 and have noticed a longer delay since I started using it a couple of months ago. My PAP2T worked well and I was able to adjust the jitter buffer to a low level to reduce delay. But the PAP2T died recently.

It is disappointing that we are unable to adjust the jitter buffer on the OBi. Maybe a new feature in the next firmware?

My jitter buffer is usually 180ms or so.

[Rambobino]

I have not had time to contact their support yet, but a feature like that in the firmware would be definitely useful.
We just have to contact them I guess, maybe they have an advice to give us already.

The weird thing is that nobody seems to have that problem? Or does not care? Or doesn't have something to compare? Maybe it happens only with voip.ms?

crazyk4952 : Have you tested the ATA with another provider?

[earthtoobi]

i used to have one-way audio, jitter buffer underrun issues early on (there are threads already on that).
all of my problems went away with  tomato firmware and the patches to Obi.
nowadays, even-though i sometimes see jitter buffer underruns/overruns, the voice quality is not affected.

[RonR]

i used to have one-way audio, jitter buffer underrun issues early on (there are threads already on that).
all of my problems went away with  tomato firmware and the patches to Obi.
nowadays, even-though i sometimes see jitter buffer underruns/overruns, the voice quality is not affected.

The issue with jitter buffer size isn't one-way audio or buffer underruns, it's latency (delays) encountered with larger buffer sizes.  The larger the jitter buffer, the more delay that's introduced.  On stable, low jitter networks, decreasing the jitter buffer size can significantly reduce the latency encountered.

[earthtoobi]

The issue with jitter buffer size isn't one-way audio or buffer underruns, it's latency (delays) encountered with larger buffer sizes.  The larger the jitter buffer, the more delay that's introduced.  On stable, low jitter networks, decreasing the jitter buffer size can significantly reduce the latency encountered.

i mentioned the above note to only say how my audio  issues were sorted out.
wrt your specific point on latency, that is what the adaptive jitter buffer is supposed to do.how do you specify on a per call basis whether it is a stable or a relatively slow network. the adaptive jitter buffer should in real time adjust size of buffer if it works as intended.

[RonR]

Like others have reported, I've had excellent results reducing the jitter buffer size in PAP2's with no adverse affects, resulting in very noticeably less latency than the OBi provides.  It would appear the adaptive algorithm in the OBi doesn't work all that well or is too generous with the buffer size for what many of us would prefer.

[LeftRight]

Just in my personal opinion,a fully adaptive jitter buffer algorithm would be preferred for majority of Obi users. I don't think that anyone can predict the network condition, thus obtaining a optimal "jitter buffer latency" for all the incoming/outgoing calls is almost impossible.

Fully agree that algorithm optimization may be needed in long term (balancing off latency and packet loss??), but I just don't see a necessity of adding parameters for the manual configuration.

[anshuman26]

I would like to note that, though the Admin mentions that it is fully adaptive, I don't think it is actually working. My buffer size ALWAYS seem to remain at 190ms and I have tested it in various network conditions.

I hope the Obi guys can take a look at this.

[LeftRight]

Why would this (190 ms) be a problem?

According to E-model (refer to one of research paper below):

http://www.recursosvoip.com/docs/english/AssessingVoIPCallQualityUsingtheE-model.pdf

If you don't see the packet loss, the latency of 190 ms means you have a estimated MOS 4.2~4.4 for G711 codec (above 4.0 for G729), which also means you would experience very good voice quality for most of your calls.

When latency goes beyond 350 ms, you might start feel voice degradation ... so I would guess you have a fairly good network environment where your OBi device sits in, and the Internet connection to the other party is also good for VoIP calls.

[anshuman26]

@leftright

You are assuming all the latency is coming from jitter buffer only. That graph gives MOS for one way total delay (part of which is delay introduced from jitter buffer). In reality, latency comes from several other factors as well.

I am not complaining about voice quality but about latency. The voice quality is great but there is some latency introduced due to this jitter. Though it is not bothersome, it is still noticeable.

For example, I did a simple test where both the Obi110 and my computer was connected to the same router. I did an echo test to Google voice both from Obi110 and from gmail on my computer and there was a noticeable difference in latency.

[earthtoobi]

in the latest firmware( 2575),notice that the jitter buffer changes based on network conditions.i have seen it go as low as 60ms.looks like there were some fixes that went in

[squalk]

[I prefer] a fully adaptive jitter buffer algorithm

Fully agree that algorithm optimization may be needed in long term (balancing off latency and packet loss??), but I just don't see a necessity of adding parameters for the manual configuration.

Add the option for manual jitter buffer but default to auto jitter buffer.  Was that so hard?

Some of us have very low latency verizon fios (fiber) [and effectively zero jitter] at home and don't need the added delay of assumptions.

Tier 4 fios does not use software rate-limiting so FULL bi-directrional saturation STILL has very low latency 

[Mango]

+1 for a configurable jitter buffer.  It's one of the few reasons to use a PAP2T instead of an OBi.