OBi1032 Voice Quality Lower than OBi202

[Charlemagne]

The voice quality on my OBi1032 is consistently lower than that of my OBi202.  Both operate on the same LAN/cable modem/router, both are on the same Voice VLAN untagged subnet, and all phone setting are in their default positions.  Both are PCMU codec prioritized, ethernet connected only - no wireless component.  Local SIP Ports and RTP ports are properly configured on the two devices.  Strong Spectrum cable modem connection 450/20.

Sharing the same Google Voice account, and the accounts of two (2) different SIP services providers, the Call Status stats are consistently good on the OBi202 (e.g., MOS 4.44, 0.00 packet loss, etc.) and noticeably inferior parallel stats on the OBI1032.  These occur regardles of the SP used, on both inbound and outbound calls.

How can a much older, lower tech ATA deliver reliably better voice quality than the newer, higher tech OBi1032.  If the 1032 would perform like the 202, I'd be happy.

Anyone know why the supposedly better phone is actually the inferior performer, and how to eliminate the voice quality gap between them?

[SteveInWA]

Hi:

There is no logical reason for the symptom you describe.  The 202 does not use "much older, lower tech".   They use the same custom ARM SoC.  The call quality should be identical.  I have tested every model of OBi IP phone ever made, along with every OBi ATA, and I have never, ever seen a call quality issue with the phones.  In fact, the call quality at the handset should be much better with the IP phone, since it is not doing a digital-to-analog conversion to send the audio out to an analog phone over the RJ-11 jack.

All I can think of, is that some Google Voice calls made on the IP phone will use the Opus codec instead of G.711u, and that codec has a variable bit-rate that could result in a problem, if the other end of your call is having signal quality issues.

Have you tried provisioning a different ITSP on one of your SPx slots, then comparing call quality?  For example, voip.ms or Callcentric?

[Charlemagne]

Thanks.  OK about the two using the same ARM SoC.  In my case, the OBi202 is probably five years older, and although I haven't done an A / B comparison, my casual observation is that the OBi1032 seems to have more configuration options (e.g., Jitter), monitoring capabilities (e. g., MOS-LQ, MOS-CQ), and of course, codec choices (e.g., Opus).

I agree that the 1032 handset quality is superior to the OBi202's analog phone handset.

My most frequent quality testing is between two different GV accounts.  Regardless of which account the call is initiate from, the Call Status results are always superior on the 202 (e.g., 10 points lower MOS on the 1032).  To make sure the differential had nothing to with the Cat 6 extension wiring, I bypassed it by connecting the 1032 directly into my D-Link gigabit PoE switch, but the problem persisted.

Yes, I have tried a different ITSP.  The results are similar on the 1032 using Callcentric, but I think their servers are in New York, and for me that is a greater distance which might affect call quality.

All of my OBi initiated calls connect through G.711u on both ends according to Call Status, which codec is in the top priority position for my GV calls.  I've also tried temporarily disabling the other codecs, and leaving Opus as the sole active codec.

Of course, G.711u is the recommended codec for Callcentric as well.

Any further suggestions to try would be appreciated.

[SteveInWA]

The feature differences you described between the phone and the ATA are firmware-based, not hardware.

If you make a pure SIP VoIP (not traversing the PTSN) call with the 1032, and if the phone on the other end supports Opus, then the call may be using Opus, not G.711u.  There are some very good articles on the web that explain how Opus works, so I'm not going to type it out here.  If the call is using Opus, then the OBi 10x2 IP phone will display a "HD" logo on the LCD.  For example, you can have an Opus HD Voice call between two IP phones via Callcentric 1777xxxxxxx extensions, because it's a pure SIP VoIP call, not traversing the PSTN.  You can do the same thing via the OBiTALK network (from one IP phone, call **9xxxxxxxxx, where the x's are the OBi number of the other phone).

If you want to compare audio quality between your IP phone and your ATA, I suggest that you try calling one OBi number from the other.

The overall audio quality of the IP phone is going to be better, because it has one less digital-to-analog conversion, and it is not constrained by the limited frequency range specified for PSTN calls.  The handset is a piece of plastic with a microphone and a speaker in it.  It is only one part of the entire system.  Since there are millions of analog telephones out in the wild, there is no way to know the design and manufacturing quality of the microphone and speaker in a random telephone.  However, the handset in the OBi IP phones are specifically designed for wideband audio along with the rest of the audio electronics in the phone, such as the amplifier, whereas the handset in the analog telephone is designed for narrowband audio.

The overall, end-to-end audio quality has many variables, including hardware, network, firmware and the human's hearing.  Audio tests, be they with a telephone, or with amplifiers, speakers, etc, are always highly subjective, and things like the volume level need to be precisely set to be equal for both devices under test.  For example, subjects may perceive louder music to sound "better" than lower volume music.

So, bottom line, without expensive test equipment, you can't make a truly valid comparison test as you described.  In my own extensive testing of OBiTALK products, if using a good quality analog telephone, sound quality on that phone plugged into a OBi 200/202 is very close to the sound quality of the IP phones, on a narrow-band G.711u call.  Sound quality on a properly functioning IP phone<-->IP phone call on a ITSP that supports Opus is fantastic.

Finally, one set of variables you can experiment with are the transmit and receive levels on the ATA, and the volume and gain settings on the IP phone.  On the phone, they are in Expert mode, under user settings-->user preferences.  If the settings are too high or two low, they can adversely impact the sound quality.

[drgeoff]

Charlemagne and SteveInWA are discussing different parameters.

In his posts Charlemagne concentrates on the network related transmission parameters - MOS, packet loss etc - saying that those computed by the 202 are better than those of the 1032.  He does not explicitly say that what he hears using the 1032 is worse than when using the 202.

OTOH SteveInWA does not really address those transmission stats but majors on the audio transducers and codecs.

[SteveInWA]

Geoff, what's your analysis of the issue?

[Charlemagne]

Thank you both for your responses.  I appreciate the additional information on the 1032, and will be interested in hearing the clarity of the Opus to Opus connection.

Since my original post, I have bypassed the Voice VLAN switch entirely, connecting directly to an unoccupied router port, pointing Voice VLAN traffic there as well.  I also did a factory reset of the 1032, configuring only SP1 with GV, with no other SPs present.  In both instances, Call Status still reported deficiencies, e.g., MOS-LQ 4.09 / Packet Loss Rate 1.27 (with that number declining during the progress of a 5 to 10 minute call to maybe 4.34 / 0.25%.  Tolerable perhaps, but not something I'd want to subject a conversation to, if I could avoid it.

I haven't limited my efforts to self-testing.  I've also called out and received calls from others.  They can hear on their end of the conversation the voice quality issues I observe concurrently on the Call Status page.  So, I find a reasonable correlation between the Call Status statistics and the voice quality perceptions.

So, I am back fundamentally to where I started, but a little more knowledgable.

Any further thoughts or ideas would be most welcomed.

[SteveInWA]

Go do some research on MOS.  The term refers to a "Mean Opinion Score", which was a method developed during the old Bell System days, well before VoIP and computer testing.  A panel of actual human listeners was used to listen to the sound quality of phone calls.  Much like the Olympic ice skating judges, each panelist would give their subjective opinion of the call quality, and then those scores would be averaged.  It is difficult to get above 4.0 on the PSTN; calls at MOS 4.0 or higher are considered good audio quality.

MOS calculated by algorithms may not represent nor agree with human opinions.  The ITU specifically includes a statement that MOS results may vary from one experiment to the next.

See:  https://en.wikipedia.org/wiki/Mean_opinion_score

Bottom line:  a) I have no idea what is causing your particular observation, and b) there are several audio parameters that can influence your perception of audio quality, most notably amplifier gain/overdrive/clipping.  If you believe that your OBi IP phone is defective, get it replaced.

[Charlemagne]

Thanks you.  I provided the MOS info not because it's conclusive or because I'm relying on it in substitution of my own judgment, but because OBi selected it for inclusion, presumably as a standardized measurement, and as a reported variable to assist in analysis.  The Packet Loss Rate provided may be a more direct indicator.  Perhaps not coincidentally, when the Call Status readings are better (including among others, the subjective MOS, the Packet Loss Rate, etc.), the voice quality is better too.  The noticeable voice quality issue, observed by me and those with whom I speak by phone, naturally preceded my effort to quantify the deviations relative to the OBi202's higher quality - using those tools which OBi provides to end users.

I will try TX and RX levels too.  My experience with altering the defaults (on the 202) has not be particularly productive.

You may be right that mine is simply a defective unit.  Of course, under the circumstances it's difficult to know in advance whether replacing it would indeed solve the problem.

Thanks again for your help.

[SteveInWA]

Incidentally,  I should have mentioned:

The theoretical maximum MOS for the G.711 CODECs is 4.4.  Anything in the range from 4.0 to 4.3 is considered a "satisfied listener" score.

RE:  experimenting, if you have a friend who is patient and has as much time as you to work with test calls, the first thing to figure out is whether the perceived call quality is poor when you are the listener on your end, or if the person on the other end is listening and evaluating.  If their voice sounds good to you, but your voice sounds bad to them, then fiddle with the microphone gain settings.  There is a narrow range between the level being too soft, and being overdriven/distorted/clipped.  If it sounds bad on your side, and you have tried a few volume settings, then it's probably some other issue.

[Charlemagne]

I appreciate that additional information.  I actually temporarily achieved a 4.44 with a G.711u codec on my OBi202.  It then fell to 4.42.  Both sounded great even with the analog phone.  The maximum I've obtained with the OBi1032 is a temporary 4.34.

What you've outlined seems like a sound procedure to follow.  That is the way I'll do it.

Thank you again.

[SteveInWA]

I appreciate that additional information.  I actually temporarily achieved a 4.44 with a G.711u codec on my OBi202.  It then fell to 4.42.  Both sounded great even with the analog phone.  The maximum I've obtained with the OBi1032 is a temporary 4.34.

What you've outlined seems like a sound procedure to follow.  That is the way I'll do it.

Thank you again.

Wait -- do you really think that 4.42, 4.44 and 4.34 are significantly different?  That's just a variation within the degree of accuracy of the measurement, which is subjective to begin with. I wouldn't even look at the hundredths digit.  Remember, the OBi device is making an arbitrary calculation of a measurement that was originally defined as the subjective opinion of a bunch of human listeners.  Tiny and normal variations in jitter and packet loss are perfectly normal, unless you were using a private, managed IP network, end-to-end, which you are not doing.  Each of your calls may take a different path over the messy public internet.

No human being could hear the difference between calls with those numbers.  They'd say "hmm, they all sound good to me."  So:  don't worry about the MOS score, as long as it consistently stays in that very good range.  The only thing that matters is how it sounds to the people on the phone.

If you are curious to explore the VoIP call quality capability of your own Internet service, and the many points between you and somewhere else, then use Visualware's test software:  http://vac.visualware.com/

[Charlemagne]

Yes, I agree with all of your points.  The average 1032 MOS/G.711u is 3.50 - 3.75.

I included the maximum achieved MOS/G.711u score of 4.34 (albeit a very fleeting one) as a contrast to that of the once perfect 202 score at 4.44 (but more typically floating around maybe 4.37).  The point I was trying to make is that on the same LAN, connection, switch. etc. the 202 performs consistenly better than the 1032, even though the latter is the more sophisticated of the two.  As you point out, MOS should not be the determinating factor, and has not been in my evaluations.  I've found that my perceiption of call quality corresponds more to the rate of lost packets (which is ironically always materially lower on the 202).  Generally, MOS and Packet Loss Rate levels appear to be in rough parallel.

I too like the Visualware tests.