syllable skipping h.323 to SIP

[wpetilli]

Can someone summarize the logic on: H.323 handset (NR -- x with Intra/Inter direct audio 'y' and hairpinnning 'n' and using codec g.711 with ss 'n') calling up a SIP TG (signalling with hairpinning 'y') to SM out to a 3rd party VM (far end NR xx with intra/inter direct audio and hairpinning 'y'). Randomly getting audio quality with a syllable skipping. Also randomly happening on internal call answer, but PSTN is always fine. Where is a starting point to pinpoint why/where or if this is just bad handsets, possibly?

this is a CM6.0.1 with G650's (4 PN's..medpros..etc..)

 

[kyle555]

Typically with RTCP monitoring and a product like Prognosis or something else in your network to be able to go back and see all those stats.

Now, H.323 and SIP aren't terribly relevant to the conversation, but SIP can be set up with early media such that a call's RTP begins shuffled. H323 typically will set up a call on a DSP first and shuffle in the first few seconds.

That all being said, when it comes to voicemail, I put a SIP MM5.2 behind a gateway and made those 2 regions never shuffle to one another. It basically took a multi-site network and instead of letting any IP phone send RTP to/from the MM server and have the MM juggle all the slight differences in between the locations, it let the DSPs in a gateway do that and it did clear up minor problems similar to what you're mentioning.

Considering you probably don't have deep analytic tools available and you can't reliably reproduce to test it being fixed or not, maybe picking a certain number of users and disabling shuffling on their station forms, or on a few network regions if you have the DSP density might get you some peace and quiet and let you work by process of elimination.

 

[wpetilli]

All my phones are set for ip-ip audio and ip audio hairpinning on. The main hubs NR1 has ip hairpinning off and the remote lsp NR's are set to on. the SG to ASM has hairpinning on. I'm debating if I should just put NR1 as the far end NR, instead of a different NR. The settings are pretty much the same so I'm not sure why that's even configured. hmmm..

 

[kyle555]

Hairpinning exactly relevant here. It determines when you tie audio to the TDM bus when using DSPs in a gateway.

With shuffling on, regardless of hairpinning, a phone can send directly to a voicemail.
With shuffling off, and hairpinning on, phone goes to Medpro, medpro to voicemail, no TDM slots used on your G650
With shuffling and hairpinning off, phone goes to medpro, and ties up timeslots on the TDM bus.

I'd say put voicemail in its own NR, that NR reaches all other NRs through intervening region 1 (if the voicemail is at the physical location of region 1) and the voicemail region does no shuffling. That would force the whole network to tie down to DSPs in NR1 to hit voicemail and be an effective test to see if managing RTP streams between phones and voicemail with/without a hardware DSP in the middle has any effect on the call quality. As an added bonus, you can list measurements on your DSPs for packet loss and jitter and stuff.