CEoIP help NM-CEM-4TE1 echo for PBX

neharris · ‎10-02-2009

I've been using the NM-CEM-4TE1 on a 928 mile MPLS silver CAR path for over a year now and have recently discovered there is echo on the PABX stations when using the CEoIP Tie line. I have tuned the dejitter buffers and payload size down as far as I can without having the circuit drop permanently but am still hearing echo between stations and vmail over the link.

here are the show cem results from each end

3M side:

cem info

cem 1/0/0 is up

Line state is up

Operational state is active

Near end ip address: 172.26.26.63, udp port: 15901

Far end ip address: 172.26.26.4, udp port: 15901

IP payload size: 288

IP dscp : 0x1A

Idle pattern length: 8 , Idle Pattern: 0x55

Payload compression is enabled

Data protection is disabled

Dejitter buffer size is 27 ms

Channel clock rate is 1536000 bps

CEIP header CRC is disabled

Signaling is disabled

Physical interface is T1 channelized

Ingress packets: 2581317, dropped: 0, overruns: 0, max_ipt: 505

Egress packets: 2578785, dropped: 0, lost pkts: 6823

Egress late pkts: 5307

Egress overruns: 2, underruns: 38, ur_delay: 0, ur2pkt: 0

Egress pkts dropped by burst control: 1166

Egress corrupt pkts rcvd: 0

30 second ingress rate 1536844 bits/sec, 667 packets/sec

30 second egress rate 1536921 bits/sec, 667 packets/sec

Tx interrupts: 2579100

Reorder queue flush: 0, visited: 0, max wait window: 0

Pkt-to-pkt jitter max: 15 ms, average: 1 ms, min: 0 ms

Adaptive clock correction is 28488 ppb tuning

Mode: closed-loop

Dejitter buffer level max: 35 ms, min: 0 ms

Event history: 0x007B0058

Effective compression: 78%

Compressed pkts: 2581291

Pkts dropped by PCI burst limit: 0

Global stats for slot 1

************************

Clk derived from : Local Oscillator

Egr free buf: 256

Egr host overruns: 0

Egr unknown dest count: 342

Last unknown dest ip : 172.26.26.63, port: 15901

Last unknown dest src ip : 172.26.26.4, port: 15901

Egr process switched: 0

Egr oos: 0

Egr unknown src count: 0, last unknown src ip: 0.0.0.0, port: 0

Ingr overruns: 0

NM cpu: 41.95 41.98 41.98 41.95

50M side:

cem info

cem 3/0/0 is up

Line state is up

Operational state is active

Near end ip address: 172.26.26.4, udp port: 15901

Far end ip address: 172.26.26.63, udp port: 15901

IP payload size: 288

IP dscp : 0x1A

Idle pattern length: 8 , Idle Pattern: 0x55

Payload compression is enabled

Data protection is disabled

Dejitter buffer size is 30 ms

Channel clock rate is 1536000 bps

CEIP header CRC is disabled

Signaling is disabled

Physical interface is T1 channelized

Ingress packets: 2562822, dropped: 0, overruns: 0, max_ipt: 127

Egress packets: 2562742, dropped: 0, lost pkts: 193

Egress late pkts: 193

Egress overruns: 21, underruns: 129, ur_delay: 0, ur2pkt: 7

Egress pkts dropped by burst control: 4875

Egress corrupt pkts rcvd: 0

30 second ingress rate 1536768 bits/sec, 667 packets/sec

30 second egress rate 1536614 bits/sec, 666 packets/sec

Tx interrupts: 2557766

Reorder queue flush: 0, visited: 0, max wait window: 0

Reports: 5, Rejects: 0

Pkt-to-pkt jitter max: 22 ms, average: 1 ms, min: 0 ms

Dejitter buffer level max: 38 ms, min: 2 ms

Event history: 0x007F0058

Effective compression: 82%

Compressed pkts: 2562837

Pkts dropped by PCI burst limit: 0

Global stats for slot 3

************************

Clk derived from : Local Oscillator

Egr free buf: 256

Egr host overruns: 0, consecutive overruns: 0, restarts: 0

Egr unknown dest count: 2597

Last unknown dest ip : 172.26.26.4, port: 15901

Last unknown dest src ip : 172.26.26.63, port: 15901

Egr process switched: 0

Egr oos: 0

Egr unknown src count: 0, last unknown src ip: 0.0.0.0, port: 0

Ingr overruns: 0

NM cpu: 42.43 42.43 42.43 42.43

I know there is not a large group of CEM users but would like to solicit whatever feedback I can....thoughts?

thanks in advance

->N

paolo bevilacqua · ‎10-02-2009

I don't think CEM is introducing echo, neither can cancel if there is, because never touches the payload it transports, so you may have an issue with the endpoints, not the transport.

If you was to transport this with regular VoiP and the enhanced echo cancellation modules on vwic2, you would have got a chance to correct echo, but that is not the case with CEM.

neharris · ‎10-09-2009

I'd have to disagree with you based on empirical evidence. I base this on the fact that when I reduce the dejitter buffer the echo can almost entirely be eliminated. Obviously I'd leave the buffer small if it didn't create an underrun problem and cause constant DS1 bouncing.

You're a pretty knowledgable person re:voip, isn't there a method of building an adhoc echo canceller with the DSP farm? I'd be intersted in knowing how to build one and if they can be applied to CEoIP card traffic?

paolo bevilacqua · ‎10-09-2009

ASAIK is not possible to use cisco dspfarm for POTS to POTS echo cancellation.

Beside, even if it would, that would require a full blown gateway in addition to the CEM setup, thus pretty much denying the need for the latter.

neharris · ‎10-09-2009

1) it's not POTS to POTS. unless you're saying that Cisco considers DS1 framed traffic POTS. this is a 1536 of payload terminating at each end on a DS1 framed circuit being used in the DSX drop for a tie line between 2 PBX's.

2) That router IS already a full CCM6.1 MGCP gateway and has DSP resources in it.

3) Having a VoIP GW doesn't eliminate the depreciation schedule for the PBX, plus we have a sort of multi-tenant world since the corp is a grouping of other indivdually operating business units.

Are you saying that ASAIK is the name of the stand alone echo canceller? Do you know what a config for it looks like?

paolo bevilacqua · ‎10-09-2009

1 yes, digital pots.

2 won't help.

3 I'm just saying cisco can only do echo cancellation using dsp on a T1 card. There is also an optional extended hardware module for it on the vwic2. Check the data sheet about.