[asterisk-users] Fwd: Patton smartnode 463x (BRI) 25ms tail echo cancellation

[Gaëtan Minet]

Thanks a lot for your help.

Since we decreased tx on the patton gateway (-3db iirc) , no more echo  
report... so far. But this is a bit too early to conclude everything  
is fine now.
Now I guess we would have had similar results with the pci card we  
replaced by the gateway (Digium B410p with misdn, but the card didn't  
seem to react to gain adjustments, even after a complete asterisk and  
misdn module unload/reload... maybe I should have rebooted the whole  
server...)

We tried BRI callback some weeks ago and indeed we never got echo. We  
clearly only have echo when we call/are called from an analog line  
(legacy line or "twin" as they call residential isdn "atas" here). So  
indeed so far it seems you're right, I guess there was no point trying  
to handle far-end echo on our side, but just avoid helping the other  
side to generate that echo with too much tx gain.

We however had another problem with the Digium card when the HW EC was  
on, even in a callback situation ("crispy echo" even when used in  
callback), but no echo at all and no sound quality problem when EC was  
off. Maybe the HW EC chips also had a problem with tx gain being too  
high... We are still waiting for an answer from Digium Support, as at  
least the last 5 cards we bought are showing the same symptoms.
That's why we switched to the patton as no other pci card but  
Sangoma's or Eicon's (currently unsupported in our systems) seemed to  
have HW EC.


Regarding gains, in this callback scenario,  I guess we should try to  
have rx/tx amplitude almost perfectly balanced (while keeping tx  
around -3db as 0db causes echo) ?
In this scenario, don't some cards/drivers use a passthrough  based on  
isdn call id to optimize the channel (I remember seeing some messages  
in the logs about passthrough mode being activated when we made such  
calls) ?


Thanks

Gaetan





On 09/09/2009, at 22:32, Jorge Mendoza wrote:

> Gaëtan,
>
> The echo arise at 4w/2w conversion point, normally at the far end. Try
> to call another phone number assigned to your BRI, i.e. call back to
> your Asterisk server. If you use a Polycom phone to initiate the call
> and another to receive the call, you have the perfect link: 4 wires  
> with
> not 4w/2w conversion. Under this test, theoretically you must not have
> echo. If so, it is necessary to look elsewhere, at your Asterisk box  
> maybe.
> Modifying the rx/tx gain is a good practice too.
>
> Best Regards
>
> Jorge Mendoza
>
> Gaëtan Minet wrote:
>> Thanks !
>>
>> We installed it in the interim and have a lot of calls with far-end
>> echo :(.
>> But it seems the solution could be to reduce the TX gain on our side
>> (these are using polycom phones, and  indeed  I can see a big
>> amplitude  imbalance between tx/rx on a recording). It's under  
>> test, I
>> hope it'll solve it.
>>
>> Kind regards,
>>
>> Gaetan
>>
>>
>> On 09/09/2009, at 16:22, Jorge Mendoza wrote:
>>
>>
>>> Gaëtan,
>>>
>>> They are using as gateway to the pstn. In fact, they are remote
>>> gateways
>>> for a centralized callcenter:
>>>
>>> [pstn] «--» [BRI] «--» [internet] «--» [callcenter]
>>>
>>> Regards
>>>
>>> Jorge Mendoza
>>>
>>> Gaëtan Minet wrote:
>>>
>>>> Thanks
>>>>
>>>> Are you using these to connect isdn phones to the voip or to as a
>>>> gateway to the pstn for a voip system ?
>>>>
>>>> Kind regards
>>>>
>>>> Gaetan
>>>>
>>>>
>>>> On 08/09/2009, at 19:40, Jorge Mendoza wrote:
>>>>
>>>>
>>>>
>>>>> We have some installations with smartnode 4554, (same tail echo
>>>>> cancellation) without problems so far.
>>>>>
>>>>> Jorge Mendoza
>>>>>
>>>>> Gaëtan Minet wrote:
>>>>>
>>>>>
>>>>>> Hi
>>>>>>
>>>>>> Is anybody using these ?
>>>>>>
>>>>>> Gaetan
>>>>>>
>>>>>>
>>>>>> Begin forwarded message:
>>>>>>
>>>>>>
>>>>>>
>>>>>>> *From: *Gaëtan Minet <gminet-ml at mcit.be <mailto:gminet-
>>>>>>> ml at mcit.be>>
>>>>>>> *Date: *Sat 22 Aug 2009 16:29:42 GMT+02:00
>>>>>>> *To: *Asterisk Users Mailing List - Non-Commercial Discussion
>>>>>>> <asterisk-users at lists.digium.com
>>>>>>> <mailto:asterisk-users at lists.digium.com>>
>>>>>>> *Subject: **[asterisk-users] Patton smartnode 463x (BRI) 25ms  
>>>>>>> tail
>>>>>>> echo cancellation*
>>>>>>> *Reply-To: *Asterisk Users Mailing List - Non-Commercial
>>>>>>> Discussion
>>>>>>> <asterisk-users at lists.digium.com
>>>>>>> <mailto:asterisk-users at lists.digium.com>>
>>>>>>>
>>>>>>> Hi all
>>>>>>>
>>>>>>> We use pci/pci-e BRI cards in our installations. Due to echo
>>>>>>> problems
>>>>>>> (that was before Oslec and others), we quickly switched to cards
>>>>>>> with
>>>>>>> hardware-based EC.
>>>>>>> So we use exclusively Digium B410p cards that provide 64ms tail
>>>>>>> EC.
>>>>>>>
>>>>>>> For several reasons we'd like to switch to external BRI gateways
>>>>>>> like
>>>>>>> the Patton smartnodes (the price is getting really close to a
>>>>>>> B410p).
>>>>>>>
>>>>>>> I'm however curious about their HW EC. I see in the datasheets
>>>>>>> that it
>>>>>>> only has 25ms tail per channel (pri are 128ms, but not BRI).
>>>>>>> Are some of you using these gateway and do your experience  
>>>>>>> (many)
>>>>>>> echo problems on calls ?
>>>>>>>
>>>>>>> Our other alternative is to use sangoma cards that have 128ms HW
>>>>>>> EC
>>>>>>> and seem more stable overall, but it is yet a bit more  
>>>>>>> expensive.
>>>>>>>
>>>>>>> Thanks for your feedback.
>>>>>>>
>>>>>>> Regards,
>>>>>>> Gaetan
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>