[Asterisk-Users] T1 Hardware Echo Can

[Steve Underwood]

Steve Underwood wrote:

> Andrew Kohlsmith wrote:
>
>> On Saturday 10 July 2004 11:21, Rich Adamson wrote:
>>  
>>
>>> If you install a T1 card and an external T1 mux (with fxo cards), the
>>> echo can function already exists within the mux and/or cards. Don't
>>> really need 'another' external echo can box unless you actually
>>> purchased a T1 mux that didn't have echo can in the first place (and
>>> they do exist).
>>>   
>>
>>
>> As Steve already said, generally speaking echo cancellation hardware 
>> on T1/E1 interfaces is an option adder.  If it doesn't mention it, it 
>> doesn't have it.
>>
>>  
>>
>>> If you install a PRI (or related types of channelized T1 arrangements),
>>> you don't need an external echo can function as those interfaces are
>>> generally 4-wire to 4-wire interfaces already. If echo exists, its
>>> generally the result of other interfaces (located somewhere else) and
>>> those locations should be addressing the corrective actions needed to
>>> resolve the issue.
>>>   
>>
>>
>> You will often hear echo from the far-end hybrid, even on PRI, as I 
>> have found out.  Normal KSU/PBX systems with T1/PRI interfaces have 
>> echo cancellation hardware within the KSU itself.  I am purchasing a 
>> T1 echo canceller in order to try and eliminate the echo we hear 
>> (i.e. far-end echo) -- something I didn't think I'd need to do.  Our 
>> telco (Bell Canada) seems oblivious to any knowlege about echo 
>> cancellation for T1 within the CO, but I continue to press, because 
>> every now and again you hear a glimpse of "oh yeah we can do that, 
>> your line just wasn't engineered with that equipment" kind of blurb.  
>> :-)
>>  
>>
> I think you are missing something important about how traditional 
> telephone networks function. In the days before echo cancellation was 
> practical, it was vital to avoid the need for them. They couldn't 
> avoid the echo, so they avoided significant delays. Within almost any 
> country, the physical delay is so short the echo from the far end 
> appears as pleasant reverberation, and not nasty echo. International 
> circuits have always been a pain, as significant delay is unavoidable 
> there. It is packetising voice that really introduced delay as a broad 
> issue. First in digital cellular networks, where codecs process voice 
> in blocks, and inherently introduce at least a one block (say 20ms) 
> delay. Now VoIP broadens the issue further.

Whoops. Codecs introduce an inherent 2 block delay - one to compress and 
one to decompress. With most codecs that could be fudged a little to 
only introduce about a 1.5 block delay, but never seems to be in practice.

>
> Equipment makers specifically design traditional network equipment to 
> minimise delay. When I was developing DSP processing within the PCM 
> network I was only allowed 375us (3 samples) delay - one sample to 
> de-serialise the PCM stream, one to process it, and one to 
> re-serialise the result. Delay budgets are always set as tight as 
> possible.
>
> Bottom line: the traditional PSTN has always had echo, and it is 
> normally irrelevant. Telcos, have no need and no interest in removing it.
>
> Regards,
> Steve