UKVoIPTalk.com

In article <47696d1b$0$19307$ba620dc5@text.nova.planet.nl>,
TheFug <nosp@mplea.ze> wrote:[color=blue]
>Gordon Henderson schreef:[color=green]
>> Echo problems will exist on ALL analogue cards. (And some ISDN cards
>> too) How bad the echo is, depends on how "bad" the BT (or telewest,
>> etc.) line is. I was convinced one problem I had was caused by some BT
>> numpty using one wire from adjacent pairs from a street cab. to the
>> premises, but couldn't do anything about it. Plug in a £1.99 analogue
>> phone and it worked brilliantly. Plug in a TDM400P card and I got echo
>> echo echo. I tried 2 cards and a selection of modules. I tried all the
>>
>> Gordon[/color]
>
>And this isn't internal feedback ? because somehow input and output are
>fed into each other ?[/color]

I don't think it was an issue with the kit. After that customer didn't
want it anymore, I recycled it for another customer and it's still
working fine today. (Well, as fine as it's possible to get - there is
still echo, but it's managable)

Gordon

Gordon Henderson schreef:[color=blue]
> In article <47696d1b$0$19307$ba620dc5@text.nova.planet.nl>,
> TheFug <nosp@mplea.ze> wrote:[color=green]
>> Gordon Henderson schreef:[color=darkred]
>>> Echo problems will exist on ALL analogue cards. (And some ISDN cards
>>> too) How bad the echo is, depends on how "bad" the BT (or telewest,
>>> etc.) line is. I was convinced one problem I had was caused by some BT
>>> numpty using one wire from adjacent pairs from a street cab. to the
>>> premises, but couldn't do anything about it. Plug in a £1.99 analogue
>>> phone and it worked brilliantly. Plug in a TDM400P card and I got echo
>>> echo echo. I tried 2 cards and a selection of modules. I tried all the
>>>
>>> Gordon[/color]
>> And this isn't internal feedback ? because somehow input and output are
>> fed into each other ?[/color]
>
> I don't think it was an issue with the kit. After that customer didn't
> want it anymore, I recycled it for another customer and it's still
> working fine today. (Well, as fine as it's possible to get - there is
> still echo, but it's managable)
>
> Gordon[/color]

But by echo you mean a delay of sound, once ?
And there's also a factor if the correct codecs are used if there's a
conversion needed, this might cause also a delay...

--

The Fug.

On Fri, 28 Dec 2007 01:10:59 +0100, TheFug wrote:
[color=blue]
> Gordon Henderson schreef:[color=green]
>> In article <47696d1b$0$19307$ba620dc5@text.nova.planet.nl>, TheFug
>> <nosp@mplea.ze> wrote:[color=darkred]
>>> Gordon Henderson schreef:
>>>> Echo problems will exist on ALL analogue cards. (And some ISDN cards
>>>> too) How bad the echo is, depends on how "bad" the BT (or telewest,
>>>> etc.) line is.[/color][/color][/color]
[color=blue][color=green][color=darkred]
>>> And this isn't internal feedback ? because somehow input and output
>>> are fed into each other ?[/color][/color][/color]
[color=blue][color=green]
>> I don't think it was an issue with the kit. After that customer didn't
>> want it anymore, I recycled it for another customer and it's still
>> working fine today. (Well, as fine as it's possible to get - there is
>> still echo, but it's managable)[/color][/color]
[color=blue]
> But by echo you mean a delay of sound, once ?[/color]

I'm going to put words in Gordon's mouth here, and say "no", that's not
what he means. Echo on analogue lines can be caused by impedance mismatch
between the kit and the line, for example. This article goes into it a
bit more:

[url]http://www.voip-info.org/wiki/view/Asterisk+echo+analog+lines[/url]

Echo cancellation is expensive to fix in terms of CPU time, so it's best
prevented at source rather than cured after the fact.
[color=blue]
> And there's also a factor if the correct codecs are used if there's a
> conversion needed, this might cause also a delay...[/color]

Possibly, but I reckon that if you're running Asterisk on a machine
that's underpowered enough for codec translations to cause significant
delay, then you're going to be having a host of other problems anyway.

--
<http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm@ale.cx)
09:24:15 up 15:21, 2 users, load average: 0.24, 0.14, 0.10
2x Broadband/IT/Telecoms support positions in Newcastle city centre.
For more info call 0191 229 8870 and ask for Steve. No agencies.

In article <4774c2ae$0$507$bed64819@news.gradwell.net>,
alexd <troffasky@hotmail.com> wrote:[color=blue]
>On Fri, 28 Dec 2007 01:10:59 +0100, TheFug wrote:
>[color=green]
>> Gordon Henderson schreef:[color=darkred]
>>> In article <47696d1b$0$19307$ba620dc5@text.nova.planet.nl>, TheFug
>>> <nosp@mplea.ze> wrote:
>>>> Gordon Henderson schreef:
>>>>> Echo problems will exist on ALL analogue cards. (And some ISDN cards
>>>>> too) How bad the echo is, depends on how "bad" the BT (or telewest,
>>>>> etc.) line is.[/color][/color]
>[color=green][color=darkred]
>>>> And this isn't internal feedback ? because somehow input and output
>>>> are fed into each other ?[/color][/color]
>[color=green][color=darkred]
>>> I don't think it was an issue with the kit. After that customer didn't
>>> want it anymore, I recycled it for another customer and it's still
>>> working fine today. (Well, as fine as it's possible to get - there is
>>> still echo, but it's managable)[/color][/color]
>[color=green]
>> But by echo you mean a delay of sound, once ?[/color]
>
>I'm going to put words in Gordon's mouth here, and say "no", that's not
>what he means. Echo on analogue lines can be caused by impedance mismatch
>between the kit and the line, for example. This article goes into it a
>bit more:
>
>[url]http://www.voip-info.org/wiki/view/Asterisk+echo+analog+lines[/url]
>
>Echo cancellation is expensive to fix in terms of CPU time, so it's best
>prevented at source rather than cured after the fact.[/color]

That's basically it... And you can google until you're blue in the face
for other echo issues too (as I've done in the past) I think it's just
luck... My home BT line works OK using Digiums (licensed) high
performance echo canceller, but I couldn't even get that to make an
improvement on a customers line. My current strategy is to get customers
migrated over to IP provision as soon as possible, if they have analogue
equipment. (Or stick to ISDN installs, which while still having an echo
issue, is much more managable)

It boils down to 2 things: One is the quality of the far-end, and it's
ability to not inject what it hears back into what it sends back, so a
laptop using built-in microphone and speakers is about the worst case,
speaker phones slightly better (although the good ones do their own echo
signal processing and cancellation before it hits the wires), and the
decision taken many many years ago to only use 2 wires to carry both
sent and recieved voice data, rather than 4 wires. To investigate the
latter, google for things like 'hybrid' and impedance matching (along
with echo) Good luck!
[color=blue][color=green]
>> And there's also a factor if the correct codecs are used if there's a
>> conversion needed, this might cause also a delay...[/color]
>
>Possibly, but I reckon that if you're running Asterisk on a machine
>that's underpowered enough for codec translations to cause significant
>delay, then you're going to be having a host of other problems anyway.[/color]

I use a 1GHz CPUs. (occasionally 1.3GHz units for 60+ extensions) They
can quite hapilly manage 10 simultaneous transcodes of g729 to g711,
and feed this via an IAX trunk over a good business quality ADSL line,
so handling "native" g711 over analogue channels is practically no load
at all. (Although I've not done any proper performance measurements with
Digiums licensed high-performance echo canceller, but casual observations
have shown that 2 simultaneous analogue calls don't introduce any
significant load)

Hope everyone's having a jolly festive time!

Gordon

On Fri, 28 Dec 2007 10:38:58 +0000, Gordon Henderson wrote:
[color=blue]
> (Or stick to ISDN installs, which while still having an echo
> issue, is much more managable)[/color]

What causes echo on ISDN? Presumably it's not the line, as a distorted
digital signal would lead to no calls at all.

--
<http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm@ale.cx)
15:06:57 up 5:01, 3 users, load average: 1.33, 1.21, 1.13
2x Broadband/IT/Telecoms support positions in Newcastle city centre.
For more info call 0191 229 8870 and ask for Steve. No agencies.

In message <47751189$0$514$bed64819@news.gradwell.net>
alexd <troffasky@hotmail.com> wrote:
[color=blue]
> On Fri, 28 Dec 2007 10:38:58 +0000, Gordon Henderson wrote:
>[color=green]
> > (Or stick to ISDN installs, which while still having an echo
> > issue, is much more managable)[/color]
>
> What causes echo on ISDN? Presumably it's not the line, as a distorted
> digital signal would lead to no calls at all.[/color]

Yes, I wondered that. It's impossible for ISDN itself to cause an
echo. There has to be something else at the far end, e.g. a
speakerphone, AFAICS.

Dave

In article <47751189$0$514$bed64819@news.gradwell.net>,
alexd <troffasky@hotmail.com> wrote:[color=blue]
>On Fri, 28 Dec 2007 10:38:58 +0000, Gordon Henderson wrote:
>[color=green]
>> (Or stick to ISDN installs, which while still having an echo
>> issue, is much more managable)[/color]
>
>What causes echo on ISDN? Presumably it's not the line, as a distorted
>digital signal would lead to no calls at all.[/color]

Just because it's ISDN doesn't mean there won't be echo )-: At some
point it'll have to be converted to analogue and if it then goes down
a 2-wire interface, or goes into rubbish hardware there will be echo
generated by the far-end, and there are some fancy (& expensive!)
BRI/PRI cards with hardware echo cancellers on-board... (Fortuantely,
I've not had to resort to using one in the UK yet!)

A lot is to do with time too - we don't notice echo if we hear the thing
we said in under 30mS (IIRC - don't quote me on this!), so using the
"normal" TDM/PSTN type network in the UK more or less guarantees this -
even if the far-end reflects it back, it ought to come back quick enough
for us to not be able to detect it... But going further afield, "ye cany
change the laws o' physics"[1] and the speed of light (or electricity
;-) starts to become noticable, so talking London to NY via cable,
will introduce a few 10's of mS each way, so if there's any feedback
at all at the far-end, and there are no echo cancellers on the line,
you will hear it, and it will be annoying!

Now convert a signal to VoIP and pump it down an ADSL line with a
ping time to the server of 20mS each way and things start to become
"interesting" too... Even without ADSL, there are still delays over IP -
if you have a local asterisk box, trying talking into a phone connected
to the Echo application - even going from phone via LAN to a local server
and back again will introduce enough of a delay for you to notice it.

Even in the non-IP world you can observe delay and echo - For amusement,
try calling from a mobile to a landline in the same room and speaking
to the person who answers it, making sure they can hear you through the
air as well as over the wires...

Gordon

[1] As a Scotsman, I'm fully licensed and entitiled to say that :)

In message <fl4tk2$1fgl$1@energise.enta.net>
Gordon Henderson <gordon+usenet@drogon.net> wrote:
[color=blue]
> In article <47751189$0$514$bed64819@news.gradwell.net>,
> alexd <troffasky@hotmail.com> wrote:[color=green]
> >On Fri, 28 Dec 2007 10:38:58 +0000, Gordon Henderson wrote:
> >[color=darkred]
> >> (Or stick to ISDN installs, which while still having an echo
> >> issue, is much more managable)[/color]
> >
> >What causes echo on ISDN? Presumably it's not the line, as a distorted
> >digital signal would lead to no calls at all.[/color]
>
> Just because it's ISDN doesn't mean there won't be echo )-: At some
> point it'll have to be converted to analogue and if it then goes down
> a 2-wire interface, or goes into rubbish hardware there will be echo
> generated by the far-end, and there are some fancy (& expensive!)
> BRI/PRI cards with hardware echo cancellers on-board... (Fortuantely,
> I've not had to resort to using one in the UK yet!)[/color]

My point is that the echo doesn't (and can't) come from the ISDN
bit; it can only come from somewhere else, some non-ISDN part of
the path.

The acoustic path from earpiece to microphone of even a cheap
handset is tiny. I tried it years ago, expecting it to be a
significant problem, but found it was truly negligible - on the
border of undetectable. So a true ISDN terminal, or ISDN with
a proprietary digital telephone system on the end, won't echo
to any perceptible extent.

[snip the rest, which I agree with]

Dave