Echo Cancellors
Grimm
information on echo cancellors and/or why an echo problem exists on voice
circuits (T1s in particular) that are carried over long distances.
I understand that the echo problem is created on the hybrid conversion
which takes place eventually somewhere from the 4wire T1 circuit to a
2wire telephone set. But I'm a little shaky on the details. I've tried
to rationalize that the voltages that exist on the 4wire line are
essentially fed back to the 2 wire circuit almost like a ground return
path, but I may way out in left field with that sentence!
Thanks all!
Grimm
Steve Lenaghan
not always perfect. The suppressor simply shuts down the return path if
you are not talking. Some of our DID circuits have pathetic echos on
long distance. Strange as the conversation goes on the circuit echo
decreases. The problem is the telco can never duplicate the echo on
their tests. I get echos on calls from as close as 125 miles.
Steve L. Southwest Tel.
Brad Reese
You may wish to investigate Newbridge Digital Echo Cancellers:
http://www.alliancedatacom.com/newbridge-mainstreet-3901-3902-digital-echo-cancellers.htm
Hope this helps.
Brad Reese
Alliance Datacom RBOC Telco Buyer's Guide
http://www.alliancedatacom.com/rboc-telco-equipment-suppliers.htm
11130 Petal Street, Suite 800
Dallas, Texas 75238
800-444-5851 Toll Free
214-503-7400 Office
ro...@my-dejanews.com
grim...@bobcat.ent.ohiou.edu (Grimm) wrote:
> Could somebody either give a quick explanation or a quick url link to some
> information on echo cancellors and/or why an echo problem exists on voice
> circuits (T1s in particular) that are carried over long distances.
>
> I understand that the echo problem is created on the hybrid conversion
> which takes place eventually somewhere from the 4wire T1 circuit to a
> 2wire telephone set. But I'm a little shaky on the details. I've tried
> to rationalize that the voltages that exist on the 4wire line are
> essentially fed back to the 2 wire circuit almost like a ground return
> path, but I may way out in left field with that sentence!
>
An excellent white paper is available at
www.tellabs.com/products/echo/echo-wp.html
Unfortunately its about 700K in size -- fortunately it explains all you ask.
good reading!
Roman
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
Brad Reese
visit Ericsson's Echo Canceller Algorithms web page URL:
http://www.ericsson.se/echo/edu/ec_a.htm
Brad Reese
Alliance Datacom RBOC Telco Buyer's Guide
http://www.alliancedatacom.com/rboc-telco-equipment-suppliers.htm
11130 Petal Street, Suite 800
Dallas, Texas 75238
800-444-5851 Toll Free
214-503-7400 Office
Brad Reese wrote:
> Hi Grimm,
>
> You may wish to investigate Newbridge Digital Echo Cancellers:
>
> http://www.alliancedatacom.com/newbridge-mainstreet-3901-3902-digital-echo-cancellers.htm
>
> Hope this helps.
>
> Brad Reese
> Alliance Datacom RBOC Telco Buyer's Guide
> http://www.alliancedatacom.com/rboc-telco-equipment-suppliers.htm
> 11130 Petal Street, Suite 800
> Dallas, Texas 75238
> 800-444-5851 Toll Free
> 214-503-7400 Office
>
> Grimm wrote:
>
> > Could somebody either give a quick explanation or a quick url link to some
> > information on echo cancellors and/or why an echo problem exists on voice
> > circuits (T1s in particular) that are carried over long distances.
> >
> > I understand that the echo problem is created on the hybrid conversion
> > which takes place eventually somewhere from the 4wire T1 circuit to a
> > 2wire telephone set. But I'm a little shaky on the details. I've tried
> > to rationalize that the voltages that exist on the 4wire line are
> > essentially fed back to the 2 wire circuit almost like a ground return
> > path, but I may way out in left field with that sentence!
> >
> > Thanks all!
> > Grimm
Brad Reese
Al Varney
Grimm <grim...@bobcat.ent.ohiou.edu> wrote:
>Could somebody either give a quick explanation or a quick url link to some
>information on echo cancellors and/or why an echo problem exists on voice
>circuits (T1s in particular) that are carried over long distances.
T1s are not involved in this issue -- it happens over ANY transport,
because the problem is time (delay) and echo at the end points.
>I understand that the echo problem is created on the hybrid conversion
>which takes place eventually somewhere from the 4wire T1 circuit to a
>2wire telephone set. But I'm a little shaky on the details. I've tried
>to rationalize that the voltages that exist on the 4wire line are
>essentially fed back to the 2 wire circuit almost like a ground return
>path, but I may way out in left field with that sentence!
The ground return path is not too far from the explanation. Maybe
a better viewpoint is that the hybrid is simultaneously taking speech
FROM the 2-wire circuit and applying speech TO that circuit. It's
difficult not to have some of the "TO" speech mix with the "FROM" speech.
I've attached a Floyd Davidson old article that explains the "hybrid"
problem in ASCII form. Pay close attention to the fact the incoming
frequencies transferred to the 2-wire circuit will find their way into
the outgoing circuit -- this cannot be avoided, but the "balance network"
is supposed to EXACTLY compensate for this. But there is little perfection
here, and attempting to have a simple balance network handle every type of
customer line yields typical echo return losses of about 10 dB.
If you want to put some real research into this, see "Time Delay and
Echo Effects" from the November 1963 Bell System Technical Journal.
Appendix B of that paper offers a proof from 1950 that no automatic linear
mechanism can "self-balance" a hybrid circuit, and thus active non-linear
elements (echo cancellers) are the reasonable means available to compensate
for 2-wire telephone interfaces.
Four-wire telephones such as ISDN sets can eliminate the hybrid interface
(at that end), but some lower level of echo still exists -- even if it's just
the air coupling of earpiece to mouthpiece. Beyond about 400 ms one-way delay,
the delay itself becomes a problem in human interactive speech, even without
echo.
Finally, there is a good background paper on echo cancelling at the
ANSI Committee T1 Web site (www.t1.org) under Technical Reports (No. 27).
Al Varney
--------------------------------------------------------------------
From fl...@polarnet.com Mon Feb 17 15:27:05 CST 1997
From: fl...@polarnet.com (Floyd Davidson)
Newsgroups: comp.dcom.telecom.tech
Subject: Re: Full or half duplex
Date: 9 Feb 1997 01:06:07 GMT
Reply-To: fl...@polarnet.com
NNTP-Posting-Host: tanana.polarnet.com
John Stewart <jste...@umiami.ir.miami.edu> wrote:
> While we are on this subject, does anyone have a good explanation for
>how it is possible to have a full duplex voice conversation on a single pair
>of wires? Why don't the incoming and outgoing signals interfere with one
>another? I've been wondering about this for years.
>
>John
The "interference" occurs in the listeners head! Hearing both
our own voice and the distant party's voice at the same time is
actually required to make us comfortable, but we can only
tolerate it if the volume of both is about the same. If we
can't hear ourselves at all it is very annoying and causes
people to speak very unevenly, with halts and starts and
interrupts. If we can hear ourselves very loud in relation to
the other person it causes people to not talk as often and to
talk too quietly. In either case people tend to concentrate on
the sound of their voice more than on the conversation.
When the levels are adjusted correctly, people tend to be
unaware of how they sound, and concentrate on what is being
said.
So, the problem with full duplex voice over a telephone line is
that if a person's voice is fed directly to the speaker in the
telephone handset it will be very loud compared to the voice of
the distant party because the distant party's signal was
originally the same level, but after going through a switching
system and many miles of cable, it is much lower in volume than
it was. That is generally about 9 dB of loss, and it is enough
to make a telephone conversation very annoying if it isn't
compensated for in some way that reduces the volume of one's own
voice in the handset speaker.
That magic is provided by what is called a "hybrid transformer".
Here is the basic concept of a telephone set:
+---------+ distant party level: -12 dBm
| | local party level: -12 dBm
| speaker |----<--------------------+
| | |
+---------+ +---------+ +-----+-----+ <=== RX LEVEL: -9 dBm
| | | |
| bal net +---+ HYBRID +--------//----O 2-W TEL LINE
| net | | |
+---------+ +---------+ +-----+-----+ ====> TX LEVEL: 0 dBm
| | |
| micro- |---->--------------------+
| phone | local party level: +3 dBm
+---------+
The trick is to send the microphone signal to the speaker with
about 15 dB of loss, but to send it to the line with only 3 dB
of loss, while at the same time sending the signal from the line
to the speaker with 3 dB of loss. Note that directly across the
hybrid is a high loss, and from one port to the next one is a
low loss.
I can draw a schematic of one way to do that. It is not the
most common way, though the principles are the same for all of
them, but this particular method has the advantage that it is a
simple schematic and the others are just too complicated to draw
the transformer windings with an ascii character set.
|| (-----O 2-WIRE PORT O----+
O------) || ( |
) || ( |
SPEAKER ) || (-----O MICROPHONE PORT O----+
PORT ) || ( |
O------) || ( |
|| (-----O BAL NET PORT O----+
Note that I'm assuming this is a perfect transformer with no
losses. In fact it probably has a loss of about 1 dB or so for
any signal sent through it.
If the "BAL NET PORT" and the "2-WIRE PORT" are terminated with
identical loads, then any input to the "MICROPHONE PORT" will be
equally split between the "BAL NET PORT" and the "2-WIRE PORT".
Since half of the power will go to each, there is a 3 dB loss to
each.
But the signal will be balanced out and canceled for the
"SPEAKER PORT" port! The signal will be equally applied to the
top and the bottom of the transformer winding (it will be equal
if the BAL NET and the 2-WIRE loads are actually exactly equal),
so no current will be induced into the right side winding
because current in the upper half of the left side winding will
be exactly the opposite of current in the lower half of the left
side winding.
Hence the signal from the microphone will get to the 2-wire line
with about 3 dB of loss and will not be seen at the speaker port
at all if everything is perfectly balanced.
Here is another drawing of the same circuit, using "600 ohms" as
the two loads, and using ^ and v to indicate current polarity at
a given instant in time. Equal amounts of current are split off
into the two loads, but those same currents induce opposing
currents in the output winding of the transformer.
+------------+---------------+
| ^| ^|
| +-----+-----+ +-----+-----+
| | 600 ohms | | 600 ohms |
| +-----+-----+ +-----+-----+
| | |
|v ^| ^+--)||
| | )|| v
O | ^ )||
| )||(----O
INPUT | +-----------)||( v
| | ||( no output!
O | | +------)||( ^
| | | | )||(----O
|v | | | v )||
| | | | )|| ^
| +------^----^------)||
| | |
| ^| ^|
+-------------------+----+
If that input is the microphone in the telephone handset, and the
output goes to the speaker in the handset, then the person would
hear no sidetone at all (assuming this is all perfectly balanced).
In real life, the hybrid in the telephone set is unbalanced just
enough to allow a nice comfortable sidetone level for the person
speaking, while a much higher level is applied to the telephone
line to be sent to the distant party.
It's all known as FM, for Fine Magic! :-)
Floyd
--
Floyd L. Davidson Salcha, Alaska fl...@tanana.polarnet.com