Date: Sun, 13 Oct 91 13:19 PDT
From: john@zygot.ati.com (John Higdon)
Organization: Green Hills and Cows
Subject: Re: Problems Conferencing on Panasonic KX-T123210?

Jack.Winslade@ivgate.omahug.org (Jack Winslade) writes:

> >>REAL<< conference bridges are amplified and equalized and deliver
> equal volume to all connections off of the bridge...

> (As a side note, I would be interested in hearing how John H. deals
> with this in his 'party line' bridge.)

There are two general ways to provide the amplification required for a
REAL conference bridge. One is switched gating used by TelLabs and
others. In such a conference, the loudest person speaking effectively
shuts off the transmit side of all the other participants. Only one
transmit is enabled at a time, determined by who is actually speaking.
This has the disadvantage of sounding like a speakerphone, but keeps
the noise level down on the bridge.

The other method is a "mix-minus" system where the transmit side of
each line is amplified and fed to each of the other receive sides, but
not its own. Now this IS a black art, but allows for a conference in
which all participants can be heard at all times, much like a normal
telephone call. This is the method that my party-line uses and it does
require carefully adjusting the level and equalization on each line.

An amusing note: the night the crossbar cut to 5ESS, the party line
bridge went completely to hell. It squealed and howled to beat the
band. I had to do a considerable amount of level and eq readjusting to
get things to quiet down.

Both conferencing methods require precision hybrids to split the two
wire line to a four wire domain.

> Even the 'real' conference bridges are sometimes quirky.  We have a
> System-85 at work with 'real' conferencing, and sometimes it seems
> that the conference circuitry is just on the point of oscillation --
> there is a definite echo in there.

This is obviously a "mix-minus" that needs adjustment.


        John Higdon         |   P. O. Box 7648   |   +1 408 723 1395
    john@zygot.ati.com      | San Jose, CA 95150 |       M o o !

From: Floyd Davidson <floyd@ptialaska.net>
Newsgroups: comp.dcom.telecom.tech
Subject: Re: Linking two lines together with a capacitor.
Date: 23 Feb 2000 17:56:05 -0900

gmt@weirdness.com (Mark) wrote:
>Nice idea Art, I had toyed with 3-way calling / centrex but I then
>have no ability to automatically detect hang-up on the first line - as
>there would be no dial-tone return unless the CALLED party also
>hung-up - which if it was a recorded info service for example, it
>would not do.
>
>This is why I said I could not use Centrex (3 way calling)
>
>The voice modems are alas precisely that, there is no sound card
>involved. the idea is that person a can speak to person b and vice
>versa, not to broadcast the same "message" to both callers.
>
>Thanks for your thoughts though.

I don't see how any of the other possible solutions are going to
avoid that same problem.  There is normally no supervisory
signal provided to indicate whether the distant end is onhook or
offhook.

As for just connecting the two lines together using capacitors,
wellll...  don't even bother, at best it doesn't work well and
commonly the two parties thus connected can barely hear each
other.

The reasons are fairly simple.  Plain Old Telephone Service is
designed for a telephone set that basically sends at 0 dBm and
receives something between -3 and -12 dBm (the target is -6
dBm).  That is how much loss can be expected in a typical
connection due to cables, switching, etc.  Hence if the typical
telephone line expects a 0 dBm send level from a telephone set,
but what it gets is the typical -6 dBm received signal from
another line, the distant end is going to have a hard time
hearing.  Worse yet, when both a telephone set and another line
are connected they split the signals (and maybe not evenly, but
lets assume for now that the line and the telephone set have the
same impedance and evenly split the signal power between them).
That causes another 3 dB drop in level.  Hence each of those
lines will hear the other line at 9 dB lower than normal (that
is 1/8th the normal power).

And if happens to be conferencing two worst case connections
together the two distant end receive levels might be as low as
-27 dBm instead of -6 dBm.  In that case neither end will be
able to hear the other.  Here is a simplified diagram indicating
what happens, showing levels going from Party A to Party B.

  TX 0 dBm
                    +-------+               Rx from A
  Party             | Telco |                -12 dBm
    A    0----------|       |---------------0 >---+
  Telset            |  CO   |                     |
                    +-------+                     |
                                      conference  |    Local
                                       "bridge"   +-<  Telset
                                      3 dB loss   |    TX 0 dBm
                    +-------+                     |
  Party             | Telco |                     |
    B    0----------|       |---------------0 <---+
  Telset            |  CO   |     Telset TX  -3 dBm
                    +-------+        A   TX -15 dBm
 RX from      A -27 dBm
 RX from Telset -15 dBm
 (Design target  -6 dBm)


Conference bridges with hybrids on each line to allow amplifiers
to adjust both the transmit and receive levels are commerically
available.  They are not inexpensive, but they function well.
Here is the basic layout of such a device, which gives some
indication that it is not just a simple case of splicing the two
lines together with a pair of blocking capacitors and a holding
coil.


                  |\        +-->----------+
                  | \      /3dB           |
             +-->-+  +----+ Split         |
             |    | / AMP  \              |
 Line       \|/   |/        +-->---+      |
  A   ==><===X HYB                 |      |
 (2W)       /|\     /|      +-<--+ |      +-->-+
             |     / |     /3dB  | |       6dB  \
             +--<-+  +----x Split| |       Split +--+
               AMP \ |     \     | |            /   |
                    \|      +-<--|-|-+    +-->-+    o RX
                                 | |  \  /
                                 | |   \/             Local
                                 | |   /\             Telset
                                 | |  /  \
                  |\        +-->-|-|-+    +--<-+    o TX
                  | \      /3dB  | |       6dB  \   |
             +-->-+  +----+ Split| |       Split +--+
             |    | / AMP  \     | |            /
 Line       \|/   |/        +-->-+ |      +--<-+
  B   ==><===X HYB                 |      |
 (2W)       /|\     /|      +-<----+      |
             |     / |     /3dB           |
             +--<-+  +----x Split         |
               AMP \ |     \              |
                    \|      +-<-----------+

Each AMP has 12 dB gain, each Hybrid has 3 dB loss, note that
there are 4 splitters with 3 dB loss and two with 6 dB loss.

(The "splitters" are actually hybrids, so this circuit has 4
amplifiers and 8 hybrids in it!  Plus a 4 sets of 3dB pads.)

  Floyd


--
Floyd L. Davidson                          floyd@barrow.com
Ukpeagvik (Barrow, Alaska)

From: Floyd Davidson <floyd@ptialaska.net>
Newsgroups: comp.dcom.telecom.tech
Subject: Re: Linking two lines together with a capacitor.
Date: 25 Feb 2000 01:28:56 -0900

dr.dongle@juno.com wrote:

>  I would suggest you experament with different options. Floyd
>is right you can expect some "incertion loss" but I don't
>believe it will be as much as 27 db,

27 dBm was the *level*, not the insertion loss.  You might not
understand levels on telephone systems well enough to believe
it, but if you try using the conference options on one of those
two line telehone sets it won't take long to experience at least
an occasional instance of exactly that kind of level.  (Heh, if
you attach it to the right two lines, *every* time it is used
might be such an instance!)

> thats where "holding ckts" would be an improvement over
>the transformers in the modems.

Any "holding ckts" which do not have amplification will merely
cause *more* loss.

> He is forgetting that the capacitor coupling method was the
>standard (with no amplification) for about 80 years).

Surely you know that is not true.  That was, and is, the
"standard" for cheap or homemade systems that do not
necessarily work for many conferences, or even at all on some
phone lines.  That is exactly what is used by those two
line phones.

But if you take a look at standard conferencing systems that
*must* work, such as an attendant's or operator's position, or
the conferencing ability of a testboard position, then you'll
find that *nobody* assumed the "capacitor coupling method" was
an acceptable design.  The reason is because, as my previous
article pointed out, within the design limit targets of POTS
service there will be "normal" instances where conferencing
losses will exceed 15 dB, and the *best* that can be done is a 6
dB loss with two parties or 9 dB when the conferencing party
remains connected too.  The 6 dB *for design targe values* is at
the *limit* for acceptable loss!

Remember the target for receive level is -6 dBm with a range of
from -3 to -12.  Two systems that are right on the -6 dBm target
level will experience -12 dBm levels when conferenced by the
"capacitor" method if the original called party does not stay
connected.  If all three are on line the levels will drop
another 3 dB, to to -15dBm (i.e, a 9 dB insertion loss).  That
is below the standards for acceptable.  And of course on any set
of lines with less than the target value receive level but still
within the acceptable range, the results will be less than
acceptable when conferenced.

>A comercal conferance bridge would work great but be quight
>expencive and I get the impression that you trying to do this on "the
>cheep".
> Two "wet" transformers( DC rated, one for each line perhaps salvaged
>from answering machines or cordless phones) and a bi- directional amp
>would be a good home brew solution.

Notice the "bi-directional amp".  That is in fact *exactly* what
my previous article described with a block diagram of hybrids
and amplifiers.  That is exactly what makes a commercial
conference bridge expensive too.

The difficulty in understanding the various levels and the
effects of loading which result from attaching various devices
such as telephone sets or second lines, not to mention "holding
ckts" is what hampers people in obtaining a clear picture of
which designs will work well or not.

But even lacking the theoretical understanding, certainly anyone
who has used the conference button on two line phones very often
will point out the fact that it often does not provide a useful
conference because two of the parties cannot hear each other
well enough.  That is OK for a home system where no harm is done
when it fails.  It just as clearly is not acceptable to anyone
who *depends* upon a conference bridge being functional.

For example, I travel on a regular basis and commonly need to
conduct extensive long distance business using just about
anybody's telephone that I can get access to.  Dialing up an 800
number to someone who can forward me to anywhere is very
reliable *if* the conference bridge they have is a "no loss"
bridge that provides me with identical connections to those I
would get by directly dialing the desired number.  I would not
be able to tolerate being forwarded by any of these "capacitor
coupling" designs!  Of course what I do use is the conferencing
ability of various PBX's or even our toll switching systems
(AT&T uses Meridian PBX's, and both a 4E and a DMS-200 in
Alaska, so all of them are potential means for my own call
forwarding).

  Floyd


--
Floyd L. Davidson                          floyd@barrow.com
Ukpeagvik (Barrow, Alaska)

From: Floyd Davidson <floyd@ptialaska.net>
Newsgroups: comp.dcom.telecom.tech
Subject: Re: Linking two lines together with a capacitor.
Date: 25 Feb 2000 23:25:05 -0900

dr.dongle@juno.com wrote:
>  Floyd Davidson <floyd@ptialaska.net> wrote:
>> > He is forgetting that the capacitor coupling method was the
>> >standard (with no amplification) for about 80 years).
>>
>> Surely you know that is not true.  That was, and is, the
>> "standard" for cheap or homemade systems that do not
>> necessarily work for many conferences, or even at all on some
>> phone lines.  That is exactly what is used by those two
>> line phones.
>>
>
> Perhaps I should clarify my statement.
>
> Capacitor coupling between battery feed ckts in the old mechanical
>step, panel, crossbar and all relay systems was the standard for about
>80 years.

But *never* between the ends of two POTS local loop lines, which
is what Mark wants to connect.  And that is what he reasonably
read your statement to mean, and was clearly mislead.

>> Notice the "bi-directional amp".  That is in fact *exactly* what
>> my previous article described with a block diagram of hybrids
>> and amplifiers.  That is exactly what makes a commercial
>> conference bridge expensive too.
>
> Bidirectional amplification is very feasable with out those expencive
>"hybrid" transformers, there are several possable ckts with inexpencive
>opamps and the salvaged transformers I mentioned.

Those "expensive" hybrid transformers cost the same a regular
transformers...  a quick look in a Newark catalog found some of
either type, costing $6.16 each.  It is insignificant.

But what is used is a *hybrid*, whether the hybrid is made up of
hybrid transformers, regular transformers, resistors, or uses an
opamp zero-impedance summing junction design.  Indeed,
simplicity of design and implementation is probably a great deal
more important than whether the transformers used cost $5 or $50
each because either way, their cost for one unit is a negligible
part of the entire project.  A design using transformers has the
advantage of fewer points of failure (Keep It Simple Stupid!).
The only thing less expensive and less failure prone is a
resistive bridge, but the losses involved are usually too high
to allow it to be practical.

Op amps in a zero-impedance summing junction configuration are a
technically interesting solution, but are 1) more complex, 2)
have more failure points, and 3) are more expensive.  However,
they can be configured to provide gain, which in this case is
going to be needed and therefore the extra expense etc. is
required regardless of the hybrid design.

Notice that the diagram I posted did *not* indicate that the
hybrids were transformers, just that it was a hybrid.

Once upon a time I put together a circuit exactly like
that block diagram, using Northern Telecom VF-300 modules.  A
couple of slightly modified QVF-22A 4w4wire bridge plug-in units
and a pair of QVF-6A 600 ohm 4w-term sets (these are all similar
to Tellabs plug-in modules).  The QVF-22A bridge modules do not
use transformer hybrids (each input and output port is coupled
with a repeat coil though, so the card is certainly full of
"expensive transformers".  The bridge itself is a zero-impedance
summing point bridge using op-amps.

At a few hundred dollars a whack for those modules, I probably
had the most expensive two line conference bridge of any
testboard in the whole country!  (That was in the early 80's.)
But once again, the difference is whether the bridge has to work
with virtually any two connections or whether it can fail to
work at all for some not insignificant percentage of the time.
Just like an attendants position or an operators position, a
testboard really cannot use a conference bridge that only works
some of the time.

  Floyd

--
Floyd L. Davidson                          floyd@barrow.com
Ukpeagvik (Barrow, Alaska)

From: Floyd Davidson <floyd@ptialaska.net>
Newsgroups: comp.dcom.telecom.tech
Subject: Re: Linking two lines together with a capacitor.
Date: 26 Feb 2000 04:03:56 -0900

gmt@weirdness.com (Mark) wrote:
>Floyd:
>
>Although I have not yet finished my experiments with injecting the
>output froim the SPKR on to the other units line, I am experiencing a
>howl when the gain is increased to a level where I am confident that
>there are no significant losses. I have re-looked at the hybrid
>diagram you had kindly drawn for me with ascii but cannot quite follow
>it. Is there anywhere on the web that you could suggest I look at
>which might have a *slighlty" clearer diagram with some suggested
>values. I would then like to try and make one of the systems. I am
>fortunate to have RS Components only a few miles away and they sell
>almost every type of transformer possible, from 100KVA sub station
>ones, down to diddy little microphone matching ones - and a few
>thousand in between.
>
>I said I would reveal my findings with the SPKR experiment to all -
>and I will - but at the moment it is becoming clear that unless I
>accept some compromise between gain and "tendency to howl" I might be
>better off using the hybrid cicuit that you suggested.

If you increase the gain to unity at any frequency, it will
howl.  Worse yet, even though you reduce the gain it will still
result in "ringing", and speech passed through the system will
sound much like someone hollering down a culvert under the road.

The circuit has to have more than 13 dB of *loss* to avoid that,
although it won't really be too awful until about 10 dB or so.
Clearly you have a problem when the whole point of using that
circuit is to have 6 dB of *gain*.  That is the reason a hybrid
is required!  A hybrid will normally provide about 20 dB of
isolation, which is far better than the 13 dB needed.  It varies
though because it has to be "balanced" against the telephone
line, and of course telephone lines come in many varieties.  A
compromise network will usually provide better than 13 dB on
virtually any connection.

I don't know of any online source for detailed schematics.
Hybrids come in several different varieties, so designing your
own is not too terrible difficult.  It would be much easier to
purchase commercial units and use them as building blocks, but
not nearly as cheap.

If you are anywhere near a decent library, take a look at what
they have for telecom engineering books.

If worse comes to worse I could draw you up a fairly complete
schematic and email it to you as an image file of almost any
type that suits your needs.  (It will be a few days, as I'm off
on a trip in the morning and won't be back until Monday evening
at the earliest.  Though if I get bored I might just draw
something up for you using my laptop...)

I could also scan or FAX a couple diagrams straight out of
manuals for various real equipment.  Heh, the one for a QVF-22
4w-4wire bridge would boggle your mind for a few days!  It is
very interesting technically, but is not something you would
want to try duplicating.  However, it certainly demonstrates the
right way to use op-amps in a telecom circuit, even if it is
something designed back in the 1970s!

  Floyd

--
Floyd L. Davidson                          floyd@barrow.com
Ukpeagvik (Barrow, Alaska)