Do you have a question? Post it now! No Registration Necessary
- Posted on
- Richard Freeman
September 5, 2005, 7:41 am
rate this thread
- - - - - - -- - - snipped for brevity - - - - - - - - - - -
The Echo is not due to Line reflections but rather due to Hybrid Transformer
leakage - see Sidetone which is a deliberate result of this leakage being
taken advantage of ....
you do not have to build loss into Telephone networks - it is there anyway -
it is more a case of keeping your gain down to make sure you do not get
howling due to feedback through the hybrid Transformers at each end -
Hmmm? what about Side tone ? how do these well designed networks provide
Re: Z mismatch - was Slew Rates and Mic Pre performance
Not entirely true. Most of the US Inter Exchange Network (IEN) [digital] is
designed to have a nominal 0dB loss exchange to exchange. The only overall
transmission losses occur on the lines between the exchange and the
customers. Even before the advent of digital transmission and switching
technology, the old Strowger exchange had a fraction of a dB loss through
the exchange on a local call. IEN echo however was not a problem as it was
a two wire circuit through the exchange and the cable network to the
In a closed 4 wire IEN circuit losses of approximately 7dB occur across the
transformer hybrids at each end of the 4 wire transmission path even under
the worst possible mismatch conditions. This effectively provides a total
of 14dB loss to the singing loop. Therefore, provided the total gains in
the singing loop don't exceed the total losses, the circuit will remain
unconditionally stable. Any difference in favour of the losses over the
gains in the singing loop is known as the Stability Margin.
Example : If the losses total 14dB (trans-hybid losses as the worst
possible condition) and the gain only 1dB in each direction of transmission
to overcome adjacent port losses in the transformer hybrids (total 2dB gain
in the singing loop), therefore:- 14dB (loss) - 2dB (gain) = 12dB stability
margin. In simple terms the closed 4 wire loop can never become unstable.
The sidetone is developed within the telephone. Older phones (pre the T200
/ T400 series) used what was known as an Anti-SideTone Induction Coil
(ASTIC) which is a purposely leaky hybrid, was designed to feedback a small
amount of the speech energy from the transmitter (microphone) to the
receiver. This sidetone is purposefully locally introduced to make the
caller think the phone was "working okay." The level of sidetone is
critical. Too much and the talker will speak softly (thinking the other
party can hear him /her loud enough). Another problem is background noise
(if loud enough) picked up by the transmitter can tend to drown out incoming
speech and make the speech unintelligible. Conversly, too low a level or
absence of sidetone tends to cause the user to speak too loudly in the false
belief the distant party can't hear them. If they shout loud enough they
could overload the A/D converter in the exchange causing distortion.
BTW. The modern telephone still achieves sidetone, but instead of a bulky
ASTIC, it is achieved with semiconductor technology.
Re: Z mismatch - was Slew Rates and Mic Pre performance
Looks like my ISPs News server is stuffed again .....
- - - - - - -- - - further snipped for brevity - - - - - - - - - - -
True however there is loss (theoretically 6dB - in reality greater than
through the Hybrid Transformer - I had meant to flesh this concept out
further yesterday but inadvertantly hit send before I had finished
this idea/posting further.
so are you starting to suspect that the Hyrbrid has something to do
echo then and not cable reflections ?
Correct - the loop will not be singing if it merely consists of two
Hybrids - ie no gain in the Transmission system - as per my earlier
that stability is not created by padding a system down (as the basic
already contains sufficent loss) but rather about keeping total gain to
Exactly the basic system - with no added gain - has a total (using your
figures) of 14dB of loss built in and is by its very nature stable (any
signs of instability in this system would be cause for celebration by
physiscts around the world).
Lets see - over 6dB (theoretical) loss each way through the hybrid ....
That sounds about right.
The Sidetone was there already as a function of the Hybrid in the
ASTIC was wired to cancel the sidetone further than a simple Hybrid did
This was only possible as the level (not really the impedance except
it affected the overall gain) of the signal between the hybrid/Astic
the receiver was both known and constant.- In a side note the ASTIC was
actually introduced (around 1939 I believe in Australia) to encourage
people to talk louder into Telephones as it provided less Sidetone than
the normal Hybrid Transformer had previously.
In reality since it is not really impedance mismatches that stop us
supressing Hybrid leakage but rather the fact that we do not know the
overall gain of a System (and hence the exact Signal level we need to
cancel out) - nor does that gain remain constant - largely due
to line length variations etc I would have to argue that in a real
world telephone network no matter how well designed is it not possible
to provide an echo free service without either a VSA (Voice Switched
Amplifier the old method
of echo supression which basically gated the signal in one direction at
time) or DSP based echo supression.
ideal Sidetone is considered to be the level we are used to hearing
speak (sorry I don't recall the figure off hand)
again my apologies I had meant to follow these ideas further before
However even well designed pure digital Transmission systems require
suppression when the round trip delay exceeds a certain amount of time
consider this to be 36mS-
http://www.iec.org/online/tutorials/acrobat/echo_cancel.pdf - which is
actually an excellent tutorial on the whole subject and one that I
highly ) due to the leakage through the hybrid. If rtd is kept below
echo does still occur however it is heard by the person speaking as
the sidetone (a variant of this effect called 'double tracking' is an
often used in recordings and live concerts to give vocals more power -
digress .....) over 36mS however this starts being noticed as a
echo and becomes a problem.
I suspect that your 'well designed network' is merely one in which no
has an RTD of over 36mS. Unfortunately due to the laws of physics this
precludes networks which have paths of over 5,700 Kms (In Australia we
approach that limit before we even consider processes such as those
the codec etc which cause additional delay). Of course when/if you add
Satellite to the equation you also add 50,000 Kms of path or 166mS and
very noticeable echo - I believe (but not working there am unable to
confirm) that India makes (or maybe made) extensive use of satellite
technology as a (relatively) cheap way of providing telecommunications.
- Alternatively ISDN type connections do not naturally have Sidetone or
suffer from echo when both ends terminate on ISDN (try any call through
tester such as an IBT 1A) - however this is due to the fact that there
fact no hybrid in such a system - anywhere! TX and RX are maintained as
completely isolated paths through the entire network. Sidetone for ISDN
phones is added deliberately by the handset manufacturer.
Bottom line is (and getting back to the point of the thread - at least
point where I joined) echo is not caused by cable reflections but
Hybrid leakage - which is often incorrectly called a 'reflection - yes
reflections do occur and this is put to good use by the pulse echo
however they do not occur at levels sufficient to cause noticeable
at VF (Voice Frequency).
Re: Z mismatch - was Slew Rates and Mic Pre performance
Trans hybrid loss in a two transformer design hybrid (4 Wire Hybrid In port
to 4 Wire Hiybrid Out port) is closer to 7dB loss where the impedance of the
termination connected to the 2 Wire Line port is an infinite mismatch to
that of the impedance of the network connected to the Balance Network port
(ie.e an infinite : 1 ratio). I have measured slightly less (6.7dB) in
proactice on some hybrids, but 7dB is closer to the norm.
The 7dB figure is derived from the fact that a minimum of 3dB loss will
occur between adjacent ports. Add a little extra (say 0.5dB) for
transformer losses and a practical adjacent port loss closely approaches
3.5dB. 3.5dB + 3.5dB = 7dB. A diagram would be easier to illustrate, but
as this is a text only NG, posting a pic is not possible.
No. Albeit that hybrid mismatch is a major contributor to echo in the 4
wire transmission path, it is not the sole cause. Mismatches due to cable
gauge changes onn a simple 2 wire circuit will produce an echo (reflection)
due to impedance mismatch at the point they join. Echo and signal
reflection are one of the same thing. Just we associate echo with long
return path delays of 35mS or more. It's purely an auditory perception
thing as delays less than 28mS are very hard for the human brain to
It would be pointless creating a 4 wire transmission system containing
hybrids if no active transmission components or ADC / DACs were involved.
Agreed. If losses exceed gains the circuit must be unconditionally stable.
Yes. Especially if transformer hybrids are used. The adjacent port loss in
each hybrid is very close to 3.5dB. As there is a hybrid at each end of the
link, the minimum losses for each direction of transmission would be close
Once again a circuit diagram would assist in the explanation. A very simple
circuit of t phone would have the transmitter and the receiver in series.
The sidetone level on the receiver would be very high.. The idea of the
ASTIC is to reduce the level of the speech current generated by the
transmitter reaching the receiver, whilst maximising transmitter signal to
line and also maximising the received line incoming speech signal reaching
the receiver. ASTICs are usually single transformer leaky hybrids. Good
examples are to be found in the old Telecom 800 series phones.
In the old phones (say 800 series and earlier) there was no gain, except in
the hearing aid version of the 800 series which featured a volume control
where the recall button is usually located.
- In a side note the ASTIC was
As I said previously. Early phones had no ASTIC and sufferred from very
high sidetone. This high sidetone level caused people to talk softer
because they believed the other party could hear them okay based purely on
the effect of the local sidetone level. This became an even bigger problem
on long distance trunk calls which suffered significantly more transmission
loss than local calls.
I beg to differ. Getting impedance matching and hybrid balancing correct
negates the requirement for echo cancellation in either the digital or
analogue domains. This is immediately apparent on facsimile and data calls
through the PSTN (not ISDN) where echo cancellation can't be used. To get
reasonable error free data throughput through the PSTN, the echo performance
of the transmission path must be reasonably good to begin with.
True. When we hold a telephone handset to our head, we cut off some of the
natural sidetone we would normally experience in non telephone coversation.
This means that we need to replace some of that lost natural sidetone with
an equivalent in the telephone. The correct level of sidetone effectively
regulates how loud we speak into the telephone. Ever noticed how spoken but
profoundly deaf people speak - quite often too loud or too soft.
The correct sidetone level in a phone should be approximately 13B below the
speaker's transmission level (local end).
The IEC tutorial takes a very simplisitic approach to the issues surrpunding
echo cancellation in typical telephone networks. It completely ignores the
issue of VF data (fax and modem calls) through the PSTN and how echo
cancellors are supposed to handle such calls.
Furthermore the document dips to mediocrity with the first sentence on Page
5 which reads " Unfortunately, the hybrid is by nature a leaky device."
What utter crap. I've measured return losses in old 1954 two transformer
hybrids which exceed 60dB! A million to 1 times power isolation is nothing
to be sneezed at. Hybrids, when correctly balanced are supposed to provide
isolation between the 4 wire transmission paths - not leakage.
You've obviously never conducted a Near End return loss measurement on a
customer's telephone line with an EDL423 Network Transmission Quality
Tester. Run a transmission test on a loaded cable (nomimal Z = 1200 ohms)
where it interfaces to a LIB7 LI in an AXE exchange (Zin = 600 ohms fixed).
Where the cable interfaces into the exchange there is a 2 : 1 impedance
mismatch. Even though the mismatch only contributes to 0.5dB additional
forward transmission loss, the return loss at this point is a mere 9.2dB - a
mismatch in anyone's book, with more than 10% of the transmitted signal
reaching this point being reflected back to the source. The only thing
reducing this reflected signal back to the customer is the transmission loss
of the cable itself which by Australian standards must be less than 6.5dB @
Beware of Alan Rutlidge
( snip far too much crap for anyone to bear)
** Alan Kendall Rutlidge of 15 Ruth Street Northbridge, WA (aka The Arse
Bandit ) is a congenital liar.
He has lied through his backside for every day of his whole life.
He first lied to his parents.
He then lied to his teachers.
He the continued lied to Telstra where they sheltered the evil sod for 26
No lie whatever is beneath or beyond him.
Rutlidge is a consummate lair.
There is no-one he will not try to deceive.
Such an individual is an, evil public menace.