1. Actually, your THL results are pretty good, although they might be improved if you optimize the chip-side driving resistor. The spec sheet for the 671-8056 calls for a 374 ohm resistor, which is the nearest MIL standard 1% value to the optimimum resistance that will provide 600 ohms on the line side.
2. Most transformers designed for North American compliance have more than
adequate bandwidth for a V.34 signal. The problem, particularly with the
33.6 kb/s
implementation, is with the Central Office codec on the line card which
contains
a multiple-order (five or six) filter. The filter starts rolling off at
around 3750 Hz,
signficantly attenuating high-frequency modem signals. For this reason, the
high-end response of the modem is generally not the limiting factor.
The low-end, below 200 Hz, is typically not a problem for a dry transformer
as
most have around 10 Hy of inductance.
3. Gack! Over the years, in talking with telecom experts, I have been told
that
the telephone line may be characterized as 600 ohms +/- 600 ohms. In
actuality, it goes beyond even that. I made some measurements of telephone
line impedance when I lived in Mankato, MN in 1984. They are available for
download (Microsoft Word format) from our web site at
http://www.midcom-inc.com/technotes/tn_abs.html#23
If you are working from behind a PBX, the impedances get even wilder. I
recall
resistive and reactive components exceeding 1.5K ohms.
3. (the second 3.): We made some measurements on a part which is very similar
to 671-8056 a few years ago where we deliberately introduced a parallel
resistor
across the transformer's chip side in an effort to reduce the effect of
varying line
impedance. The results showed that the use of a 3.25K ohms "compensation"
resistor allowed a line impedance variation between 419 and 851 ohms and
still
maintain 20 dB THL. I will fax you the one-page summary of the results.
I have asked several people to try this idea in an actual design to see if
it helped.
I have not received a response from any of them, so I can't give you any
field
test data.
Best regards,
Dave LeVasseur
Advanced Product Development Manager
Midcom, Inc.
----------
From: David Drori[SMTP:[email protected]]
Sent: Tuesday, October 08, 1996 2:45 AM
To: TREG
Subject: V.34 DAA Transhybrid Loss
Dear Treggers,
I wonder if any modem and DAA designers can give us some advice. We are
developing a DAA for a V.34 modem. Our design is based on a "dry" Midcom
671-8056 transformer, connected to the line via a 2.2 uF capacitance . The modem
itself is a proprietary design based on a DSP and analog front-end (A/D and D/A
sigma-delta converters).
The modem DSP software designer is asking us to improve the transhybrid loss
(THL) characteristic by making it flatter against frequency. He is complaining
that the THL of our design droops for low frequencies of around 150 to 400
Hertz, compared with the value for higher frequencies, when the DAA is connected
to a line comprising 600 ohms + 2.2 uF. Measurements indicate THL values of
approx. 19, 22 and 28 dB, at 100, 300 and 1000 Hertz, respectively.
We have explained that near-end DAA echo-cancelation necessarily represents a
compromise, since line impedances vary considerably, and we designed for an
"average" sort of line. According to our simulations, we can expect measured THL
responses to be even less steady than the results measured for a 600 ohm + 2.2
uF line, when the circuit is tested against a full range of line impedances. We
have stated that we could improve the response for a specific line type to the
detriment of response against other line types, but that such a step would not
be beneficial.
Since the software designer's complaint refers to low frequencies on the order
of 200 Hertz, it might be possible to increase the insertion loss in both the
transmission and reception path at these low frequencies. However, I am
concerned that this might adversely affect modem performance in other respects
(due to increased group delay, or some other factor).
Naturally, we are keen to keep the design straightforward and avoid any exotic
adaptive hardware, etc., which nobody else seems to require. It seems to us that
other modem software designers have reconciled themselves to a far-from-ideal
THL response from simple, inexpensive DAA designs, and manage to cope
satisfactorily.
With reference to the above, could anyone tell us the following:
1. Is the described THL response against frequency for our design typical of DAA
designs for V.34, or is our design particularly bad?!
2. What kind of flatness against frequency should we be aiming for, based on the
accepted state-of-the-art reached by V.34 software, and which other V.34 DSP
software has to cope with?
3. What "average" line impedance do you recommend us to choose, against which we
should design for best THL response? I have seen figures of 350 ohms in series
with a parallel combination of 1000 ohms and 0.21 uF.
3. Can anyone confirm or dispel our worries about the possible adverse effects
of deliberately increasing the insertion loss for low frequencies (below around
400 Hertz), in order to minimize THL degradation at these frequencies?
Any advice on the above will be enormously appreciated, and will make us feel
quite sure of ourselves when talking to the software designer - or cause us to
try and improve our design - as the case may be.
Best regards,
David Drori
-----------------------------------------------------------
Novarex Enterprises Ltd.,
POB 2833,
306-6, Mevo Hapartisanim Street,
Jerusalem 91028,
Israel.
Tel: +972 2 581 0995
Fax: +972 2 581 3750
-----------------------------------------------------------
<<application/ms-tnef>>
