Maker Pro
Maker Pro

models for simulating a dsl link

E

Eeyore

Jan 1, 1970
0
A couple of recent threads about the effects of DSL (micro)filters gave me the
idea of analysing their effect under various conditions in a simulation package.

I've got several examples of filters of varying designs here and I'd really like
to know how much difference and indeed what kind of differenc(es) they can make.

As some know from previous threads I'm a bit sceptical about how much they
affect the DSL signal itself.

Before I can do this I need some suitable 'models' for the various bits of
telecom gear involved. Can anyone point me in the right direction ?

Graham
 
P

Paul Hovnanian P.E.

Jan 1, 1970
0
Eeyore said:
A couple of recent threads about the effects of DSL (micro)filters gave me the
idea of analysing their effect under various conditions in a simulation package.

I've got several examples of filters of varying designs here and I'd really like
to know how much difference and indeed what kind of differenc(es) they can make.

As some know from previous threads I'm a bit sceptical about how much they
affect the DSL signal itself.

Before I can do this I need some suitable 'models' for the various bits of
telecom gear involved. Can anyone point me in the right direction ?

Unlike cable (DOCSIS) modems, DSL comes in many flavors with widely
varying performance specifications. Take a look at the Wikipedia article
for a start (http://en.wikipedia.org/wiki/DSL). I don't think there is
an industry standard like there is for cable broadband. Each
manufacturer has their own specs.

As far as the DSL filters go, since they are intended to be placed on
all of the other (non DSL) equipment on the shared line, their
performance might also depend on the nature of the telephone equipment
connected downstream of them.
 
E

Eeyore

Jan 1, 1970
0
Paul Hovnanian P.E. said:
Unlike cable (DOCSIS) modems, DSL comes in many flavors with widely
varying performance specifications. Take a look at the Wikipedia article
for a start (http://en.wikipedia.org/wiki/DSL). I don't think there is
an industry standard like there is for cable broadband. Each
manufacturer has their own specs.

That won't affect the basic physics though, which is all I want to model.

As far as the DSL filters go, since they are intended to be placed on
all of the other (non DSL) equipment on the shared line, their
performance might also depend on the nature of the telephone equipment
connected downstream of them.

That's exactly one of the things I'd like to experiment with.

Graham
 
G

Gordon Hudson

Jan 1, 1970
0
As some know from previous threads I'm a bit sceptical about how much they
affect the DSL signal itself.

You can do a simple test by unplugging the filter and putting a lead from
the ADSL modem straight into the phone socket.
Although the different quality of the cable might be as important as the
filter.

They are bound to attenuate slightly.
Even a straight through extensions lead three inches long would have some
attenuation.
Whether it makes enough of a difference is the issue.


--


Gordon Hudson || Hostroute.com Ltd
e-mail:ghudson [at] hostroute.net
http://www.hostroute.co.uk/resellers Host 5 web sites for £9 per month
http://www.nameroute.co.uk/ Domain Names with free hosting and email
http://www.myqth.co.uk/ 3000MB of web space for £29 per year
 
E

Eeyore

Jan 1, 1970
0
Gordon said:
You can do a simple test by unplugging the filter and putting a lead from
the ADSL modem straight into the phone socket.

Yes I know this kind of thing. It misses the point though. I want to *quantify*
in decibels of loss, frequency response etc the effect of such things using a
modern electronics simulation package.

Although the different quality of the cable might be as important as the
filter.

Absolutely not. If a cable wasn't twisted pair that might make a difference
though. I wouldn't call that 'quality' myself. I like to see things called by
their correct names.

Graham
 
A

Alec

Jan 1, 1970
0
If you wish to model the filters you will need to measure the inductance and
capacitance of the elements of the filter.

If you can do this you probably have the equipment to measure the real
performance so why not simply measure the performance under the conditions
you are interested in?

Alec
 
E

Eeyore

Jan 1, 1970
0
Alec said:
If you wish to model the filters you will need to measure the inductance and
capacitance of the elements of the filter.

Yes. The Capacitance values are printed on the caps btw.

If you can do this you probably have the equipment to measure the real
performance

Not from the exchange end. I don't have access to that for one thing ! The whole
entity of the line from exhange to subscriber is what I intend to simulalte.
It's quite simple as long as I can get the numbers to punch in for various
pieces of kit such as phones.

Let me put it another way. Right now I'm looking for phone schematics so I can
make a model of them. Can you point me in the direction of any ?

so why not simply measure the performance under the conditions
you are interested in?

Because I want to simulate any number of generalised (and specific) conditions
and measure the effect of individual changes.

Measuring performance the wayyou suggest is so passe these days and hoplessly
time consuming.

Graham
 
J

John

Jan 1, 1970
0
Eeyore said:
Yes. The Capacitance values are printed on the caps btw.



Not from the exchange end. I don't have access to that for one thing ! The
whole
entity of the line from exhange to subscriber is what I intend to
simulalte.
It's quite simple as long as I can get the numbers to punch in for various
pieces of kit such as phones.

Let me put it another way. Right now I'm looking for phone schematics so I
can
make a model of them. Can you point me in the direction of any ?

Try looking on http://www.telephonesuk.co.uk/ There's wiring diagrams and
schematics on there, I believe (not looked myself so don't shout at me if
there isn't :eek:)

John
 
If you wish to model the filters you will need to measure the inductance and
capacitance of the elements of the filter.

If you can do this you probably have the equipment to measure the real
performance so why not simply measure the performance under the conditions
you are interested in?

You can measure the physical parameters R, L and C and put these
into an analysis package to get the passband, attenuation, phase shift
and other parameters, expressed in compatible units such as db.

This won't be much help because the filter assumes input and
output impedance parameters that are laboratory design concepts,
that don't match real life, so your analysis package graphs won't
match real life. In fact you may find that what appears to be a
cheapo sub-standard filter gives better results because it creates
a better match between the incoming lines impedance, and the
modems input impedance.

Like a lot of HF electronic design you can impedance match (often with
difficulty) at a specific frequency, but the match rarely holds over
a range of frequencies where the source and load are complex,
and frequency dependent, numbers.
Complex numbers in this sense are numbers/vectors with real and
imaginary components, where you get into hyperbolic functions.

Like your TV antenna should be 75 ohms across its frequency range,
and your TV's antenna socket should be 75 ohms across the
full range of TV channel frequencies, it's a pie in the sky concept.
In the heyday of CB radio it was all about matching, many and
various, were the weird and wonderful combinations of coils and
capacitors, pie, half pie, etc.
 
E

Eeyore

Jan 1, 1970
0
You can measure the physical parameters R, L and C and put these
into an analysis package to get the passband, attenuation, phase shift
and other parameters, expressed in compatible units such as db.
Correct.


This won't be much help because the filter assumes input and
output impedance parameters that are laboratory design concepts,
that don't match real life, so your analysis package graphs won't
match real life. In fact you may find that what appears to be a
cheapo sub-standard filter gives better results because it creates
a better match between the incoming lines impedance, and the
modems input impedance.

But *does* the filter (did the filter designer) assume that ? It certainly doesn't
have to be that way. That's why I fancy simulating some.

Like a lot of HF electronic design you can impedance match (often with
difficulty) at a specific frequency, but the match rarely holds over
a range of frequencies where the source and load are complex,
and frequency dependent, numbers.
Complex numbers in this sense are numbers/vectors with real and
imaginary components, where you get into hyperbolic functions.

Like your TV antenna should be 75 ohms across its frequency range,
and your TV's antenna socket should be 75 ohms across the
full range of TV channel frequencies, it's a pie in the sky concept.
In the heyday of CB radio it was all about matching, many and
various, were the weird and wonderful combinations of coils and
capacitors, pie, half pie, etc.

At the lowish frequencies used for ADSL I'm not clear how much that 'matching' is
relevant. It certainly won't be a big issue at ~ 30kHz for example, rather more of
one at 300kHz and 1MHz..

Graham
 
J

Joel Kolstad

Jan 1, 1970
0
In the heyday of CB radio it was all about matching, many and
various, were the weird and wonderful combinations of coils and
capacitors, pie, half pie, etc.

All of that is still around... it's quite visible with the HF amateur radio
guys, and of course in something like a cell phone it's all there but just
much, much tinier.

When you're talking about transmitters, not only is proper matching desirable
from the point of view of getting as much signal into the air as possible,
it's also important in that serious mismatches can create voltages that simply
blow up the transmitter's finals. :) (At least with solid state devices --
tubes apparently are much more immune to this.)
 
R

Rich Grise

Jan 1, 1970
0
That won't affect the basic physics though, which is all I want to model.


That's exactly one of the things I'd like to experiment with.

Find someone nearby with DSL, and ask them if they have a spare filter. My
DSL provider gave me a handful of filters - just see if you can get
somebody to give you one, and hack it and find out what's actually inside
the thing.

By Occam's razor, it's probably just an ordinary LC low-pass filter,
probably balanced and impedance-matched to the line(s).

Good Luck!
Rich
 
But *does* the filter (did the filter designer) assume that ? It certainly doesn't
have to be that way. That's why I fancy simulating some.

What you are up against is that an ADSL filter designer has to
start from a set of R, L, C parameters that create the impedence
of an incoming exchange line.
The problem is that someone a quarter mile from the exchange
sees a different incoming line impedence to someone 2 miles
from the exchange.
In the early days of telephony loading coils were used to level
the frequency response of lines over the speech band, don't
know if they are still used. In essence twisted pair street
cables are very capacitive, and loading coils were inductive
to compensate.
You have to assume what comprises a typical ADSL line.
Then you have to hope that the typical ADSL modem matches
the line, and would be a good match when plugged in without
a filter.
Having made these sweeping assumptions you now decide
how sharp the filter needs to be, and what configuration to
use.
You end up with a filter that hopefully does as good a job
as possible in most situations.
At the lowish frequencies used for ADSL I'm not clear how much that 'matching' is
relevant. It certainly won't be a big issue at ~ 30kHz for example, rather more of
one at 300kHz and 1MHz..

Matching is always important because the launch energy at
the source has to go somewhere, it can't just be destroyed and
disappear, "conservation of energy" and if it's not absorbed
it is likely in 200 pair street cables, to appear as noise in
adjacent (more than likely) ADSL pairs.

Unfortunately a lot of street cables in the UK are immediate
post war, though there is still a bit of paper insulated, lead
overall, still around. Never designed for ADSL.

These cables were never speced for Near End Cross Talk
(NEXT), if we were starting again they would be, but then
if we were starting again fibre to the kerb would be a better
model.

But if you wan't to dig deeper
http://whatis.techtarget.com/definition/0,,sid9_gci1059969,00.html
is a good starting point for understanding what is involved.
 
M

Michael A. Terrell

Jan 1, 1970
0
Paul Hovnanian P.E. said:
DSL modems aren't supposed to be used downstream of DSL filters. DSL
filters are supposed to be put in line with everything _except_ the
modem.


Paul, some of the filters have two output phone jacks. Filtered, for
a telephone, and unfiltered for the DSL modem.


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida
 
P

Paul Hovnanian P.E.

Jan 1, 1970
0
Gordon said:
You can do a simple test by unplugging the filter and putting a lead from
the ADSL modem straight into the phone socket.
Although the different quality of the cable might be as important as the
filter.

DSL modems aren't supposed to be used downstream of DSL filters. DSL
filters are supposed to be put in line with everything _except_ the
modem.
 
E

Eeyore

Jan 1, 1970
0
Rich said:
Find someone nearby with DSL, and ask them if they have a spare filter.

I talk to myself and find I have 3 spare filters ! :~)

My DSL provider gave me a handful of filters - just see if you can get
somebody to give you one, and hack it and find out what's actually inside
the thing.

2 of them are basically simple 2nd order LC filters feeding the POTS socket. The
L consists of 2 windings on one core to retain line symmetry. They also include
the 'ring capacitor' too, a UK thing AFAIK. The bell circuit runs as a 3rd wire
inside the house here. Doing it that way avoided 'bell tinkle' in the old days
of pulse dialling.

By Occam's razor, it's probably just an ordinary LC low-pass filter,

See above. The third is stuffed full of bits. 3 x 2-section inductors, 5 caps,
and 2 Rs.

probably balanced and impedance-matched to the line(s).

Not so sure about impedance matched. What impedance is a phone anyway, never
mind 4 in parallel ( the max load for a line and a single filter is *supposed*
to cope ) ?

Graham

p.s. *what impedance is a phone anyway* ? Please see thread title ! Off-hook,
on-hook I need data !
 
E

Eeyore

Jan 1, 1970
0
What you are up against is that an ADSL filter designer has to
start from a set of R, L, C parameters that create the impedence
of an incoming exchange line.

'Create the impedance'. Can you elaborate on that ? I don't think it's that at all
unless I'm very mistaken. In any case the impedance of a filter will be all over the
place with frequency.

The problem is that someone a quarter mile from the exchange
sees a different incoming line impedence to someone 2 miles
from the exchange.

More like a different resistance actually. At *audio* frequencies the short run of cable
from the exchange doesn't really impose any characteristic impedance value.

At the DSL frequencies the issue will come into play but the DSL filter's not in that
path so the DSL modem / router has to deal with that one just for itself.

In the early days of telephony loading coils were used to level the frequency response
of lines over the speech band, don't know if they are still used.

I've heard of them but know almost nothing more about them. I gather DSL won't work if
there's one on the line. I suspect also they may have been there to help with the volume
level on long lines.

In essence twisted pair street cables are very capacitive, and loading coils were
inductive to compensate.

OK. Easy to simulate again if so inclined.

You have to assume what comprises a typical ADSL line.

Yup. Or several examples of same from 100m from the exchange to several km.

Then you have to hope that the typical ADSL modem matches
the line, and would be a good match when plugged in without
a filter.

They are apparently.

Having made these sweeping assumptions you now decide
how sharp the filter needs to be, and what configuration to
use.

I'm not making that decision. I have some examples here and they vary considerably.
That's the objective of this exercise. To see just how much difference a DSL filter can
make and in what ways.

You end up with a filter that hopefully does as good a job
as possible in most situations.

relevant. It certainly won't be a big issue at ~ 30kHz for example, rather more > of
one at 300kHz and 1MHz..

Matching is always important because the launch energy at
the source has to go somewhere, it can't just be destroyed and
disappear,

Only when you're talking about a transmission line. POTS telephone local loop circuits
are *NOT* transmision lines.

"conservation of energy" and if it's not absorbed
it is likely in 200 pair street cables, to appear as noise in
adjacent (more than likely) ADSL pairs.

Unfortunately a lot of street cables in the UK are immediate
post war, though there is still a bit of paper insulated, lead
overall, still around. Never designed for ADSL.

Can't take all that into account too. I'm interested in what's likely to have been
around for say 30-40 yrs.

These cables were never speced for Near End Cross Talk
(NEXT), if we were starting again they would be, but then
if we were starting again fibre to the kerb would be a better
model.
Yes.


But if you wan't to dig deeper
http://whatis.techtarget.com/definition/0,,sid9_gci1059969,00.html
is a good starting point for understanding what is involved.

Thanks.

Graham
 
E

Eeyore

Jan 1, 1970
0
Michael A. Terrell said:
Paul, some of the filters have two output phone jacks. Filtered, for
a telephone, and unfiltered for the DSL modem.

Yes. The modem connection is simply a pass-through.

Graham
 
See above. The third is stuffed full of bits. 3 x 2-section inductors, 5 caps,
and 2 Rs.



Not so sure about impedance matched. What impedance is a phone anyway, never
mind 4 in parallel ( the max load for a line and a single filter is *supposed*
to cope ) ?

Although it may be a pass through, what is shunted across the line
is important. Like a recent post here, where sync only occured when
the phone was lifted. What happens on the phone side will be reflected
onto the ADSL side.

Tlelephone twisted pair has a characteristic impedence, for
high spec cables the mfr usually quotes it, along with figures
for NEXT and other line characteristics.
You can do it the hard way and start from the dielectric
constant of the insulation, the wire spacing (thickness of
the insulation), and the wire diameter.
Old timers will remember the Royal Core of Signals
bible The Handbook of Line Transmission, and the
various BT books.
There are also test sets for measuring line parameters.

You can measure anything, it's really a matter of
determining what you need to measure, and how it
relates to the real world.
 
E

Eeyore

Jan 1, 1970
0
mind 4 in parallel ( the max load for a line and a single filter is *supposed*

Although it may be a pass through, what is shunted across the line
is important. Like a recent post here, where sync only occured when
the phone was lifted. What happens on the phone side will be reflected
onto the ADSL side.

But not at all frequencies. I don't deny it must have an effect of some kind of
course.

Tlelephone twisted pair has a characteristic impedence,

Not at the lengths used for the local loop at audio frequencies. This is a total
fallacy.

for high spec cables the mfr usually quotes it, along with figures
for NEXT and other line characteristics.

Twisted pair stuff has a characteristic impedance of around 100 ohms (when it
becomes relevant which it isn't at audio frequencies in the local loop).

You can do it the hard way and start from the dielectric
constant of the insulation, the wire spacing (thickness of
the insulation), and the wire diameter.

That is indeed what matters. As I said, it's about 100 ohms for LLU wiring.

Graham
 
Top