TCI Cobra interconnects against Chord Chameleon
 

In article , John Phillips
wrote:


The real-world effects seemed, I think, to be mainly due to the cable's
lumped R and L parameters interacting with the 'speaker's impedance
curve. An ESL-57 'speaker model (with thanks to Jim Lesurf's web site)
was an interesting illustration of some extreme differences in frequency
response that can occur between different cables.

Should say I took that model of the 57 from someone else. I suspect that,
as with other QUAD speakers, the actual impedance varied when they made
issue changes to tweak the speakers during the long period they were on
sale.

Slainte,

Jim

--
Change 'noise' to 'jcgl' if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

TCI Cobra interconnects against Chord Chameleon
 

In article , Don
Pearce
wrote:


No, cables don't have lumped parameters - they have distributed
parameters - that's what makes them cables.

Probably better to say they don't "have" either lumped or distributed
parameters. We use those to describe what they do. :-)

But the point is this; wherever the lumped equivalent parameters matter,
you will get a better answer from the distributed model.

Again, this rather depends on what you mean by "better". Problem here is
that if you really want accurate results then you have to include the
frequency dependencies of all four RCLG values.


Speakers are particularly interesting in that there are many cables
available, some of which (from Goertz) have inductance and capacitane
which together come down to around 8 ohms as a distributed impedance.
These cables, despite having enormous capacitance, are essentially
ruler-flat in frequency.

The last sentence is meaningless as it has no reference to the conditions
of use. A 'cable' doesn't have a "ruler flat" response. What you get
depends on the system. And as soon as the load doesn't match the cable
perfectly, the source matters as well. Problem here is that domestic LS
uses a voltage assertion approach with loads that have values that vary all
over the shop.

Slainte,

Jim

--
Change 'noise' to 'jcgl' if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

TCI Cobra interconnects against Chord Chameleon
 

In article , Don
Pearce
wrote:
Eiron wrote:


Are you saying that Goertz speaker cables act as transmission lines
with a characteristic impedance of 8 ohms? If so, you should also
state the low corner frequency below which the characteristic
impedance rises, and the attenuation per unit length.


The point about that corner is that it is a phenomenon of the low end,
where it makes essentially no difference.

That would rather depend on what actual frequency was meant by "low end".

Were it something that happened as you went up in frequency it would
matter much more. As for attenuation per unit length, does it ever
matter in a domestic setting? If there is a little more, just nudge the
volume control. Anyway, to achieve the low characteristic impedance you
need a fair amount of copper, so I'm guessing they aren't too bad.

Actually, what may matter is the geometry, not so much the amount of
copper. Although for clear reasons, tiny cross sections of wire tend to
mean a high value for R', which then disrupts the attempt to get a level
characteristic.

However since LSs aren't in general 8 Ohm loads...

Slainte,

Jim

--
Change 'noise' to 'jcgl' if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

TCI Cobra interconnects against Chord Chameleon
 

In article , John Phillips
wrote:
On 2008-07-15, Don Pearce wrote:
John Phillips wrote:
Do you have particular example where you think the lumped and
transmission line models will differ in a cable-speaker application
to a degree that is practically relevant?

Not handy. But it is clear that the lumped version of the model is a
lowpass filter, which a cable simply isn't.

The lumped version presumes a 'short' cable in wavelength terms. For audio
frequencies and runs of a few metres this is generally quite a decent
assumption.

[snip]

The problem comes when you are working in matched conditions - like
the Goertz cable, and for the real cable, like the distributed model
there is no turndown. The lumped approximation fails. You are left
with the question of working out the degree of failure in other
conditions.

I am concerned about your use of "matched conditions" WRT the Goertz
cable. Matched conditions don't exist in an amp-cable-speaker system.
Even if you have an "8-ohm" speaker, and an 8-ohm Goertz cable.

1. The amplifier output impedance is not 8 ohms - maybe 10 mOhms for a
good example.

That may not affect response if the load *is* matched to the cable. Just
affects the nominal efficiency.

[snip]

... If I have two pieces of
cable, one of 10 microhenries and 100 pF per metre, and another of 20
microhenries and 200 pF per metre, they will perform identically.

I think not in a mismatched system.

Indeed.

They certainly perform differently in the quick frequency response
simulation I just did of a 10 mOhm impedance amplifier driving 5 metres
of the two different cables (10-section lumped model or a single section
- same result), each terminated with the netlist above. When I get home
I will check to see if I made an error (but I can't spare the time just
now).

FWIW You should get similar results if working out the reflection
coefficient behaviour using a transmission line model. For such short
cables at audio frequencies the two methods tend to yeald very similar
results.

Slainte,

Jim

--
Change 'noise' to 'jcgl' if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

TCI Cobra interconnects against Chord Chameleon
 

John Phillips wrote:
On 2008-07-15, Don Pearce wrote:
John Phillips wrote:
Do you have particular example where you think the lumped and transmission
line models will differ in a cable-speaker application to a degree that is
practically relevant?
Not handy. But it is clear that the lumped version of the model is a
lowpass filter, which a cable simply isn't. The area we are dealing with
here is that part of the filter where the flat slope is just starting to
turn down marginally before it gets to the true corner. Now for all
practical purposes, that slope is a good approximation of what happens
in a true cable model when it is mismatched. The problem comes when you
are working in matched conditions - like the Goertz cable, and for the
real cable, like the distributed model there is no turndown. The lumped
approximation fails. You are left with the question of working out the
degree of failure in other conditions.

I am concerned about your use of "matched conditions" WRT the Goertz
cable. Matched conditions don't exist in an amp-cable-speaker system.
Even if you have an "8-ohm" speaker, and an 8-ohm Goertz cable.


No, I'm not being pedant about match. What I'm saying is that a
nordinary speaker cable is about 110 ohms or so - a long, long way from
the impedance of the speaker. This results in the lead looking like a
series inductor. If the impedance of the lead can be brought down to the
same ballpark as the speaker, that inductance largely disappears. That
makes the response flatter.

1. The amplifier output impedance is not 8 ohms - maybe 10 mOhms for
a good example.


Not important. Hypothetically, if you could have an 8 ohm purely
resistive speaker, and joined it to the amp with an 8 ohm cable as long
as you wished, the amplifier would see a purely resistive 8 ohm load,
just as if the speaker had been joined directly to it.

2. I agree that a cable can have an 8-ohm impedance (most are, in
practice much higher, though).

3. The loudspeaker is (almost) never flat 8 ohms. For example, try
this netlist which simulates the ESL-57 in your simulator:

.subckt SPEAKER 1 9
* Quad ESL57 model
L1 1 31 20e-6
R1 31 9 80
L2 31 9 80e-3
L3 31 32 50e-3
R2 32 9 80
C1 31 9 4e-6
R3 31 33 15
R4 33 34 15
C2 34 9 2.5e-6
C3 33 34 5.6e-6
L4 31 35 0.3e-3
R5 35 36 15
C4 36 9 35e-6
.ends

And I can provide netlists for a more conventional loudspeaker (my
"8-ohm" Proac D15s) which show similar wide variations in impedance
with frequency.

... If I have two pieces of
cable, one of 10 microhenries and 100 pF per metre, and another of 20
microhenries and 200 pF per metre, they will perform identically.

I think not in a mismatched system. They certainly perform differently in
the quick frequency response simulation I just did of a 10 mOhm impedance
amplifier driving 5 metres of the two different cables (10-section
lumped model or a single section - same result), each terminated with the
netlist above. When I get home I will check to see if I made an error
(but I can't spare the time just now).


They will in a lumped model. In real life they don't.

d

TCI Cobra interconnects against Chord Chameleon
 

Jim Lesurf wrote:
In article Mr6dnYcrJMOI_OHVnZ2dneKdnZydnZ2d@plusnet, Don Pearce
wrote:
Jim Lesurf wrote:
In article , Don

1) That the cable impedance (and EM wave velocity) is set by the L'
and C' prime values. Thus ignoring R' and G'. If you actually examine
the situations for LS cables at *audio* frequencies the actual values
for Zc and V are generally very different to those you get simply by
using L' and C'. Sometimes orders of magnitude different. And strongly
frequency dependent.

You are right about the effect consider R' and G', but they are
unimportant for the following reason - they only become factors that
affect the cable impedance at low frequencies, at which they make no
difference.

The snag being that "low frequencies" here generally covers the bulk of the
audio band when one looks at the kinds of cables people use.

See, for example

http://jcgl.orpheusweb.co.uk/temp/cableZV.gif

Think of the above as a 'sneak preview' of some of the results that HFN
should be publishing soon. :-) It shows the impedance and EM wave
velocity values for a random selection of cables. ( did actually model
rather more cables than this, but decided to just use some. Otherwise the
graph had so many lines they became impossible to disentangle by eye!)

The lines with the red squares show an example of the "8 Ohm" woven cables.
You can see the actual impedance is well above 8 Ohms over much of the
frequency range.

The above was from some historic published data. A later article will
publish some new measurements I have done which give similar sorts of
values and results.

G' probably doesn't matter much for most cases. But L' may well, whilst
with unmatched use most cables can be well modelled using R' and L' and
ignoring C' as well as G'. This is the result I get from comparing models
with measured results. It stems from the general rule that the source and
load impedance is usually well below the cable characteristic impedance,
and the cable is short in wavelength terms. This is for audio frequencies,
of course. I got some other results from modelling and measuring at
ultrasonic/RF frequencies. But the first article focusses on the audio
range, and leaves high frequencies to a later article. :-)


Once you are up into the region where cable parameters can
cause unflatness, they are second order effects, and the hf model which
only considers L and C is just fine.

2) You assume matched operation. But with domestic LS use this is
generally far from being so. When you examine more practical
situations the behaviour is very different to a matched case - which
would be almost impossible to arrange at audio due to the strong
frequency dependence of the impedance of the cables.


No, I'm not assuming matched operation, although there are cables which
come close for loudspeakers.

It seems to be a rare situation as in general both the cable impedance and
the LS impedance are frequency dependent, using with quite distinct
patterns of variation. And when the operation isn't matched then it tends
to turn out as I have described, based both on models and measurements.


What I'm saying is that a model which only looks at L or C in isolation
will always give wrong answers.

Rather depends on what you mean by "wrong". No model will give "right"
answers in all cases if you require absolute precision. But the reality
seems to be that in the audio band for LS cables in domestic situations you
get accurate results if you use R' and L' and ignore C' in general.


People talk about cables being capacitive on the basis of a high
pF/metre figure. This is nonsense; for a cable to be capacitive it must
have a characteristic impedance lower than the load impedance -
something which almost never happens with speaker cables, which in 99%
of cases will be inductive.

Indeed. Hence the points I have been making.

Been doing a lot of both modelling and measurements on this recently.
All being well, the detailed results will be appearing in HFN soon in
a series of articles. But the basic situation is that with LS cables
the primary effects are due to L' and R'. C' may have some effect on
amp stability.


Is this true at 20kHz? I can't remember my analysis results in detail,
but I seem to think it wasn't so.

Have a look at the gif whose URL I gave above for some examples. The "8
Ohm" cable comes close around 20kHz, but most other cables that most people
will be using come no-where near.

I was quite surprised by some of the other results I got from
measurements. Food for thought. So I have been re-thinking some of my
ideas about LS cables. But my base position is still that - when using
a decent amplifier - that low R' and L' are sensible, and that C'
doesn't matter much. And that working in terms of using L' and C' as
a 'pair' probably doesn't tell you much about the audio behaviour.


Not so sure I'm with you there.

Perhaps when you see the fuller results in the articles you will be
pursuaded. :-)

Not sure if the first article will be in the issue of HFN due in a week or
so. But with luck it will be published them. There are two other articles
that follow on from that at present, and in the pipeline. I found doing
them threw up some results I hadn't always expected, but I am not sure it
ended up changing my mind.

Slainte,

Jim


OK - this is all on hold for the moment. Interested though why you are
including EM field velocity - is there some point to be made?

d

TCI Cobra interconnects against Chord Chameleon
 

On 2008-07-15, Jim Lesurf wrote:
In article , Don
Pearce
wrote:
Speakers are particularly interesting in that there are many cables
available, some of which (from Goertz) have inductance and capacitane
which together come down to around 8 ohms as a distributed impedance.
These cables, despite having enormous capacitance, are essentially
ruler-flat in frequency.

The last sentence is meaningless as it has no reference to the conditions
of use. A 'cable' doesn't have a "ruler flat" response. What you get
depends on the system. And as soon as the load doesn't match the cable
perfectly, the source matters as well. Problem here is that domestic LS
uses a voltage assertion approach with loads that have values that vary all
over the shop.

However my simulations suggest Don is quite right in the sense that a LS
cable with a lower nominal impedance interacts less with most loudspeaker
loads and in general results in a flatter frequency response for the
voltage transferred from the amplifier to the LS.

This does assume the amplifier is stable driving the load.

And just how the LS and the room then deal with that input voltage is,
of course, another matter which may make that flatter frequency response
less relevant.

--
John Phillips

TCI Cobra interconnects against Chord Chameleon
 

Jim Lesurf wrote:


... If I have two pieces of
cable, one of 10 microhenries and 100 pF per metre, and another of 20
microhenries and 200 pF per metre, they will perform identically.

I think not in a mismatched system.

Indeed.


No, the degree of match doesn't make any difference. In RF, you can take
a piece of PE coax, a microstrip, an edgeline, some Andrew's Heliax -
they all have totally different inductive and capacitive values. But if
they were put in a black box with only connectors visible, there is not
a single measurement you could make that would tell you which is which.
All you need to specify them completely is the characteristic impedance
and electrical length (I'm ignoring resistive differences of course).

d

TCI Cobra interconnects against Chord Chameleon
 

Jim Lesurf wrote:
In article , John Phillips
wrote:
On 2008-07-15, Don Pearce wrote:
John Phillips wrote:
Do you have particular example where you think the lumped and
transmission line models will differ in a cable-speaker application
to a degree that is practically relevant?
Not handy. But it is clear that the lumped version of the model is a
lowpass filter, which a cable simply isn't.

The lumped version presumes a 'short' cable in wavelength terms. For audio
frequencies and runs of a few metres this is generally quite a decent
assumption.


Ah but how short - that is the question. Here is another graph, this
time of the loss of 10 feet of Monster cable into an 8 ohm resistive
load. There are three traces, one in which the lumped elements are all
in one piece, another where they have been split into 20 equal sections
and yet another with 30 sections. Now, which one is "right"? The
legends tell you which is which.

http://81.174.169.10/odds/sections.gif

d

TCI Cobra interconnects against Chord Chameleon
 

On 2008-07-15, Don Pearce wrote:
Jim Lesurf wrote:


... If I have two pieces of
cable, one of 10 microhenries and 100 pF per metre, and another of 20
microhenries and 200 pF per metre, they will perform identically.

I think not in a mismatched system.

Indeed.


No, the degree of match doesn't make any difference. In RF, you can take
a piece of PE coax, a microstrip, an edgeline, some Andrew's Heliax -
they all have totally different inductive and capacitive values. But if
they were put in a black box with only connectors visible, there is not
a single measurement you could make that would tell you which is which.
All you need to specify them completely is the characteristic impedance
and electrical length (I'm ignoring resistive differences of course).

Inside the black box:

- 1 metre of 100pf/m (10uH/m - 316-ohm) cable open circuit at the far end;

- 1 metre of 200pf/m (20uH/m - 316-ohm) cable open circuit at the far end.

Are you saying a capacitance meter operating at (for example) 1 kHz will
measure the same capacitance value? What value will it measure?

--
John Phillips