BACKGROUND: There is always discussion on how to ground a shield. The answer lies in what the
worst-case noise situation is. It isn’t always the same answer. Do you leave one end open or
ends? If you ground only one ends, which end?
We can look at how the interference behaves to answer these questions and they follow understood electrical properties within swept frequency regions.
BODY: The first step is to understand what a shield is doing, and how. In the simplest terms, a shield creates two separate “electrical” environments, one on each side of the shield. One side is measured as a RATIO of the field’s intensity relative to the other side, in dB.
The shields we are working with are ONLY effective with an EM wave that is predominantly “E” field in nature versus “B” field, or magnetic. Magnetic fields are not shielded or blocked with conductive shields, but need a shield material that blocks magnetic flux lines or a low permeability material…think “material a magnet will stick to.” Those kinds of material allow magnetic flux lines an easier path than through air. We can capture and re-route the flux lines in low permeability materials. For this discussion, we will look at mostly electric field shields, stuff that conducts electricity.
What is an ideal electrical shield? It is a shield that 100% blocks electrical energy at the surface of the shield,
that has infinitely low resistance. Shields don’t have infinitely low resistance and they block electrical
energy at differing impedance based on the skin depth of the shield at a specific frequency.
We can measure the effectiveness of a shield across frequency with a transfer impedance plot. This is measured in milli-ohm/meter. It describes the resistance we can expect a shield to have, and thus the ratio of the fields energy in the shield based on the CURRENT the resistance causes to flow at that frequency…and it is not linear.
The graph below shows the frequency-dependent nature of a shield. A perfect shield would have NO RESISTANCE and both ends would be identical and thus seem like a SINGLE point of reference to a flow of current. Since we have zero resistance across the shield, we can’t have current flow caused by the shields. We CAN have current flow between the two points connected at the ends of the shield. In a “perfect” world the GROUND at both ends is the same potential and thus, an ideal shield has ZERO current flow, and is a measure of the POTENTIAL on one side of the shield relative to the other, in dB.
Since we don’t have a perfect world, Transfer Impedance describes what to expect at frequencies based on the shield’s impedance, and how that shield resistance creates a CURRENT flow and thus a voltage (shield resistance times the shield’s impedance = a voltage). When we have different resistances at each end of a shield we have current flow.
Bonded Foil +60% braid
Quad Shield 60% +40% Braid
Bonded Foil +95% Braid
The chart below is what JUST the shield impedance looks like for a set length of cable at lower frequencies. We see the same non-linear behavior of shield and frequencies.
A SEED (Shield Effectiveness Evaluation Device per IEC 61196-1) test shows the dB relationship to a Lower Shield Impedance. Series number 5 with a 95% coverage 45-degree braid and Duofoil tape is clearly superior.
Okay, we can see a shield is not perfect, and not linear. So how does this say what to do with each end of a shield? We have to weigh the CHOICE of HOW the shield WORKS to decide our fate.
– If you have ideal grounds and meet IEEE bus bar grounding (see the 568C.2 or later grounding specifications) limits it means BOTH ends can be grounded and the shield current will inductively couple less interference than the shield ATTENUATES through its material composition.
– If we have severe ground differential, we can induce a strong current in the shield that CAN, if the shield’s resistance is higher, induce noise into the core that is WORSE than if we disconnect one end. We convert our shield to an antenna, not a shield!
– An antenna does NOT create two separate environments between them with the ratio of one
measured to the other. One end of the antenna is infinite impedance (the open end) with the other end at ground.
The antenna “wire” is as close to zero impedance as possible in order to NOT
attenuate the antenna’s signal going to ground. The signal won’t go to the open end, but seeks the lowest potential in the circuit.
We trade the noise caused by a poor ground resistance potential between shield ends for the induced noise in an antenna’s wire parallel to the signal wires that induces a voltage based on the antenna resistance. In an antenna type ground, it is best to ground the SEND end, as the SIGNAL on the internal wires is as LARGE as possible relative to the antenna signal, improving the signal to noise between the two.
An often-ignored aspect of shields is HOW to ground one at lower frequency versus RF. There is a big difference and again, it is based on the shield characteristics at each frequency.
The charts below are derived from TWO slightly differing MODELS of RF shield inductive reactance resistance. I have this paper for those interested. But, the data is the same message in that as frequencies increase, the shield reactance goes UP. This necessitates a FULL 360 ground at the shield termination point in RF circuits. This is why good RF connectors are fully capturing the shield all the way around the cable. On your RF digital cables, use 360 degree grounds for the best true shielding.
SUMMARY – Most
systems will have proper GROUND differentials between them and thus have near
ZERO shield current. The shield relative to the signal wires will ATTENUATE outside interference. When you have poor grounds,
it may be beneficial to unhook the “receiver” end of the shield and hope
that the induced antenna current voltage is less severe than the induced voltage caused by differential
shield ground potentials. This should be a SECOND choice, not the first. A properly working shield, by
design, has a KNOWN shield dB rating that can be trusted in a proper electrical circuit.
An antenna ground’s induced voltage onto the cable is not fully described and is dependent on the GROUND proximity point and shield’s distance from the signal wires. In severe situations, it may be the best choice to mitigate noise to the lowest possible reference value as it is pretty hard to REMOVE a shield already on a cable. Some, such as coaxial cables, can’t be mitigated and need to be properly designed to EXCEED the ground differential by several orders of magnitude so as not to aggravate any ground differential.
ICONOCLAST will use double ground interconnect shields and proper DCR RCA grounds.
Power cables should also use grounds
at BOTH ends if you have a proper GROUND plane resistance such that ZERO current flows and thus you have ZERO induced voltage from differential current. An
antenna type ground CREATES a differential in each end of the “antenna” by design (one end is ideally infinity the other is ideally zero) and is thus a second choice if you have known ground issues.
One last note, those heavy 10 AWG power or more cables, may provide benefit as they induce less ground differential resistance than smaller power cords as the ground wires are larger. The circuit may not need the power delivery of a larger cable, but the lower ground resistance values may be of benefit on longer runs in marginal power grid situations. This will improve a shield’s current to nearer zero across frequencies. The dB isolation numbers values are for a proper shielded system with IEEE and TIA compliant shield differentials.
The Transfer Impedance numbers are between two-reference point probes on a shield, and DO NOT need the ground potential differences for characterizations.
you have spent plenty on cables you may well wonder WHY these cables are
physically as they are. If care is taken to adhere to fundamentals, there are
very good reasons for a physical design in audio cable, of both high
(interconnect) and low (speaker) input impedance types. If we look at all the
fundamental electricals through the audio band, is it any wonder every cable
doesn’t sound different? Let’s see why that might be, and no magic need apply
throughout this analysis.
is happening in audio frequency ranges?
What exactly are we “moving”
with zero distortion?
Current and Phase
Electromagnetic wave propagation differences with respect to frequency.
Impedance and matching to a load at audio.
Capacitance and Inductance with respect to frequency.
6.0 Cable Capacitive and Inductive reactance properties rise and decay time distortions.
Current normalization / skin effect.
AC resistance changes and frequency.
Cable symmetry issues.
Attenuation at audio.
12.0 Passive low pass filter effects.
we look at pure tones; sinewaves, square waves, frequency and TIME are
interchangeable. Math says that this is so, and there isn’t anything new that
explains that away. When we add TIME based distortion to the sound delivery
system our ears are quick to “hear” the deterioration in fidelity based on
frequency arrival time and phase coherence more than amplitude limitations
(attenuation). How much is a cable responsible
for this? The superposition of the 12
listed distortions (and there are more) are much more significant than any one
taken on its own. There is truth to the concept that slew rates, or how fast a
system responds (wider bandwidths), affect performance. A square wave is but a
multiplicity of sine waves. Mathematically every frequency’s characteristics,
at every point in a cable can be predicted.
Cable is far from perfect at moving electromagnetic wave through the
audio band, however well we can calculate the accumulating TIME based
distortion as the electromagnetic wave moves down the cable. Better designs
minimize those distortions and place more or less emphasis on each one
depending on the designer engineer’s concept of audible influences. The fact
remains, cable design is still driven by the DESIGN needed to reach the R, L
and C values with minimal influences on tertiary elements. Can you hear a more
fully optimized design? This is why we present these designs for audition.
1.0 ELECTROMAGNETIC WAVE
The issue –
What do we actually LISTEN to on a cable? What is the “root” reason to be for a
exist to move the “signal” from one place to another, but few really consider
WHAT that signal is. The signal we “use” is the electromagnetic wave moving
down the cable at the group velocity of propagation of the dielectric. OK, what
did I just say? Imagine our wire surrounded by a donut with a hole in the
middle! The electromagnetic wave is this donut. There is an ELECTRIC (E-field) around
our wire too, but this field is attached to the donut radially, and ninety
degrees orthogonally to the donut’s circumference. To make the E-field, take a
bunch of tooth picks and stick them all around the outside of the donut, that’s
Now we have two imaginary waves, one low frequency and one high, sitting there. To MOVE that field, electron flow starts it happening. To keep it simple let’s distort our wire to be a TUBE full of marbles (electrons) that has an inside diameter the same as the marble’s diameter. To make the magnetic field move, and drag along the E-field with it, we apply an electromotive force (electrons / marbles) to the tube. When a marble is inserted into the end of the tube, the marble at the opposite end pops out as fast as the marble can be inserted into the send end of the tube. This “speed” is determined by the velocity of propagation of the dielectric, or the tube in our case. Something funny happens with the magnetic field though; it follows the PROGRESSION of the electron (marble) flow. When the marble is half way into the send end of the tube, our donut with all our toothpicks (the B and E fields) is halfway down the cable already! When the marble is inserted all the way in at the send end, the B and E fields are at the END of the cable. So the “signal” we use travels at the VP (velocity of propagation) of the cable, and NOT the speed of the electrons at all. Those move very slowly compared to the electromagnetic B and E fields. Now we have the donut at the end of the cable. But, we won’t ever see a baker’s dozen, or zillions more moving electrons appear at the same time at the opposite end of the cable if we carry more than one frequency concurrently since every frequency has a different VP through the audio band. All individual frequencies will have significant arrival time “distortion” between frequencies. In other words, every marble that represents a frequency in my example is inserted at a different speed (Velocity of Propagation) depending on the frequency the marble represents. Ideal cable should move a signal (now we know it is the B and E fields) down a wire at the same speed and shape at all frequencies. It doesn’t.
2.0 Voltage and
The issue – Current and voltage are locked into a phase shifted
The reactive properties of inductance and capacitance are
responsible for a ninety degree time based shift in all electronics, not just
cable. There is a common ditty about the current to voltage phase relationship
that goes like this; “ELI the ICE man”. It is a memory tool to remember that
voltage (E) leads current (I) in an inductor (L) and that current (I) leads
voltage (V) in a capacitor (C).
Why is this? A capacitor has to charge with applied current
to reach a steady state voltage, so as the voltage potential increases the
current drops. The current has to be there BEFORE the voltage potential hence
current leads voltage in a capacitor.
An inductor resists current change when voltage is applied.
Current reaches a steady state over TIME with applied voltage, so as the
current potential increases the voltage drops. The voltage has to be there
BEFORE the current potential hence voltage leads current in an inductor.
These two locked-in relationships lead to all sorts of other
TIME based issues in cable and circuits. They are the variables that constitute
PHASE in an impedance trace, for instance, and reactive TIME CONSTANTS that
we’ll cover later in the paper.
3.0 VELOCITY OF
The issue – VP varies the arrival time of signals moving down a cable.
Signals should ideally leave and arrive at the same time and shape as they are
sent at all frequencies.
Audio is in an electromagnetic transition band. This is the
elephant in the room. It prevents cable from EVER being perfectly accurate when
moving low frequency electromagnetic waves. The propagation constant, the speed
at which the electromagnetic wave / signal moves down the wire’s outer
circumference, and not IN the wire, is determined by the dielectric material
that the electromagnetic wave is predominantly traveling through. We can
measure this effect directly and indirectly.
At RF, where life is way more consistent for cables, we can
calculate the velocity from the DELAY equation. For Ethernet cables the following
equation is used;
Delay EQUATION at RF
The delay equation uses FREQUENCY. This is a TIME based value so it tells us that we have arrival time issues as the frequency changes, and less so at RF, and WAY more so at audio frequencies. The table illustrates the slow erosion of speed as we reduce the RF frequency. A little change is evident but audio frequencies see much more change.
Delay values measured at RF (MHz)
Actual data shows what audio cables do; the impedance RISES as we go LOWER in
frequency, by a lot. This is because the DELAY / VP factor drops, and adds
TIMING issues to signal delivery.
Above are actual traces of how ICONOCLAST performs across
the audio frequency band vs. typical zip cord speaker wire (1313A) and out to RF,
to prove a point. The impedance increases considerably below the RF frequency
reference values. Those 87% and 90% VP factors we love to “hear”, high VP, are
clearly not valid in the audio band.
How significant is the VP change? In the example above we
drop from ~110,000,000 m/Sec @ 20 KHz to ~5,000,000 m/sec @ 20 Hz or a factor
of 22 times slower through the audio band.
To make matters worse, it is a LOG function so it is not
linear. This is what physics has thrown into the design process. Can we hear this change? Attenuation at audio
is a passive linear variable and considered to be insignificant (keep your
cables short) but every variable keeps adding up to the overall actual
Notice that the cable’s impedance, made for audio not RF, flattens
out @ ~ 50 ohms above 100,000 Hz (see the table below for the actual values).
Just because something has an “impedance” (real and reactive L and C component)
does not mean it is a transmission line.
Look at the low-frequency range. Isn’t cable supposed to be
the same at all frequencies or the same TIME base? The velocity constant at a
frequency is TIME, so the fact that we see a difference indicates a non-
linearity across the usable audio band. The
problem is that thing called propagation velocity (VP) or the speed that
information travels at differing frequencies in the cable.
The equation at audio compared to RF is more complex
(wouldn’t you know it!);
Z = sqrt((R+j*2*pi*f*L)/(G+j*2*pi*f*C))
resistance (R )
conductance (G )
Using the general simplified RF equation, where all the
extra stuff in the complicated impedance equation at audio goes to a one or a
zero and drops out, we are left with; 101670 / (Capacitance x Velocity) =
Impedance. At RF for ICONOCLAST speaker cable;
101670/(VP*45pF/ft) = 50 ohms @ RF
Solving for velocity of propagation we see it is no higher
than 45% at RF. This isn’t RF cable, and the design changes necessary for audio
are what ICONOCLAST is after. We need to
ideally FLATTEN the VP curve for audio cables to better time align the signal
in the frequency range where we use it.
The calculated graphs using a 75-ohm coaxial cable below
show that VP change as we go lower and lower in frequency. Look at the
IMPEDANCE at audio frequencies shoot way up, and the VP drop like a rock in a
pond. Notice, too, that VP begins to flatten out at 100,000 Hz, just like the
charts above on ICONOCLAST. This is real stuff, and it won’t go away…you have
to MANAGE it to a balance in each cable.
What does our measured data
show that corresponds to the theoretical chart above? Below we see several
BELDEN products measured VP drop considerably from RF to, and through, the
audio band. And, the measured values are near the exact same values I will
calculate from measurement on ICONOCLAST; ~ 5% VP to 50% VP between 20Hz to 20
WHAT CABLE Velocity REALLY DOES THROUGH
THE SWEPT FREQUENCY
The impedance goes up as we go lower in frequency because
the velocity keeps going down, and the alternative variable, capacitance, just
sits there (we’ll get to that soon). We have a differential in signal velocity
across the audio band. Also notice that typical 1313A ZIP cord behaves much
worse than ICONOCLAST™, rising to double the ICONOCLAST reference impedance
value. Be warned, audio cable does NOT respond to impedance matching like
Speaker cables are theoretically designed to be much lower
impedance, and terminate into reactive 2-16 ohm loads, and some point way north
of 16 ohms. Interconnect cable is
terminated into “high” impedance resistive loads of 47K to 120K or higher, and
should be much higher theoretical impedance than speaker cable, and the graphs
above show exactly that.
It is good to see impedance matching to a load, but other
variables are in play, and impedance matching isn’t meaningful or practical at
these frequencies and impedances. Good designs usually address ALL parameters,
Interconnect and speaker cables, with VERY low audio range
VP values show a much faster VP in the RF band. The values of 87% VP @ RF are
NOT really correct for WHERE the cable is used, but “sounds” exciting.
What do we see at RF on an ICONOCLAST interconnect cable? We can calculate what we measured in the graphs above. We can use a grossly simplified equation to predict the VP based on capacitance measurements;
101670/(11.0 pF/ft * VP) = 105 ohms @ RF
Solving for VP we get a value of 88%, using the measured values of 1 KHz referenced capacitance. This VP factor will DROP considerably in the audio range to much LESS than that. Imaginary values (L and C) stay the same from 1KHz to RF frequencies so VP is changing;
If we take the VP reduction factor of a coaxial cable into
the audio band @ 22 X lower, we see; 87% / 22 = 3.9% @ 100 Hz. Close to the
same answer in our rough calculation.
The data shows a 13X to 20X or so DECREASE in cable speed as
we drop in frequency. Signal arrival times are NOT staying in perfect symmetry
relative to the input start point. The AMPLITUDE may be near the same, but the
TIMING is certainly not. Arguments persist as to how long the cable needs to be
to her the arrival time coherence.
4.0 IMPEDANCE AT
The issue – All cables should terminate into their characteristic
impedance (not really true at audio). At audio, the cable isn’t a fixed
impedance, or even really an “impedance”. Interconnects see a resistive
“infinite” load, but not speaker cables, which see a highly reactive low
Impedance is a REACTIVE vector value. This is a dead
giveaway that we’ll have to deal with Dv/Dt stuff. All cables are a wire that
is in series with an inductor and a capacitor to ground. All three R, L and C,
keep getting bigger the longer the cable on a bulk value basis. The impedance
is a VECTOR sum of the REAL part and the IMAGINARY part. The PHASE is created
by the imaginary part of the impedance vector value. The impedance values
aren’t the same for all frequencies (see the 1 KHz and 1000 KHz chart below)
since VP keeps changing, and this is a component of the impedance value. Since
the impedance is a vector sum magnitude ratio, it stays constant for each frequency
point no matter how long the cable is. R, L and C increase proportionally.
Reactive Change with Frequency
Most of us kind of know that we are supposed to match the
impedance to the load for the best transfer of energy. We are actually only terminating
the resistive component we call “impedance” to the load; a resistor in the case
of interconnects, or a speaker load for low-impedance speaker cables. There is a reactive component that is also at
issue for good signal transfer. That reactive (usually capacitive) part of the
Impedance vector magnitude diminishes the transfer of energy in time. Audio is
not RF, so this matched resistor to resistor ideal isn’t exactly correct
anymore, even for high impedance interconnects. The physics of the velocity of
propagation make impedance matching impossible at audio as does the wavelength,
which is far, far too long to react like a true “impedance” vector.
For transmission line effects to be a factor, the cable
length also has to be at least 10X or more the quarter wave length of the
frequency of interest. This relates to the fact that a voltage change has to
happen BEFORE it gets to the end of the cable and audio speaker cables transit
times are too fast, even @ 50% VP, for this to happen.
A cable can have impedance (real and imaginary values), but
it is largely irrelevant to true load matching. There can be a signal
reflection based on the CUT length of the cable relative to the speaker. This
simple reflection can be absorbed with a ZOBEL network across the speaker
terminals if it induces amplifier oscillations. But, low cap cables are benign
to amplifiers, even with this simple length defined reflection. The cable will sound the same with or without
the network as the parallel circuit is not in the signal path. The tertiary
effect of better amplifier stability is what improves the sound with too high
At RF, a signal is “used” efficiently only when two like
resistive loads see each other. RF cables are designed so that the cable impedance
matches the restive termination load. Audio cables don’t work like this at such
low frequencies since we can never transmission-line “impedance” match to a
load with short passive cables. But, the “work” done across the load STILL has
to be resistive. The imaginary components of a vector (Impedance is a vector
sum of the real and imaginary components) store and release energy since they
are composed of reactive variables; Capacitance and Inductance, both variables,
are store and release variables of voltage and current respectively. Short
cables still have reactance.
We can see what happens at RF. The graph below shows actual
cable data of what is called Return Loss. The return loss, RL, represents the
“reflected” signal that does not transfer to the load for an RF Ethernet cable.
RL= the imaginary part that can’t do work till it is “real” or resistive.
Notice that we see several RL values “dead nuts” on 100-ohms from a low of -55
dB to a high of ~ -22dB. WHY are the RL variables not all the same? The impedance
shows 100-ohms for all those RL values. The impedance at every frequency has a
different reactance due to a lot of things too complicated to explain today.
Simply put, at the frequencies with the lowest imaginary component, more energy
is transferred to the load. In our example, if the impedance is above or below
100 ohm, and more or less reactive, the RL is decidedly worse. This is the
cause of the FAN shaped graph that we see below.
Audio cables aren’t used at RF, though, and suffer from
simple reflections more than load matching ones. This isn’t bad thing, as the
critical attributes at RF aren’t restricting what we need to do in the audio
band for better signal quality. We don’t need to worry about minute wire
diameter fluctuations that cause the above graphed RL reflections. Audio
wavelengths are too long to see the diameter variation issues so designers can
work with geometries that may not be ideal at RF, but are far more useful for
coherence adjustments in the audio band. Those adjustments still have to be
real, of course, and measured or calculated with accepted standards.
Audio speaker cable with AC signals is terminated into a
load that is resistive and reactive. Alternating current reacts to the imaginary
circuit cable values and regulates how fast, and when, we can get work out of
the cable. Some early cables were so reactive that amplifiers would shut off
using them. Even though our cable is not a true impedance we do have reactive
Interconnects see an “infinite” ideal resistive load;
47K-ohm on up, and speaker cables see a very low, and varying, reactive input
impedance (the impedance of all loudspeakers changes with frequency).
Speaker cables are
CURRENT signal devices that are designed to transfer power to an
electromechanical motor. And, a motor that constantly “changes its spots” at
every frequency as does the cable. The “argument” between the speaker EMF and
cable is complex.
are VOLTAGE signal devices terminating into a HIGH impedance resistor. We want
to transfer the signal shape and amplitude to a load. To avoid distortion(s) we
don’t want the cable or the load to mess with the transmit circuit, but they
Audio cables are way too short to be transmission lines,
needing at least 10X the wavelength inside the dielectric to be a true
transmission line. Even 20 KHz is way too long a wave length to match that
definition. We DO have simple reflections off the LOAD (speaker itself) that
cable can’t manage as the load varies with frequency. This is very different
than RF where I can make a cable nearly look like the load, minimizing
reflections. I said “nearly” as all cables exhibit reactance, a TIME based
storage of energy. Audio cables have significant measured time based
propagation error due to VP and now we add-in a rise time error from reactance.
The reactance of cable can be used to calculate “time constants”. At audio
every frequency is associated with a different constant value. We’ll look at
time constants later.
Zobel networks have been used to good effect to dampen the
cable to speaker load variation, but they are estimations of where the two are most aggressively
reflective. A Zobel network is a passive means to connect two differing but
fixed characteristic impedance lines with a resistive value. Neither the cable
nor the speaker are linear loads making it an approximation as to where to tune
the Zobel network.
Compared to our “typical” Belden cable (blue trace),
ICONOCLAST is flatter (orange trace) in velocity change as we go lower in the
theoretical impedance. This is more the result of a higher, but still low,
capacitance between the two designs. Lower inductance was preferred over
The table data below is REAL and represent what even really
good cables do through the audio band. The physics of the propagation delay
match the measurements.
WHAT CABLE IMPEDANCE REALLY DOES
The interconnect tables follow and yes, they too show time
5.0 CAPACITANCE AND
The issue – What do the reactive
variables do with respect to frequency?
Capacitance and inductance are essentially FLAT with
frequency. Yep, capacitance and inductance are, interestingly, the same from
near DC to the “sky is near the limit” frequencies. Capacitance is set by the dielectric,
assuming it is a linear dielectric material, and some aren’t (PVC).
Measurements show that stable dielectrics offer frequency linear capacitance.
Inductance is set by the distance between the wires and the loop area; it isn’t
changed by the dielectric at all. These two values are always steady Eddies,
but their time based effects on current and voltage change with frequency.
Here is ICONOCLAST speaker cable that shows L and C
and it is FLAT with frequency using
Teflon® as the dielectric.
The choice of what plastic to use sets the dielectric
constant. You want stability with respect to frequency. Teflon® has the lowest
dielectric of any SOLID plastic and thus the lowest capacitance with the
thinnest walls of any material and, it is durable. It costs a LOT to buy and
process, too. Cost isn’t why we use it,
Plastics aren’t magic for capacitance, that is just the way
it is. You want to pick the lowest dielectric constant value not just for low
capacitance, but to help offset the change in the dielectric constant with
respect to frequency. PVC dielectrics are far worse in linearity with respect
to frequency, and the slope is not the same everywhere. The chart and graph below assumes a set wall
thickness and changes to the dielectric material alone. We can alter the WALL
thickness based on the dielectric constant to get a given capacitance between
two wires. Double the dielectric constant means doubling the wall for the same
capacitance. Use the cheap stuff then? Sure, but more wall thickness increases
loop area (space between the wires) which increases inductance! Oops, we’re not
going to get zero cable reactance that way! A wire in a vacuum inside a braid
ground would be the smallest size with lowest capacitance you can realistically
see. This design would also have the lowest inductance since the loop area
would be at a minimum with the vacuum acting as a low dielectric material.
We can calculate the effects of the dielectric and
capacitance using a shorthand RF formula 101670 / C *V. We fixed the reactive
impedance to a set value, so for a fixed wall of insulation, the capacitance
rises as the dielectric constant is higher.
Since we know that the capacitance value is flat with frequency, this
applies to the audio band as well. Better dielectrics for a given wall mean
lower capacitance. This has nothing to do with Inductance, which is related to
the magnetic field lines. Inductance is related to the distance between
conductive surfaces, the less the better and field cancellation…if any.
Impedance = 100 ohms
Velocity = 1 / SQRT (E)
Capacitance = 101670 / (impedance * VP)
6.0 INDUCTIVE AND
CAPACITIVE REACTANCE VARIABLES, XL AND XC.
The issue – all cables store and release energy (current or voltage)
reactively to the frequency being electromagnetically moved through the wire,
adding time based distortion.
Look at the Impedance / Phase trace shown above on Part 3,
Velocity of Propagation Issues. Notice that the PHASE on BOTH cables changes.
PHASE include reactive components that TIME shift the signal’s ability of a
signal to become resistive (If the phase trace hits “0” the circuit is
resistive and has no reactive component).
A capacitor looks like an OPEN to DC or to very low
frequency AC voltage changes. The cable is very “reactive” to voltage changes
at lower frequencies. As you go up in frequency the cable’s bulk capacitance
looks more and more like a SHORT circuit. The cable becomes more “resistive”
looking with less reactance to voltage change. The trace explains why we can
use 75-ohms and 100-ohm loads for RF cables, they look “mostly” resistive at
Using a high impedance probe to measure the cable’s
reactance produces the traces that you see. There is little current flow into
the cable and, this is essentially how interconnects are used. The terminating
load is VERY high impedance limiting current flow. When you put a voltage
across a capacitor (our cable) it sends a momentary inrush of current to try to
fill the capacitor. Output devices loading the circuit are ideally super low
impedance to allow for this “inrush current”. Cables with lower capacitance
mitigate the inrush current issue. Current LEADS voltage in a capacitor so
there is a TIME shift caused by the cable.
Speaker cables are differing in that we don’t measure them
like they are used; terminated into what is essentially a short circuit, the
speaker. The large current flow in speaker cables responds to reductions in
INDUCTANCE. Inductors resist current flow changes and that’s what speaker
cables are trying to “move”. Voltage leads current in an inductive circuit and
again, we see a TIME shift caused by cable but the opposite reactive variable,
inductance verses capacitance, than the interconnect cable.
Also consider that in speaker cables, the most reactive
region is exactly where speaker’s impedance is also the most reactive, too. We
want is a cable that is purely resistive but that’s impossible since a cable is
a vector of capacitance and inductance.
Can we look at this another way? Yes, we can. If we examine
the capacitive reactance equations below, and stick in the values at DC (F=0)
and infinity frequency (remove F) and see what the results are we get the same
answer; reactance is high at low frequencies and lower as you go up in
Xc= ½ * pi * F * C
XL = 2 * pi * F * L
The inductive reactance is the opposite, it looks much
smaller at DC (F=0) than at higher frequencies (F= infinity). An inductor is a
SHORT at low frequencies and an OPEN at higher frequencies. Fortunately speaker
cables are relatively lower frequency making things less severe than at RF.
Cables, and all circuits, have capacitive and inductive
reactance. Capacitive reactance resists voltage change and inductive reactance
resists current change. They are both frequency dependent.
The TIME it takes to CHANGE the signal applied against a
reactive load is measured in TIME CONSTANTS. It takes about 5 to 6 time
constants to reach steady state amplitude. Our signal is also distorted the
longer it takes to reach steady state amplitude so it may get nearly as big
(we’ll pretend attenuation isn’t an issue), but it isn’t the same SHAPE. Don’t
forget, every frequency is associated with a different time constant, and the
decay or removal of the signal is the inverse. It takes TIME for the signal to
bleed away to zero and this alters the decay signal.
As frequency changes, so do the reactive variables the determine a cable’s reactive performance.
At the very high end of the
graph below, we see simply SQRT (L/C). At the low end the simple reactance
(denominator) enter in.
7.0 SKIN EFFECT
The issue – Current magnitude
normalization at audio frequencies. Is this real?
There are several ways to calculate skin depth, and they all
will yield the same answer. Impedance / RL can be derived from several
inter-related factors and so can skin depth. It is real, and it can be managed
to control phase distortion.
We all know about skin effect, but WHAT exactly is it doing
at audio frequencies and is it real? Yes, skin effect is real at audio and all
industry accepted calculations show that it is. The definition of skin depth is
the point inside a wire where the current decreases to 37% the surface current
magnitude. Skin depth is always the same depth of penetration no matter the
wire size. Skin depth will vary based on the material’s electromagnetic
properties and the frequency of the signal. For audio we calculate @ 20 KHz.
Looking at COPPER, we
would calculate 461um (0.0181” depth).
If the skin depth at a given frequency is 10-mil on a
100-mil wire the 37% current point is well near the wire’s surface, it’s just
10-mil away in 50-mil radius. If we
halve the wire size, the current magnitude is larger through more and more of
the wire. Each time we decrease the wire size, the larger the current magnitude
becomes across the wire relative to surface current. In our 18-mil skin depth
wire example above, the current in the “center” of a 36-mil wire will see 37%
the magnitude of the surface current. Making wire smaller will INCREASE the
current magnitude in the wire’s center to be closer and closer to the surface
current in the wire at higher frequencies.
AC resistance involves FREQUENCY which is a TIME based
RAC= (RDC) (k) SQRT (Freq)
K is a wire gauge factor that involves skin depth.
Freq is in MHz.
The internal wire impedance (AC resistance) is driven by the
INTERNAL magnetic field’s relationship to inductance. Inductors RESIST
instantaneous current flow and have higher “resistance” as AC frequency goes
up. Current flows in the least resistive part of the wire as frequency goes up,
so it reaches the surface where the self-wire inductance is nearest to zero.
Once we flatten the velocity change as best we can with a
good dielectric design, we need to ALSO time align the effects of the
dielectric at ALL frequencies using SMALL wires. Small wire improves arrival
times as it forces the effects of the composite dielectric speed to be more
uniform, as best we can, at all frequencies. This counters the skin effect
problem that moves the current density magnitude to the surface of the wire as
frequency goes up. Smaller wire increases the current magnitude (arrow length)
in the wire center region to make it more efficient at time alignment.
One BIG wire
More SMALL wires
Even if we have the SAME current magnitude throughout the
wire at all frequencies (impossible unless our wire is one atom in size) the
velocity of propagation of the electromagnetic wave energy is STILL different
at every frequency! But the ears say if we MANAGE the problems, our cables can
sound much better. I took the time to measure all of this and flattened the
impedance trace as much as I could. The VP changes less with frequency the
flatter the impedance curve. Capacitance stays the same at all frequencies, so
this VP is therefore changing less the more consistent. Smaller wires are more consistent
dielectrically at all frequencies.
Bigger wires will cause even more signal speed change
relative to frequency because each electron’s is far smaller inside the wire.
Each magnetic field contribution changes velocity the closer or farther that
electron is away from the dielectric material. When a current is applied
(electrons start moving) an inner wire located high frequency current mode
travels slower than the same frequency signal on the outer wire surface and all
these current modes are superimposed one on top of the other. This is called
Not all signals at the same frequency arrive at the same
time, it depends on WHERE they traveled (MODE path) through the wire and what the velocity of propagation is from the
geometric perspective. The lower in frequency you go the less you can change
the group delay since the current density through the wire is more and more
The overall magnetic field is a summation and superposition
of ALL the moving electrons, the whole “group”. This is also why air is often
used in interconnecting cables to mitigate the dielectric’s impact on the
signal, and why you see more small wires in speaker cables. Electromagnetic
field uniformity in the dielectric is important. The overall audible
improvements are more debated. But, there is science involved in the
The issue – dielectrics can impact weak electromagnetic signals
disproportionately. Electromagnetic fields are squared law fields, and are most
influenced by dielectrics nearest the wire. Weaker electromagnetic fields are
most susceptible to dielectric distortions and the group velocity is mostly set
by the strongest signals dielectric medium.
Using too many small wires splits up the current and starts
to allow the dielectric to influence the sound more and more, negating the
“advantage” of dielectric uniformity. The electromagnetic field is strongest
nearest the wire, decreasing with the square of the distance moving out away
from the wire. The electromagnetic signal moves from being “in” the dielectrics
to being around it. The signal propagation speed is an average of ALL the
dielectrics, with the material the stronger fields reside in have the most
influence on the average of the “group”.
Four-fifths or more of the current magnitude at audio is
below 3 KHz. Some call this the spectral power density, or roughly where the
most energy is being placed. The electromagnetic energy does not STOP in the
plastic or air. It emanates out in an inverse LOG power decay THROUGH all the
materials it encounters along the way. The predominant material VP effect
occurs CLOSEST to the wire. The smaller the signal (interconnect cables) the
bigger the effect of the immediate dielectric nearest the wire.
Weaker signals will be impacted by the dielectric’s effects
more than stronger ones, as they decay to far weaker signals moving away from
the wire. The speed is more and more determined by the dielectric near the wire
as we go up in frequency. Interconnects see little of the plastic out away from
the bare wires as the field decays so quickly, but, the smaller the
electromagnetic signal, the MORE it is influenced by the material nearest the
wire. That superposition of materials SLOWS the signal (air to plastic) or
speeds it up (plastic to air) relative to just the initial material’s
We can see this in actual practice as the “group” velocity
of all the materials on Ethernet cable shoes a value SLIGHTLY higher than the
dielectric (66%) itself, and measures 71%. The signal is in the “air”, a good
dielectric” and this influences the overall signal speed.
ICONOCLAST interconnect design switches this around, and
puts the AIR nearest the wire, where the signal strength is highest. This
negates the outer plastic dielectric’s contribution to the group velocity, so
we see a higher 87% value at RF. This translates to lower capacitance number
where we use the cable in audio applications.
The VP speed variation caused by the “composite” velocity is
complicated by the fact that the LOWER in frequency you measure, Mother
Nature’s devious plan slows everything and this time shifts the audio band.
We can’t change the fully diffusion coupled (same magnitude
current through the wire) low frequencies, so we try to time align the faster
upper frequencies. At RF the upper frequencies are “on” the wire surface so the
dielectrics affect them nearly 100%. At RF this is fine because it is near the
same VP at all RF frequencies. At audio, we want to move most of the high
frequencies AWAY from the dielectric so the speed is closest to the lower
frequencies. We already know that the VP
is faster the higher in frequency we go so this messes up the signal arrival
times. The only good way to slow the
upper frequency magnetic field is to make the wire smaller so less energy is
JUST at the wire surface nearest the dielectric. More current is “in” the wire
versus “on” the wire based on skin depth.
Can we overdo field current normalization? What if we could
make a wire one atom wide? Now, the impact of the DIELECTRIC is as big as it
will ever be and with a really, really small current in each wire. The total
current will be the sum of all the wires we want to use in parallel. The more wire
you use, the smaller the current in each wire. Current is the number of
electrons past a point with respect to time. Well, we have ONE tiny electron
moving in each “wire” and THAT is as small a current as you can have! Model a
weak signal, and the electromagnetic wave is so weak it never really leaves the
dielectric, whatever material the dielectric is. The dielectric better be
really decent as it is hugely involved in capacitive rise time (calculated
capacitive reactance rise times constants) signal arrival time (velocity of
At very high frequencies, and if the wire is infinitely big,
we see ONLY the dielectric as the current is at the surface (skin effect).
Likewise if the wire is infinitely small we AGAIN see JUST the dielectric (no
skin effect can happen). Between the extremes of wire size, somewhere, we can
alter the arrival time of the upper frequencies with wire diameter and
Interconnects are much easier, but not real easy, as they
terminate into a high resistance, nearly open looking circuit. The reflections
off a CONSISTENT resistive load of 47K-120Kohm aren’t as bad as the mismatch
speaker cables experience as BOTH the cable AND the load are in constant flux.
Worse, the speakers change by design! The seemingly high measured impedance
slope of RCA or XLR interconnects in the initial graphs aren’t as bad as they
seem. Not only are the “impedances” not real at audio but you have far bigger
issues with the non-linearity of cables loading the output devices in your
preamplifier. Trying to match ideal infinite input impedance on RCA or XLR
cable would mean tiny capacitance
values. We go as LOW in capacitance as we can to allow the output devices to
see an easy load. This is way we shoot for keeping capacitance reasonably low.
Massive signals in the speaker cable are less impacted by
the “composite” dielectric speed. The electromagnetic field will travel at an
“average” of all the stuff it is moving through, so the better the “average”
material is that the electromagnetic field is in, the FASTER the signal
travels, and the less TIME the signals have to become separated as they travel
down the cable. This is the time and distance story problem.
PC’s stopped using FLAT cables because the signal arrival
TIME differential got to be too high. They went to SERIAL digital designs, and
re-clock the data from memory. This at first seems counterintuitive, adding the
re-clocking circuit, but unless the TIME can be managed, you’re screwed. Faster
is better but I’d take SLOWER and the SAME in an instant! This is the “keep
cable shorter” thing, but to be LONGER we have to be FASTER, too, if time
errors are to be kept low. Mother Nature says we get a raw deal in the audio
band verses RF.
In speaker cable, the stronger low frequency electromagnetic
waves emanate into the air through the plastic dielectric more than the higher
frequency signals so they are theoretically aided by the air around the wire
(superimposed dielectric value) more than the weaker high frequencies that see
more of the slower plastic dielectric. But the VP erosion as we drop in
frequency eats-up that advantage in the low-end. It’s there, but small. The
problem is that the low frequencies still drop in speed way more than the air’s
addition to the overall speed. Seeing more air as we go lower in frequency
speeds the signal up relative to the faster high frequencies and offsets some
of the problem…but it never aligns it away to zero. The VP still marches slower
and slower as we go lower and lower in infrequency. Arrival times are more
important than SPEED down the wire.
The highest frequency carried in the speaker cable is most
fragile, but compared to interconnect cable, it is relatively robust. The high impedance interconnect cables are
yet another problem. ALL the signals are VERY, VERY low current electromagnetic
field energy states. Here, I need the BEST material possible to time align the
energy field “whipping” (slowly whipping) down the wire; air. The VP is the
inverse of the dielectric constant so we want a fast dielectric and the lowest
associated capacitance it can also provide. This is why I HAVE to use AIR core
designs to properly time align the energy AND use SMALL wires to better
distribute the dielectric’s effects at ALL frequencies nearest to the same
composite velocity. The third leg is to decrease output device capacitive
loading. Air helps mitigates velocity variation across the frequency band that
is the bane of audio signal transmission. It incidentally also pushes UP the
“impedance” to better match the load, the opposite of a speaker cable. I’d be
wary of that improvement as we need to be aware that audio isn’t a transmission
It seems counterintuitive to use air, as it speeds up the
higher frequencies relative to the lower frequencies (makes the difference
worse) but the capacitive reactance influences rise time error if you let it
get too high. The propagation time and the rise time need to be balanced,
somehow. There is no perfect solution.
We call, it “sound quality” when we use the cable, but it
really is the arrival time alignment of all the signals. The human brain hears
superimposed time alignment and amplitude preservation first, everything else a
distant second. The argument is: does this make a difference?
CABLE RCA to XLR MATCHING
The issue – Changes in electromagnetic properties between interconnect
cables types can alter the ideal “tone” that was intended.
An often ignored issue is, what do you do with a really good
sounding RCA cable? Why not make a really good sounding XLR that’s the same
reactive measured design? Most RCA to XLR cables never match. ICONOCLAST is no
accident. I purposefully designed the RCA and XLR to be the exact same reactive
match and thus the same “quality” of sound through the channel. The above
impedance chart that we saw earlier shows both the RCA and XLR. Look closely,
they are electromagnetic buddies.
Does that make a difference? If you have a very good RCA
design, it sure can’t hurt to start there on the XLR!
10.0 CABLE SYMMETRY
The issue – how to make complex cable’s cross section look like one
simple wire electrically, and every wire sound the same?
Matching multiple wires into a complex structure isn’t easy
to do well. The ideal cable is one wire that is exactly the same as the
opposite polarity wire. To meet other objectives, we usually have several
More small wires will make a nice big capacitor (wires with
a dielectric between them) and trash reactive signal conversion to resistance
products. Inductance will inversely follow capacitance, messing up the current
delivery, and to get BOTH intrinsically LOW, you can’t go “whole hog” on the
opposite variable. The two variables are tied together inversely. Rats! A
suitable compromise must be reached? Yes, audio is a compromise, as we are
seeing. BOTH L and C need to be low in value and good design manages this.
Trade-offs for better sound can, and should, be logically explainable.
The less understood variable is Inductance. This variable is
a big contributor to more wires. We all think, “capacitance” for audio. Realize
that if we had NO inductance, we could separate the wires as much as we wanted
and eventually have no capacitance (outer space actually has capacitance, so
that’s impossible, too). The reduction of the magnetic fields by proper cable
geometry reduces the inductance, allowing a larger wire center-to-center
distance for low capacitance. The LOWER
the electromagnetic field, the LARGER the loop area can be (lower capacitance)
for a given inductance and vice versa. Too many cables ignore getting the
electromagnetic field as low as possible. The higher the current (speaker
cables) the more field energy you need to eliminate. Wires with low electromagnetic fields and
small loop area have the lowest inductance.
ICONOCLAST’s speaker cable design balances out the wires’ proximity to
one another so as to not “rob Peter to pay Paul”. The unique weave pattern
increases the average wire C-C (Center to Center) distance creating a wire
pattern that CANCELS the electromagnetic field while increasing the average
spacing for low capacitance. Cancelling the field energy allows me to also
lower inductance which would be impossible to do with JUST wire spacing for
capacitance alone. No magic need apply.
EVERY wire in a cable has to be the same wire if you use
superposition of the electromagnetic fields traveling down each wire. This is
why symmetrical cable designs are used to efficiently remove reactive time
alignment issues. I measured the reactive time based issues on other designs
and they all came up short. Capacitance and Inductance have to be the same on
EVERY wire to as tight a manufacturing standard as is possible. Multiple, and
differing wire sizes are too complex to align things nearly as well. The
signal’s SPEED has to be best matched at all frequencies and not just the
physical wire length. The wire’s “signal length” is the problem. Use too many
non-symmetrical, differing sized wires and this is all but near impossible to
do with all the variables involved. I call this type of mixed wire cable,
“cable in a cable”. The effect is a kindergarten lunchroom in the dark; a mess.
In passive cable, you can’t force the highs to go in the
small wire and the lows in a bigger wire, and adjust the wire lengths to offset
the VP changes. The ENTIRE spectrum goes into EVERY wire, so now we compound
the time based issues. Only active electronics can separate the spectrum, and
that’s a problem too.
Does ICONOCLAST remove the “cables in a cable” problem? Only
one way to find out and that is to MEASURE them. The data is showing each
polarity with 12 two wire BONDED pairs, 24 wires in each polarity, and 48 total
wires in each cable;
175.3815 pF (X) +175.3815 pF = 180.1803 pF
175.3815 pF (X) + 171.4507 pF = 171.4507 pF
X = +2.74 % and – 2.29% variation between wires, or, they
are ~ 97.5% the same.
I would say yes, I got it right.
For ICONOCLAST speaker cable I set my design goal at no more
than 50 pF on capacitance and 0.1 uH/foot an inductance (45 pF and 0.08 uH
typical). On the interconnect I set the goal at 12.5 pF and 0.16 uH (12.0 pF
and 0.15 uH typical). This is WITH
connectivity and tested to prove it.
The complex electromagnetic designs of the RCA, XLR and
speaker cables allow ICONOCLAST to exist. The RCA is the most pure electromagnetic
equation that I have to work with and defines the interconnect cable problem.
How do we reach the greatness that a PROPERLY designed RCA does in the XLR
design (matched impedance / phase)? How can I convert the small signal world of
the RCA and XLR into the large current world in the speaker cable (low
inductance with still low capacitance)?
11.0 ATTENUATION At AUDIO
issue – is it mostly LOG linear so we can’t hear it?
If it is true that we can’t hear LINEAR attenuation
(measured Rs values say there is non-linearity) or TIME based issues in audio
cables, WHAT are we hearing with optimized designs, i.e. those that try to get
L and C to near ZERO as we can and with low time based issues? The design goal
difference in ICONOCLAST is TIME based and I’m not so sure that the
inaudibility of difference values of 5-10 micro seconds is correct. Linear
attenuation, I agree, is MUCH harder for the ear to pick out in typical cable
lengths. I said LINEAR LOG type decay.
Rest assured, if there is snake oil in these products it
sure looks like physics to me. All the above data is measured and real. The
question remains, WHY do the cables SOUND so much better if TIME based issues
aren’t audible? WHAT are we hearing, then? The reactive TIME altering L and C
along with the VP change with respect to frequency seem to be the difference in
cables, and audibly so. Linear attenuation can’t account for the differences.
Series resistance says that that factor isn’t as linear as we’d like, either.
There is a measurable difference in cables resistance across the audio band.
Is attenuation linear? I measured the Rs (series
resistance), with respect to frequency, of ICONOCLAST and saw a significant
CHANGE in attenuation with high quality R, L and C. Look at standard 1313A
speaker, 10 AWG Zip cord style cable (red trace). ICONOCLAST flattens resistive
non-linearity artifact, and the interconnects are both flat to 20 KHz human
hearing test point. Still, look at the
UNITS; it isn’t a wall of lost energy above 20 KHz.
12.0 Low Pass Filter
The issue – Cable is a low pass filter, and rolls off the frequency at
the frequency of the filter’s cut-off; Fc.
How does this change what we hear?
I saved this one for dead last since it was even overlooked
on my categorization of issues with audio cables. You’ll see why in a moment.
Here is the basic circuit. There is actually a SMALL
inductor in series with the resistor but notice that it doesn’t appear in the
equation that defines how the filter will behave, and is omitted. There are
circuits that involve larger inductors, and unless we have a resistor to
ground, they won’t apply to “cable” filters. Well, decent cable anyway.
The capacitor is a reactive device, like I’ve mentioned
before, so its properties change with frequency as does an inductor. A
capacitor eventually looks like a short to ground (capacitive reactance value
keeps changing) at higher frequencies so the signal energy takes the path of
least resistance through the capacitor to ground. R is in Ohms when capacitance
is in picofarads (pF).
The good thing about almost ALL
audio cables is that the roll-off properties of the filter are WAY above the
audio band. Yes, a first order filter will change the PHASE at the -3dB
attenuation point by 45 degrees, and time based distortions are more audible
than the roll-off attenuation. First order filter attenuation nor phase changes
are going to be an issue, theoretically.
Typical ICONOCLAST™ R, L and C Variables
The RCA shield “goes away” as
it is such a low resistance in series with the center wire, leaving essentially
the center wire DCR.
The XLR uses TWO 25 AWG wires
in parallel for each polarity, so the resistance is HALF the two wires, or
about the same as a ~22 AWG wire.
Calculating Fc we arrive at;
15.5 GHz for the 5 foot RCA.
36.4 GHz for the 5 foot XLR.
15 GHz for the 10 foot Speaker Cable.
The real problem with cable is that it can load down the
output op-amps with too high capacitance and change the frequency response and
possibly phase response. Some really high capacitance or high inductance
speaker cables can bug the heck out of power amplifier output stages, too. But
these problems aren’t filter problems, but bulk capacitive or inductive loading
problems on the output circuits.
All circuits “push back” below their operating region into
the pass band but a rule of thumb is to keep the fc pass band 10X or
more above the circuit’s operating frequency. We surely are meeting that
requirement with any decent cable, even zip cord.
Many outside this sub-discipline of engineering will STILL
insist that electromagnetic field time management and time alignment are not
important, and that only the bulk R, L and C matter. The ear is a time domain
instrument and readily time aligns the signal to the natural world we live in.
EVERY effort was made to pay attention to TIME domain issues in audio cables
and attenuation non-linear artifact. There are a myriad of ways to lose track
of TIME, and an audio cable is not a good place to make mistakes.
Consider all the measured and factual information above on
cable design and then ask yourself why cables sound different. Why wouldn’t
they sound different given how complex
it all is? True, poorly made cables all fall into a bunch of warm and
soft sounding products. Elevate the engineering and they indeed measure
different. The above is 100% true for ALL cables, if I may add. If I
mischaracterized a topic then, of course, only my cables are affected! All the
cable designs in the ICONOCLAST line are under US patents.
I hope my cables bring years of enjoyment to you, and NEVER
a feeling of complacency in what was provided to enhance your hobby’s (mine
too!) pleasure. The search is constant to try to align TIME based issues to
arrive at the best sound possible. The bad layers of the onion can’t be
removed, but the order and thicknesses can be altered. Signal coherence is both
arrival time and amplitude time dependent. Passive cable won’t allow
perfection, just a lot of hard work to manage the ill effects that Mother Nature
threw our way.
“Sound Design Creates Sound Performance”, and this means
driving down all measurable variables to the lowest possible balance we can
achieve. Does this make better sounding
Galen Gareis Principal Product Engineer ICONOCLAST Design Engineer
We tend to believe
that SHIELDED cables are superior to UNSHIELDED cables but the opposite is true
from a signal transmission evaluation. Why we feel SHIELDED is better is
because we overestimate NOISE ingress (outside the cable into the cable)
Cable electrical is determined by the primarily REACTIVE
variables that change signal shape arrival times. SHIELDING is to be considered
a necessity if, and only if, the ingress noise is more damaging than the time
based errors and physical size shielding imposes on cables. Why even have
shields if it doesn’t HELP improve the signal integrity from one end of the
cable to the other?
Capacitance is derived by the relationship of the shield to
the signal conductors in cable. The shield is usually at GROUND potential to be
a low impedance path for noise, so far so good. The bad news is that the CLOSER
a shield is to the signal wires, the more the cable varies per unit length in
measured electrical values of capacitance. It isn’t the same cable all along
its length from shield geometry variation, and the variation is much more
aggressive the closer the shield is to the signal wires. Capacitance, and thus
also inductance, change with smaller physical changes in the cable.
If you want to keep cable size SMALL, a shield means much
higher CAPACITANCE. And, a smaller size WITH that higher capacitance means a
larger per unit length variation in measured electrical. Even with AIR as a dielectric, we will see
much higher capacitance, and have a harder time controlling it with shields, so
we better need one for the function of the cable, and where it is used.
The following calculated table shows that the DIELECTRIC
in-between the shield and the signal wire can REDUCE capacitance, but only to a
point. It cannot remove the shield to conductor physical variation, which is
built into the DESIGN, good or bad.
How bad is the actual variation between the shield and
signal wire? This exact question was discussed when ultra high-speed
communications cables were being developed. Do we control the center-to-center
spacing in a BONDED PAIR over all else, or do we control the shield spacing and
geometry AROUND that bonded pair? BOTH will influence the final impedance, and
its variation. Which is really the bigger problem? Can we make better cables
managing what really makes the biggest difference, and reserve the less
aggressive physical attribute for higher performance requirements? This can
make the AVERAGE level of performance much higher at a much lower cost than
blindly trying to manage every variable all the time without a firm reference
to the cable’s final electrical values and variations.
Here is that exact analysis;
To demonstrate the
effectiveness of conductor center to center (C-C) in an ISTP cable, the example
below shows a change of C-C from 0.055” to 0.072”, holding a constant 0.061”
insulation diameter. This simulates a conductor with poor concentricity within
a well-controlled and constant insulation diameter. The impedance is nominally
102 ohms with a 0.061” C-C spacing and changes ever so slightly as the
conductors are spaced closer, or farther apart.
The shield inside dimension is a constant 0.122”. Under these
circumstances, the impedance goes from just over 101 ohms to just over 97 ohms.
A total impedance spread of about 4 ohms.
The significance of
the calculations is the relative insensitivity of impedance value with changing
C-C spacing compared to the variation in diameter of the shield, both of which
affect impedance variation with frequency. The impedance versus shield spacing
graph shows how severe the impedance change is with ISTP shield inside diameter
(I.D.) changes. Just a 20 mil change (0.120”-0.140”) moves the impedance almost
14 ohms. Our specifications allow only a 15-ohm swing.
The control of the effective shield diameter is three and
one-half times more sensitive than the C-C spacing of the conductors in ISTP
cables. Or, shield tape control is much more important than insulation
centering or backtwisting to compensate for off-center conductors. Also notice
that the closer the conductors move towards the shield in the IMPEDANCE VS
CONDUCTOR SPACING chart, the more Zo changes. When the conductors are 0.055” to
0.065” C-C, the impedance varies by less than one ohm. In contrast, when the
conductors are near the shield in the 0.065” to 0.072” C-C range, the impedance
changes 4 ohms. Unless your C-C is well out of spec (we have a 0.01” variation
with little change in impedance in this example) good shield dimensions are
much more important.
In contrast to ISTP cables, the UTP cable example shows
how profound the impedance impact is when the C-C changes just 11 mils compared
to 17 mils in the ISTP example above.
Where the ISTP cable had about a 4-ohm swing, the UTP cable has a 60-ohm
swing! In UTP cable, ground plane consistency is inherently stable because it’s
the metallic area around the cable which, under normal circumstances, is
perceived to be infinitely far away by the cable, too far to effect the
electrical to any significant degree. So the crucial variable in UTP cable for
consistent impedance is the strict control of C-C. This is why Belden’s
patented bonded pair technology is so important in UTP cable designs.
Impedance is, after all, a function of the Inductance,
capacitance and dielectric values. The impedance variation, and even at each
frequency in the audio band, changes with the dielectric and the spacing.
A cable with NO SHIELD, sometimes called UTP, does have a
reference ground “around” the cable, the environment. But, the capacitive and /
or inductive coupling are so far away that changes in the “reference” are
Shields have to pull their weight in signal integrity
improvements compared to cables used without a shield. If we have no external
noise, SHIELDS ARE WORSE than no shields! The math of cable electrical
stability firmly squares that up, per the data shown above.
This forces the consideration of NOISE. It even considers
HOW noise is transferred into (ingress) a cable, and even if the cable itself
is the source of NOISE for other external devices (EGRESS).
First, let’s be super straightforward about this from a
25,000-foot view. The closer a shield is, the capacitance value is high, and it
varies the most around the average value. Knowing that the proximity a shield
has around the signal wire can really upset the cable’s uniformity of
electrical, and how uniform we can engineer them, would we not want to use
designs that NATURALLY calculate an advantage to use with shield? Yes, we
To keep this easy, look at coaxial cables. This technology
HAS TO HAVE a shield to work. A signal wire surrounded by a shield. The signal
waveform travels along the wire surface, and under the shield surface and
inside the dielectric as a TEM (Transverse Electromagnetic Wave) wave. The more
perfectly round the inner surface of the shield and the outer surface of the
signal wire, the lower the capacitive and inductive variation and thus a lower
For signal transmission, we use 75-ohm cable (77-ohm is the ideal) and for
power 50-ohm Cables (30-ohm is the
ideal). Approximately 53.5-ohm military RG cables came about because it is the
mean between 33 and 77. If we freeze the materials we use to make the cable
(same plastics and metals) we will see that a 75-ohm cable has a larger
dielectric layer (lower capacitance) than the 50-ohm cable.
This is nice, since the farther away the shield is, the less
a given VARIATION of the shield changes the electrical stability. Reactive
variation impacts small voltage signals far more than larger 50-ohm power cable
applications with much more robust signal levels.
In a 50-ohm power type cable, we have a shield that is far
closer to the signal wire. This seems like a problem and it is, but the SIZE of
the signal is vastly larger than the NOISE. We can overcome the noise with a
larger signal, and even the return loss caused by more variable impedance can
also be mitigated with the size of the signal on power type coaxial cable.
This is simply the signal to the noise reference working in
our advantage in each design.
Voltage signal cables, dB or dBm, need shields farther
away (higher impedance) and it so happens this is the case with 75-ohm cables,
reducing capacitive coupling of noise.
Power signal cables, often in WATTS, need closer
shields for energy transfer (lower impedance) but this allows more capacitive
noise coupling. 50-ohm cables use more robust signals to overcome the noise.
This is like a low impedance speaker cable’s signal WAY over the terrestrial
There is NO EXCEPTION, lower impedance cables are much more
subjective to NOISE than higher impedance cables with the same noise ingress. We must fit the signal levels to the
impedance for ideal overall performance. 75-ohm cables are far better for
low-level signals as they capacitively couple less noise, as the DISTANCE to
the shield is larger.
To put the signal in perspective to NOISE, look at the table below.
Digital data cable go 100 meters / 328 feet with
over 23 dB of attenuation at 100 MHz and with ZERO errors due to external
noise, with UTP designs. Audio cables go mere feet, and yes we seem to want to
be the underdogs of signal integrity but we aren’t, and that’s a really good
The integrity that even a MC phono cartridge’s
0.35mV signal represents to the noise is in our favor. The robust signal even covers up POORLY made
SHIELDED cables. Do the shield really right, and it can help some RF, but
usually in a good unbalanced RCA system a RF bleed capacitor routes RF to
ground through the cap somewhere in the ground.
Coaxial cables need shields to work, and they need
shields to be super low DCR to prevent ground loop differential currents
between devices. The GROUND is shared in coaxial cables at uneven ground
reference points. RCA grounds have resistive differences. This can cause signal
bleed between channels. A BIG part of an audio coaxial cable shield is to
mitigate ground potential differences, and not to “shield” ingress.
A balanced XLR uses a SHIELD, yes, but it is NOT a part of the signal path, and
each right and left signal doesn’t share the virtual ground between the
differential voltage signals. Each amp has its isolated virtual SIGNAL ONLY
ground reference. There can be no inductive or capacitive coupling of right and
left channels. Unhook the GROUND on a XLR and it will work, with MAYBE a
slightly higher SN ratio. The outer shield simply knocks down the noise ingress
at RF, if any is there, so UNBALANCE in the pairs mitigates to a lower residual
value. One-percent unbalance of a small signal is better than one-percent of a
This is the true advantage of XLR cables over RCA.
Both have good RF noise immunity with the XLR having far superior signal
channel isolation and…low frequency noise isolation.
Since an XLR FLOATS the virtual ground independent
from any other signal, noise is the same on each leg, so it cancels. We see the
“difference” of each leg as the signal, which doesn’t change potential. This
includes magnetic and electric fields. Coaxial cables can’t shield magnetic
fields since copper is “invisible” to 60 Hz magnetic interference.
ICONOCLAST™ uses SHIELDS, but the WAY we use shields
insures geometric consistency to the signal wires. Care was taken to insure a
good BALANCE within the XLR signal wires so even if a shield is broken, little
performance impact will be measured;
Capacitance @ 1 kHz per ELP 423, Agilent E4980 Precision LCR Meter,
Belden 4TP Cap/Ind Test Fixture, all tests performed on a 20ft specimen.
Pr to Pr(star quad) – 10.4113 pF/ft
UnBalanced: Pr 1 to Shld – 401.9868 pF/20ft
Pr 2 to Shld – 405.9738
Cap UnBal ((diff/max) * 100) – 0.98%
Requirement – 3% maximum
SHIELD TRANSFER IMPEDANCE – This is
a measure of the cable’s shield impedance in milli-ohm/meter. The lower the
transfer impedance at a specification frequency the better the shield at that
frequency. It is frequency and design dependent. The current traveling in the
shield times the transfer impedance produces a interference voltage product to
ground in the shield, E=I*R where R is the transfer impedance.
SUMMARY – Shields have to be considered relative to noise and the
resulting S/N ratio since the application of a shield is ALWAYS a negative
variable to signal integrity that has to be over weighed by true noise
mitigation. If noise is paramount over the signal, then shields are a necessary
requirement. If shields are a part of the signal path, then the noise they
generate has to be mitigated with shield DCR.
ICONOCLAST uses shields properly,
and insures that the negative influences are geometrically stabilized, and
measured for performance in both RCA (DCR) and XLR (unbalance percentage).