XLR Design Brief

1.0 Conductors
1.1 Copper Size
2.0 Dielectric Materials
3.0 Dielectric geometry
4.0 Shield Material and design considerations
5.0 Jacket design and material considerations

1.0 Conductors.

  1. Copper Size.

BOTH of the copper conductor and size considerations were answered when we started the RCA cable. We don’t want to change the current coherence with a differing conductor diameter if we are to mirror the reactive variables, too. We need the same exact wire to shield reactive L and C parameters in each cable in the end configuration design. The geometry of each cable is entirely different so how to you do that? That is, assuming you want to match the RCA and XLR properties and maintain the same signal quality…and we certainly do.  There is no reason to copy a bad sounding RCA cable when designing an XLR, so the RCA is designed FIRST.

  • Dielectric material(s).

One difference in the XLR is that we are going to use FOUR wires in a star quad configuration. (Note: in our “Gen 2” XLR product, there are sixteen wires — four wires making a star quad in place of each single wire in the design shown below.  For more detail, see the last paper in this series.)  Four wire XLR cables use two cross-connected wires for each polarity, which doubles-up the wire AWG for lower attenuation. Two 25 AWG have the DCR of a single 22 AWG yet has way better signal coherence by using smaller wire.

I could have used a cheaper and easier two wire XLR design but the inductive and signal coherence benefits of a star quad are too good to pass-up. If I can get the materials and quad design to achieve a high enough level of performance it is a better cable design.

Star quads have a higher degree of CMRR (Common Mode Rejection Ratio) when properly signal balanced. There are three primary reasons for this;

  • The two or four wire stranding “twist”.
  • The differential encoding.
  • The outer shield properties, but only at RF frequencies.

Two wires of a star quad are a “positive” voltage, and two wires are a “negative” voltage (180 degrees out of phase), hence the term “balanced”.  If the cable were a teeter-totter, it would sit level. Some call this differential mode since each signal is equal but different.

Differential Mode Transmission
Perfect Wire Balance Equals Less Noise

In the example above we show two wires, but the system is the same in a star quad. The signal we WANT is encoded as +2 volts and -2 volts. The noise can’t “change its spots” relative the cable’s twisted pairs and shows up as the same voltage on each wire, +1V noise in this example. The TWIST ratio helps make sure that the wires see the noise the same amount of time and this is vital to the function of the circuit.

Here is where the balance is so important; the signal IDEALLY becomes the superposition of ALL the voltages, or +3 volts and -1 volt. No more, no less. The signal voltages are STILL exactly 4 volts “apart” from each other; +2 to -2 with no noise and +3 to -1 volts with the noise.  The signals are fed into a difference amplifier that, you guessed it, looks at the “difference” between the two voltages and see’s 4 volts with, or without, the noise. The noise is absent in a perfect world at the difference amplifier’s output.

In order to do this, every wire has to be presented to the noise in the exact same way via the cable twist and has to be the same length so the signal stays TIME aligned down the wire and has to have the same attenuation. The difference amplifiers need to be nulled perfectly between gain halves. Believe it or not, this gets done really well with good quality products.

The control tolerance of the copper is 0.0005”, so attenuation issues are mitigated and CUB (Capacitance UnBalance) tests insure we see MIL standard quality in the finished cable. All quality types of copper can be used in the XLR design. It is the overall structure that is the most “magic” and not as much the copper itself, although the copper draw process does influence the sound.

We have several variables that aren’t present in a coaxial cable design to contend with;

  • CUB, Capacitance Unbalance or, each wire shows a differing capacitance to ground.
  • DCR unbalance, each wire has to be the same DCR.
  • CMRR remainder, the differential signs have to NULL to the exact same point neither above nor below reference ground.
  1. Dielectric geometry.

Lots of words, time for a picture;


The above CAD drawing is what we have inside our XLR design so far (well, I ignored two wires in the drawing).

Remember I wanted to make L and C reactive variables EXACTLY the same for each cable with EXACTLY the same wire size and draw science? What else do we know? I also said that CAPACITANCE is sensitive to the distance to a conductive plate area, and that means ALL the way around the wire. The coaxial cable is easy; we purposefully put a ground around the wire at a known distance that defines the capacitance ground plane reference distance and inductive loop area.

In the coaxial cable, the center of the wire to the inside of the tube is 0.098” / 2 = .049”.  Ok, so what? This is what. The capacitance is a squared law property and predominantly sees the ground closest to the wire. The shield on the opposite side of the XLR cable, to a first approximation, falls a way. We actually measure the capacitance BETWEEN the two cross paired wires but the ground location still influences the capacitance. Also, we have four wires that are capacitors.

This doesn’t “sound” good, does it? We have four times as many wires and all have capacitance. Somehow this is supposed to come out around 12 pF/foot (with connectors), same as the RCA!

Now for the inductance part, L. Inductance is loop area defined. It could care less about the dielectric, but the graph above shows a HUGE ~0.170” loop area! How is THAT going to get to the 0.15 uH/foot inductance of the coaxial cable? I could make something up, but that isn’t as neat as what’s really going on.

To get capacitance as low as I need it to be to match the coaxial cables, I use DISTANCE between the wires. And yes, this DIRECTLY sets what the inductance will do…hold on a minute.  By using AIR, I can set the C-C of the wires to meet my capacitance target needed for the final tested value with two cross wires connected and tested between them. AIR lessens this distance for a given value of C so I can also manage inductance now. For inductance, L, the smaller the wire loop area the better for a given value of total capacitance. Air gets me far closer than any other dielectric.

How much air? Well, EXACTLY the same as the coaxial cables! How do we do that? The standard answer is, “very carefully”. Let’s look at a drawing;


Still, so what? Yep, I agree, until we compare this area to the area in in the RCA cable air dielectric; 0.00754 in2. OK it isn’t exact; I missed by ~0.000009” in2.  I use the exact same thread design around each identical wire so it’s all the same area in the chamber as in the RCA.

Let’s do some reality checking as to what it SHOULD be based on MEASUREMENTS and calculations.

  • We have the EXACT (can I say that as close as it is?) same velocity of propagation based on the composite (air and plastic inside the ground plane) dielectric; 87% at RF reference.
  • I measured the IMPEDANCE at RF @ 100 ohms, same as the coaxial cable.
  • The dielectric constant can be calculated and from that the VP, VP = 1/ SQRT (E).
  •  And from that composite dielectric I also know what the capacitance has to be.

Capacitance (remember that chart on dielectric value and capacitance earlier?) is directly linked to the group dielectric constant. I know VP, and I know the impedance, so I can calculate the capacitance and then get the dielectric constant from that.

101670 / (C * 87) = 100 ohm

C = 11.68 pF/foot.

What does the cable actually measure on capacitance? The chart below shows 11.767 pF/foot. Notice that the capacitance values between each of any two wires has to be ~ 5 pF/foot to “double-up” the two wires capacitance and still to arrive at a final ~11 pF/foot! Yep, that’s LOW capacitance. Capacitance adds in parallel so this is a significant issue when a design uses four wires.

Below is the measured and calculated imbalance of the capacitance between 1-3 and 2-4 cross wires’ conductors as a “pair”; 2.02% unbalance, very low.

We seem to have the capacitance and VP looking much like the coaxial cable. Remember, measurements include ALL the approximations in the soup.

So what about inductance with that WAY larger loop area? Isn’t that going to really kill this thing?  No, because of some properties of magnetic fields. Magnetic fields CANCEL if they see each other in OPPOSITE directions. Inductance is the “reactance” or “resistance” to instantaneous flow of current. If we can REDUCE the magnetic field lines, we can directly reduce the measured inductance.

We also know from the basic equations that DISTANCE between the two wires is important. Keeping BOTH distance and magnetic field line magnitude small lowers inductance, and removes the noise.

The picture below shows what’s going on…sort of. For now, we’ll pretend the field’s ONLY go “inwards”, or inside the wire, and stop there (they don’t). If the lines that extend outside each wire do the OPPOSITE as the field INSIDE the wires, they reinforce the field!  It is generally accepted that the flux lines concentrate substantially BETWEEN the wires.

If we draw arrows that represent the DIRECTION of the circumferential magnetic field waves AROUND each wire we get what is shown below for a NOISE signal hitting the wire. We have TWO different voltage polarities so we have TWO different current directions for the SIGNAL, but the NOISE is the SAME direction in all the wires.

If you grasp a wire with all four of your fingers, and point your right hand THUMB in the CURRENT direction, your fingers will point in the field’s circumferential direction around each wire. The arrows are a “part” of the noise current field lines “inside” the four-wire group.

NOISE FIELDS (all the same direction)

Where the arrows are OPPOSITE each other in direction between any two wires, the field lines cancel. For NOISE every field theoretically cancels. ADJACENT or ACROSS from any TWO wires we induce field cancellation with a star quad design.

For the SIGNAL, we now have TWO equal but opposite current directions.

This allows larger wire-to-wire spacing in order to lower capacitance and also keeps inductance low. Inductance is managed with field line cancellation geometry.

Now we know why I didn’t use a two-wire system, you can’t manage CMRR.

Let’s look at the situation for the signal. Below is a simple picture of the field cancellation between four wires with opposite polarities wired as a star quad.

SIGNAL FIELDS (opposite directions)

   Minus = Current INTO the page (CW rotation)

   Plus = Current OUT of the page (CCW rotation)          

Reduce Signal Loop Area to Reduce Inductance

What do we see? The all the signal field lines DO NOT cancel. Adjacent wires reinforce, and opposites wires cancel. Reducing loop area is the best way to manage inductance because we can’t cancel all of the field lines, only some of them. This theoretical field relationship limits the ability to reduce capacitance for a given inductance. Using low dielectric constant materials to lower capacitance (Air!) allows closer spacing needed for low inductance.

Is there a design that can, in theory do BOTH, reduce signal and noise fields to “zero”? Within the limits of DESIGN, yes there is. The ICONOCLAST series II reduces both noise and signal field cancellation. The wires, in practice aren’t EXCATLY the same distance apart and EXACTLY the same resistance, so we say “in theory”. But, reducing the nose to 1000 or more times less and reducing the inductance 27% is indeed achievable.

So, after all that explaining, how does the star quad ICONOCLAST cable measure up? Tests at 1 KHz show the following values below. The inductance between the two cross wire pairs of the star quad are 0.15 uH/foot inductance…same as the RCA.

So what does the “reactive” picture look like comparing the RCA and XLR? How close are they to being the same? This swept test is the real deal. There are no approximations to fudge.

What we see above is impedance / phase for the XLR and RCA superimposed one on top the other. Note that there are four separate lines. We have two identical cables with exceptionally low reactive variables.

  • Shield material and design considerations.

There is yet one last thing to consider in the XLR design; the outer shield. A 95% BC (Bare Copper) braid is used. Audio cables are not RF designs, and the braid shield will NOT shield low frequency magnetic interference. The CMRR of the XLR is going to do that for us. Excellent CUB, DCR unbalance and twist ratio all aid CMRR. The braid DOES knock down RFI by 80 dB, so that’s a given. The shield isolation @ RF mitigates NULL balance at high frequencies only.

Like it or not, 20-20K is a predominantly magnetic field frequency range where the B-fields decay at a ratio of 1/x^3. DISTANCE is the best solution for isolation of cables with magnetic properties.

5.0 Jacket design and material considerations.

All ICONOCLAST cables use FEP as the jacket to reduce UV sensitivity, plasticizer migration and provide chemical resistance.  The cables are designed to last decades.

I hope that this design summary of ICONOCLAST RCA and XLR interconnect cables shows how important good design is for ALL your audio cables, and that every manufacturer has to manage all the same variables to produce these results. There is little “magic” in the design of good cables. There are indeed tertiary variables that we can’t measure, but those should not influence the ones we can measure, or at least not excessively so. Mother Nature abhors complexity, so the better managed the known variables in a cable are, the more properly it may highlight “unknowns.”  To put it another way, the more we put knowns into their proper place, the better we may distinguish the effects of the unknown. Wire draw science, for instance, can be heard better, and more fairly, in a superior electromagnetic design.

Belden appreciates your interest in how quality interconnects are made, and how / why ICONOCLAST RCA and XLR cables were physically derived as you see them in their production form. We have no special sauce or magic in our products, and I think that the cables perform as well as they do BECAUSE we did not design around “unknowns” and then make it appear as though we had unique influence on those unknowns in the design.

Truly low R, L and C cables are difficult to make when consideration is given to all three variables to manage them in a truly balanced fashion. The designs can be frustratingly simple looking but hard to manufacture, as processes are pushed to the limits of current capabilities. Belden’s focus is to make real measured values as low, and properly balanced, as we can. ICONOCLAST interconnects represent the pinnacle of low frequency measurements and electrical balance between the RCA and XLR (same electromagnetic properties).

The next design analysis will look at the SPEAKER cable.

Leave a Reply

Your email address will not be published. Required fields are marked *