Quality by Design

Introduction

Sound systems are made up from a number of different electronic components or sub systems which are connected together using what we refer to as cables. Even today, the vast majority of cables are made using copper wire. When the sound system is relatively small - like a HiFi system for the home - then the use of cables appears to be pretty simple. But as systems get physically large and/or complicated, interconnections using copper wire become complex. If we are going to be successful at designing and building an individual piece of electronic equipment, and/or a complete system, we must understand the behaviour of electric currents and cables - first.

Cable detail

As we can see from Keith Armstrong's diagram "Overall Equivalent Circuit for Ground Loops" - cables turn out to be rather more complex than they look from the outside. All conductors in the cable exhibit series resistance and series inductance. There is also a parallel capacitance effect between each signal conductor and the common conductor, caused by the insulation material used in the cable construction. All this gives rise to a series of filters which - even on their own - can produce a rather tragic effect on an alternating signal as the length of the cable increases. In other words, any long cable behaves like a band-pass filter network.

There is another major problem with the parasitic capacitance between each conductor and the cable shield. Unfortunately, the values of each parasitic capacitor are not equal. Thus, as the cable gets longer, the high frequency CMRR (common-mode-rejection-ratio) deteriorates and high frequency interference couples with the signal pair.

If all that was not enough - at each connector interface (source and load) - inductance in the terminations produce mutual coupling with the conductor that is common to both circuits. Mutual coupling occurs because the arrangement of the inductances produces electromagnetic behaviour similar to that of a transformer. In Keith's diagram, the common conductor is complex. Having a complex common conductor is normal for large sound systems, so solutions for minimising the transformer action at connector interfaces have already been developed. Indeed, solutions to all of the above problems are readily available, which is one of the reasons for this paper. So - here is the main point of this introduction: unless we understand the physics of interconnections and cables, there is no point in considering the design of individual high performance circuits.

Minimizing cable transformer action at input and output ports

At power frequencies (50/60Hz), cable inductance gives rise to potential differences between any two system nodes. This means that even the relatively heavy-duty SAFETY EARTH wire between two pieces of apparatus in the system, has more inductance than it has resistance at power frequencies. LF considerations So, although each piece of equipment is connected to EARTH - they are not necessarily going to be at the same potential. This is because the inductance in the earth/ground cable between any two system nodes gives rise to a voltage difference between the two pieces of equipment (Vg1-Vg2, in the illustration) . Now - it is widely accepted that the conductive enclosures used to house electronic equipment must be connected to a protective earth/ground conductor for personnel safety. Another fact is that for the shield of a cable to be effective at protecting the signal conductor(s) from RFI, the shield must be connected to the conductive enclosure at both ends (for the cable shield to be able to provide protection - RF current must flow in the shielding conductor)! When we make such connections between two pieces of equipment, a low frequency current will flow in the loop formed by these connections (Loop "L in the illustration). The fact that a loop current is flowing between the two system nodes, does not mean that interference will occur. If the loop current flows in the equipment enclosures, or in a common conductor designed into each device, no interference will be produced. Hum or buzz can only occur if the loop current gets inside either device, where mutual coupling can occur. In fact, equipment that is sensitive to hum or buzz has merely been designed incorrectly! Quite simply - if the cable shield conductor connection (pin 1 on an XLR connector) - bypasses the common conductor (usually the chassis) and connects directly to the internal circuit common on the pcb (printed circuit board), then current flowing in the cable shield conductor will be injected into the internal circuit. This will result in hum, buzz and RF demodulation. This type of design fault is referred to as the "Pin 1 Problem".

Note that even if we "lift" the shield connection at one end of a cable interconnection - high frequency loop currents still flow between the two system nodes. This is because of the parasitic capacitances between each signal conductor and the common conductor, formed by the insulation required to stop the signal conductors from shorting-out. So, in a poorly designed piece of equipment, "lifting" the shield at one end may solve the hum problem, but RFI will still be injected into the internal circuit.

Practical solutions for equipment designers

AES 48 diagram

The solution to the "pin 1 problem" for equipment manufacturers is rather simple:

Most of this information is now formalized in the AES Standard: AES 48. However, many of us have been using these techniques since 1995. In that year, the AES published a series of papers on interference and noise suppression, in their Journal No.43, Vol. 6. In particular, the paper by Niel Muncy featured the accompanying common-bonding illustration, which proves to be the basis for designing electronic equipment with excellent immunity to interference and noise.

Note that all reference conductors (cable shields on input and output ports, power 0V wires, the safety earth connection, etc), are bonded to the same continuous conductive element in the structure of the device. Ideally, the "continuous conductive element" should be the case or chassis of the device. However, excellent results can also be obtained by using a single conductive panel. A good example of this is the PCI card used in computers. All input and output connectors are mounted on a single metal panel mounted at one end of the circuit board. This panel is bonded to the ground plain on the circuit board, which - in turn - gets bonded to the chassis of the computer when the card is plugged in and secured. Similarly, the same single panel technique can be used in non-conductive rack units. The wiring of rack units is discussed later.

A modular backplane

In mounting all input and output connectors on a common conductive panel, we have created a structure known to radio frequency engineers as a RF backplane. Radio frequency currents tend only to flow in the skin of the conductive panel, and this is often referred to as the skin effect. Even a modular common conductive panel provides excellent RF protection for the internal circuits and presents a very much lower impedance to low frequency current flowing in the structure, than can be achieved by using individual wires or pcb traces.

In fact, a large conductive area approximates to an equipotential plane. The most desirable quality of an equipotential plane is that there should be no potential difference between any two points that are geographically separate. Thus, connections to each point can be regarded as connections to the same potential. In the context of system immunity to interference, no noise voltages exist between these points and therefore no interference is injected into a system that is "earthed" only to these two points.

Such a perfect equipotential structure is in fact impossible, but it can be approximated by a large highly conductive area of metal, as shown in the adjacent illustration of the rear of a CADAC F-Type console. Here, the overall design makes use of a conductive metal structure to support the electronics; a conductive metal enclosure to provide electromagnetic shielding for the electronic circuits inside; a modular but highly conductive low impedance return path for all input and output cable shield connections. Note that the connectors for all input and outputs (including power) are mounted on the modular back-plane.

Input and output filtering

io filtering

RFI filters are required at all input and output ports, including data i/o, a.c. and d.c. power ports. Here, we concentrate on the filter topology for analogue audio signal i/o.

One of the most important aspects of interference control is to understand the distinction between the possible modes of coupling. The basis for this distinction is the idea that two separate circuit paths can coexist in the same set of conductors. One of these is the circuit that was intended by the designer - signal send and return, along which the desired signal currents flow "differentially", that is - in opposition to each other. The second path is the parasitic circuit that is formed between this desired circuit and the structure within which it is located. This is known as the "common-mode" circuit, because the currents in the conductors all flow in the same direction. Thus, when we are designing a filter, it must be designed to deal with the mode that is most relevant to the interference that is present; conversely, a filter that is attenuating the wrong mode will be ineffective, no matter how good it is. Traditionally, we know that two types of filter topologies are available to us: common-mode, for suppressing interfering currents that flow in the same direction in both conductors, and differential-mode, for interfering currents that flow differentially (one way down a signal conductor, the opposite direction on its return or alternate balanced signal). In practice, we find that both types of filter are required at input and output ports (see references [1], [2], [3], [5], [6], [8]). Up until quite recently, the cost of combined CM and DM filters was quite prohibitive. But since the introduction of EMC regulations, a wide range of low cost filter components have become available from manufacturers serving the general electronics industry. One of the most useful components to become available for use with audio signal circuits is illustrated, and is generally referred to as a Common-mode-choke or Common-mode-filter (CMF).

CMF characteristics

The CM choke itself comprises two inductors wound in the same direction on a common ferrite core. The magnetic fluxes in the core generated by the send and return signal currents cancel out to a very high degree, so the choke's impedance to differential-mode signals is very low. But to currents travelling in the same direction (common-mode currents) the fluxes in the core add together so the CM choke presents high impedance to common-mode currents. Various values of inductance (and thus common-mode impedance) are available, and combining this high CM impedance with the relatively low impedances of capacitors and other types of circuits results in attenuation, hence filtering, of the unwanted CM interference or noise. The differential-mode fluxes do not entirely cancel out, so at radio frequencies the CM choke also provides some DM impedance, helping to filter out the unwanted DM interference and noise. Typical filter response characteristics are also shown. Note that for each common-mode characteristic, there is an equivalent differential-mode curve shown as a dashed line.

Generally, the differential-mode performance of the filter increases with the value of filter capacitors shown in the figure as "C". Practical i/o filter networks using this configuration are currently in use with all analogue audio products supplied by CADAC Electronics PLC. Typical values for the components are as follows:

The practical network shown has been found to provide excellent suppression of CM-to-DM conversion products in correctly shield-bonded cables used to interconnect both analogue and digital-audio input and output circuits over the frequency range 250kHz to 1GHz. The network has also been fundamental in assuring that all audio products supplied by the company comply with the EMC Directive's standard for the immunity of professional audio equipment [7], which is a mandatory legal requirement in Europe - and is good audio engineering practice in any case because compliance with this standard has been found to lead to reduced audio-frequency noise levels [5] & [6]. Note that the good RF performance of any filter can easily be compromised by poor design of wiring, PCB layout, and general assembly [6]. These issues are not covered here, but there are many references in the EMC literature that deal with them. Without proper shield-bonding, the simple "CM choke and capacitors" filter described above would be inadequate to achieve compliance with the EMC immunity test standards required by the EMC Directive. Cascading two or more such filters might be adequate, but the costs and PCB space required would be unacceptable - it makes much better economic sense to combine affordable filtering with affordable shield-bonding as discussed above.

Shield bonding and filtering in Digital equipment

Controlling processor emissions

D16 insides

Unfortunately, digital electronics produces very large amounts of high frequency emissions, so shielding, shield bonding and filtering are absolutely essential to the success of a design.

A typical digital rack unit is shown in the adjacent illustration. Most of the really noisy electronics, which includes the main processor and DSP chips, is purposely designed into the the main board. The main board is made multi-layer, so that one whole layer can be used as a "ground-plane" - the purpose of which is to help control the high levels of high frequency emissions from clocking the data at high speeds. When the main board is mounted close to the rack chassis, with the main board ground-plain bonded directly to the chassis - we are able to minimise the level of the high frequency emissions inside and outside of the chassis. Controlling internally generated RFI is essential for the equipment design - rather than for EMC considerations - because the analogue electronics is inside the same container.

Internally generated RFI

To measure the extent of the internally generated RFI, we need an EMI probe connected to a spectrum analyser. In the left hand illustration, the main board ground plain is insulated from the rack chassis. Even though the main board ground-plain is connected to safety earth, high frequency emissions are pretty hot inside the box. When the main board ground plain is bonded to the rack chassis, the level of the clock and its harmonics are dramatically reduced.

Once more, the relatively large area conducting surface of the equipment chassis used as the common-bonding conductor minimises internally generated (emissions) and externally generated (immunity) RFI.

Power Supply Filtering

Conducted Emissions

We often forget that every electronic component in a sound system is directly connected to every other electronic component in the system via the AC power wiring. If any component puts interference noise onto the AC power wiring, every other component in the system will be effected!

The adjacent figure shows the interference put back onto the AC power wiring by a poorly designed power supply unit and mixing console. Needless-to-say, this equipment has no AC input or DC output filtering. Interference put back onto the AC supply wiring up to 1MHz is caused by the switching pulses from the AC-to-DC converter (the bridge rectifier). Interference above 1MHz is generated by the electronics in the console. The closely spaced spikes reveal that the electronics have a large digital content.

Note that the thick black line is the EMC Emissions pass mark, and the overall bandwidth of the display is 150kHz to 30MHz. This is a typical example of RFI generated internally by legacy equipment (pre-EMC regulations) or by equipment from manufacturers that have not done any formal EMI testing of their designs. The figure shows failure to comply with EMC regulations

Designing for low noise and compliance

There are three sets of filters required for a power supply unit:

AC input filters are now readily available from most electronic component suppliers. Double-pole filter designs are the most effective, such as the FN series by Schaffner. Such filters are able to minimise the ingress of interference already on the AC supply line, and minimise interference generated by the equipment from getting out onto the AC supply line.

Snubber capacitors are required across each junction of the bridge rectifier. This is to absorb the switching spikes as each diode conducts. A typical value for the capacitors is 0.1uF, but with a working voltage of 8 x transformer-secondary-voltage.

D.C. output filter topology can be a simple single-pole network for a linear psu design, but can become quite complex, if a switch-mode circuit unit is employed.

When all three types of filter are employed in a PSU design, there is no problem in passing the conducted emissions test.

EMC Filters
















Understanding the difference between balanced and unbalanced interconnections

The unbalanced case

Unbalanced interconnection

We only ever need two conductors to transfer an electronic signal from one place to another. Current generated by any source, travels along the "Send" Conductor to the load, and then returns to the source using the "return" conductor. The actual signal current flows in a complete loop. Originating from the current source, the current flows through the load and back to the source, as shown in part (a) of the accompanying figure. But we know, in the case of our unbalanced audio cable, that the send conductor is physically inside the return conductor. We call this special type of cable structure coaxial (see part (b) of the illustration). Never-the-less, the two conductors in a coaxial cable behave "electrically" in the same way as a pair of individual wires. In fact, this type of cable construction does have a special property. The coaxial construction produces some immunity to low frequency interference. The immunity is excellent if the cable length is relatively short - say less than 4 m. However, any conductor that is close to a varying electromagnetic disturbance, will be subject to interference induced by varying electromagnetic fields.

In an unbalanced interconnection, induced noise current (interference) flows in the same phase as the wanted signal current (the dotted loop in part (a) of the figure). This means that interference will always be present in the source and the load. We cannot escape from it. Levels of interference in short unbalanced cables is relatively small and can often be ignored. But as cable lengths increase, the signal-to-noise-ratio soon deteriorates to a point where interference is continuously present at the output. A good rule of thumb, for high performance audio systems, is to keep unbalanced cable runs to less than 5m.

The balanced case

Balanced interconnection

In a balanced interconnection, there are still only two signal conductors, but the signal-to-noise performance is quite different. We create this difference by designing-in components that change the polarity between the "Send" and "Return" conductors. The "Send" conductor is made positive with respect to signal common, and the "Return" conductor negative. The signal current flows in the normal loop between the source and the load (continuous line) , but interference currents now flow in two different loops, that go directly to the signal common conductor. In other words, the interference current induced in the "Send" Conductor is 180 degrees out-of-phase with the interference current induced in the "Return" conductor. At the output of the balanced interconnection (the secondary winding of T2, in the illustration), any interference induced in either conductor is very nearly cancelled out.

In our example interconnection circuit, transformers are specified as the polarity coding and de-coding elements. The use of transformers has several advantages over any other form of polarity coding/de-coding element, network, or circuit:

Unfortunately, transformers are rarely designed-in by equipment manufactures these days. This is partly due to their high cost, and partly because of the fact that many excellent electronically balanced input and output circuits have become available. However, if true galvanic isolation between two system nodes is a necessity, then the transformer can often be the lowest cost option.

Bonding at system nodes

System interconnections

When we are designing a modern audio system, we can expect a relatively harsh RFI environment: computers, digital control systems, phase control lighting, cell telephones, etc.. This means that me must use shielded cables for all interconnections, and that in each case the shield termination must be bonded at both ends (remember that the shield can only protect the signal pair from RFI when the RF current flows in the shield conductor). Further, this means that the equipment that we specify at each system node must have input and output connector terminations designed to AES48. If it is absolutely necessary to use legacy or poorly designed equipment at a particular system node, then the interconnections from/to the previous/next node must be modified to the AES48 standard.

Analogue and digital audio signals can use either single shielded twisted pair cable, such as Beldon "F" series; 1696A , or multi-core cable consisting of shielded twisted pairs plus an overall braid shield, like Van Damm FC/OFC. These modern cables have very low inter-conductor capacitance with good balance, in the order of 43pF/m. This gives good bandwidth and minimises the affect of SCIN (Shield Current Induced Noise).

Minimising the loop current

Diverting loop current

In large sound systems, the value of the loop current flowing in an individual cable shield is usually negligible (see references [5], [6] and "Bonding cables at both ends to reduce noise", on the home page). This is because the total loop current is spread across all of the shield conductors in the cable bundle. If there are 8 cables in the bundle, then each cable will carry 1/8th of the overall current, and so on. However, if the loop current is too high in a cable bundle, the IEC have developed a method for diverting most of the current. The technique is known as "Parallel Earth Conductor" (PEC).

With respect to dealing with large value current flow, a cable shield conductor is not good because the DC resistance per metre is too high. The ideal sort of conductor for carrying large values of current would be a cable made from multiple strands of oxygen-free copper. And this is how the PEC technique works. A PEC is made the same length as the multi-core or bundle of signal cables; it is laid such that the PEC follows the same path as the cable bundle; the PEC is bonded to the common-conductor (chassis) at each end; typically, some 80% of the loop current is diverted away from the signal cable shields and flows in the PEC.

But the most important aspect of the PEC technique is:

Cable trucking as PEC

Using the cable trunking as the PEC

When we are designing a brand new installation, then we have the opportunity to optimise the concept of the PEC. Instead of using copper wire, we can use large area metal structure between each system node as the PEC. A large area metal structure gives us:

...and therefore...

When we use the cable trunking structure between nodes as the PEC, we can forget about the effect of loop currents.

Rack units

Rack bonding

Most rack systems used in live sound are made from non-conductive materials, such as MDF or plywood, so there is no chance that the enclosure can provide EMI shielding. This need not be a problem for interference control.

The solution is once more, very simple:

*If more than one panel is used, bond all panels to the common bonding network.

More detailed notes on wiring up equipment that is designed with the "pin 1 problem" can be found in "A Practical Interference-free Audio System" (part 1) on the home page.

Rack unit with a conductive enclosure

Optimizing rack bonding

When a rack unit is constructed using conductive material, then even better results can be achieved with respect to noise immunity.

Where possible, connect the protective earth/ground conductor to the common bonding network for safety reasons.

Note that the illustration shows the plan-view of the rack.

References

  1. JAES Vol. 43, No. 6.
  2. Radio Frequency Susceptibility of Capacitor Microphones - Jim Brown and David Josephson - Preprint Number: 5720
  3. Common-Mode to Differential-Mode Conversion in Shielded Twisted-pair Cables (Shield-Current-Induced Noise) - Jim Brown and Bill Whitlock - Preprint Number: 5747.
  4. IEC 61000-2-5: Electromagnetic compatibility (EMC). Environment. Classification of electromagnetic environments
  5. A practical Interference Free Audio System (Part 1) - Tony Waldron - EMC & Compliance Journal (www.compliance-club.com - archive section.
  6. EMC for Systems and Installations - Tim Williams and Keith Armstrong - Butterworth-Heinemann
  7. EN 55103-2: Electromagnetic Compatibility Part 2 - Immunity
  8. High power microwave effects on electronic components - Gunnar Göransson, IEEE 1999 EMC Symposium, Seattle, August 1999 (pages 543-548 in symposium record).

email Tony: twaudio@hotmail.com ; cell 'phone: +44 (0)7932 863670; postal address: 24 Knoll Rise, Luton, LU2 7JA, UK.

Valid HTML 4.01 Transitional copyright © Tony Waldron 2005
This site is sponsored by FragrantSword