While there is no doubt that computer analysis is one of the most valuable tools that the analog designer has acquired in the last decade or so, there is equally no doubt that analog circuit models are not perfect and must be verified with hardware. If the initial test circuit or "breadboard" is not correctly constructed it may suffer from malfunctions which are not the fault of the design but of the physical structure of the breadboard itself. This chapter considers the art of successful breadboarding of high performance analog circuits.
The successful breadboarding of an analog circuit which has been analysed to death in its design phase, has the reputation of being a black art which can only be acquired by the highly talented at the price of infinite study and probably a human sacrifice or two. Analog circuitry actually obeys the very simple laws we learned in the nursery: Ohm's Law; Kirchoff's Law, Lenz's Law and Faraday's Laws. The problem, however, lies in Murphy's Law.
Murphy's Law is the subject of many engineering jokes, but, in its simplest form "What can go wrong - will!" it states the simple truth that physical laws do not cease to operate just because we have overlooked or ignored them. If we adopt a systematic approach to breadboard construction it is possible to consider likely causes of circuit malfunction without wasting very much time.
In this chapter we shall consider some simple issues which are likely to affect the success of analog breadboards, namely resistance (including skin effect), capacitance, inductance (both self inductance and mutual inductance), noise, and the effects of careless current routing. We shall then discuss a breadboarding technique which allows us to minimise the problems we have discussed.
As an applications engineer I shall be relieved when room temperature superconductors are finally invented, as too many engineers suppose that they are already available, and that copper is one of them. The assumption that any two points connected by copper are at the same potential completely overlooks the fact that copper is resistive and its resistance is often large enough to affect analog and R.F. circuitry (although it is rarely important in digital circuits).
The diagram in Fig. 2 shows the effect of copper resistance at DC and L.F. At H.F. matters are complicated by "skin effect". Inductive effects cause H.F. currents to flow only in the surface of conductors. The skin depth (defined as the depth at which the current density has dropped to 1/e of its value at the surface) at a frequency f is
where µ is the permittivity of the conductor, and σ is its conductivity in ohm metres.
henry/metre except for magnetic materials, where µ=4µrπ×10-7
is the relative permittivity). For the purposes of resistance calculation in cases where the skin depth is less than one fifth the conductor thickness we can assume that all the H.F. current flows in a layer the thickness of the skin depth, and is uniformly distributed.
Skin effect has the effect of increasing the resistance of conductors at quite modest frequencies and must be considered when deciding if the resistance of wires or PC tracks will affect a circuit's performance. (It also affects the behaviour of resistors at H.F.)
Good H.F. analog design must incorporate stray capacitance. Wherever two conductors are separated by a dielectric there is capacitance. The formulae for parallel wires, concentric spheres and cylinders, and other more exotic structures may be found in any textbook but the commonest structure, found on all PCBs, is the parallel plate capacitor.
When stray capacitance appears as parasitic capacity to ground it can be minimised by careful layout and routing, and incorporated into the design. Where stray capacity couples a signal where it is not wanted the effect may be minimised by design but often must be cured by the use of a Faraday shield.
A Faraday shield consists of a grounded conductor placed between the signal source and its destination. It is surprising how often designers omit to ground a Faraday shield, which almost invariably makes the problem worse. It is good practice to ground any substantial conductor which would otherwise float: crystal cans, coil cans, metal semiconductor packages, and structural metal.
If inductance is to be minimised the lead and PC track length of capacitors must be kept as small as possible. This does not mean just generally "short", but that the inductance in the actual circuit function must be minimal. Fig. 6 shows both a common mistake (the leads of the capacitor C1 are short, but the decoupling path for IC1 is very long); and the correct way to decouple an IC (IC2 is decoupled by C2 with a very short decoupling path).
Any length of conductor has inductance and it can matter. In free space a 1 cm length of conductor has inductance of 7-10 nH (depending on diameter), which represents an impedance of 4-6 Ω at 100 MHz. This may be large enough to be troublesome, but badly routed conductors can be far worse as they form, in effect, single turn coils with quite substantial inductance.
If two such coils are close to each other we must consider their
mutual inductance as well as their self inductance. A change of current in one will induce an EMF in the other. Defining the problem, of course, at once suggests cures: reducing the area of the coils by more careful layout, and increasing their separation. Both will reduce mutual inductance, and reducing area reduces self inductance too.
It is possible to reduce inductive coupling by means of shields. At L.F. shields of mu-metal are necessary (and expensive, heavy and vulnerable to shock, which causes loss of permittivity) but at H.F. a continuous Faraday shield (mesh will not work so well here) blocks magnetic fields too, provided that the skin depth at the frequency of interest is much less than the thickness of the shield. In breadboards a piece of copper clad board, soldered at right angles to the ground plane, can make an excellent HF magnetic shield, as well as being a Faraday shield.
Magnetic fields are dipole fields, and therefore the field strength diminishes with the cube of the distance. This means that quite modest separation increases attenuation a lot. In many cases physical distance is all that is necessary to reduce magnetic coupling to acceptable levels.
Kirchoff's Law suggests that return currents in ground are as important as signal currents in signal leads. We find here another example of the "superconductor assumption" too many engineers believe that all points marked with a ground symbol on the circuit diagram are at the same potential. In practice ground conductors have resistance and inductance and potential differences. It is for this reason that such breadboarding techniques as matrix board, prototype boards (the ones where you poke component leads into holes where they are gripped by phosphor bronze contacts) and wire wrap have such poor performance as analog prototyping systems.
The best analog breadboard arrangement uses a "ground plane" a layer of continuous conductor (usually copper clad board). A ground plane has minimal resistance and inductance, but its impedance may still be too great at high currents or high frequencies. Sometimes a break in a ground plane can configure currents so that they do not interfere with each other, sometimes physical separation of different subsystems is sufficient.
It is often easy to deduce where currents flow in a ground plane, but in complex systems it may be difficult. Breadboards are rarely that complex, but if necessary it is possible to measure differential voltages of as little as 5 µV on a ground plane. At DC and L.F. this is done by using an instrumentation amplifier with a gain of 1,000 to drive an oscilloscope working at 5 ΩV/cm. The sensitivity at the input terminals of the inamp is 5 ΩV/cm, there will be some noise present on the oscilloscope trace, but it is quite possible to measure ground voltages of the order of 1 ΩV with such simple equipment. It is important to allow a path for the bias current of the inamp, but its common mode rejection is so good that this bias path is not critical.
The upper frequency of most inamps is 25-50 KHz (the AD830 is an exception it works up to 50 MHz at low gains, but not at X1,000). Above LF a better technique is to use a broadband transmission line transformer to remove common mode signals. Such a transformer has little or no voltage gain, so the signal is best displayed on a spectrum analyser, with µV sensitivity, rather than on an oscilloscope which only has 5 mV or so.
The final issue we must consider before discussing the actual techniques of breadboarding is decoupling. The power supplies of HF circuits must be short circuited together and to ground at all frequencies above DC. (DC short circuits are undesirable for reasons which I shall not bother to discuss.) At low frequencies the impedance of supply lines is (or should be) low and so decoupling can be accomplished by relatively few electrolytic capacitors which will not generally need to be very close to the parts of the circuit they are decoupling, and so may be shared among several. (The exception to this is where a component draws a large LF current, when a local, dedicated, electrolytic capacitor should be used.)
At HF we cannot ignore the impedance of supply leads (as we have already seen in Fig. 6) and ICs must be individually decoupled with low inductance capacitors having short leads and PC tracks. Even 2 3 mm of extra lead/track length may make the difference between the success and failure of a circuit layout.
Where the HF currents of a circuit are mostly internal (as is the case with many ADCs) it is sufficient that we short circuit its supplies at HF so that it sees its supplies as stiff voltage sources at all frequencies. When it is driving a load the decoupling must be arranged to ensure that the total loop in which the load current flows is as small as possible. Fig. 15 shows an emitter follower without supply decoupling the HF current in the load must flow through the power supply to return to the output stage (remember that Kirchoff's Law says, in effect, that currents must flow in circles). Fig. 16 shows the same circuit with proper supply decoupling.
This principle is easy enough to apply if the load is adjacent to the circuit driving it. Where the load must be remote it is much more difficult, but there are solutions. These include transformer isolation and the use of a transmission line. If the signal contains no DC or LF components it may be isolated with a transformer close to the driver. Such an arrangement is shown in Fig. 17. (The nature of the connection from the transformer to the load may present its own problems but supply decoupling is not one of them.)
A correctly terminated transmission line constrains HF signal currents so that, to the supply decoupling capacitors, the load appears to be adjacent to the driver. Even if the line is not precisely terminated, it will constrain the majority of the return current and is frequently sufficient to prevent ground current problems.
Having considered issues of resistance, capacitance and inductance it is clear that breadboards must be designed to minimise the adverse effects of these phenomena. The basic principle of a breadboard is that it is a temporary structure, designed to test the performance of a circuit or system, and must therefore be easy to modify.
There are many commercial breadboarding systems, but almost all of them are designed to facilitate the breadboarding of digital systems, where noise immunities are hundreds of millivolts or more. (We shall discuss the exception to this generality later.) Matrix board (Veroboard, etc.), wire wrap, and plug in breadboard systems (Bimboard, etc.) are, without exception, unsuitable for high performance or high frequency analog breadboarding. They have too high resistance, inductance and capacitance. Even the use of IC sockets is inadvisable. (All analog engineers should practice the art of unsoldering until they can remove an IC from a breadboard [or a plated through PCB] without any possibility of damage to the board or the device solder wicks and solder suckers are helpful in accomplishing this.)
The most practical technique for analog breadboarding uses a copper clad board as a ground plane. The ground pins of the components are soldered directly to the plane and the other components are wired together above it. This allows HF decoupling paths to be very short indeed. All lead lengths should be as short as possible, and signal routing should separate high level and low level signals. Ideally the layout should be similar to the layout to be used on the final PCB.
Pieces of copper clad may be soldered at right angles to the main ground plane to provide screening, or circuitry may be constructed on both sides of the board (with connections through holes) with the board itself providing screening. In this case the board will need legs to protect the components on the underside from being crushed.
When the components of a breadboard of this type are wired point to point in the air (a type of construction strongly advocated by Robert A. Pease of National Semiconductor [Reference 1] and sometimes known as "bird's nest" construction) there is always the risk of the circuitry being crushed and resulting short circuits, also if the circuitry rises high above the ground plane the screening effectof the ground plane is diminished and interaction between different parts of the circuit is more likely. Nevertheless the technique is very practical and widely used because the circuit may so easily be modified.
However there is a commercial breadboarding system which has most of the advantages of "bird's nest over a ground plane" (robust ground, screening, ease of circuit alteration, low capacitance and low inductance) and several additional advantages:- it is rigid, components are close to the ground plane, and where necessary node capacitances and line impedances can be calculated easily. This system is made by Wainwright Instruments in the town of Andechs in Bavaria and is available in Europe as "Mini-Mount" and in the USA (where the trademark "Mini-Mount" is the property of another company) as "Solder Mount". [Reference 2.] (There is also a monastery at Andechs where they brew what is arguably the best beer in Germany.)
Solder-Mount consists of small pieces of PCB with etched patterns on one side and contact adhesive on the other. They are stuck to the ground plane and components are soldered to them. They are available in a wide variety of patterns, including ready made pads for IC packages of all sizes from 8 pin SOICs to 64 pin DILs, strips with solder pads at intervals (which intervals range from .040" to .25", the range includes strips with 0.1" pad spacing which may be used to mount DIL devices), strips with conductors of the correct width to form microstrip transmission lines (50 Ω, 60 Ω, 75 Ω or 100 Ω) when mounted on the ground plane, and a variety of pads for mounting various other components. A few of the many types of Solder-Mount are shown in Fig. 20.
The main advantage of Solder-Mount construction over "bird's nest" is that the resulting circuit is far more rigid, and, if desired, may be made far smaller (the latest Solder-Mounts are for surface mount devices and allow the construction of breadboards scarcely larger than the final PCB, although it is generally more convenient if the prototype is somewhat larger). Solder-Mount is sufficiently durable that it may be used for small quantity production as well as prototyping: two pieces of equipment I have built with Solder-Mount have been in service now for over twenty years.
Fig. 21 shows several examples of breadboards built with Solder Mount. They are all HF circuits, but the technique is equally suitable for the construction of high resolution LF analog circuitry. A particularly convenient feature of Solder Mount at VHF is the ease with which it is possible to make a transmission line.
If a conductor runs over a ground plane it forms a microstrip transmission line. Solder Mount has strips which form microstrip lines when mounted on a ground plane (they are available with impedances of 50 Ω, 60 Ω, 75 Ω, and 100 Ω). These strips may be used as transmission lines, for impedance matching, or simply as power buses. (Glass fibre/epoxy PCB is somewhat lossy at VHF and UHF, but the losses will probably be tolerable if microstrip runs are short.)
It is important to realise that current flow in a microstrip transmission line is constrained by inductive effects. The signal current flows only on the side of the conductor next to the ground plane (its skin depth is calculated in the normal way) and the return current flows only directly beneath the signal conductor, not in the entire ground plane (skin effect naturally limits this current, too, to one side of the ground plane). This is helpful in separating ground currents, but increases the resistance of the circuit.
It is clear that breaks in the ground plane under a microstrip line will force the return current to flow around the break, increasing impedance. Even worse, if the break is made to allow two HF circuits to cross, the two signals will interact. Such breaks should be avoided if at all possible. The best way to enable two HF conductors on a ground plane to cross without interaction is to keep the ground plane continuous and use a microstrip
on the other side of the ground plane to carry one of the signals past the other (drill a hole through the ground plane to go to the other side of the board). If the skin depth is much less than the ground plane thickness the interaction of ground currents will be negligible.
It is not possible in a short note to discuss all the intricacies of successful analog breadboard construction, but we have seen that the basic principle is to remember all the laws of nature which apply and consider their effects on the design.
In addition to the considerations of resistance, skin effect, capacitance, inductance and ground current, it is important to configure systems so that sensitive circuitry is separated from noise sources and so that the noise coupling mechanisms we have described (common resistance/inductance, stray capacitance, and mutual inductance) have minimal opportunity to degrade system performance. ("Noise" in this contect means a signal we want [or which somebody wants] in a place where we don't want it, not natural noise like thermal, shot or popcorn noise.) The general rule is to have a signal path which is roughly linear, so that outputs are physically separated from inputs and logic and high level external signals only appear where they are needed.
Thoughtful layout is important, but in many cases screening may be necessary as well.
A final consideration is the power supply. Switching power supplies are ubiquitous because of their low cost, high efficiency and reliability, and small size. But they can be a major source of HF noise, both broadband and at frequencies harmonically related to their switching frequency. This noise can couple into sensitive circuitry by all the means we have discussed, and extreme care is necessary to prevent switching supplies from ruining system performance.
Prototypes and breadboards frequently use linear supplies or even batteries, but if a breadboard is to be representative of its final version it should be powered from the same type of supply. At some time during development, however, it is interesting (and frightening, and helpful) to replace the switching supply with a battery and observe the difference in system performance.
Reference 1.||Robert A. Pease "Troubleshooting Analog Circuits" Butterworth-Heinemann 1991|
|Reference 2.||Wainwright Instruments Inc., 7770 Regents Rd., #113 Suite 371, San Diego, CA 92122
(619) 558 1057 | Fax: (619) 558 1019
Wainwright Instruments GmbH, Widdersbergerstraße, 14, D-8138 Andechs-Frieding, Germany
(+49) 8152-3162 | Fax: (+49) 8152-5174
LATE NEWS - Mini-Mount is to be discontinued in Europe in May 2006 - get it NOW if you want it! I do not know if it will continue to be available in the USA.
15 March 1994
Updated 5 December 2005
Eur.Ing. James M. Bryant, MIEE
Head of European Applications