Breadboarding & Prototyping

Intuition

A strip of five sockets

A breadboard is a plastic block full of holes with spring clips underneath, and the clips are what you are actually wiring to. Five holes in a line share one clip. Push two component leads into any two of those five and they are connected, without solder and without a wire between them. Five chairs at one table: whichever chair you take, you are sitting at the same table as everyone else.

The rest of the board follows from that. Short strips run in two banks, and a channel down the middle separates them so a dual-in-line package can straddle it with each row of pins landing in strips of its own. Along the outside edges run the rails, and a rail is one clip as well, except that it stretches the whole length of the board instead of five holes.

The dimensions are not arbitrary. Holes sit on a 2.54 mm pitch, the 0.1 inch grid the entire through-hole component world is built to, so a resistor's lead spacing, an integrated circuit's pin spacing and the board's holes all agree without anyone having to arrange it. A strip runs 10.16 mm from its first socket to its fifth, all 5 of them one node. The channel is 7.62 mm across. Count the strips in the segment drawn below, add the rails, and there are 36 separate nodes to wire between.

The holes you can see are not the connections. The clips underneath are, and they run the short way in the banks and the long way in the rails.

Practitioner

Laying out a circuit you can still read

A board rewards a habit or two and punishes improvisation. The layout worth having is the one you can still trace three days later with a meter in your hand.

1. Power first. Bring the supply to the rails before anything else goes in, and settle which rail is which. Most boards print a red line and a blue one along the edges; use them the same way every time.
2. Chips across the channel. Seat every dual-in-line package straddling the centre gap, pin 1 at the same end each time, so a probe never has to guess.
3. Short jumpers, flat to the board. A long wire arching over the circuit is awkward to trace and behaves as an aerial. Cut solid-core wire to length instead of reaching for a pre-formed lead of the wrong size.
4. One node, one strip. Where a node needs more connections than a strip has holes, run a jumper to a second strip rather than forcing two leads into one hole.
5. Check whether the rails split. Many boards break each rail into halves partway along, with a gap in the printed line as the only warning. A jumper across that gap costs nothing.
6. Leave room to probe. Every node you expect to measure wants one hole in its strip kept free for a probe tip.

None of that touches the electrical question, which is whether the rails carry what the circuit draws. They do, up to a point, and the point is worth working out rather than guessing at.

Take a load some way along the rail from where the supply enters. Three resistances sit in that path: the contact where the supply lead grips its clip, the rail's own metal between there and the load, and the contact at the far end where the load lead grips. Call one contact 12 mΩ and the strip 3.0 mΩ per socket of run. Both are illustrative figures for the arithmetic below rather than a specification for any board you can buy.

Steady current is the easy case, and this is the easy case drawn.

For most bench work that settles it. A logic chip drawing the same average current in short bursts asks a different question, and Layer 4 answers that one.

Engineer

The parts the board adds to your circuit

Every joint is a resistance

A soldered joint is a metal-to-metal bond. A breadboard contact is a spring pressing on a lead, and that pressure is the entire connection. It works well — and it does not work perfectly.

A part with one lead in one strip and its other lead in a second strip meets two contacts, so 24 mΩ before the part itself. A jumper between two strips meets two more, plus its own wire. Copper is the easy part to be exact about: at a resistivity of 16.8 nΩ·m, a 100 mm length of 0.644 mm solid wire has a cross-section of 0.326 square millimetres, and

gives 5.16 mΩ for the wire. The jumper link end to end is 29.16 mΩ, and the wire is the smaller half of it. Soldering the same wire at both ends, at an illustrative 1.5 mΩ per joint, brings the link to 8.16 mΩ. Soldering basics covers what makes a joint that good.

A logarithmic axis, so no bar height here stands in proportion to any other. The wire is barely involved — the contacts do all of it.

At the currents most prototypes draw, none of that is a problem. The problem is that a contact does not stay at 12 mΩ forever. Clips lose their spring, plating oxidises, and a hole that has taken a component lead too thick for it never grips a thin one properly again. Suppose one contact in the supply path has drifted to 1.20 Ω.

At the same 240 mA that one contact drops 288 mV and dissipates 69.1 mW, both of them arithmetic on the resistance rather than a measurement. There is the intermittent breadboard fault in a line — nothing is broken, everything is connected, and a quarter of a volt has gone somewhere you cannot see. Press on a jumper and the circuit recovers, which is a diagnosis, not a repair.

The strip next door

Any two pieces of metal a few millimetres apart, running parallel for a centimetre, make a capacitor. A small one — the figure this lesson works with is 3.5 pF between one strip and its neighbour, illustrative rather than measured. A strip has a neighbour on each side, so

puts 7.0 pF on every strip before anything is plugged into it. Spread one signal across 4 strips, which a circuit with a few jumpers does without trying, and the node carries 28.0 pF. Let the same signal sprawl over 8 and it carries 56.0 pF.

Whether that matters depends on what drives it. A source with 50 Ω of output impedance sees

which is 1.40 ns for the tidy net and 2.80 ns for the sprawling one, and

turns those into 3.08 ns and 6.16 ns of rise time contributed by the board alone.

Put a real edge through it. Model the driver's own transition as a straight ramp reaching the rail in 10.0 ns, so its 10 to 90 per cent time is the 8.0 ns a datasheet would quote. Solve that ramp against the row capacitance and the edge arrives at 8.39 ns across the tidy net and at 10.05 ns across the sprawling one. The first is a rounding error. The second has lost a fifth of the edge to the board itself.

Same driver, same edge, three different amounts of strip hanging off it.

Seen in the frequency domain, the same capacitance is a low-pass filter nobody designed:

113.7 MHz for the tidy net, 56.84 MHz for the sprawling one. Both sit far above anything a beginner's prototype runs at, and that is the honest reason breadboards work as well as they do. What they do not sit far above is the spectrum inside a fast edge, which a logic family puts there whether the clock is fast or not. Pulse response and rise time works out the relation between an edge and the bandwidth it needs.

The corner moves down every time a signal picks up another strip.

Professional

Loop area, and when to stop breadboarding

Layers 2 and 3 both treated the supply as a pair of conductors with resistance. At the instant a chip switches, it is something else.

Current returns. Whatever leaves the positive rail comes back along the negative one, and the path between them encloses an area. That area has inductance. The figure this lesson uses for it is 1.0 nH/mm of loop, a rule of thumb rather than a specification, and the arithmetic below inherits that status.

Put a decoupling capacitor 2 columns from the chip and the loop out to it and back measures 25.40 mm, worth 25.4 nH. Park the same part at the end of the rail, 24 columns away, and the loop becomes 137.2 mm and 137.2 nH.

Inductance only sends a bill when the current changes. Switch 240 mA in the 8.0 ns edge from Layer 3 and the rate is 30.0 MA/s. The short loop then develops 762 mV and the long one 4.11 V, both of them from the loop length and the stated inductance per millimetre rather than from an instrument. On a 5.0 V rail the second figure is more than four fifths of the supply, which in practice means the rail sags and rings rather than holding while the arithmetic asks for the impossible. Every chip sharing that rail sees it.

The shading is the area the return current has to go round. Only the capacitor moved between the two panels.

The capacitor and its loop are also a series resonant circuit, which is the part that catches people out. With 100 nF across the rails,

gives 19.84 Mrad/s for the short loop, or 3.16 MHz in ordinary frequency, against 1.36 MHz for the long one. Above its own resonance the pairing behaves as an inductor, so the far placement gives up at less than half the frequency the near one reaches. Placement beats capacitance here — and placement is what a breadboard gives you least of.

The board is honest about DC and about audio, and honest enough about a few megahertz. It stops being honest when a circuit contains an edge fast enough to notice the row capacitance, or a supply loop wide enough to notice the switching current. A node quiet enough to notice the strip next door is the third case, and that one catches analogue work rather than digital. Radio circuits, switching regulators, crystal oscillators and analogue front ends hunting microvolts all want a soldered prototype, on stripboard or on a plated-through board where the return path can run under the signal instead of around the outside of it.

There is a measurement trap folded into all of this. A scope probe on a breadboard node adds its own capacitance to the strip's, so a marginal circuit gets worse the moment you look at it, and a circuit that only behaves with the probe attached has told you something. Oscilloscope probes and 10x attenuation has the arithmetic for that, and using a multimeter covers the DC half of the same job.

One habit repays itself constantly and costs nothing: photograph the board before pulling it apart. A prototype that worked last month always contains one jumper you cannot account for, and a phone picture beats the schematic you meant to draw.