EC-13.3 Circuits and Devices on the Bench

What this is and why it exists

This stage is where a reader first meets the gap between a circuit on paper and a circuit on a board.

The gap is always in the same three places. The operating point, the source and load impedances, and the parts of the model that were ignored. Working through it once with an amplifier makes every later analog measurement faster.

It also establishes an order of operations that saves a great deal of time. Bias before signal, always.

The vocabulary

• Operating point — the steady voltages and currents with no signal applied.
• Bias — the arrangement that sets the operating point.
• Gain — the ratio of output amplitude to input amplitude at a given frequency.
• Midband — the range of frequencies over which gain is flat.
• Bandwidth — the range over which the gain stays within a stated fraction of its midband value.
• Tolerance — the manufacturing spread of a component's value.

The mental model

Building from a schematic means checking each connection against it before power is applied. A wiring error found by inspection costs a minute, and one found by smoke costs a part and a session.

The habit worth building is to trace each net on the schematic and on the board in turn, marking the schematic as you go. It feels slow the first three times and then becomes fast.

Now the rule that organises everything else in this topic. Measure the operating point first, before any signal is applied.

Take the direct voltage at every node and compare it with the design. If the bias is wrong, everything measured afterwards is a measurement of a different circuit from the one you designed. No amount of signal work will reveal that. A transistor biased into saturation will still produce an output, and it will be the wrong one.

This single ordering accounts for more saved time in analog laboratory work than any other habit.

Gain is measured wrongly in two specific ways and both are avoidable.

The first is reading the generator's display instead of the actual input. The display assumes a load the circuit probably does not present, so the input is not what it says. Measure the input at the circuit, on the same instrument, at the same time as the output.

The second is measuring at a frequency outside the flat region. A gain quoted without a frequency is meaningless if the circuit rolls off nearby. The answer will differ from the calculation for a reason unrelated to the design.

Bandwidth should be found by sweeping rather than assumed from a formula. Raise the frequency until the output amplitude falls to about seven tenths of its midband value, and record that frequency.

Doing it by hand once, point by point, makes every later frequency response plot meaningful. It also shows how gradual the roll-off is, which no single number conveys.

Loading is the next lesson and it is where the measurement module and this one meet. Connecting a probe or a load changes the circuit, and the change is largest where the impedance is highest.

Measure with and without a load, predict the difference from the output impedance, and check. Turning an error into a prediction is what makes it a result rather than a nuisance.

Finally, accounting for the difference. A measurement ten per cent from the prediction is normal and needs an explanation rather than a shrug.

Four candidates cover nearly all of it. Component tolerance, which is often five per cent per part and compounds. A parameter left out of the model, such as a transistor's early effect. Loading by the instrument. And temperature, particularly on anything carrying current.

Naming which one, and checking, is the actual laboratory work.

What you should now be able to explain or do

• Build from a schematic and verify the wiring before applying power.
• Measure and check the operating point before looking at any signal.
• Measure gain correctly, at the circuit and within the flat region.
• Find a bandwidth by sweeping and describe the shape of the roll-off.
• Predict and then measure the effect of loading the output.
• Account for a difference between measurement and calculation from four usual causes.

Check yourself