Build a Driver Board

One board, built from a schematic you can read, tested against a document you wrote, and kept. You will use it in Class 12 and again in the capstone.

What you are building#

A four-channel isolated low-side switch. In plain terms: four logic inputs, four outputs that can each switch several amps of DC load, with the control side electrically separate from the load side.

It is a real thing. Boards exactly like this drive practical lamps in scenery, solenoid catches, smoke machine triggers, small motors and LED tape all over this industry.

Reading the schematic, stage by stage#

Each channel is five parts in a chain. Understand the chain once and you understand all four.

1. Input and pull-down. The signal arrives from a microcontroller or a control system. A 10 kΩ resistor from the input to ground is the pull-down, and it exists to answer one question: what does this output do when nothing is connected to the input?

Without it, the input floats, picks up interference and switches randomly. With it, a disconnected input is a definite off. A floating input is not off, it is undefined, and undefined behaviour in a show is the thing you are being trained to prevent.

2. The optocoupler. The input current lights an internal LED through its own series resistor. On the far side, a phototransistor conducts. There is no electrical path across it: that gap is the whole point of the board.

3. Gate resistor and gate pull-down. A small series resistor, 100 to 220 Ω, limits the current spike into the MOSFET's gate capacitance. A 10 kΩ pull-down from gate to source guarantees the MOSFET is off when nothing is driving it — the same argument as stage one, applied at the dangerous end.

4. The MOSFET. Logic-level, low R_DS(on), switching the low side of the load: the load connects to the positive rail, and the MOSFET connects the load's other end to ground.

Low-side switching is simpler because the gate reference is ground. Its cost is that the load sits at supply potential even when off, which is a real safety consideration and one of the things your documentation must state.

5. Flyback diode and indicator. A diode across the output terminals, cathode to the positive rail, catching the inductive kick from any load. An LED and its resistor across the output so you can see the channel's state without a meter.

The bill of materials#

Two decisions in that table are worth arguing about, and we will: why a Schottky rather than a 1N4007, and why 220 Ω rather than 1 kΩ on the optocoupler. Come with an opinion.

Build order, and why it is this order#

Build order is not arbitrary. It is chosen so that each stage can be tested before the next one hides it.

1. Lowest components first. Resistors and diodes lie flat and are held by the board when you turn it over. Tall parts last, or the short ones fall out.
2. Passives before semiconductors. Semiconductors are heat-sensitive and static-sensitive, and you do the most fumbling early.
3. Test the supply before fitting anything active. Fit the terminals and the decoupling, apply power, and confirm the rails are where you think. If a rail is wrong, you have destroyed nothing.
4. One channel, complete, tested. Then the other three. Building all four and testing at the end means a mistake is repeated four times before you find it.
5. Indicators last. They are the easiest to fit wrong and the least important.

The habits that separate a good bench from a slow one#

• Check orientation before you solder, on every polarised part. Diodes, LEDs, electrolytics, optocouplers, MOSFETs. Desoldering an eight-pin part without lifting a pad is a genuinely difficult operation and there is no reason to learn it today.
• Clip leads after each joint, not at the end. A clipping that falls onto a finished board and bridges two pads is a fault you will spend an hour on.
• Inspect after every ten joints, under the loupe. Finding a bridge at joint 12 costs a minute; finding it at joint 60 costs an hour.
• Do not power a board you have not inspected. Look at the underside with the loupe, at an angle, with a light behind it. Bridges are visible from one angle and invisible from another.

Testing: the part that is actually assessed#

Anybody can assemble a board. What is being assessed today is whether you can prove it works, in an order that finds faults safely.

Test 1 — Dead tests, before any power#

With the board unpowered and nothing connected:

• Resistance across the supply terminals. Expect high, at least tens of kilohms, settling upward as the electrolytic charges from the meter. A low reading here means a short, and powering it would turn a five minute fix into a destroyed board.
• Resistance between control ground and load ground. Expect open. This is the isolation test, and it is the single most important measurement on the board.
• Continuity from each MOSFET drain to its output terminal. Expect near zero.
• Resistance from each gate to source. Expect about 10 kΩ, which proves the pull-down is fitted.

Test 2 — Powered, no load, no input#

• Supply current with everything off: a few milliamps at most. Anything above that means a channel is partly on.
• Every output should read the full supply voltage relative to load ground, because the MOSFET is off and the load path is open.
• Every gate should read near zero.

Test 3 — One channel at a time, with a load#

Fit a resistive load. Apply the control input. Confirm:

• The LED lights.
• The output pulls down to near zero.
• V_DS while on is small. Calculate R_DS(on) from V_DS and the current and compare it to the datasheet. If it is far higher, the gate is under-driven and you have a problem to find now rather than in a performance.
• The MOSFET is cool after two minutes. Warm is a question. Hot is a fault.

Test 4 — The flyback test#

Swap the resistive load for a relay coil. Scope across the MOSFET's drain and source, and switch it.

Photograph the trace. Then lift the flyback diode, repeat, and photograph again. You did this on breadboard in Class 5; on your own board it means something different, because now it is your design decision being validated.

Test 5 — Isolation, at voltage#

The technician tests the isolation barrier with an insulation tester at 500 V. You watch. A board that fails this test is not a board with a fault, it is a board that is dangerous, and the difference in language matters.

Document it#

Every board leaves this room with a single sheet, and the sheet is marked. In the industry this sheet is the difference between a piece of equipment that can be used on a production and a mystery object in a flight case.

It carries:

1. What it is and what it is for. Two sentences.
2. Ratings. Supply voltage range, maximum current per channel, maximum total, isolation rating.
3. The pinout, drawn, with the connector orientation shown.
4. The behaviour on loss of control signal. Yours fails to off, because of the pull-downs. Say so explicitly: this is the sentence a production manager actually reads.
5. Test results, with the numbers you measured, dated and initialled.
6. Known limitations. Yours has at least two: the load sits at supply potential when off, and there is no over-current protection. Write them down. A limitation you have named is engineering; the same limitation undocumented is a defect.

Where the parts go, and why it is not arbitrary#

The schematic says what connects to what. The layout says where the current actually flows, and on a board switching several amps that is a separate question with its own answers.

Four rules, in the order they matter on this board:

Keep the high-current loop small and fat. Supply positive, through the load terminal, through the MOSFET, back to supply negative. That loop carries five amps switching in microseconds, and its physical area is the antenna radiating interference into everything else. Short and wide.

Keep the control ground away from the load ground. They are not connected on this board, and the layout has to make that visible. A copper pour that quietly bridges under the optocoupler defeats the isolation while every component is still working perfectly.

Decoupling goes at the chip, not at the connector. A 100 nF capacitor eight centimetres from the part it is decoupling is decoupling the track, not the part.

Put the heat where it can leave. The MOSFET's tab is its thermal path. Copper attached to it is a heatsink; a tab in free air with a via under it is not.

For veroboard and tag strip this becomes: plan the current path before you cut a track, and run the high-current return as its own copper rather than sharing it with the control ground.

The sheet that goes in the case#

Every board leaves this room with one side of A4, and it is marked. In the industry this sheet is the difference between a piece of equipment that can be used on a production and a mystery object in a flight case.

Ten points, and point eight is where the marks are won and lost:

1. What it is and what it is for. Two sentences.
2. Ratings: supply range, current per channel, maximum total, isolation rating.
3. Pinout, drawn, with the connector orientation shown.
4. Behaviour on power-up.
5. Behaviour on loss of the control signal.
6. Fault conditions detected, and the indicator code for each.
7. Reset procedure.
8. Known limitations. Yours has at least two: the load sits at supply potential when off, and there is no over-current protection.
9. Test results, as numbers, dated and initialled.
10. Who to contact.

A limitation you have named is engineering. The same limitation undocumented is a defect. A production manager who reads "the load sits at supply potential when off" can plan around it. One who discovers it is having a different conversation, at a worse time, about you.

At the bench#

Three hours and twenty minutes, essentially all of it building and testing.

Block A — Read and check (25 min)#

Compare your pre-class resistor calculations with the group. Walk the schematic stage by stage. Identify, before anybody picks up an iron, which single component's omission would make the board dangerous rather than merely broken.

Block B — Build (100 min)#

In the order above. Channel one tested before channels two to four are populated.

Block C — Test and record (50 min)#

All five tests, all results written down as numbers rather than ticks. "Pass" is not a measurement.

Block D — Break somebody else's, then fix your own (25 min)#

Swap boards. The technician has introduced one fault into each: a reversed diode, a missing pull-down, a bridged pad, an under-driven gate. Find it using the test sequence, not by looking at the board, and write down which test caught it.

Then get your own board back and re-run the test that would have caught the fault you were given.

Numbers from this class#

Self test with model answers

Answer it yourself first, out loud or on paper, then open the answer. Reading the answer without attempting the question teaches you almost nothing.

1. 1Which single omitted component would make your driver board dangerous rather than merely broken, and why?
2. 2Why is the control-ground to load-ground resistance the most important measurement on the board?
3. 3Why test one channel completely before populating the other three?
4. 4Your board’s documentation must state the behaviour on loss of control signal. What is yours, and why does it matter more than the ratings?
5. 5The dead test across the supply terminals reads 15 Ω. What do you do, and what would happen if you did not?

Note. These are model answers, not the only correct ones. If yours differs and you can defend it with the arithmetic and a measurement, that is worth more than matching the wording.