Half-Split: Divide and Conquer
Where you are: Module 6 — Isolation Techniques · Lesson 6.1 · Based on Chapter 6 (§6.1, Figure 6.1) of Dr. Tahseen Al Doori's Troubleshooting Essentials.
What you'll be able to do
- Explain what it means to isolate a fault and name the book's six isolation techniques.
- Apply the half-split method to a chain of components, recording each test and its result.
- Calculate the maximum number of half-split tests for a chain of a given length, showing your reasoning.
- Recognise the half-split idea in other techniques used later in the course.
Watch: Half-Split: Divide and Conquer
The video animates the book's Figure 6.1: eight components between a power socket and a cloud server, one of them faulty, and three tests that corner it. It then shows why the method scales so well. As you watch, think about:
- Where is the first test placed, and why there?
- After each test, which half is kept and which is discarded?
- How many tests would a chain of sixteen components need?
From quick wins to isolation
When the five quick-win checks (Module 5) do not resolve a problem, the obvious causes have been ruled out and the next task is to isolate the fault. The book defines this as shrinking the area in which the fault could lie until only one cause remains. Isolation is how Steps 2 and 3 of the seven-step method (theory and test) are carried out efficiently on a system with many parts.
Chapter 6 describes six isolation techniques that apply to almost any system:
Half-split
Test the midpoint; keep the half that holds the fault. (This lesson.)
Substitution
Swap a suspect part for a known-good one. (Lesson 6.2.)
Comparison
Find the difference from an identical working system. (Lesson 6.2.)
Following the path
Bottom-up, top-down or follow the flow. (Lesson 6.3.)
Remove and restore
Strip to the minimum, then restore items one at a time. (Lesson 6.2.)
Rollback
Undo the last change. (Lesson 6.2.)
The half-split method
Half-split, also called divide and conquer, works on any system that can be seen as a chain: a series of components or stages that a signal, power or data passes through in order. The glossary defines it as testing the midpoint of a system to find which half contains the fault.
- Divide the system into two halves.
- Test the point between them: is the signal correct at that point?
- Decide. If the first half works (the signal is correct at the midpoint), the fault must lie in the second half. If the signal is already missing at the midpoint, the fault is in the first half.
- Repeat on the half that still contains the fault, until only one component remains.
Because each test removes half of the remaining possibilities, the method is extremely efficient.
Worked example: Figure 6.1, test by test
The book's Figure 6.1 shows a chain of eight components between a user's power supply and a server on the internet:
Power → PC → NIC → Cable → Switch → Router → ISP → Server
(A NIC, or network interface card, is the network adapter inside the PC; the ISP is the internet service provider.) (Fictional framing:) a desktop PC cannot reach the company's cloud server, other PCs in the office can, and the quick-win checks changed nothing. One of the eight components is faulty.
| Test | Where the test is made | Result | Suspects remaining |
|---|---|---|---|
| — | Before testing | — | All 8: Power, PC, NIC, Cable, Switch, Router, ISP, Server |
| 1 | The midpoint: the Switch | Fails | The left half, up to and including the Switch: Router, ISP and Server are cleared |
| 2 | The midpoint of what is left: the NIC | Passes | 2: Cable, Switch (everything up to the NIC works) |
| 3 | The Cable | Fails | 1: Cable |
Three tests, eight components, one fault located, exactly as Figure 6.1 shows. Notice the rule behind each step: a test that passes clears everything up to that point, and a test that fails tells you the fault lies at or before it. In practice, you would then confirm the diagnosis by swapping in a known-good cable (substitution, Lesson 6.2) before declaring the cause found.
The maths: why it is so efficient
Each test halves the number of suspects, so the question "how many tests, at most?" becomes "how many times must I halve the chain to reach one?"
| Components | Halvings | Maximum tests |
|---|---|---|
| 2 | 2 → 1 | 1 |
| 8 | 8 → 4 → 2 → 1 | 3 |
| 16 | 16 → 8 → 4 → 2 → 1 | 4 |
| 1,000 | 1,000 → 500 → 250 → 125 → 63 → 32 → 16 → 8 → 4 → 2 → 1 | 10 |
A compact way to say this: the maximum number of tests is the smallest whole number t for which 2 multiplied by itself t times is at least the number of components. Since 2 × 2 × 2 = 8, eight components need 3 tests; since 2 multiplied by itself ten times is 1,024, a chain of 1,000 needs only 10. (Mathematically, this is log₂ of the number of components, rounded up; the book's figure notes log₂ 8 = 3.) When a half has an odd number of items, as with 125 above, one "half" is a little larger; round up and carry on.
Compare checking components one at a time from one end. In the worst case you might test every component but the last before you are sure: 7 tests for 8 components, 15 for 16, and 999 for 1,000, against half-split's 3, 4 and 10. The longer the chain, the bigger the saving.
Using half-split well
- You need a test point. Half-split depends on being able to check the signal somewhere between the halves: a light on a switch port, a test page, a command, a meter reading. Choose test points that are safe to reach; never open equipment to create one unless you are trained and authorised.
- The halves need not be exact. If the true midpoint is hard to test, pick the nearest point that is easy and safe. You lose a little efficiency, not correctness.
- Record every test. Note where you tested and what you found. The book's lab asks for exactly this, and Step 7 of the method depends on it. If a result surprises you, your notes show where the reasoning went.
- One fault at a time. The method assumes a single fault. If the results contradict each other, there may be more than one problem, or an intermittent one; treat the first confirmed fault, retest, and continue.
You will meet half-split again
The book returns to this idea repeatedly. In Chapter 9, asking whether one device or all devices are affected "halves the problem immediately". In Chapter 7, the five-command connectivity sequence tests one more part of the network path at each step, which the book describes as the half-split method applied to the network path. And in Chapter 11, printing a test page from the printer itself separates printer faults from computer or network faults.
Try it
Goal: practise half-split as a paper exercise.
- Draw a chain of 16 boxes and label them A to P. Ask a partner to choose one box secretly as "faulty" (or choose by rolling dice and recording it out of sight).
- Find it using half-split. For each test, name the test point (for example "between H and I") and ask your partner "is the signal OK here?". Your partner answers "OK" if the faulty box lies after that point, or "missing" if it lies at or before it.
- Record each test, its result and the remaining suspects in a table like the one in the worked example.
- Repeat three times with different faulty boxes.
You have finished when you have located the faulty box in every round using no more than four tests, and your table shows the suspects halving after each test.
Knowledge check
What is the main purpose of isolation techniques in troubleshooting?
Using half-split, what is the maximum number of tests needed for a chain of 16 components?
In a chain of eight components, the signal is correct at the midpoint. What do you conclude?
About how many half-split tests, at most, would a chain of 1,000 components need?
Answers explained
- A. The book defines isolation as shrinking the area where the fault could lie until only one cause remains. Replacing everything is guesswork, documentation is Step 7, and ordered restarts are a quick win.
- C. Each test halves the suspects: 16 → 8 → 4 → 2 → 1, which is four tests. Sixteen and eight describe one-by-one checking or a single halving; two tests only cover four components.
- B. If the first half works, the fault must lie in the second half. The system still fails, so there is a fault, and testing every component individually throws away the information you just gained.
- A. Ten halvings cover up to 1,024 components. 500 is one halving; 999 is the worst case of checking one by one.
Key takeaways
- Isolation shrinks the area where a fault can lie until only one cause remains; the book gives six techniques.
- Half-split: test the midpoint, keep the half that still holds the fault, and repeat.
- Each test halves the suspects: 8 components need at most 3 tests, 16 need 4, and 1,000 need only 10.
- Record every test point and result; half-split reappears in network, printer and connectivity troubleshooting.
Sources
- Dr. Tahseen Al Doori — Troubleshooting Essentials, University Edition (2026), Chapter 6 introduction, §6.1 and Figure 6.1; Chapter 6 Review Question 1; Glossary (Half-split); cross-references to §7.3, §9.1 and §11.1
