Gains:
- Ability to produce scripts that clean and visualize signal, log and telemetry data with Python, with AI support
- Ability to recognize data quality, overlearning and extrapolation risks when building simple prediction and anomaly models
- Ability to verify AI-generated code and result with unit control, reference data, and engineering plausibility
Module Exam
1. An intern asks the AI chat tool for the required supply current for an LNA (low noise amplifier) and writes the resulting value of '15 mA' directly into the power budget table. What is the fundamental mistake in this approach?
- A) The current value must come from the data sheet and operating condition of the selected component; The number generated by the AI may be an unsourced fabrication ✔
- B) It should have been asked to AI in µA instead of mA
- C) The value is correct, only the unit spelling should have been corrected
- D) It is enough to ask the AI the same question three times and take the average
Explanation: The language model does not know the datasheet of the specific component you have selected; produces a number that seems most likely (hallucination). The supply current comes only from the data sheet and operating condition of the part used; AI output cannot enter the table without validation.
2. What does 'risk-based classification' enable in the use of AI in electronics/communications work?
- A) Reducing the subscription fee of the AI tool
- B) Determine the role of AI and the mandatory depth of verification according to the risk level of the task ✔
- C) Prompts should be written shorter
- D) Automatically delegate all decisions to AI
Explanation: Not every mission is at the same risk level. Classifying the task as low/medium/high/critical determines which output can be freely used and which requires measurement, simulation and qualified engineer approval.
3. What does the Nyquist sampling criterion say in signal processing and what happens if it is violated?
- A) The sampling frequency must be at least twice the highest frequency; Otherwise, aliasing will occur and the signal will be irreversibly degraded ✔
- B) The lower the sampling frequency, the higher the resolution
- C) Sampling is only required for analog signals, not digital
- D) The Nyquist criterion only determines the noise level
Explanation: To accurately recover a signal, the sampling frequency must be at least twice the highest frequency in the signal. If this condition is not met, high frequencies appear as low frequencies; This is called aliasing and the data is permanently corrupted.
4. What is the best approach when having AI suggest a PCB layout?
- A) Sending the AI suggested layout directly to production
- B) Evaluate the proposal as a checklist/startup to be verified by DRC, simulation and prototype measurement ✔
- C) Copying AI text without ever opening the data sheets
- D) Ignoring signal integrity rules because AI rejects them
Description: AI can recall general layout and signal integrity principles (short return path, reference plane, decoupling capacitor placement); however, the final layout is verified by DRC (design rule checking), simulation and prototype measurement. The AI output is a checklist and starting point.
5. What does 'EIRP' mean in RF link budget?
- A) Effective radiated power, which combines the power of the transmitter and the antenna gain by deducting losses ✔
- B) Absolute value of the noise floor at the receiver
- C) Physical size of the antenna
- D) Characteristic impedance of the cable
Explanation: EIRP (Effective Isotropic Radiated Power) is the effective power radiated by an isotropic antenna, combined with the output power of the transmitter and the antenna gain, minus the cable/connector losses. Link budget and regulatory limits largely depend on the EIRP; the final value is verified by measurement and regulation.
6. What does it mean when AI finds an 'anomaly' in temperature and current data in predictive maintenance?
- A) Proof that the equipment has definitely failed
- B) A warning that the data deviates from normal; Not a definitive fault diagnosis, but a sign that needs to be confirmed by field inspection ✔
- C) Order that the equipment should be automatically shut down
- D) Guarantee that data was recorded incorrectly
Explanation: An anomaly is a deviation of data from normal behavior and may indicate a malfunction; but it is not a definitive diagnosis. The decision to disable or replace a card/equipment is made with physical control on site and the approval of the operations manager.
7. Which is mandatory for security/stability before loading an AI-written firmware interrupt service routine (ISR) code into hardware?
- A) Deploy directly to devices in the field without reading the code
- B) Just check that it is compiled and install it
- C) Reading the code, passing it through static analysis, checking shared data/timing security and testing it on hardware ✔
- D) Relying on the small number of lines of the output
Explanation: AI code may skip protecting shared variables (volatile/atomicity), put long/blocking operation in ISR, or use wrong register. It is essential to read the code, run it through static analysis, and test it in real conditions on hardware; ISR can be security-critical.
8. What is the typical relationship between BER (bit error rate) and SNR (signal-to-noise ratio) in digital communications and what should AI remember when interpreting it?
- A) The two always increase together
- B) There is no relationship between them
- C) Generally, as SNR increases, BER decreases; actual values depend on modulation, channel and coding and must be verified by simulation/measurement ✔
- D) The relationship is constant and has the same numerical value in all systems
Explanation: Generally, as SNR increases, BER decreases (error decreases as the signal becomes stronger relative to noise). AI can explain this general trend and theoretical curves, but actual values depend on modulation type, channel, coding and hardware and must be verified by simulation/measurement; The general curve does not replace the system-specific truth.
9. What is the most appropriate use when solving an I2C communication problem (device not responding) with AI?
- A) List possible causes in a list of hypotheses; Making a definitive diagnosis with analyzer/oscilloscope measurement ✔
- B) Accept the first reason given by the AI as a definitive fault and replace the hardware.
- C) Loading the code suggested by the AI many times without ever measuring the line
- D) Skipping analyzer measurement and relying solely on AI prediction
Description: AI can extract possible causes (wrong address, lack of pull-up resistor, speed mismatch, lack of common ground, contention) as a systematic list of hypotheses. However, definitive diagnosis is made by measuring the line with a logic analyzer or oscilloscope; The AI output is a search map, not a definitive diagnosis.
10. In cellular network optimization, what is the most accurate attitude towards a 'deterioration' sign that the AI finds in the RSRP/SINR trend?
- A) Changing live network parameters directly based on AI output
- B) Ignore the warning and continue watching
- C) Verify the root cause through field measurement/drive testing and network engineer analysis, then decide ✔
- D) Relying only on AI and canceling field measurement
Explanation: It may early signal a possible coverage/interference issue in the AI KPI trend; however, the root cause (antenna tilt, neighboring cell interference, hardware failure) is determined by field measurement/drive testing and analysis by the network engineer. The AI output is a scan alert, not a parameter change decision.
11. What is the most accurate approach to dealing with a solution suggested by AI (e.g. adding ferrite beads) in an EMC/EMI study?
- A) Declaring the product compatible directly after applying the AI recommendation
- B) Consider the suggestion as a possible solution hypothesis and verify it with measurements according to the standard limit in an accredited laboratory. ✔
- C) Fixing the standard limits with the values given by the AI
- D) Skipping measurement and relying only on simulation
Description: AI can recall possible interference paths and typical solutions (filtering, shielding, grounding, cable management). However, the decision on suitability is made by measurement in an accredited laboratory according to the standard limit (e.g. CISPR/EN); The AI output is an initial hypothesis, not a statement of fitness.
12. What is the most robust way to prevent AI from getting a made-up (hallucinatory) reference when you ask it for the part/clause number of a standard such as 3GPP or IEEE 802.11?
- A) Relying on the item number and table value given by AI
- B) Asking the same question again with different words
- C) Looking at a forum post instead of the standard
- D) Opening and confirming each reference from the current official text of the standard (correct Release/version) and obtaining authorized approval when necessary ✔
Explanation: Even if the language model correctly knows the name of the standard, it may hallucinate the section number, table value, and text; Additionally, values vary between releases. Each reference must be opened and confirmed from the current official text of the standard and, where necessary, authorized person approval must be obtained.
13. What is the most critical check before running a Python signal analysis script written by AI?
- A) Reading the code and verifying the sampling/unit matching, testing with a known reference signal and checking the plausibility of the result ✔
- B) Running the code directly on the entire data set without reading it
- C) Just seeing that there is no mistake and accepting the result
- D) Relying on the graphic of the output to be visually beautiful
Description: AI code may incorrectly assume the sampling frequency, mix units (Hz/kHz, V/mV), skip windowing/normalization, or mismatch columns. It is essential to read the code, verify the unit and sampling parameters, test with a known reference signal, and check whether the result is physically plausible.
14. What is the best course of action in terms of privacy and intellectual property when providing design files or customer network data to a cloud-based AI tool?
- A) Pasting all raw design/mesh data as is
- B) Anonymize context and clean sensitive design/customer/measurement data and use corporate, policy-compliant tooling ✔
- C) Assuming that providing real customer and IP information is essential for result quality
- D) Ignoring privacy and intellectual property policy at all
Description: Schematic/PCB files, firmware source, customer topology and subscriber/measurement data are sensitive in terms of trade secret and legislation (KVKK/GDPR). It is necessary to anonymize the context, clear real IP/MAC/coordinate/client names, choose corporate and policy compliant tool and comply with company policy.