Gains:
- Ability to break down a circuit problem into assumptions, formulas and unit steps and ask the AI in a structured way
- Ability to verify AI recommended resistance, capacitance and gain values against standard series and power limits
- Ability to catch arithmetic and topology errors by cross-checking AI output with SPICE or hand calculation
Circuit analysis is the backbone of electrical and electronic engineering: starting from Ohm's and Kirchhoff's laws and extending to dividers, filters, amplifiers and supply circuits. AI can be a powerful aid in these tasks; It creates formulas, generates the first number, suggests alternative topologies, and even writes a SPICE netlist. But AI doesn't "understand" a circuit; completes the pattern. Therefore, the resistance value it recommends may not be in the standard series, it may exceed the power limit, or it may set up the topology incorrectly. In this unit, we will see how to ask the circuit problem to the AI in a structured way, how to verify the output with hand calculation and SPICE, and the most common pitfalls.
Setting the Problem Correctly: Assumption, Formula, Unit
Before throwing a circuit problem at AI, break it down into three parts: knowns (inputs), wanted (output), and constraints (power, voltage, standard series). This distinction both clarifies the prompt and makes the output controllable.
A classic example: series resistor to drive an LED from a 5 V supply. Known: supply 5 V, LED forward voltage V_F ≈ 2.0 V, desired current I = 10 mA. Search for: R. The formula is simple:
R = (V_supply − V_F) / IR = (5.0 − 2.0) / 0.010 = 300 ΩPower loss of the resistor: P = I² · R = (0.010)² · 300 = 0.03 W → 1/8 W is sufficient
300 Ω is available in the E24 standard series. If not, we would round to the nearest standard value (e.g. 330 Ω) and check how the current changes: I = (5−2)/330 ≈ 9.1 mA. This is the exact flow to follow when validating AI output: formula → number → standard series → power check.
Tip: When you get a resistance or capacitance value from AI, always ask two things: "Is this value available in the E12/E24/E96 series?" and “Is the power/voltage rating of this component sufficient?” AI often returns ideal, non-produced values.
Weak Prompt / Strong Prompt
WEAK:"Calculate resistance for LED."(Result: which LED, which supply, which current is uncertain; a single number.)STRONG:"Calculate series resistance to drive a red LED at 10 mA from a 5 V supply. LED forward voltage is 2.0 V. Show SEPARATELY:1) Resistance formula and result (Ω)2) Nearest E24 standard value3) Actual current with that standard value4) Power dissipation of the resistor and recommended rating "Write units in all steps."
Verifying Resistance, Capacity, and Gain Values
The most common mistake of AI is that it returns values that are mathematically “correct” but unusable in practice. Three control areas:
Standard serial control. Resistors and capacitors are produced in E-series (E12, E24, E96...). If AI says 4737 Ω, you decide whether to use 4.7 kΩ (E24) or 4.75 kΩ (E96) based on tolerance.
Power and voltage rating. The power a resistor dissipates is as important as its value. A capacitor's voltage withstand is as critical as its capacity. AI skips these most of the time.
Gain and bandwidth. In an op-amp circuit, AI can give the gain as R_f/R_in but ignore the gain-bandwidth product (GBW) limit. If you want 100x gain with an op-amp with 1 MHz GBW, the bandwidth is only 10 kHz.
component
given by AI
Additional control
Resistance
Value (Ω)
E-series, power loss (I²R)
capacitor
Value (F)
Tensile strength, type (ceramic/electrolytic)
Op-amp gain
R_f/R_in
GBW, slew rate, feeding interval
Filter
cut-off frequency
Actual component tolerance, load effect
Cross-Validation with SPICE and Hand Calculation
Before trusting an AI-generated circuit, test it in two independent ways: hand calculation (rough check) and SPICE simulation (detailed check). The AI can even write you the SPICE netlist; but it is your job to run netlist and interpret the result.
Example of an RC low pass filter: R = 1.6 kΩ, C = 100 nF. Cutting frequency:
f_c = 1 / (2π · R · C)f_c = 1 / (2π · 1600 · 100e-9)f_c ≈ 995 Hz ≈ 1 kHz
If you scan the same circuit with AC analysis in SPICE and see that the −3 dB point is at ~1 kHz, the two paths overlap and confidence increases. If they do not overlap, there is an error in either the hand calculation or the netlist; You chase both.
* AI recommended RC filter netlist (check!)V1 in 0 AC 1R1 in out 1.6kC1 out 0 100n.ac dec 100 10 100k.end
Attention: Visually follow the connection of node names in the SPICE netlist written by the AI. The most common error is connecting a component to the wrong node or forgetting the ground (0) node. Even if the netlist appears to "work" it may be simulating the wrong circuit.
Catching Topology Errors
Sometimes the AI gives the component values correctly but sets up the topology incorrectly: connecting the feedback leg backwards, driving an op-amp from the wrong input, or forgetting the supply bypass capacitor. The way to catch these is to redraw the circuit with your own hand and follow it node by node. “That's what the AI said” does not justify a topology; You must verify the operating logic of the circuit.
Mini Case
Analog designer Kaan has an inverting op-amp circuit designed to amplify a sensor signal. AI recommends R_in = 1 kΩ and R_f = 220 kΩ; gain −220. Kaan looks at the datasheet: GBW of the selected op-amp is 1 MHz. At 220 gain the bandwidth is only left at ~4.5 kHz, whereas it needs to pass the signal up to 20 kHz. Kaan either chooses an op-amp with a higher GBW or divides the gain into two stages. AI got the math right; But Kaan reached the physical limit. Lesson: AI's gain formula may be correct, but you control the actual limits of the component from the datasheet.
Common Mistakes
- Using the value given by AI regardless of whether it is in the standard series or not.
- Ignoring resistor power dissipation and capacitor voltage withstand.
- Calculating the op-amp gain and forgetting the GBW/slew rate limit.
- Trusting the result without visually verifying the SPICE netlist.
- Bypassing the ground node or bypass capacitor.
- Accepting the topology without questioning, "This is how AI built it."
In summary
- Problem structured into AI by separating the circuit problem into assumption, sought and constraint.
- Verify each resistance/capacitance value with standard series, power and voltage rating.
- Check physical limits such as GBW, slew rate and feed gap in gain calculations.
- Verify the AI output by hand calculation and SPICE in two independent ways.
- Observe the netlist and topology node by node; The most common error is incorrect connection.
- Even if the math is correct, the engineer verifies the physical boundary and topology.
Application task
Choose a simple circuit (RC filter, voltage divider or inverting amplifier). Ask the AI for component values and a SPICE netlist. Then: (1) manually calculate the cutoff frequency/gain/output, (2) run the netlist in a SPICE tool, (3) compare the two results and whether the selected components are in the standard series. If you find a mismatch, identify its source (formula, netlist binding, or value rounding) and fix it.