Passive Selection and Package Choice

• The E series and why odd values existStandard values are logarithmically spaced so that tolerance bands just touch. E24 (5%) has 24 values per decade; E96 (1%) has 96. This is why 4.7 k exists and 5 k does not. Designing with non-standard values means paying for a special order or accepting a substitution you did not choose.E12 = 10% steps, E24 = 5%, E48 = 2%, E96 = 1%, E192 = 0.5%. Stick to E24 unless precision demands otherwise.
• Imperial package codes and what they cost you0805, 0603, 0402, 0201 describe length and width in hundredths of an inch. Smaller means denser boards and lower parasitic inductance, but lower power and voltage rating, tighter placement tolerance and harder hand assembly. 0603 is the usual sweet spot for hand-assembled prototypes; 0402 for dense production.Typical resistor power: 0805 = 125 mW, 0603 = 100 mW, 0402 = 62.5 mW, 0201 = 50 mW.
• Resistor voltage rating and pulse ratingResistors have a maximum working voltage independent of power — an 0402 is typically rated 50 V, so a 1 M resistor across 200 V fails even though it dissipates only 40 mW. High-voltage dividers must be built from series chains for exactly this reason. Pulse rating is a separate spec again.
• Temperature coefficient and driftA thick-film resistor is typically 100 to 200 ppm/C; thin film is 10 to 25 ppm/C. Over a 100 C range a 100 ppm/C part drifts 1%, which silently destroys the accuracy of a 0.1% divider. Matched pairs and resistor networks exist because ratio stability is often what you actually need.
• Capacitor selection beyond capacitanceDielectric class, rated voltage, DC bias derating, temperature coefficient, ESR, ESL and self-resonant frequency all matter, and they trade against package size. The habit to build: never specify a capacitor as just '10 uF'. Specify 10 uF / 25 V / X7R / 0805 and know why each field is what it is.
• Standardise your libraryA design that uses 47 distinct resistor values costs more to buy, stock and assemble than one that uses 12. Pick a preferred-value list and a preferred package, and deviate only with a reason. This is a real, measurable cost saving and it is entirely under your control.
• Land patterns and IPC density levelsIPC-7351 defines three density levels for the same part: Most (M, largest pads, best for hand soldering and wave), Nominal (N), and Least (L, densest). KiCad's libraries include _HandSolder variants of common footprints, which are the M-level pattern. Choosing the wrong one makes assembly harder or the board bigger than it needs to be.

ProblemA 48 V nominal battery (up to 58 V charging) must be measured by a 3.3 V ADC with better than 1% total accuracy over -20 to +60 C. Choose the divider.

1. 1
   Ratio: 58 V must map to at most 3.2 V, so the divider ratio is at least 18.1:1. Choose 20:1 for headroom: 58 V -> 2.9 V.
2. 2
   Divider current: this is a battery monitor, so standing current is a battery drain. Target 50 uA at 48 V, giving a total resistance of about 960 k. Use 950 k top and 50 k bottom for exactly 20:1.
3. 3
   E96 values: 953 k and 49.9 k give a ratio of 20.10:1. Close enough, and the exact ratio is calibrated in firmware anyway.
4. 4
   Voltage rating check: the top resistor sees nearly the full 58 V. An 0402 rated 50 V working voltage FAILS. An 0603 is typically rated 75 V, which passes but with little margin; an 0805 at 150 V is comfortable. Alternatively split the top leg into two series 475 k 0402s, halving the voltage across each — this is the standard high-voltage divider technique.
5. 5
   Power check: P_top = V^2/R = 58^2 / 953k = 3.5 mW. Trivial for any package. Voltage, not power, is the binding constraint here — the point of the exercise.
6. 6
   Tolerance: with two 1% resistors the worst-case ratio error is about 2%, which fails the 1% requirement. Options: (a) use 0.1% parts, (b) use a matched resistor network, or (c) calibrate in firmware at production and store the correction. Option (c) is usually cheapest.
7. 7
   Tempco: this is the error calibration cannot remove. Two independent 100 ppm/C thick-film parts can drift in opposite directions: over an 80 C range that is up to 1.6% of ratio error. This alone busts the budget. Choose 25 ppm/C thin film (worst case 0.4%) or a matched network with a specified ratio tempco of 5 ppm/C.
8. 8
   Impedance check: 953 k || 49.9 k = 47.4 k of source impedance at the ADC. Most SAR ADCs want under 10 k. Add a buffer, or add an RC (e.g. 1 k + 100 nF) and lengthen the sampling time — check the ADC datasheet's acquisition time requirement against the RC.
9. 9
   Final: 2 x 475 k 0402 1% 25 ppm in series (top), 49.9 k 0402 1% 25 ppm (bottom), 1 k + 100 nF RC to the ADC, firmware gain calibration.

AnswerTwo series 475 k thin-film 0402s for the top leg (voltage rating), 49.9 k thin film bottom, 25 ppm/C for tempco, firmware calibration for initial tolerance, and an RC plus long acquisition time for the ADC source impedance. Value alone would have got three of those five wrong.

Tool: KiCad footprint library and the built-in PCB Calculator

• Open KiCad and launch the Footprint Editor from the main window.
• Load Resistor_SMD:R_0805_2012Metric and note the pad dimensions and courtyard. Then load R_0805_2012Metric_HandSolder and compare. Measure the pad length difference with the ruler tool.
• Load R_0402_1005Metric and R_0201_0603Metric. Zoom to the same scale as the 0805 and look at all three together. This is the size trade you are making.
• Open the KiCad PCB Calculator (a separate tool in the KiCad launcher). Go to the 'E-Series' tab. Enter a target resistance and see the closest E24/E96 combinations, including two-resistor series and parallel combinations.
• Go to the 'Regulators' tab and compute a divider for an adjustable regulator's feedback pin. Note that it gives you E-series-friendly answers.
• Go to the 'Track Width' tab. Enter 1 A and a 10 C rise for an external layer and note the required width. You will use this in layout.
• Go to the 'Electrical Spacing' tab and find the required clearance for 58 V under IPC-2221 B4 conditions. This is the other half of the high-voltage divider problem.

• KiCad PCB Calculator documentationopen licencedoc~30 minFree and open. The E-series, track width, electrical spacing and regulator tabs are all directly useful for this skill.
• E series of preferred numbersopen licencereference~15 minPrint the E24 and E96 tables and keep them at the bench.
• Vishay / Yageo thick vs thin film resistor selection notesfreereference~45 minCompare a thick-film and a thin-film family datasheet directly: tempco, tolerance, working voltage, pulse handling and price. The differences are stark.
• IPC-7351 land pattern naming overviewfreereference~30 minUnderstand the M/N/L density levels and the naming convention. KiCad's _HandSolder suffix corresponds to the most permissive level.

Produce a standardised passive component library for a small product family, then apply it to a real specification exercise that forces at least three non-obvious package or grade decisions.

Deliverables

• A preferred-parts list: a documented set of resistor values (state which E series and why), capacitor values with dielectric and voltage, and a default package for each class, with a written justification for the defaults covering assembly method, board density and cost.
• Three specification exercises worked in full, each of which must be decided by something other than the value: (1) a divider from a rail above 50 V, (2) a decoupling capacitor for a 1.8 V core rail where DC bias derating changes the answer, (3) a current-sense shunt where power dissipation and tempco both bind.
• For each exercise: the constraint that binds, the calculation showing it binds, the chosen part with full specification (value / tolerance / tempco / voltage / package / dielectric where applicable), and a named real manufacturer part number.
• A 'rejected options' note for each exercise: the obvious choice a beginner would make and the specific number that rules it out.
• A KiCad symbol library or a spreadsheet encoding the preferred-parts list in a form you could actually reuse on the next project.
• A count of distinct part numbers your preferred list would require for a hypothetical 60-component board, compared to an unconstrained design.

Acceptance criteria — how you know it is good enough

• The high-voltage divider exercise correctly identifies working voltage, not power, as the binding constraint.
• The decoupling exercise quotes a real DC bias derating curve and shows the effective capacitance at operating bias.
• The shunt exercise includes both a power calculation and a tempco-driven accuracy calculation.
• Every chosen part is fully specified across all relevant fields, not just value.
• The preferred-parts list has a stated rationale, not just a list.

If you want to push further

• Add a pulse-rating analysis to the shunt exercise for a fault condition, and choose a part rated for it.
• Price your preferred list at three volumes (10, 100, 1000 boards) and show the effect of standardisation on cost.

• I choose values from the E series appropriate to the tolerance I need.
• I check resistor working voltage as well as power, and know when to use a series chain.
• I specify capacitors with dielectric, voltage and package, never value alone.
• I account for tempco when ratio accuracy matters over temperature.
• I choose package size deliberately against assembly method and board density.
• I standardise on a preferred-parts list and can explain the cost argument for it.