Spartanz

Sketch → Circuit turns a photo of a hand-drawn circuit into a circuit that is actually simulated, and tells you what is wrong with it in plain language. The problem is a gap every electronics student falls into. You work a circuit out on paper, because paper is where you think. To find out whether it works, you have to build it a second time — clicking parts into a simulator, or wiring a breadboard. Redrawing takes longer than drawing did, so most people skip simulation and go straight to hardware. When it is wrong, the feedback is a dead LED or a hot wire, and nobody explains why. In a class of sixty there is one teacher. You photograph the page instead. Muse Spark reads the symbols, the handwriting and the values. SAM 3.1 segments the parts and the wires. We recover connectivity by walking the ink on the page rather than guessing, canonicalise it into a netlist, and solve it with a nonlinear DC operating-point solver (modified nodal analysis with Newton-Raphson and real diode models). A rule engine then says, for example: "D1 has no current-limiting resistor. At 9V it will draw roughly 438mA, about 22x its 20mA rating. Add a 470Ω resistor." What it does beyond reading the drawing: - Reads connections you wrote as a pin table, checks them against what you drew, and flags contradictions instead of silently choosing one. - Knows Arduino Uno and Raspberry Pi pin limits, so it catches the GPIO pin you are about to damage. - Lets you describe a part in words — "a 5V relay with one normally open contact" — and builds it from the solver's own device primitives. Every generated part ships with bench tests that our solver runs for real; a failing test is sent back with the measured numbers to be corrected. - Is editable: drag parts, wire pins, delete, rotate — in the schematic or on the 3D board — and everything re-solves. - Pairs with your phone by QR so you can photograph the page with the better camera. Measured, honestly: 100% component identification on our test sketches; 3/3 exact netlists; the solver agrees with hand calculation to within 0.1%; 23 hand-computed device-model cases pass; zero geometric intersections across 18 circuits in 3D; zero overlap or crossing conflicts across 19 saved schematics. That is a small test set that we drew ourselves. It is not a benchmark, and we have not yet tested it on real phone photographs in poor lighting.