Mads Rudolph · CV

Profile

5th-semester electrical engineering student at DTU who finds faults at the bench and fixes them: a 130 µm board bow that broke PCB milling, a floating ground that would have killed a supply, a control loop identified against the wrong plant. I design and fabricate PCBs in KiCad on my own mill and laser, write firmware in C for ATmega microcontrollers, and verify with a scope and tests. Available for an internship from late January 2027, with electives aimed at analog, power electronics and electroacoustics.

Projects as experience

• SRM-CAM — CAM tool for a Roland SRM-20 PCB mill

  Solo · in daily use at the DTU Ballerup fab lab
  • Diagnosed why milled isolation traces skipped copper on a 130 µm bowed board: a 3-point routine could only fit a plane. Replaced it with a probed grid and bilinear leveling so every Z follows the real surface.
  • Reverse-engineered the mill’s SPI remote header with an Arduino: audited Roland’s library (5 of 17 commands in use, 1 of 64 status bits read), decoded the full status word, and added a STOP that interrupts a move in flight.
  • Brought double-sided registration to under 0.1 mm after the flip with a swept fit-test coupon for dowel-hole compensation; 374 commits, ~1,000 tests, Windows installer and Linux AppImage built in CI.

• Vinyl ADC — discrete 3rd-order delta-sigma audio ADC

  Solo · working; rev D boards sponsored by PCBWay
  • Designed a stereo 48 kHz converter from TL07x op-amps, an LM311 and 74HC logic, sized to the DTU parts shop: 1.536 MHz clock, OSR 32, ~68 dB SNR in simulation.
  • SPICE caught a charge pump drawn backwards (the −5 V rail simulated at +3.4 V) that every connectivity check had passed; found that comparator delay, not op-amp bandwidth, limits the clock.
  • Split one unroutable single-sided board (45 wire bridges) into four milled boards with a script that proves the split matches the reference netlist.
  • Measured 67.9 dB dynamic range on the milled boards against ~68 dB simulated, then redrew it as a four-layer Raspberry Pi HAT made by PCBWay (sponsored): about 8 dB more A-weighted range recording a turntable, on the Pi’s own supply.

• Sub crossover — measured line-level low-pass for a Bose bass module

  Solo · bring-up in progress
  • Characterised the module before designing: 8.9 kΩ aux input, 0 ms latency, and a 63–203 Hz bandpass that killed the planned 50 Hz Linkwitz-Riley and moved the crossover to 85–190 Hz.
  • Designed a mono-summing Sallen-Key low-pass on one TL074 with three switched corners, every part from the DTU shop; caught a missing DC bias path and a 6 V output offset in review.
  • Scripted, gated bring-up on an Analog Discovery 3: three corners within 0.2 dB of the as-built model, 58 µV rms noise (instrument-limited), 3.52 V rms in before 1% THD; a fitted capacitor model predicted held-out settings to 0.7%.

• Korad KD3005D — hidden UART reverse-engineered

  Solo
  • Found the unpopulated remote header on a bench supply and decoded its protocol with a logic analyzer: 9600 8N1, poll-only ASCII, 100–200 ms command pacing.
  • Found with a multimeter that the supply’s ground floats, which would have destroyed the first board plugged into it, and designed a small isolated ESP32 carrier board for it instead.

• VLF metal detector — DFT phase detection on ATmega328P

  Group of 4 · DTU 34621
  • Single-bin DFT on 8 kHz coherent samples reads amplitude and phase of the receive coil to separate ferrous from non-ferrous targets; MOSFET H-bridge transmitter and a bucking-coil head.
  • Designed the power path for endurance: MCU sleeps between DFT windows. Bench discharge test: 164 min continuous on one charge, pack sagging 8.88 V to 7.76 V.

• REGBOT Balance — cascaded control for a self-balancing robot

  Group · DTU 34722
  • Traced a sluggish inner loop to system identification done wheels-up (motor pole ~6× too fast). Re-identified on the floor and re-designed all four loops: rise time 27× faster, 2 m move error from 10.7 cm to 3.6 cm.

• KiCad-Autoplace — placement and routing for single-sided milled boards

  Solo, for the DTU Ballerup workshop
  • Electron + Python desktop app that scores several placement candidates (wirelength, crossings, overlaps) and routes the winner with FreeRouting; used to mill real boards.

All 16 projects →

Skills

Education

• B.Eng. in Electrical Engineering (Elektroteknologi) Sep 2024 — thesis planned autumn 2027
  DTU — Technical University of Denmark, Kgs. Lyngby

  115 ECTS completed. Electives in analog IC design, power electronics, electroacoustics, acoustics and EMC. Mandatory internship spring 2027, bachelor thesis autumn 2027.

• STX (upper secondary), Mathematics A / Physics A 2019 — 2022
  Sønderborg Statsskole

Employment

Service and seasonal work alongside my studies. What it shows: I turn up, keep my head in a busy shift, and communicate clearly with guests and colleagues in Danish and English.

• Waiter, Hotel D'Angleterre, Copenhagen Dec 2024 — present
• Bartender / waiter, Visselulles Vinbar, Sønderborg May 2023 — Aug 2024
• Line cook, Mommark Marina Jun 2021 — Jun 2024
• Production worker, JKS at Danfoss, Nordborg Aug 2022 — Nov 2023
• Ski instructor (CSIA Level 1), Sun Peaks Resort, Canada Dec 2022 — Apr 2023

Languages

Danish (native), English (fluent).