← Back to hardware

LPA: Light Plate Apparatus

One question, six answers

Which wavelength and intensity gives us the response we need?

Six conditions. One run. Controlled light.

Loading 0%

Summary

Twenty‑four cultures under six defined light conditions at once: green light to induce protectant production, red to halt it.

It took three design‑build‑test‑learn cycles. Prototype 2 measured 40 LEDs one by one: the green parts varied by 26% and the red by 22%, and matching them in fours brought each group within 0.4–1.8%. Matching alone did not separate the intensity tiers, so the LEDs are only matched to each other and the low, mid and high levels are set by PWM in code. Prototype 3 enclosed the array; overnight on a shaker a post snapped and the white holder leaked light between wells.

The instrument is built and characterized. Its first culture run, on 19 September 2026, grew in all six arms (3.1), but light leaked between wells, so it is not a light‑dose result. The fixes from that run are not yet re‑tested, the intensity test the array was built for has not been done, and every reading on this page is in lux, which is weighted for human vision and is not a biological light dose until it is converted to photon flux.

Array
4×6
Cultures
24
Conditions
6
Intensity levels
3

Results

LPA: Light Plate Apparatus

The LPA is a 24‑culture LED array designed to test how wavelength and light intensity affect our CcaS/CcaR optogenetic system in engineered Bacillus subtilis. It was designed around B. subtilis 168 and first ran on B. subtilis WB800, the protease‑deficient strain the project moved to in October (bioreactor, 2.2). Red and green LEDs are arranged across six light conditions, with four replicate cultures per condition.

Through three design–build–test–learn cycles, we improved LED consistency, reduced light leakage between wells, and moved intensity control into software using PWM. The final enclosed design provides controlled, repeatable illumination for optogenetic testing.

Our optogenetic system depends on light intensity and wavelength. We therefore built the Light Plate Apparatus to expose multiple samples to defined red and green light conditions while minimizing variation between wells.

Intensity is controlled through pulse‑width modulation from an Arduino, with one MOSFET assigned to each intensity tier. Once a condition is set, the output remains fixed throughout the run so the LPA provides repeatable illumination rather than adjusting light dynamically.

The array itself, six columns of four with the MOSFET that drives each pair, is drawn in 4.2 Light Layout.

Engineering Cycles

Three design–build–test–learn cycles, from the first array on a breadboard to the enclosed, software‑controlled instrument. The three prototypes follow in order.

DBTL cycle 1 of 3

2.1 DBTL Cycle 1 – Dual‑Wavelength Array

One green and one red LED per column

  1. DDesignA green and a red LED per column
  2. BBuildFirst array on a breadboard
  3. TTestRed and green switch separately
  4. LLearnLED output was uneven

Design

Each column carries a green and a red LED, so the wavelength can be switched quickly to exercise on and off control of protectant production. Nothing is calibrated at this stage: the aim was to confirm that the array drives both channels under software control.

Build

The first LPA on a desk: a black printed LED holder with a few green and red LEDs lit, wired across to a breadboard and a microcontroller board.

Each column carries a dual LED, so its output can be switched from green to red. Wire‑management prints were included from the start, since the harness is otherwise difficult to route at this density.

Test

The first dual‑wavelength array powered up and both LED channels switched on and off under software control. The light output was inconsistent across wells, which matters because an intensity comparison needs equal output within a condition.

Learn

The first prototype confirmed that the dual‑wavelength concept was feasible, but it also revealed a source of variation that had to be addressed before the light conditions could be compared reliably.

LED output was inconsistent

Individual LEDs in different wells produced different light intensities even when driven at the same setting. This meant that our low, mid, and high intensity levels were not yet consistent across the array, making comparisons between conditions unreliable.

DBTL cycle 2 of 3

2.2 DBTL Cycle 2 – Matched Output

Every LED measured, matched in fours

  1. DDesignTiers from matched LEDs
  2. BBuildEach LED measured, then rewired
  3. TTest40 LEDs sorted into groups
  4. LLearnMatch LEDs; set tiers by PWM

Design

Prototype 2 addressed the variation between wells. We measured every LED individually and grouped LEDs of similar output into matched sets.

Build

Every LED was driven at 5 V and measured individually on a light meter in a printed sensor fixture, so each reading was taken at the same distance and orientation.

A round black printed fixture on a desk with one green LED lit at its centre, wired to the bench, the light meter's case behind it.
Measure Each LED was measured individually at a fixed distance and angle.
A row of loose LEDs laid along the edge of a desk, each pair of leads carrying an orange paper tag.
Tag Each LED was labeled so its measurement could be traced.
The black LED holder standing upright with green and red LEDs lit, tagged red and black leads running from it to a breadboard in front.
Rewire LEDs with similar outputs were grouped into matched columns.

Test

We measured all 20 green LEDs and all 20 red LEDs individually at 5 V, then sorted them into matched groups of four based on output. After rewiring, the six channels formed the low, mid, and high conditions for each wavelength.

Green LEDs at 5 V, klux
Row12345
A2.542.192.33 (low)2.40 (high)2.40 (high)
B2.37 (mid)2.112.34 (low)2.482.33 (low)
C2.012.37 (mid)2.38 (high)2.42 (high)2.36 (mid)
D2.33 (low)2.052.35 (mid)2.452.20
Red LEDs at 5 V, lux
Row12345
A431 (high)371424 (high)393 (low)438
B400380430 (high)397 (low)360
C410 (mid)410 (mid)360416 (mid)397 (low)
D411 (mid)390 (low)360388428 (high)
The six channels after rewiring, in klux: the four LEDs of each matched group, from the readings above
Green lowRed lowGreen midRed midGreen highRed high
2.330.3902.350.4102.380.424
2.330.3932.360.4102.400.428
2.330.3972.370.4112.400.430
2.340.3972.370.4162.420.431

Learn

The spread is real, and matching tightens it Every LED from the two tables above on one axis (top), then the twelve kept per color magnified (bottom). Hollow dots are the eight per color that were not used.

Matching reduced the variation within each group. Before sorting, the green LEDs varied by 26% and the red by 22%; after matching, the four LEDs within each group differed by about 0.4–1.7% for green and 1.5–1.8% for red.

This showed that LED matching should be used to make each column internally consistent, while the actual low, mid, and high intensity levels should be set through PWM duty cycle rather than LED selection.

Prototype 2 left three limitations: the electronics were exposed, the intensity settings were adjusted by hand, and the culture tubes had no fixed position relative to their LEDs.

DBTL cycle 3 of 3

2.3 DBTL Cycle 3 – Enclosed Software‑Controlled Instrument

Enclosed, with intensity set in code

  1. DDesignEnclosed, with intensity in code
  2. BBuildOne enclosed, printed unit
  3. TTestShaker run: a post snapped
  4. LLearnFour screws, black tube holder

Design

Prototype 3 kept the matched LED groups from Prototype 2, but focused on making the LPA easier to operate and reproduce through four key changes.

01 · LPA_Box

Enclosed electronicsA printed outer frame houses the LED wiring and perf board so the electronics are protected and contained.

02 · Arduino code

Intensity in softwareLow, mid, and high light levels are set through PWM values in code rather than by manual adjustment.

03 · Left_C, Right_C

Tool‑free assemblyPrinted slider components allow the LPA to be assembled and disassembled without tools.

04 · Tube_Holder

Fixed tube position and light shieldingThe tube holder keeps each culture at the same distance from its LED, improving repeatability between runs, and covers the tubes from ambient light, so the effect of the green and red light is isolated and the data can be relied on.

Software duty cycle per tier Arduino Uno R4 WiFi PWM pins MOSFET ×3 Green + red column low mid high green induces · red halts 6 conditions
Drive chain The duty cycle for each tier is fixed in software before the run; the Arduino’s PWM pins drive three MOSFETs, one per tier, and each MOSFET switches a green and a red column of four. Nothing in the light path changes during a run, so each culture receives the same illumination from start to finish.

Build

The LPA assembled on a desk: a white 24-well tube holder seated in the top of a black printed housing, one lit tube standing in a corner well. Beside the housing, the Arduino sits on the desk on a clear base, with a USB-C cable and three jumper wires tagged P9, P10 and P11 running to the housing.
Prototype 3, assembled The white tube holder, the electronics and the housing combined into one compact unit.

Test

During the overnight shaker test one of the printed connection posts on the bottom plate broke off, and several wires came loose and were resoldered.

Close-up of the LPA on a lab shaker: the lower edge of the black housing with a screw, the grey bottom plate beneath it with white debris spread around a torn hole, a black rubber mat, and the shaker's blue front panel with a dial marked 60 and 80 and a strip of yellow tape by its knob.
Shaker damage Printed material around the connection point failed during the overnight run.
A grey printed bottom plate with a cone-shaped base; fingers hold the snapped-off cylindrical post just above it, a screw through its top, with the perf board, three MOSFETs and dust on the plate behind.
Snapped post One of the bottom-plate connection posts broke off at its base.

Learn

Reinforce the base at four points

The overnight run showed that concentrating the load on two mounting points was not mechanically sufficient, so all four screw positions on the bottom plate were used and the joints reinforced.

Block light between neighboring wells

The white tube holder allowed light to scatter into adjacent wells, so it was replaced with a black holder and the wall thickness increased from 2 loops to 4.

Results

3.1 First Culture Run, 19 September 2026

On its first culture run, the LPA grew B. subtilis WB800 under all six light arms to an OD600 of 0.91–1.01 at 24 h, close to the fitted plateau of 1.005.

The Run

On 19 September 2026, B. subtilis WB800 grew in the LPA under its six light arms, red and green each at a low, medium and high intensity (R1–R3 and G1–G3), and was read on a plate reader at five time points up to 24 hours.

Twelve of the 24 wells were loaded: one tube in each red arm and three in each green arm; the other wells stayed empty. By the procedure on the team wiki’s Measurement page, each tube held 2 mL of LB inoculated with the culture and was incubated at 37 °C with shaking; the Digital Twin page notes that the plate’s temperature was not recorded during the run.

This is the overnight shaker run of prototype 3 (2.3): a bottom‑plate post snapped and light leaked through the white tube holder into neighboring wells, so the arms cannot be compared with each other.

The LPA running inside an incubator: twelve capped culture tubes glowing green and red in the white tube holder on the black housing, the Arduino and tagged wires beside it, on a shaker.
The LPA running on 19 September 2026. Prototype 3 with its white tube holder, on the shaker.
Observed OD600 at 24 h, by light arm
ArmLightObserved OD600 at 24 h
R1Red, low0.967
R2Red, medium0.910
R3Red, high0.967
G1Green, low1.014
G2Green, medium0.975
G3Green, high0.958
The Growth Fit

A logistic model fitted the pooled readings better than a logistic with a lag phase (by AICc), with r = 0.299 ± 0.017 h−1 (a doubling time of 139 min) and a ceiling K = 1.005, at an NRMSE of 1.9% and R² 0.997.

The chart is drawn by script; its readings are in the table under it.

Logistic growth fit of B. subtilis WB800 on the LPA, 19 September 2026. OD600 (background subtracted) of the green‑light and red‑light wells at 2, 4, 6, 8 and 24 h, with the pooled logistic fit and its 5–95% band; four 6 h readings are excluded from the fit, and there are no samples between 8 and 24 h. From the team wiki’s Digital Twin page.
The readings behind the chartOD600, background subtracted, by arm and time
OD600, background subtracted, 19 September 2026
TimeR1red, lowR2red, mediumR3red, highG1green, lowG2green, mediumG3green, high
2 h0.0470.0470.0420.057, 0.028, 0.0500.048, 0.054, 0.0480.045, 0.047, 0.042
4 h0.0560.0550.0610.063, 0.064, 0.0600.064, 0.090, 0.0670.062, 0.060, 0.061
6 h0.1290.1350.1420.353 (excluded), 0.324 (excluded), 0.1360.226 (excluded), 0.136, 0.1490.236 (excluded), 0.129, 0.150
8 h0.1920.1850.1970.234, 0.178, 0.1910.231, 0.225, 0.2000.210, 0.221, 0.210
24 h0.9670.9100.9670.997, 1.024, 1.0221.000, 0.949, 0.9770.941, 0.961, 0.972

One tube per red arm and three per green arm, in tube order; × marks the four 6 h readings excluded from the fit. Readings from the team wiki’s Measurement page, less the blank its Digital Twin figure subtracts (0.061 at 2 h, 0.059 after); the fit and its band from the Digital Twin page.

The team’s digital twin uses this fit as its growth prior. Leaving each arm out in turn and forecasting its 24 h OD600 from the first eight hours, the error was at most 7% with the prior and 29 to 68% without it (Digital Twin page).

Final Design & Specifications

The final LPA combines what the three cycles established: LEDs matched by measured output, intensity set by PWM in code, enclosed electronics, a fixed tube position and a black holder that shields each well.

It lights 24 cultures under six fixed conditions, two wavelengths at three intensities with four replicates each, from one Arduino and three MOSFETs.

4.1 Specifications

The LPA, assembled, at a three-quarter view: the black tube holder on the black housing, on its grey base plate, with the grey sliders at the sides.
Array
4 × 6 on an 18 mm pitch; 24 cultures
Light conditions
6: green and red, each at low, mid and high; 4 replicate tubes per condition
Channels
Green and red
LEDs
24 in PCV‑220 sockets, matched in fours from 40 measured one by one (within 0.4–1.8%)
Intensity control
PWM per tier, set in code before the run and held constant: pins 9, 10, 11 at analogWrite 5, 31, 255 (about 2%, 12%, 100%)
Controller
Arduino Uno R4 WiFi
Drivers
3 × IRLZ44N logic‑level MOSFETs, one per tier; each switches a green and a red column
Tube bores
24 × 17 mm, 1 mm wall between bores, 4 wall loops, black PLA
Footprint
211 × 164 mm; 217 mm across the sliders
Height
108 mm without the tube holder; 188 mm with it
Printed parts
6, in black and grey PLA
Assembly
Printed sliders through the housing carry both holders; 4 × M6, 4 × #10‑32 and 4 × M3 screws (5.1); only a screwdriver is needed
Cost
US$65.52 (NT$2,097): US$44.06 parts + US$21.46 filament (5.1)
The parts, numberedas in the model and the STL downloads

Printed parts 1–6

  1. 1Base plate (Bottom_Plate)Closes the bottom and carries the perf board on four stands; screwed to the housing at four points.
  2. 2Left slider (Left_C)Slides in through the housing’s side and carries the LED and tube holders.
  3. 3Right slider (Right_C)The same, on the other side.
  4. 4Housing (LPA_Box)Encloses the LEDs and the electronics from outside light.
  5. 5LED holder (LED_Holder)Holds the 24 LEDs in their sockets on the 18 mm grid.
  6. 6Tube holder (Tube_Holder)Holds each tube over its own LED; black, 4 wall loops, to stop light crossing.

Bought part 7

  1. 7Perf boardCarries the three MOSFETs under the LED holder.
The final LPA on a desk: a black printed tube holder with 24 round bores seated in the top of the black housing, the slider tab on its side.
Final build The black tube holder that replaced the white one, seated on the housing.

4.2 Light Layout

The array, from above

The array from above: six columns of four wells on an 18 mm grid. From the left, a green and a red column at the low tier, then at the mid tier, then at the high tier, drawn brighter as the tier rises. One column, outlined, is one light condition: four matched LEDs under four tubes. Under the array, three MOSFETs, one per tier, each wired to its green and its red column.LowMidHighOne condition4 matched LEDsunder 4 tubesMOSFET ×3one per tier

One well, in section

Section through one well, to scale. The LED sits in its socket in the LED holder; above it the culture tube stands in a 17 mm bore of the black tube holder. Its light (green) goes up into the culture. Light leaving at an angle meets the 1 mm black wall to the next bore and stops there.17 mm boreCulture tube1 mm wall,black PLA, 4 loopsStray light,stopped at the wallLED in its socketLED holder
Six-condition LED array. Red and green channels are tested at low, mid, and high intensity, with four replicate cultures per condition.

Each column of four is one light condition. A tier is a green column and a red column driven by the same MOSFET, so the two wavelengths at one tier share one duty cycle. The duty cycles are fixed in code before the run, so every tube gets the same light from start to finish.

Why green and red?three reasons
  1. Green turns production on. In the CcaS/CcaR system, green light activates the CcaS sensor, which switches on CcaR and, through it, the production genes.
  2. Red turns it off. Red light switches CcaS back to its inactive form.
  3. One array asks both questions. With both colors at three intensities, one run shows how strongly production follows the light and whether red stops it.
Why match the LEDs and still use PWM?matching evens, PWM sets

Matching cut the spread inside each group from 26% (green) and 22% (red) to 0.4–1.8%, but left the three groups too close together to serve as tiers (2.2). Matching therefore equalizes the LEDs within a condition, and PWM sets the intensity between conditions.

4.3 Design Decisions

The LPA's requirements and how the final design meets them
RequirementHow the final design meets it
The same light in every tube of a conditionEvery LED measured one by one; the four in each column matched within 0.4–1.8%.
Three clearly different intensitiesEach tier’s brightness set by PWM duty cycle in code, through its own MOSFET.
Fixed illumination for the whole runDuty cycles set before the run and not changed; no feedback in the light path.
No outside lightLEDs and electronics enclosed in the housing; the tube holder closes the top.
No light from the neighboring wellBlack tube holder with 4 wall loops, replacing the white one with 2 that leaked; not yet re‑tested.
The same distance every runThe tube holder seats each tube at a fixed height over its own LED.
Survives a night on a shakerBase plate held at four screw points, not two; solder joints reinforced; not yet re‑tested.
Simple to assemble and openThe sliders slide in through the housing and carry both holders; twelve screws in three sizes hold the rest, so only a screwdriver is needed (5.3).

Reproduction

5.1 Materials List

Parts priced from the team’s cost table; filament from the slicer. Costs in US$, converted at NT$32 to the dollar.

Materials list
ComponentPart number / specQtyUnit (US$)Total (US$)Total (NT$)
Green LED5 V, built‑in resistor120.202.4077
Red LED5 V, built‑in resistor120.202.4077
LED socketsPCV‑220240.102.4077
MOSFETIRLZ44N, logic‑level31.003.0096
MicrocontrollerArduino Uno R4 WiFi127.5027.50880
Perf board160 × 100 mm11.161.1637
Jumper wire20 AWG, 10 cm120.101.2038
ScrewsM6 × 45 mm, stainless40.522.0867
Screws#10‑32 × ½ in, stainless40.230.9229
ScrewsM3 × 10 mm, stainless40.140.5618
Heat‑set insertsM3, brass40.110.4414
Parts subtotal44.061,410
FilamentBambu PLA Matte, 933.4 g (5.2)22.99/kg21.46687
Total65.522,097

Screws and inserts are priced per unit from packs: M3 heat‑set inserts, US$10.99 per 100; M3 × 10 mm stainless screws, US$13.50 per 100; M6 × 45 mm stainless socket head screws, US$51.89 per 100; #10‑32 × ½ in stainless, US$23.29 per 100. Four M6 screws hold the tube holder and the LED holder to the sliders, four #10‑32 the base plate to the housing, and four M3 the perf board, into the inserts in the base plate’s stands.

Matching needs spares. Twelve of each color go into the array, but they were picked from twenty of each (2.2, Test). Buying 20 + 20 to match from costs US$8.00 in LEDs instead of US$4.80.

Equipment

Light meter, 5 V bench power supply, 3D printer, soldering iron, calipers, a screwdriver for the assembly, and the printed sensor fixture used for LED matching.

5.2 Design Files

The LPA’s six printed parts and its perf board, numbered as in its 3D model (4.1). Each STL downloads under a descriptive name, in millimeters and in the instrument’s own coordinates, so the parts open already assembled. The six printed parts take 933.4 g of Bambu PLA Matte (US$22.99/kg), US$21.46 in all; each card gives its print’s filament and weight from the slicer.

The seven STLssix printed parts and the perf board
  1. 1
    Base plate

    Closes the bottom and carries the perf board on four stands; screwed to the housing at four points.

    STL 196 KB

    Grey PLA · 125.7 g · US$2.89

  2. 2
    Left slider

    Slides in through the housing’s side and carries the LED and tube holders.

    STL 140 KB

    Black / grey PLA · one plate with the LED holder and the right slider: 174.1 g, US$4.00 for the three

  3. 3
    Right slider

    The same, on the other side.

    STL 143 KB

    Black / grey PLA · one plate with the LED holder and the left slider: 174.1 g, US$4.00 for the three

  4. 4
    Housing

    Encloses the LEDs and the electronics from outside light.

    STL 159 KB

    Black PLA · 383.6 g · US$8.82

  5. 5
    LED holder

    Holds the 24 LEDs in their sockets on the 18 mm grid.

    STL 196 KB

    Black / grey PLA · one plate with both sliders: 174.1 g, US$4.00 for the three

  6. 6
    Tube holder

    Holds each tube over its own LED; black, 4 wall loops, to stop light crossing.

    STL 196 KB

    Black PLA, 4 wall loops (2.3, Learn) · 250.0 g · US$5.75

  7. 7
    Perf board

    Bought, not printed: carries the three MOSFETs under the LED holder. Its STL is for checking the fit.

    STL 47 KB

Run sketch, LPA.ino

For the Arduino Uno R4 WiFi: holds the three intensity tiers by PWM on pins 9, 10 and 11 for the whole run.

INO 369 B

5.3 Assembly Guide

LPA assembly guide

Six stages, from printing the parts to loading the run sketch, with the parts list and a check after each. 10 pages.

PDF 382 KB

Safety. The LPA runs on 5 V: unplug the Arduino and switch off the bench supply before changing any wiring, and do not look straight into the LEDs at full brightness (the guide’s “Before you start” page). The cultures are B. subtilis, handled at BSL‑1 as set out on the team wiki’s Safety page.

References

Illuminating a plate of cultures from below with per‑well control is the Light Plate Apparatus pattern from the optogenetics literature1, which our LPA follows and is named after. What we added is the dual‑wavelength column layout and LEDs matched by measured output, in an enclosure sized for our own tubes.

The biology is the CcaS/CcaR green/red switch, also from the literature, not designed here; it was characterized in B. subtilis by Castillo‑Hair and colleagues2. This page reports light in lux and does not convert it to photon flux, so our tiers are not placed on that paper’s dose scale.

  1. Gerhardt KP, Olson EJ, Castillo‑Hair SM, et al. An open‑hardware platform for optogenetics and photobiology. Scientific Reports 6, 35363 (2016). doi:10.1038/srep35363
  2. Castillo‑Hair SM, Baerman EA, Fujita M, Igoshin OA, Tabor JJ. Optogenetic control of Bacillus subtilis gene expression. Nature Communications 10, 3099 (2019). doi:10.1038/s41467-019-10906-6