Arduino bench & real-world scaling

What our small experimental circuit measured on three connectors (run 2), how each quantity is scaled to a representative charger, and how that compares with published charger figures.

Experimental circuit Measured on our Arduino prototypeReal-world estimate Calculated from the prototype by a documented formulaCommercial reference Published by a standard or manufacturer

Arduino bench & real-world scaling

What our small experimental circuit measured on three connectors (run 2), how each quantity is scaled to a representative charger, and how that compares with published charger figures.

Experimental circuit Measured on our Arduino prototypeReal-world estimate Calculated from the prototype by a documented formulaCommercial reference Published by a standard or manufacturer

What the bench shows

Each connector needs its own calibration
Observed:
The connectors have different loads (C1 493 Ω, C2 350 Ω, C3 388 Ω) and different healthy levels (C1 3.455 V, C2 3.067 V, C3 3.193 V): a 12 % spread.
Evidence:
Load = median V ÷ I; healthy level V_ref = median V ÷ (1 − degradation/100) — the log's own definition, 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level.
Why it matters:
Reading all three against one shared healthy level would misstate degradation by about 1.7 points (median). A voltage only means something relative to that connector's own healthy level.
Next step:
Record each connector's healthy voltage when it is installed (a commissioning sweep) and judge later readings against it. EVOLT's model does exactly this.
The run sweeps every connector through its full range, repeatedly
Observed:
Each connector goes from near-healthy to an open circuit and back: C1 every 15 s, C2 every 20 s, C3 every 22.5 s. Open circuit (0 V) takes up C1 22 %, C2 19 %, C3 12 % of the readings.
Evidence:
127 time steps every 0.5 s from 1.2 s to 64.2 s; a sweep is counted from one open-circuit episode to the next.
Why it matters:
Repeated full-range sweeps are good calibration data: the whole voltage–degradation curve is seen several times on every connector, so the model can be tested on sweeps it never saw. They are a controlled stress pattern, not natural wear, so they give no wear rate.
Next step:
Add a slow, one-way wear test (many mating cycles, readings over hours or days) to measure how fast degradation actually builds up.
At charger scale, a 50 % bench state is a serious fault
Observed:
On C1, scaled to DC 50 kW · 400 V, 125 A, the 50 % state would lose 200 V and turn 12.5 kW into heat in the connector.
Evidence:
Method A: same resistance ratio as the bench (extra 493 Ω on a 493 Ω load), × k_R 6.50e-3; heat = I² × R per path.
Why it matters:
A real contact wasting this much would overheat long before it reached this state. The bench shows the direction and shape of the effect (more resistance → lower voltage), not the size of real wear.
Next step:
Treat the bench as a test of the detection idea. Model gradual wear with small, stepped resistances.
Early real-world wear is below the bench's resolution
Observed:
A real contact only moves C1's reading by one 4.9 mV ADC step at about 10 × the new-contact maximum (5 mΩ) in this scenario.
Evidence:
Method B: R = 0.5 mΩ (datasheet maximum) × wear multiple, mapped back with ÷ k_R and the bench's voltage divider.
Why it matters:
Gradual, real-world wear changes resistance by milliohms; a 10-bit ADC on this circuit cannot see the early stage of it.
Next step:
Measure contact resistance directly with a four-wire (milliohm) method, or amplify the voltage across the connector before the ADC.
Degradation is calculated from voltage, not measured separately
Observed:
Logged degradation matches 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level within 0.49 points on every connector, and current follows V ÷ R_load.
Evidence:
Largest |logged − formula| over conducting readings, using each connector's measured V_ref; median V ÷ I per connector.
Why it matters:
The bench has one independent signal — voltage. Any model trained on its degradation column rediscovers this definition; what the data does show is that the reference must be per connector.
Next step:
Add an independent current sensor and a temperature sensor on each connector.

Three connectors, side by side

Three connectors (C1–C3), each with its own fixed load, logged every 0.5 s from 1.2 s to 64.2 s. Each connector is swept repeatedly from near-healthy to an open circuit and back. Each connector is summarised from its own readings; nothing is averaged across connectors.

Experimental circuit

Parameter	C1	C2	C3
Readings (conducting)	127 (99)	127 (103)	127 (112)
Load resistance (median V ÷ I)

Why the model uses relative voltage

The same readings twice. Against raw voltage the three connectors form three different lines; against voltage relative to each connector's healthy level they fall on one.

Experimental circuit

From the bench to a real charger

Pick a representative charger, a bench connector and a level of the bench's degradation index. Every number below is recalculated from the stored readings.

k_V voltage: × 115.8; 400 V ÷ C1 healthy V_ref
k_I current: × 17,824; 125 A ÷ C1 healthy mA

Heat in the connector

DC 50 kW · 400 V, 125 A. Log scale: the bench at 50 % against real contacts at the same current.

Share of power lost

Percent of delivered power. The bench share is dimensionless and carries over unchanged.

What real wear looks like (Method B)

Start from the datasheet maximum for a new contact (0.5 mΩ), multiply by illustrative wear levels, and apply DC 50 kW · 400 V, 125 A. Then map back: would the bench see it?

Contact resistance	Drop per contact	Heat per contact	All 2 contacts	Share of delivered	Bench-equivalent drop	Visible on bench?
0.5 mΩ (datasheet max, new)	62.5 mV	7.81 W	15.63 W	0.031%	0.54 mV	No (ADC step 4.88 mV)
1 mΩ (× 2, illustrative)	125 mV	15.63 W	31.25 W	0.063%	1.079 mV	No (ADC step 4.88 mV)
2.5 mΩ (× 5, illustrative)	312.5 mV	39.06 W	78.13 W	0.156%	2.697 mV	No (ADC step 4.88 mV)
5 mΩ (× 10, illustrative)	625 mV	78.13 W	156.25 W

Try the bench model

Turns a bench voltage into the bench's degradation index with the saved, tested relative-voltage model — experimental-scale inputs only.

Model: degradation = 100.11 − 100.23 × (V ÷ V_ref), fitted on 314 readings; held-out error 0.09 points (MAE). Re-fitted from the database on this page load: identical.

Assumptions and calibration still needed

Circuit values (each connector's load and healthy level V_ref) are measured from the readings. The supply voltage itself was not logged; V_ref stands in for it.
Degradation % is the log's own definition, 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level. It is not scaled; a level is only used to pick a bench voltage.
The run sweeps each connector repeatedly between near-healthy and open circuit. It is a controlled stress pattern, so it gives no rate of wear over time.
Charger operating points are representative nominal values (4 scenarios), checked against IEC 62196-1 ratings — not measurements of a specific charger.
AC power factor 0.99; balanced three-phase (neutral ≈ 0 A); session length 1 h for energy.
Wear multiples (1, 2, 5, 10, 20 ×) are illustrative: no verified field rate of contact wear was found.
To calibrate further: log the supply voltage, add an independent current sensor, measure contact resistance with a four-wire method, and fit a thermal resistance (K/W) with a thermocouple.

Scaling version 2026.10-2. Run bench-run-2, stored from database/data/arduino-bench-run-2.csv (SHA-256 7b27b8eb9bcc…), converted cell for cell from database/data/raw/arduino-bench-run-2.xlsx.

References used

Only figures that appear on these pages are used as commercial references.

IEC 62196-1:2022 — Plugs, socket-outlets, vehicle connectors and vehicle inlets, Part 1 (abstract) — DER-Lab standards summary; accessed 2026-10-26. 690 V AC · 250 A · 1 500 V DC · 800 A
MIDA CCS 2 charging cable / plug, DC 1000 V 250 A — product specification — MIDA (manufacturer); accessed 2026-10-26. 0.5 mΩ max · < 50 K (terminal) · > 10 000 · DC 1000 V, 250 A
HARTING charging cable CCS Type 2 — technical data — HARTING (manufacturer); accessed 2026-10-26. 60 A – 200 A, 1000 V DC (power contacts) · > 10 000 · 90 °C
ADAC tested five EVs for charging losses — The Next Web (reporting ADAC tests), 19 Aug 2026; accessed 2026-10-26. 5.1 % – 7.0 % · 12.7 % – 24.2 % · 5 % – 15 % (1–4 % warm battery)

What the bench shows

Each connector needs its own calibration
Observed:
The connectors have different loads (C1 493 Ω, C2 350 Ω, C3 388 Ω) and different healthy levels (C1 3.455 V, C2 3.067 V, C3 3.193 V): a 12 % spread.
Evidence:
Load = median V ÷ I; healthy level V_ref = median V ÷ (1 − degradation/100) — the log's own definition, 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level.
Why it matters:
Reading all three against one shared healthy level would misstate degradation by about 1.7 points (median). A voltage only means something relative to that connector's own healthy level.
Next step:
Record each connector's healthy voltage when it is installed (a commissioning sweep) and judge later readings against it. EVOLT's model does exactly this.
The run sweeps every connector through its full range, repeatedly
Observed:
Each connector goes from near-healthy to an open circuit and back: C1 every 15 s, C2 every 20 s, C3 every 22.5 s. Open circuit (0 V) takes up C1 22 %, C2 19 %, C3 12 % of the readings.
Evidence:
127 time steps every 0.5 s from 1.2 s to 64.2 s; a sweep is counted from one open-circuit episode to the next.
Why it matters:
Repeated full-range sweeps are good calibration data: the whole voltage–degradation curve is seen several times on every connector, so the model can be tested on sweeps it never saw. They are a controlled stress pattern, not natural wear, so they give no wear rate.
Next step:
Add a slow, one-way wear test (many mating cycles, readings over hours or days) to measure how fast degradation actually builds up.
At charger scale, a 50 % bench state is a serious fault
Observed:
On C1, scaled to DC 50 kW · 400 V, 125 A, the 50 % state would lose 200 V and turn 12.5 kW into heat in the connector.
Evidence:
Method A: same resistance ratio as the bench (extra 493 Ω on a 493 Ω load), × k_R 6.50e-3; heat = I² × R per path.
Why it matters:
A real contact wasting this much would overheat long before it reached this state. The bench shows the direction and shape of the effect (more resistance → lower voltage), not the size of real wear.
Next step:
Treat the bench as a test of the detection idea. Model gradual wear with small, stepped resistances.
Early real-world wear is below the bench's resolution
Observed:
A real contact only moves C1's reading by one 4.9 mV ADC step at about 10 × the new-contact maximum (5 mΩ) in this scenario.
Evidence:
Method B: R = 0.5 mΩ (datasheet maximum) × wear multiple, mapped back with ÷ k_R and the bench's voltage divider.
Why it matters:
Gradual, real-world wear changes resistance by milliohms; a 10-bit ADC on this circuit cannot see the early stage of it.
Next step:
Measure contact resistance directly with a four-wire (milliohm) method, or amplify the voltage across the connector before the ADC.
Degradation is calculated from voltage, not measured separately
Observed:
Logged degradation matches 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level within 0.49 points on every connector, and current follows V ÷ R_load.
Evidence:
Largest |logged − formula| over conducting readings, using each connector's measured V_ref; median V ÷ I per connector.
Why it matters:
The bench has one independent signal — voltage. Any model trained on its degradation column rediscovers this definition; what the data does show is that the reference must be per connector.
Next step:
Add an independent current sensor and a temperature sensor on each connector.

Three connectors, side by side

Experimental circuit

Parameter	C1	C2	C3
Readings (conducting)	127 (99)	127 (103)	127 (112)
Load resistance (median V ÷ I)

Why the model uses relative voltage

The same readings twice. Against raw voltage the three connectors form three different lines; against voltage relative to each connector's healthy level they fall on one.

Experimental circuit

From the bench to a real charger

Pick a representative charger, a bench connector and a level of the bench's degradation index. Every number below is recalculated from the stored readings.

k_V voltage: × 115.8; 400 V ÷ C1 healthy V_ref
k_I current: × 17,824; 125 A ÷ C1 healthy mA

Heat in the connector

DC 50 kW · 400 V, 125 A. Log scale: the bench at 50 % against real contacts at the same current.

Share of power lost

Percent of delivered power. The bench share is dimensionless and carries over unchanged.

What real wear looks like (Method B)

Start from the datasheet maximum for a new contact (0.5 mΩ), multiply by illustrative wear levels, and apply DC 50 kW · 400 V, 125 A. Then map back: would the bench see it?

Contact resistance	Drop per contact	Heat per contact	All 2 contacts	Share of delivered	Bench-equivalent drop	Visible on bench?
0.5 mΩ (datasheet max, new)	62.5 mV	7.81 W	15.63 W	0.031%	0.54 mV	No (ADC step 4.88 mV)
1 mΩ (× 2, illustrative)	125 mV	15.63 W	31.25 W	0.063%	1.079 mV	No (ADC step 4.88 mV)
2.5 mΩ (× 5, illustrative)	312.5 mV	39.06 W	78.13 W	0.156%	2.697 mV	No (ADC step 4.88 mV)
5 mΩ (× 10, illustrative)	625 mV	78.13 W	156.25 W

Try the bench model

Turns a bench voltage into the bench's degradation index with the saved, tested relative-voltage model — experimental-scale inputs only.

Model: degradation = 100.11 − 100.23 × (V ÷ V_ref), fitted on 314 readings; held-out error 0.09 points (MAE). Re-fitted from the database on this page load: identical.

Assumptions and calibration still needed

Circuit values (each connector's load and healthy level V_ref) are measured from the readings. The supply voltage itself was not logged; V_ref stands in for it.
Degradation % is the log's own definition, 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level. It is not scaled; a level is only used to pick a bench voltage.
The run sweeps each connector repeatedly between near-healthy and open circuit. It is a controlled stress pattern, so it gives no rate of wear over time.
Charger operating points are representative nominal values (4 scenarios), checked against IEC 62196-1 ratings — not measurements of a specific charger.
AC power factor 0.99; balanced three-phase (neutral ≈ 0 A); session length 1 h for energy.
Wear multiples (1, 2, 5, 10, 20 ×) are illustrative: no verified field rate of contact wear was found.
To calibrate further: log the supply voltage, add an independent current sensor, measure contact resistance with a four-wire method, and fit a thermal resistance (K/W) with a thermocouple.

Scaling version 2026.10-2. Run bench-run-2, stored from database/data/arduino-bench-run-2.csv (SHA-256 7b27b8eb9bcc…), converted cell for cell from database/data/raw/arduino-bench-run-2.xlsx.

References used

Only figures that appear on these pages are used as commercial references.

IEC 62196-1:2022 — Plugs, socket-outlets, vehicle connectors and vehicle inlets, Part 1 (abstract) — DER-Lab standards summary; accessed 2026-10-26. 690 V AC · 250 A · 1 500 V DC · 800 A
MIDA CCS 2 charging cable / plug, DC 1000 V 250 A — product specification — MIDA (manufacturer); accessed 2026-10-26. 0.5 mΩ max · < 50 K (terminal) · > 10 000 · DC 1000 V, 250 A
HARTING charging cable CCS Type 2 — technical data — HARTING (manufacturer); accessed 2026-10-26. 60 A – 200 A, 1000 V DC (power contacts) · > 10 000 · 90 °C
ADAC tested five EVs for charging losses — The Next Web (reporting ADAC tests), 19 Aug 2026; accessed 2026-10-26. 5.1 % – 7.0 % · 12.7 % – 24.2 % · 5 % – 15 % (1–4 % warm battery)

Healthy level V_ref (0 %)	3.455 V	3.067 V	3.193 V

Healthy current (V_ref ÷ load)	7.01 mA	8.76 mA	8.23 mA

Highest reading	3.38 V (2 %)	3.03 V (1 %)	3.15 V (1 %)

Voltage at 50 %	1.727 V	1.533 V	1.597 V

Extra resistance at 50 %	493 Ω	350 Ω	388 Ω

Sweeps to open circuit (period)	5 (every 15 s)	3 (every 20 s)	3 (every 22.5 s)

Degradation logged: mean / median	62 % / 65 %	59.7 % / 61 %	57.4 % / 59 %

Largest gap to the log's formula	0.34 pts	0.35 pts	0.49 pts

Near healthy (< 10 %)	8	9	11

Moderate (10–40 %)	32	31	30

Severe (≥ 40 %, still conducting)	59	63	71

Open circuit (0 V, 100 % logged)	28	24	15

Parameter	Experimental circuit Arduino reading	Scaling method / factor	Real-world estimate Scaled estimate	Commercial reference Commercial reference	Difference
Voltage across one current path (healthy, 0 %)	3.455 VThe connector's own reference V_ref, from the log's definition	× k_V = 115.8 (400 V ÷ 3.455 V)	400 V	IEC 62196-1 maximum DC: 1,500 V	27 % of the DC maximum
Current in one conductor (healthy, 0 %)	7.013 mAV_ref ÷ R_load	× k_I = 17,824 (125 A ÷ 7.01 mA)	125 A	IEC 62196-1 max 800 A DC; HARTING CCS2 cable 60–200 A; MIDA CCS2 plug 250 A	16 % of the IEC maximum
Load resistance (healthy path)	492.593 ΩMedian V ÷ I of readings ≥ 1 mA (middle 50 %: 490–496 Ω)	× k_R = 6.50e-3 (k_V ÷ k_I)	3.2 Ω	—	—
Extra series resistance at 50 % degradation	492.593 ΩR_L × (V_ref ÷ V − 1), V = 1.727 V at 50 %	Method A: same resistance ratio as the bench (extra ÷ load), ≈ × k_R	3.2 Ω	New CCS2 contact: ≤ 0.5 mΩ (MIDA datasheet)	6,400 × the new-contact maximum
Voltage lost in the connector at 50 %	1.727 V	Method A: × k_V	200 V	At ≤ 0.5 mΩ and 125 A: ≤ 62.5 mV per contact (I × R)	3,200 × the new-contact drop
Heat generated in the connector at 50 %	6.057 mWI² × R_extra with I = 3.51 mA	Method A: I² × R per path, 1 path (k_P = 2,063,840)	12,500 W	New contact at 125 A: ≤ 7.81 W per contact (I² × 0.5 mΩ)	1,600 × a new contact
Share of power lost in the connector	50 %	Dimensionless — unchanged by scaling	50 %	ADAC: 5–15 % (DC fast charging, whole chain)	3.3 × the top of the whole-chain range
Energy lost in the connector per 1 h session	6.057 mWhbench heat held for 1 h	Method A heat × 1 h (time is chosen, not scaled)	12.5 kWh	ADAC whole-chain loss on 50 kWh: 2.5–7.5 kWh	—
Contact temperature rise	—Not measured on the bench	Needs a calibrated thermal resistance (K/W); not estimated	—No estimate without thermal calibration	Terminal temperature rise < 50 K (MIDA datasheet)	—
Degradation index (bench definition)	50 %Chosen level; defined as 100 × (1 − V ÷ V_ref), V_ref = the connector's own healthy level (V_ref 3.455 V)	Not scaled: it is a property of the bench's voltage reference, not of charger wear	—No real-world equivalent	No verified field degradation rate was found	—