Project Details

Note : first analyses have been done with the valuable help of our friend Mistral AI

Chapter 1 : Choice of Mechanical Architecture and components

a/ Architecture

First thing is to determine which interaction will be between the bench and the DUT. It seems that two interesting possibilities should be available : first could be to insert a pneumatic T-tap between the monitor and the cuff, to inject directly the pneumatic stimulus inside the monitor. Second possibility is to create an artificial arm inside the cuff.

• T-tap between the monitor and the cuff: rejected — the device would be tested outside its original configuration (modified circuit, extra volumes, bench leak risks). Despites its simplicity and better reliability, the test will probably not take into account the cuff.

• Artificial arm placed inside the cuff: retained. The monitor and its cuff stay strictly original and operate in a closed loop, exactly as on a patient. Oscillations are born inside the cuff and detected by the device's own sensor. Reproducibility comes from the geometric and mechanical stability of the artificial arm. This reproduces the real physics of oscillometry — the device is tested in its nominal domain, the one of its clinical validation.

Here we have to notice that we may not have to represent a full realistic arm. The main goal is to respect maximum repeatability over time, in order to detect sensor drift over the years.

b/ Artificial arm material

Candidates examined: PET bottle (too rigid — 3× too fast deflation), latex balloon (replaced every 6 months — fails the 10-year requirement), platinum silicone bulb (gas permeability 10-25× butyl), Omron cuff bladder (flat shape), EPDM membrane accumulator (bulky, costly). Retained: 700c butyl bike inner tube — compliant (~0.1-0.3 mL/mmHg, comparable to an arm segment), the most stable and gas-tight elastomer, no creep when stored uninflated, original Schrader valve for sealing.

Form-holding textile sleeve: braided polyamide sleeve Ø 80-100 mm (Techflex F6 type): the geometry is set by the sleeve, not the elastomer. No foam by default (the most perishable part — the "zero consumable" architecture avoids it). Optional EPDM foam (not PE): if the ABP signal looks too "square" (rise/fall time ratio > 1.8, computed by the firmware), add one 3 mm closed-cell EPDM wrap between sleeve and cuff, and log the change. Markers: cuff position on the sleeve and number of Velcro notches, identical at every session.

c/ Drive choice to modelize dynamic blood pressure : syringe + eccentric cam (retained)

• Pure positional drive: every oscillation is a deterministic piston position; amplitude is fixed by a mechanical stop, not by software.
• No elastic part in the drive chain: stepper → cam → bearing → rigid piston.
• Durability: stepper + bearings + screws >> 10 years (silicone grease is maintenance, not replacement). 60 rpm = one cam cycle per second = 1 Hz pseudo-heartbeats (60 bpm).

Rejected alternatives: rotary servo (backlash, hysteresis), linear servo, peristaltic pump (aging tube), proportional valve (cost, pilot air).

Chapter 2 : Pneumatics — sizing

a/ Static and dynamic pressures

• Internal static pressure: at the start of each session the butyl arm is inflated to P_int = 100 mmHg (mean arterial pressure of the nominal 120/80 case). Storage between sessions at 0 mmHg (deflated, Schrader valve closed): no permanent stress, no creep, no thermal drift of a pressurized volume to monitor. This creates the real physics of oscillometry: envelope maximum centered on 100 mmHg, with shoulders toward systolic and diastolic — exactly like a real arm.
• Dynamic pressure: the cam injects a pulsed ΔV, transmitted to the cuff through the butyl wall + sleeve (+ optional EPDM foam). Target: 2-4 mmHg at the cuff.
• Pressurization: automatic, syringe in pump mode (§ 3), motor stopped at 100 ± 2 mmHg read by the ABP sensor, cross-checked with the aneroid gauge (stopcock S2). Ambient temperature logged every session (~0.4 mmHg/°C gas drift).

b/ Calculations

• 1 m butyl tube folded accordion-style: ~16 cm × Ø 80 mm (~6 strands), V_arm ≈ 120-160 mL.
• Boyle measurement (retained value): inject a known ΔV (10 mL calibration syringe on port S2), read ΔP on the ABP sensor: V_arm = P_abs × ΔV / ΔP, average of three injections; effective all-inclusive volume, logged on the arm's card.
• Total volume ≈ 155 mL (arm ~150 + tubing ≤ 4 + fittings ~1).
• Internal oscillation (isothermal): ΔP_arm = 860 × 2 / 155 ≈ 11 mmHg for a 2 mL cam stroke.
• Transmission: k = C_wall / (C_cuff + C_wrap) ≈ 0.15-0.4 → 1.7-4.5 mmHg at the cuff (target 2-4 ✓). k is measured at commissioning and the cam stroke adjusted (typically 2 mL); once the stop is locked, k is constant by construction.
• Device deflation: effective gas volume ≈ 135 mL, compliance comparable to a real arm → nominal deflation speed.

Waveform: the butyl wall + sleeve naturally round the edges — no additional pneumatic RC filter. "Squareness" is continuously quantified by the firmware (rise/fall ratio; alert if > 1.8 → EPDM foam).

Chapter 3 : Limits and critical analysis

1. A single "patient": one nominal point (100 mmHg, 60 bpm, one amplitude). We look for drift, not the complete error map. Possible extension: a second tube sealed at 120 mmHg ("hypertensive patient").
2. Mechanical coupling not computable: everything is protocolized to make the transmission constant, but it is unknown in absolute value — the bench measures relative evolutions, never absolute accuracy.
3. The standard drifts too: mitigations = accredited recalibration every 2 years (APB sensor) + cross-control chart with the aneroid gauge (two independent standards watching each other).
4. Butyl still ages, slowly: stored at 0 mmHg it lasts well beyond 10 years; the leak test detects any micro-leak; visual inspection at every session (cracks on the folds).
5. Check valves = the only soft active parts: doubly isolated during measurement (S1 + S4), their drift only affects inflation speed.
6. Thermal sensitivity: ~0.4 mmHg/°C during a session — temperature logged, ±3 °C range vs the reference session.
7. No clinical validation: the bench answers "has it drifted?", never "is it accurate?". Keep the annual comparison with a validated monitor.
8. Device firmware: an algorithm update can change the response without hardware drift — log the device firmware version in every session sheet.
9. Strengths: no elastic consumable in the drive chain (metal, bearings, air); mechanically defined stimulus (stop + cam), reproducible by construction; two cross-checked standards; the assembly under test operates strictly as original.