Same-capability glove cost reduction
Target: one-third of the complete delivered cost for one right-hand glove. Keep 20 independent local pressure and vibration locations, acquisition quality/rates, both ISM330DHCX sensors, continuous camera, six-hour operation, SD, Bluetooth/Wi-Fi/USB, charge protection, button, indicator, keyed assembly and glove feel. The locked requirements preserve the full brief. The compact cuff and tracking/texture demonstration requirements remain open engineering work in the baseline.
Current result: the first cost screen and a native analog alternative simulation are complete. A cheaper complete design and a threefold saving are still unestablished. Both complete delivered totals remain unknown. The canonical hardware is unchanged; the selected-cost and manufacturing gates remain open.
Decisions from this pass
| Change | Evidence | Decision |
|---|---|---|
| Replace a complete Alpha 5851 industrial wire spool with soft 30 AWG silicone wire | One Alpha 100 ft spool observed at $143.30; two Adafruit 3164 50 ft spools at $9.90 total. Adafruit lists 0.8 mm OD and 331 Ω/km; its nominal OD and gauge fit the published GH contact envelope. | Prioritize a factory harness qualification. $133.40 is a goods-price difference, not an approved or delivered saving. Preserve labels, lead mapping, actual cut lengths, crimp pull, strain relief, bundle fit, leakage/capacitance and movement-noise checks. |
| Reduce expensive one-off component loading | Actual PCB: 561 placements, 107 MPNs, 183 front/378 back; JLC's public Standard loading rate produces a $163.71 scenario for 107 distinct types before other charges. | Consolidate equivalent part types and compare complete factory quotes. The scenario is not a quote; exact billable groups may differ. |
| Change to Economic assembly | JLC requires single-sided placement, supported 2/4/6-layer stack and at least 0.5 mm BGA pitch. Current PCB is double-sided, eight-layer and includes a 0.4 mm BGA. | A layer-count change alone cannot deliver that saving. A smaller, less complex layout must also preserve all electrical and assembly checks. |
| Remove bank followers and final analog multiplexer | Four native full-step cases completed with original TI amplifier models. Three fail existing acquisition waits; worst pressure residual is 5.262 mV versus a 45.8 µV limit. | Reject this candidate. Book $0 savings. Keep the buffered topology. |
| Substitute a cheaper 6 Ah battery | Existing battery-input power includes converter and external path loss. The candidate's nominal energy yields 5.7106 h after the same reserve. | Reject at the unchanged load. Preserve six-hour capture. |
| Use pre-terminated TE LDT0-028K piezo strips | Commercial stock/price exists, but 13 × 25 mm backed film replaces 10 × 14 mm bare film and increases thickness. | Exclude from a same-construction saving until fit, feel and transfer response satisfy the existing requirements. |
| Re-source ADC and C0G filter capacitors | Exact 16-bit ADC variants and same-value dielectric requirements identified; 1 MSPS extended-temperature option should be checked before considering a slower member. | Continue exact sourcing; retain timing and signal performance. Candidate delivered prices and integration checks remain open. |
The wire sources are Adafruit 3164 and Mouser's matching part. Assembly observations use the JLC price schedule and capability table, checked 23 October 2026. Price, stock and delivery must be reconfirmed for the actual basket.
Cost basis
The regenerated current CSV subset is $54.5377 for 35 historical JLC-priced MPN rows / 414 placements. This is narrower than the older $54.61 prose snapshot: it excludes unpriced components and separately captured Comet prices. cost-review.json records the exact CSV hash and each included row. Neither subtotal is a complete cuff cost.
The primary metric is cash needed for one usable glove. A supplier's mandatory two-PCBA order, entire sheet/spool/syringe, shipping, tax, factory preparation, flashing and testing all count. Spare material has residual value, but allocating it to imaginary future gloves does not reduce today's payment. A factory can charge for consumed material from its own stock only when its actual quotation offers that service.
Keep currencies and VAT basis separate until a dated conversion and destination quote exist. The quote comparison template leaves missing amounts null. The calculator rejects incomplete coverage, mixed-currency totals and unsupported zero-cost lines. Public fee formulas and hypothetical component reductions remain scenarios, not invoices.
Engineering continuation
- Obtain comparable complete factory pricing for the baseline and the same-capability alternative, including minimum quantities and supplier-provided harness material. Ask your chosen manufacturer for the same explicit scope and qualification boundaries.
- Complete the prepared silicone-wire harness definition and verify electrical/mechanical compatibility. Re-source the unchanged analog parts and trace each accepted alternative to its exact datasheet/footprint.
- Reduce distinct MPNs, bespoke mechanical pieces and costly assembly features only where the resulting circuit, layout and enclosure meet the existing requirements. A lower price has no authority to waive thermal, fault, fit or signal findings.
- Re-run schematic/pin checks, routing/DFM, analog/power simulations, target firmware build, concurrency/transport checks and enclosure clearance against the actual alternative files. Perform physical signal, thermal, six-hour endurance and fit acceptance on the first article.
- Compare complete quote totals. If one-third cannot be reached, retain the lowest demonstrated cost passing the same requirements and report its real ratio. Current qualification status keeps these gates open.
Evidence and reproduction
- Tactile materials and harness, source observations.
- Electronics review, native shared-COM rejection.
- Power and packaging, battery calculation.
- Cost calculator, public assembly fees, generated arithmetic.
Run python3 hardware/glove-v4/value-engineering/build_cost_review.py from the repository root. It binds the comparison to the current requirement/component hashes and counts physical PCB pads and placement sides. The native simulation runner and compressed waveforms are retained separately; they establish model behavior under documented assumptions, not physical glove measurements.