Engineering guide · Medical wearables
ECG, EEG and EMG Front-End Design Guide for Medical Wearables (2026)
A wearable biosignal front end has to resolve signals of a few microvolts (EEG) to a few millivolts (ECG, surface EMG) in the presence of electrode offsets of up to a few hundred millivolts and mains interference, which is why most designs start from an integrated biopotential AFE such as TI’s ADS1299 (8-channel EEG), ADS1293 or ADS1292R (ECG), or an instrumentation-amplifier chain around a part like the INA333. Turning it into a medical product adds IEC 60601-1 safety, IEC 60601-1-2 EMC, a particular standard such as IEC 60601-2-47 for ambulatory ECG, and IEC 62304 software processes; as a planning range, allow 9–18 months and $120,000–$400,000+ to a regulatory submission.
Key facts
| Signal | Typical amplitude | Useful bandwidth | Typical front end | Particular standard |
|---|---|---|---|---|
| ECG | About 0.5–4 mV | 0.05–150 Hz diagnostic; narrower for monitoring | ADS1293, ADS1292R, ADS1298-class | IEC 60601-2-25, -2-27, -2-47 |
| EEG | About 10–100 µV | About 0.5–70 Hz (wider for research) | ADS1299 | IEC 60601-2-26 |
| Surface EMG | Tens of µV to a few mV | About 20–500 Hz | ADS1298/ADS1299-class or INA333 chain | IEC 60601-2-40 |
Amplitudes and bandwidths are typical textbook ranges; your requirements depend on the clinical claim. AFE figures come from the manufacturer’s product pages (sources at the end). Cost and timeline ranges are Rapid Circuitry planning estimates, consistent with the regulated-device range in our cost benchmarks; clinical studies are budgeted separately.
Choosing the analog front end
| Part | Channels / resolution | Noise (datasheet) | Power | Best for |
|---|---|---|---|---|
| ADS1299 | 8 ch, 24-bit, 250 SPS–16 kSPS | 1 µVpp (70 Hz BW) | Moderate; power-down modes | EEG, high-channel EMG, research-grade ECG |
| ADS1293 | 3 ch, 24-bit, up to 25.6 ksps | 7 µVpp (40 Hz BW) | 0.3 mW per channel | Low-power multi-lead ECG, pace detection |
| ADS1292R | 2 ch, 24-bit, 125 SPS–8 kSPS | 8 µVpp (150 Hz BW, G = 6) | 335 µW per channel | Single-lead ECG plus respiration |
| INA333 + ADC | 1 channel per amplifier | 50 nV/√Hz; 1 µVpp 0.1–10 Hz | 50 µA quiescent | Few channels, custom analog filtering, EMG envelopes |
Integrated AFEs include the pieces every biopotential design needs — programmable gain, right-leg or bias drive, lead-off detection, test signals — and they make the noise performance repeatable. A discrete instrumentation amplifier chain still makes sense when you need very few channels, analog processing such as an EMG envelope for a simple prosthetic controller, or the lowest BOM cost. Other vendors, including Analog Devices, offer comparable ECG AFEs; compare them on the same criteria.
Two of our projects show both routes. The 16-channel wearable EEG headset uses one ADS1299 per ear so each side digitises its own electrodes, while the EMG sensor front end for a bionic hand uses a discrete amplifier chain with an envelope output for proportional control. For a slim fitness EMG wearable, the wireless EMG sensor pairs an ADS1292 with a BLE module.
Noise: building a budget you can defend
- Start from the requirement: for EEG, microvolt-level noise across the band of interest; for monitoring ECG, far less stringent but with strong mains rejection and baseline stability.
- Add the contributions: AFE input-referred noise, input filter resistors (thermal noise), electrode–skin interface noise (much higher for dry electrodes), and ADC quantisation at the chosen gain.
- Common-mode rejection in the real system is limited by electrode impedance mismatch, not by the datasheet CMRR; bias drive and shielding recover much of it.
- Measure with inputs shorted, then with an electrode phantom, then on subjects at rest and in motion; report which figures are measured and under what conditions.
- Choose sample rates and digital filters to the clinical bandwidth; do not over-filter diagnostic ECG, where standards constrain the frequency response.
The electrode interface
| Electrode | Contact impedance | Trade-offs |
|---|---|---|
| Wet gel Ag/AgCl | Low and stable | Best signal quality; gel dries and irritates skin over days |
| Dry metal or conductive polymer | High and variable | Comfortable and reusable; more noise and motion artefact |
| Active (buffered) electrode | Buffer at the electrode | Tolerates high impedance and cable movement; more power and wiring |
| Textile electrode | High, changes with sweat and pressure | Garment integration; demanding signal processing |
- Tolerate electrode DC offsets of a few hundred millivolts without saturating at the chosen gain.
- Keep input bias currents low; with dry electrodes they create offsets and noise.
- Implement lead-off detection (DC or AC) so the system knows when contact is lost.
- Protect inputs against ESD and, where the use environment requires it, defibrillation, without adding leakage or noise.
- Skin-contact materials need biocompatibility evaluation under the ISO 10993 series.
Isolation and IEC 60601-1
IEC 60601-1 (edition 3.2, 2005 with amendments 2012 and 2020) sets basic safety for medical electrical equipment. For biosignal wearables the key concepts are:
- Applied part type: most body-surface ECG, EEG and EMG electrodes are BF (body floating) applied parts; CF (cardiac floating) applies to direct cardiac contact.
- Leakage currents: patient leakage limits in normal condition are 10 µA for CF and 100 µA AC for BF applied parts, with higher single-fault limits.
- Means of protection: anything mains-connected needs two means of patient protection (2 MOPP) between mains and the patient, which drives isolation barriers, creepage and clearance.
- Charging while worn: a battery-powered wearable is simple while on battery; the risk appears when it is charged or connected by USB. Either prevent use while charging or use a medical-grade isolated supply and isolated data path.
- Home use: products used at home also follow IEC 60601-1-11.
Our IEC 60601 explainer covers the standard family in more depth.
EMC and wireless: IEC 60601-1-2
IEC 60601-1-2 (edition 4.1, 2014 with amendment 2020) sets emission and immunity requirements based on the intended use environment. Home-healthcare devices face higher immunity levels than professional ones, including ESD of ±8 kV contact and ±15 kV air and proximity fields from mobile phones and other RF transmitters. For a biosignal device, the question is not only “does it survive?” but “does it keep its essential performance?” — for example, no false arrhythmia alarms under interference. The Bluetooth radio also brings radio-approval and cybersecurity requirements (see our EU and US market-access guide).
Software: IEC 62304 and cybersecurity
Firmware, the companion app and any cloud algorithms that are part of the device follow IEC 62304 (2006, amended 2015): a software development plan, requirements, architecture, detailed design and verification scaled to the software safety class (A, B or C), management of third-party software (SOUP), and maintenance and problem-resolution processes. A second edition of IEC 62304 has been in ballot during 2026; check its status when you plan a new programme. In the US, devices with software that connect to networks are also subject to the FDA’s cybersecurity requirements for “cyber devices”, including a software bill of materials in the premarket submission. See our IEC 62304 explainer.
Regulatory pathways at a high level
| Market | Typical route | Notes |
|---|---|---|
| United States (FDA) | Many ECG and EEG devices are Class II, cleared through 510(k) against a predicate device | QMSR (ISO 13485:2016 incorporated) effective 2 February 2026; 510(k) user fee $26,067 / $6,517 small business (FY2026), $28,653 / $7,163 (FY2027) |
| European Union (MDR 2017/745) | Active diagnostic and monitoring devices are usually Class IIa or IIb under Rule 10; software under Rule 11 | Notified body assessment and clinical evaluation; a new device must meet the MDR in full |
| India (CDSCO) | Licence under the Medical Devices Rules, 2017, by risk class A–D | State authority for Class A/B manufacture, CDSCO for C/D and imports; test licences for clinical units |
Wellness products that make no medical claims can fall outside these regimes, but the claims on your packaging, website and app decide that, not the hardware. Our medical device FDA compliance white paper and India certification guide cover the pathways in more detail.
From research prototype to clinical-study units
Many biosignal products start as a lab prototype on development kits. Turning it into 20–100 units for a clinical pilot means:
- Freezing requirements and the intended use, and writing a risk analysis per ISO 14971.
- Redesigning the electronics for isolation, leakage, EMC and battery safety, and using biocompatible skin-contact materials.
- Controlled firmware builds with traceable versions on every unit, and a data pipeline that preserves raw signals for analysis.
- Pre-compliance safety and EMC testing, and the documentation your ethics committee and regulator will ask for.
- A small, controlled build with incoming inspection, functional test and serial-number traceability.
Keep what the science depends on — electrode montage, sampling, filtering and algorithms — and change it only with a documented comparison against the prototype, so the redesign does not break the research results. Our wearable EEG headset case study shows the step before this one: a research-alpha design package, revised with the client’s engineers until they accepted it as their baseline.
Typical stages, costs and timeline
| Stage | Typical time | Planning cost (India-based team) |
|---|---|---|
| Feasibility on evaluation hardware; requirements and risk analysis | 4–8 weeks | $8,000–$25,000 |
| Custom AFE and system board, firmware, app (EVT) | 8–12 weeks | $25,000–$70,000 |
| Design iteration, pre-compliance, verification documentation (DVT) | 12–20 weeks | $40,000–$120,000 |
| Formal IEC 60601 testing support, design history file completion | 8–16 weeks | $30,000–$130,000+ incl. lab fees |
| Total to regulatory submission | 9–18 months | $120,000–$400,000+ |
How Rapid Circuitry works on biosignal devices
We design biosignal front ends, low-power wearable electronics, firmware and companion apps for medical device and healthcare IoT clients, with processes aligned to ISO 13485 and IEC 62304 and designs prepared for certification testing at accredited labs. Rapid Circuitry does not hold ISO 13485 certification; we work under our clients’ quality systems. Related work includes our cardiac remote-monitoring wearable and the EMG-controlled prosthetic arm case studies, a 20 mm neonatal SpO₂ and heart-rate wearable built on the MAX86161 optical (PPG) front end, and our wearable devices service.
Frequently asked questions
What is involved in designing a custom 8-channel EEG board with an ADS1299 and dry electrodes?
The ADS1299 provides eight simultaneous 24-bit channels with about 1 µVpp input-referred noise in a 70 Hz bandwidth, plus bias drive and lead-off detection, so the hard work is around it: input protection and filtering that do not add noise or leakage, active (buffered) electrodes or very high input impedance to cope with dry-electrode impedance, a low-noise isolated or battery-only power design, careful layout and shielding, motion-artefact handling, and firmware for sampling, impedance checks and streaming. Plan a feasibility phase on evaluation hardware, then two board iterations, and validate noise with shorted inputs and on real subjects.
Can we use our Arduino or ESP32 prototype in a clinical study, or do we need a proper redesign?
Usually you need a redesign. A clinical study needs devices whose safety and performance are documented: patient isolation and leakage currents per IEC 60601-1, EMC behaviour, risk management per ISO 14971, biocompatible skin-contact materials, controlled firmware and traceable units. Development boards and hobby modules are not designed or documented for that. Ethics committees and regulators (for example an FDA IDE decision for significant-risk studies, or a CDSCO permission in India) will ask for this evidence.
How much does it cost and how long does it take to develop Class II medical device electronics and software up to a 510(k) submission?
As a planning range, electronics, firmware and app development for a Class II biosignal wearable, including the design-control documentation and IEC 60601 test preparation, typically costs $120,000–$400,000+ with an India-based team and takes 9–18 months to a 510(k) submission. Test-lab fees, clinical studies and FDA user fees are extra; the FDA’s 510(k) user fee is $26,067 standard or $6,517 for qualifying small businesses in fiscal year 2026, rising to $28,653 and $7,163 from 1 October 2026.
Does outsourcing firmware to a vendor that is not ISO 13485 certified create problems for our FDA submission?
Not by itself. The legal manufacturer is responsible for the quality system; since 2 February 2026 the FDA’s Quality Management System Regulation incorporates ISO 13485:2016 by reference, and it requires you to control suppliers through purchasing controls and your design controls. A supplier without its own certificate can deliver IEC 62304-conformant software if it works under your procedures or equivalent documented processes, and you audit it. Rapid Circuitry, for example, is not ISO 13485 certified; our processes are aligned with ISO 13485 and IEC 62304 and we work under the client’s QMS.
How long does IEC 60601 testing take and what does it cost?
For a battery-powered wearable, formal IEC 60601-1 safety and IEC 60601-1-2 EMC testing, plus the relevant particular standard, typically takes 4–10 weeks of lab time once samples and the risk file are ready, and more if the product fails and needs changes. Lab fees vary widely by lab, scope and number of standards; ask several accredited labs for quotes and run pre-compliance first to avoid paying twice.
What is the difference between BF and CF applied parts?
Both are floating applied parts isolated from earth. CF (cardiac floating) parts are for direct cardiac application and have the strictest leakage limits — 10 µA patient leakage in normal condition — whereas BF (body floating) parts, used for most body-surface ECG, EEG and EMG electrodes, allow 100 µA AC in normal condition. Most wearable biosignal products are designed as BF applied parts, often defibrillation-proof where the use environment requires it.
Which front end suits an EMG armband for prosthetic control?
For a multi-channel armband, a multi-channel biopotential ADC such as an ADS1298-class or ADS1299 device gives synchronous raw EMG for pattern recognition. For a few channels driving simple proportional control, a discrete chain with an instrumentation amplifier such as the INA333, band-pass filtering around 20–500 Hz and an envelope detector can be smaller and cheaper. Either way, electrode placement, skin contact under movement and sweat dominate performance.
Sources
- Texas Instruments — ADS1299 — 8-channel, 24-bit, 1 µVpp noise (70 Hz BW), −110 dB CMRR
- Texas Instruments — ADS1293 — 3-channel, 24-bit, 0.3 mW/channel, 7 µVpp noise (40 Hz BW)
- Texas Instruments — ADS1292R — 2-channel, 24-bit, respiration impedance, 335 µW/channel
- Texas Instruments — INA333 — 50 µA instrumentation amplifier, 25 µV max offset
- IEC — IEC 60601-1:2005+AMD1:2012+AMD2:2020 — basic safety and essential performance, edition 3.2
- IEC — IEC 60601-1-2:2014+AMD1:2020 — EMC collateral standard, edition 4.1
- IEC — IEC 60601-2-47:2012 — ambulatory ECG systems
- ISO — IEC 62304:2006 — medical device software life-cycle processes
- FDA — Quality Management System Regulation (QMSR) — effective 2 February 2026; incorporates ISO 13485:2016
- FDA — MDUFA user fees — FY2026 and FY2027 510(k) fees
- EUR-Lex — Medical Device Regulation (EU) 2017/745 — EU MDR
- CDSCO — Medical devices and diagnostics — Medical Devices Rules, 2017 (checked September 2026)
This guide is general engineering information, not regulatory advice. Cost and timeline ranges are Rapid Circuitry planning estimates.