Nano/Micro-Power Reference for Battery IoT & Wearables
← Back to: Voltage / Current References
Nano/Micro-Power Reference Overview
A nano/micro-power reference is a precision voltage source designed for battery-powered and energy-harvesting systems. It typically draws quiescent current from tens of nA up to a few µA while keeping usable accuracy and drift, enabling long-life IoT nodes, wearables and sensor hubs with duty-cycled wake and measurement windows.
What makes nano/micro-power references different?
Unlike classic precision references that run at tens or hundreds of µA, nano/micro-power devices are biased to draw only nA–µA in steady state. They trade some bandwidth, noise and startup speed to keep long battery life while still providing a stable reference node for ADCs, comparators and supervisors in ultra-low-power systems.
Typical use cases
- Coin-cell sensor beacon: multi-year lifetime from a single cell, briefly waking the reference and ADC before each broadcast.
- Wearable heart-rate band: continuous or burst sampling close to skin, where reference Iq must stay in the µA region to meet battery targets.
- Energy-harvesting weather node: reference wakes only when harvested energy and sampling windows permit, with slow VIN ramps and long sleeps.
Key spec snapshot
- Quiescent current
- From tens of nA up to a few µA in steady state, depending on accuracy and startup targets.
- Output voltage
- Common fixed outputs around 1.2–3.3 V, sometimes with trimmed or ratiometric options.
- Input range
- Input ranges typically cover coin-cell and single Li-ion rails, for example 1.7–5.5 V.
- Temperature grades
- Consumer, industrial and automotive grades, with −40 °C up to +125 °C on selected families.
- Typical applications
- Battery IoT nodes, wearables, sensor hubs, data loggers and energy-harvesting endpoints.
Key Traits & Trade-Offs of Nano/Micro-Power References
Nano and micro-power references are tuned for nA–µA quiescent current rather than maximum dynamic performance. They keep a usable level of accuracy and drift for battery-powered IoT nodes and wearables, but trade away some noise, bandwidth, startup speed and load drive compared with classic precision bandgap or buried-Zener devices. This section makes those trade-offs explicit so you can judge where these devices fit and where they do not.
How pushing Iq down changes behaviour
In a classic precision reference, tens or hundreds of microamps bias the bandgap core and amplifier so that noise, bandwidth, line and load regulation and startup time all look comfortably “stiff.” When bias current is pulled down into the nA–µA region, the same circuits become slower and more delicate: noise rises, PSRR falls, and output drive becomes weak, especially into capacitive or distributed loads.
In low-duty-cycle IoT systems this is often acceptable, because the reference only needs to hold a stable node for a short sampling window. The payoff is much lower average current: the reference can sleep most of the time and wake just long enough to support an ADC conversion or comparator threshold.
High-level comparison
| Trait | Nano/Micro-Power Reference | Classic Precision Reference |
|---|---|---|
| Quiescent current (Iq) | Tens of nA up to a few µA in steady state, often duty-cycled or disabled between conversions. | Typically tens to a few hundreds of µA, usually kept always-on for best stability and noise. |
| Startup & settling | Longer and more VIN-dependent startup; may require warm-up conversions and generous settling windows after enable. | Generally fast and repeatable startup with short settling even under load, suitable for frequent or continuous conversions. |
| Noise & bandwidth | Higher 0.1–10 Hz noise and limited bandwidth; best suited for modest-resolution ADCs and slow measurement loops. | Lower noise and wider bandwidth, preferred for precision metering, fast control loops and high-resolution data converters. |
| Accuracy & drift | Accuracy and TC are usually “good enough” but not best-in-class; long-term drift may dominate over absolute accuracy in tiny systems. | Tighter initial accuracy and lower temperature coefficients available, with families optimised for calibration-grade performance. |
| Load drive capability | Weak drive; intended for high-impedance ADC or comparator inputs. Large capacitors, long traces or multiple loads can cause droop or instability. | Stronger drive and better load regulation; can tolerate bigger capacitors, heavier loads and complex distribution networks. |
When nano/micro-power references are a good fit
- Battery or energy-harvesting IoT nodes and wearables where multi-year lifetime or very small storage capacitors dominate the budget.
- Measurement chains using 10–12 bit ADCs with modest sampling rates, where noise and bandwidth requirements are not extreme.
- Designs that can duty-cycle the reference, keeping it enabled only during brief sampling windows and fully off for long sleeps.
- Reference nodes that drive mostly high-impedance inputs with short, well-guarded PCB traces and minimal capacitive loading.
When you should avoid nano/micro-power references
- High-resolution metering, precision DAC references or RF/IF chains that demand very low noise and fast settling under dynamic load.
- Reference rails that must drive large capacitors, long cables or multiple distributed analog blocks without a dedicated buffer amplifier.
- Control loops with tight startup timing where long, temperature-dependent reference settling cannot be tolerated or calibrated out.
- Systems where the reference must stay on continuously and power is available from strong rails, making ultra-low Iq less valuable than precision.
From a sourcing point of view, nano/micro-power references often come as sub-families of larger reference portfolios. When comparing vendors, always check how Iq, startup time, noise and temperature limits are specified and tested, as datasheets may use different operating points or “typical use” conditions.
Key Traits & Trade-Offs
Nano and micro-power references deliberately push quiescent current into the nA–µA region. The reward is multi-year lifetime in coin-cell or energy-harvesting systems, but the price is higher noise, slower startup and settling, weaker load drive and, in some families, looser accuracy or temperature drift than classic precision references. This section makes those trade-offs explicit.
Trait-by-trait comparison
| Trait | Classic Precision Reference | Nano/Micro-Power Reference |
|---|---|---|
| Noise | Low 0.1–10 Hz and broadband noise with ample bandwidth; preferred for precision metering and fast control loops. | Higher low-frequency noise and limited bandwidth; acceptable for modest-resolution ADCs and slow measurement chains. |
| Startup time | Short, repeatable startup and settling across VIN and temperature, often negligible in system timing budgets. | Longer, more VIN- and load-dependent startup; may require warm-up conversions and explicit tSETTLE margins. |
| Load capability | Stronger output stage and tighter load regulation; can drive larger capacitors, several loads and moderate wiring. | Weak drive intended for high-impedance inputs; large CLOAD, long traces or multiple sinks can cause droop or instability. |
| Accuracy & TC | Tighter initial accuracy options and lower temperature coefficients; variants optimised for calibration-grade references. | “Good enough” accuracy and TC for many IoT nodes; long-term drift and temperature corners may dominate the error budget. |
| Quiescent current (Iq) | Tens to a few hundreds of µA, normally kept always-on; power cost is acceptable in mains or high-capacity battery systems. | Tens of nA up to a few µA, often duty-cycled or disabled between conversions; enables multi-year life on small cells. |
Design perspective: when you can afford the trade-offs
- Sampling rates are low and resolution is in the 8–12 bit range, so some extra noise and slower settling are acceptable.
- The reference node only drives ADC or comparator inputs through short, well-guarded traces with minimal capacitive loading.
- Firmware already runs in duty-cycled bursts and can easily schedule enable, warm-up conversions and measurement windows.
- Battery life, coin-cell capacity or harvested energy budget is the primary constraint, and saving tens of µA matters more than absolute noise.
Sourcing perspective: what to watch in datasheets
- Startup time and settling are sometimes buried in graphs or application notes rather than the main electrical tables.
- Noise and drift specs may use different bandwidths, averaging methods or test circuits between vendors, complicating one-to-one comparisons.
- Low-Iq variants can sit in a different family or suffix from standard references, with different temperature grades and lead times.
- When preparing a BOM, always capture acceptable Iq range, maximum startup time and noise/drift expectations to avoid surprises at sample stage.
Internal Architecture & Bias Techniques
High-level block structure
A nano/micro-power reference still looks like a familiar bandgap-based architecture: a startup and bias block feeds a low-current bandgap core, which in turn is watched by an error amplifier and compensation network, and optionally buffered before driving the VREF output node. The difference lies in how aggressively each element is starved of current and, in some devices, only woken for brief periods.
Bias techniques that push Iq into nA–µA
- Subthreshold bias: transistors operate below threshold so that tiny currents set the bandgap and amplifier operating points, at the cost of larger device mismatch and stronger process and temperature dependence.
- High-R current mirrors: extremely large resistors or pseudo-resistors mirror leakage-level currents into the core, reducing Iq but increasing sensitivity to leakage and parasitic capacitances.
- Dynamic or switched bias: extra current is injected only during startup or while sampling; in deep sleep the amplifier and buffer are almost fully off.
How bias choices show up in real-world behaviour
- Startup becomes more sensitive to VIN ramp rate and output capacitance, and can vary with temperature much more than in high-Iq devices.
- Output response to load or ADC sampling transients slows down; the reference may droop or ring if CLOAD is large or wiring is long.
- PSRR and line regulation degrade at higher frequencies, so nano-power references prefer quiet rails or extra filtering ahead of them.
- Startup and bias blocks are tuned for very small currents, so VIN ramp and output capacitance strongly affect startup time.
- The low-I bandgap and error amplifier provide a stable reference node but have limited bandwidth and PSRR.
- Dynamic bias controller raises current only during startup or sampling, then lets the core and buffer fall back to nano/micro-amp levels.
Iq, Noise & Dynamic Operating Modes
Datasheets usually quote quiescent current under static conditions, with the reference enabled and settled. In real IoT firmware, the reference is often duty-cycled, so average current depends on how long it stays active, how deep the sleep state really is, and how much extra current flows during startup, ramp and sampling bursts.
A simple way to estimate average reference current over a periodic measurement cycle is:
Iq_avg = (Iq_active × t_active + Iq_sleep × t_sleep) / T_cycle
Here T_cycle is the full measurement period. Iq_active captures the steady-state plus startup and sampling peaks
while the reference is on, and Iq_sleep captures the leakage or deep-sleep current when it is disabled.
Static versus dynamic Iq
The “Iq” line in a datasheet is usually a static number: the reference is enabled, VIN is in its nominal range, the output is settled and the load is minimal. Real firmware adds dynamic pieces on top of that static current: extra current while startup circuits fire, additional peaks while the amplifier slews and while ADC sampling capacitors tug on VREF.
When the reference is duty-cycled, the static Iq becomes the active component in the equation above.
Peaks during startup and sampling stretch t_active and effectively raise Iq_active,
while leakage in the off-state sets Iq_sleep. Accurate average current numbers therefore need both timing and waveform knowledge,
not just the static table entry.
Noise behaviour as Iq is reduced
Pushing the bias current down shrinks amplifier bandwidth and raises effective 1/f noise. At very low Iq the reference behaves more like a slow, noisy node: its spectrum is dominated by low-frequency noise and it takes longer to settle after disturbances.
In sampled systems, the practical impact is that ADC conversions should be scheduled inside a quiet window where VREF is fully settled. Averaging multiple samples reduces broadband noise, but flicker noise and slow drift still leak into long integration periods, especially in duty-cycled designs with long intervals between updates.
Common dynamic operating modes
- Always-on reference: simplest firmware and best noise and settling behaviour; average Iq is essentially the static value, suitable when power budget is comfortable.
- Duty-cycled reference: enable only around each measurement, wait for startup and settling, then convert and disable; cuts average Iq sharply but adds timing complexity and startup spread.
- Burst-sampling mode: keep the reference on for a short burst while multiple channels or samples are acquired, then disable for a longer sleep; useful when ADC or sensor readout overhead dominates.
Practical tips:
-
Budget explicit
t_startupandt_settlebefore sampling. - Measure real Iq waveforms on the bench to calibrate the model.
- Group ADC channels into bursts to amortise startup overhead.
Startup, Wake Timing & Sampling Windows
From a firmware point of view, the critical question is not “what is the typical startup time” but “after I assert enable, how long until every conversion is reliably within the error budget?” Nano/micro-power references respond more slowly to VIN and load, so the first one or two conversions can be biased low or noisy.
What controls startup time?
- VIN level & ramp rate: slow or marginal rails delay startup circuits and can leave the bandgap core hovering near its trip point for longer.
- Output capacitance: large COUT or long traces take more time and current to charge; in nano-power devices this directly stretches tSTARTUP and tSETTLE.
- Temperature: cold corners slow subthreshold and leakage-based bias circuits; startup time can easily double between room and −40 °C.
- Previous state: restarting from a partially discharged output node behaves differently from a fully cold start or a short glitch.
First-conversion error checklist
- Assume the very first ADC conversion after enable is suspect: VREF may still be ramping or settling under load.
- ADC sampling capacitors interact with the reference node; in nano-power devices their charge pulses cause visible droop or undershoot.
- Plan to discard one or more warm-up conversions in firmware, especially at cold temperature or when VIN ramps slowly.
- Characterise “first-good” conversion index on the bench across VIN, COUT and temperature, then hard-code a conservative margin.
Planning the wake & sampling sequence
- Enable the reference: assert EN or power gate while MCU wakes and configures the ADC and sensors.
- Wait for tREF_SETTLE: insert a fixed delay that covers datasheet startup plus extra margin for COUT and worst-case temperature.
- Perform warm-up conversions: take one or more dummy samples and discard them to flush early transients from the reference node.
- Acquire useful samples: collect the N conversions that will be averaged or filtered for the measurement result.
- Disable the reference: once all conversions are complete, gate EN low again to return to the low-leakage state.
Quick timing rules of thumb
- Reserve at least 2–3× the datasheet startup time as the initial tREF_SETTLE budget for nano-power parts.
- At cold temperature or with large COUT, expand the delay further; verify with worst-case bench measurements.
- For 10–12 bit measurements, discarding the first 1–2 conversions is often enough; higher resolution may need more warm-up samples.
- Keep ADC sampling windows clustered while the reference is on, rather than scattering them, to avoid frequent re-start penalties.
Design-In for Coin-Cell & Single-Cell Li-Ion Systems
This section walks through how to pick a nano/micro-power reference for a coin-cell or single-cell Li-ion board, then connect it cleanly into ADCs, comparators and LDOs without breaking the power or accuracy budget.
Fix VREF, required ADC resolution and full temperature range. This sets accuracy, TC and noise requirements before you look at Iq.
Use the duty-cycle model from the Iq section to estimate Iqavg over the real wake/sleep schedule, including startup and sampling peaks.
Compare datasheet tstartup/tsettle (with margin) to the time between MCU wake and “first good” ADC conversion in your system.
If VREF must drive several ADC channels, long traces or comparator thresholds, consider a low-power buffer op amp instead of loading the reference directly.
Confirm package height, pinout, ESD rating and surge/EMC robustness fit the board and enclosure, especially for wearables and exposed sensor nodes.
Integration patterns with ADCs & comparators
In simple IoT nodes, VREF usually feeds one internal ADC. In more complex boards it may fan out to multiple channels and comparators. A nano/micro-power reference prefers short, high-impedance connections.
- Single ADC, single reference: connect VREF directly to the MCU’s reference pin, keep traces short and shielded, and use the minimum recommended COUT.
- Multiple ADC channels sharing VREF: route VREF as a short star point near the MCU, avoid stubs and large capacitances, and schedule conversions back-to-back while VREF is on.
- Comparator thresholds derived from VREF: derive thresholds using high-value resistor dividers close to the comparator; keep the divider leakage small versus reference bias and implement hysteresis on the comparator side, not by loading VREF harder.
Common pitfalls in nano/micro-power designs
- Oversized output capacitor: “just in case” capacitors make startup very slow in nano-power references and can cause unstable ramps.
- Long, contaminated traces: routing VREF across the board introduces leakage, noise pickup and coupling into other rails.
- Too many series elements: stacking dividers, jumpers or muxes on the reference node creates complex impedance and leakage paths that the tiny bias current cannot tolerate.
- Ignoring enable timing: enabling VREF too late in the wake sequence leaves no time for settling and warm-up conversions.
Example nano/micro-power reference choices (for coin-cell / Li-ion)
| Brand | Part number | Key traits | Why it fits nano/micro-power IoT |
|---|---|---|---|
| Texas Instruments | REF35 family | Series reference, nanopower (~650 nA typ), ±0.05 % initial accuracy, ~12 ppm/°C drift, up to 10 mA load capability.:contentReference[oaicite:0]{index=0} | Good balance of ultra-low Iq, strong accuracy and decent output drive; EN pin makes duty-cycling easy, ideal as main VREF for MCU ADCs in coin-cell or single-cell Li-ion designs. |
| Analog Devices (Maxim) | MAX6006 | Ultra-low-power shunt reference, guaranteed <1 µA operating current, 0.2 % initial accuracy, up to 30 ppm/°C TC, tiny SOT-23 package with 1.25–3.0 V options.:contentReference[oaicite:1]{index=1} | Simple 2-pin shunt topology works well as a reference on sensor front-ends or comparator thresholds where VIN range is wide; ultra-low Iq and small package suit space-limited IoT nodes. |
| Specialised nano-power vendor | Ai Linear nano-power VREFs | Dedicated nano-power portfolio with voltage references targeting ≤350 nA bias currents and other ultra-low-power analog blocks.:contentReference[oaicite:2]{index=2} | Useful when system-level power budget is extremely tight and you want a matched family of nano-power blocks (reference, ADC, bias sources) for long-life metering or sensing. |
When preparing a BOM, you can treat these part numbers as a reference candidate pool: first filter by VREF, Iq and accuracy, then screen by price, lead time and automotive qualification (AEC-Q100) requirements.
Layout, Leakage & Guarding for nA-Level References
At nano/micro-amp bias levels, tiny leakage currents on the PCB translate directly into millivolt-level errors on VREF. A “harmless” 1 GΩ leakage path from a 1.8 V reference node pulls about 1.8 nA; if the reference can only spare a few nanoamps, that leakage becomes a significant fraction of its bias budget.
As a first approximation, you can model leakage as a parallel resistor Rleak: the leakage current is roughly I = VREF / Rleak. When this current is on the same order as the reference’s own bias current, it will noticeably pull the output up or down.
Typical leakage sources around VREF
- PCB surface contamination: fingerprints, humidity and ionic residues create high-value resistive films along the board surface.
- Flux & solder residues: incomplete cleaning leaves moisture-sensitive paths between VREF and neighbouring nets or ground.
- High-impedance inputs: ADC and comparator inputs use very small geometries; their ESD and protection structures have finite leakage.
- Connectors & cables: long sensor cables and exposed connectors collect dirt and moisture, effectively adding Rleak from VREF to other nodes.
Guarding & keep-out practices
- Use a guard ring driven at VREF potential around the high-impedance node on the top layer to intercept leakage paths.
- Define a keep-out zone: avoid vias, unrelated traces and polygons under or near the VREF node to minimise surface leakage.
- Control solder mask openings so that only the necessary pads are exposed; avoid large bare-copper areas near VREF.
- Keep reference traces short, straight and away from switching nodes or RF lines to reduce capacitive coupling and noise injection.
Production checklist for nA-level references
- Verify cleaning process (water-based, semi-aqueous or no-clean) actually delivers GΩ-level insulation in humidity tests.
- Choose PCB materials and finishes with good insulation resistance for the expected humidity and contamination levels.
- Include environmental stress tests (damp heat, temperature cycling) to catch slow leakage paths that only appear after soaking.
- For field returns, measure leakage around the VREF node directly (megohmmeter) rather than only swapping devices.
Validation, Characterization & Long-Term Drift
Nano/micro-power references must be validated both on the bench and in real duty-cycled operation. The goal is to prove that startup, noise, average Iq and long-term drift remain within budget across VIN, temperature, load and aging.
Bench characterization
- Sweep Iq versus VIN and temperature: test at VINmin, VINnom, VINmax across −40 °C, 25 °C and 85 °C (or the relevant range).
- Measure startup time versus output capacitor: compare recommended COUT to “over-sized” cases that stress nano-power startup.
- Characterize 0.1–10 Hz noise: use a low-noise preamp, narrowband filter and FFT or time-domain RMS calculation to capture flicker noise.
- Log VREF at each corner while toggling EN to detect any metastable or stuck-start conditions.
Application-level tests
- Instrument the real firmware duty-cycle: measure average Iq with a current probe or sense resistor over many wake/sleep cycles.
- Test worst-case droop: wake multiple nodes or ADC channels at once and check reference sag while the system is in its heaviest activity.
- Run noise and accuracy checks with full sensor chain connected, not just the bare reference on the bench.
- Verify that discard counts for “warm-up conversions” are sufficient under real system timing, not only ideal bench delays.
Long-term drift & aging plan
- Storage at elevated temperature (for example, 85 °C for 1000 hours) with periodic VREF measurements to estimate long-term drift.
- Thermal cycling across the operating range to detect hysteresis and mechanical stress effects on VREF.
- Duty-cycled endurance: run representative wake/sleep patterns over days or weeks and track any systematic drift or shift in startup behaviour.
- Record all results in a validation matrix so that coverage across VIN, temperature, duty-cycle and load is visible at a glance.
Example validation matrix coverage
| Temperature | VINmin | VINnom | VINmax |
|---|---|---|---|
| −40 °C | DC + duty-cycled test (light & heavy load) | DC + duty-cycled test (light & heavy load) | DC + duty-cycled test (light & heavy load) |
| 25 °C | DC + duty-cycled test (typical load) | DC + duty-cycled test (typical & worst load) | DC + duty-cycled test (typical load) |
| 85 °C | DC + duty-cycled test (light & heavy load) | DC + duty-cycled test (light & heavy load) | DC + duty-cycled test (light & heavy load) |
BOM & Procurement Notes for Small-Batch IoT
For small-batch buyers, the key is to capture enough engineering detail about the reference so that sourcing can pick the right family, avoid late surprises on startup, Iq or drift, and keep a second source in view where possible.
BOM essentials (fields to capture)
- VREF value: nominal reference voltage and acceptable tolerance.
- Iq target & duty-cycle pattern: static Iq, expected average Iq and basic wake/sleep profile.
- TC & drift limits: maximum temperature coefficient and long-term drift the system can tolerate.
- Temperature grade: commercial, industrial or extended/automotive range.
- Package & height: SOT-23, SC70, WLP, maximum height allowed by enclosure.
- Automotive grade? explicitly mark if AEC-Q100 or similar is required.
- Startup constraints: maximum startup and settle time budget, plus any special enable logic.
Series & supply risks to watch
- Lifecycle & EOL risk: some low-power references are “niche” families; check lifecycle status and design-in newer series where possible.
- Regional stock & MOQ: verify minimum order quantities and typical lead times in your target warehouse region.
- Grade price gaps: automotive or extended-temperature variants can be significantly more expensive and less available than industrial grade.
- Package supply skew: one package (e.g., SOT-23) may be plentiful while another (e.g., WLP) has long lead times.
- Hidden specs: startup, noise and drift may be in graphs or app notes; ask engineering to flag any “must-have” behavioural details.
Submit BOM for nano/micro-power references
Share your BOM and constraints so we can shortlist suitable nano/micro-power references and, where possible, suggest second-source options.
- Include your primary and fallback brands (if any) and acceptable alternates.
- Provide a target cost range per reference and expected annual volume.
- Attach schematics or notes if startup timing, noise or drift are especially sensitive.
We use your data only to prepare part suggestions and sourcing options for this project.
Example reference families to consider
The shortlist below is only an illustration of nano/micro-power reference families that often appear in IoT and low-power designs; your final choice should follow the selection and validation steps above.
| Brand | Example family / PN | Why it is relevant for nano/micro-power BOMs |
|---|---|---|
| Texas Instruments | REF35 / REF35-Q1 | Nanopower series reference family with precise specifications and both standard and automotive-grade options, suited for main ADC VREF in battery-powered systems. |
| Analog Devices (Maxim) | MAX6006–MAX6009 | Ultra-low-power shunt references in tiny SOT-23 packages; useful where the reference must work across a wide VIN range with very small bias currents. |
| Specialised nano-power vendor | Ai Linear voltage references | Dedicated nano-power reference portfolio focused on sub-microamp bias; can be interesting for extremely aggressive lifetime and power budgets. |
Request a Quote
FAQs on Nano/Micro-Power References
The questions below reflect common design, validation and procurement issues for nano/micro-power references in IoT, wearables and other long-life systems. Each answer is designed to be copy-ready for PAA/JSON-LD and traceable back to a specific section of this page.
How low can Iq go before noise and TC become unacceptable for a 12-bit ADC?
For a 12-bit ADC, you normally want reference noise and drift comfortably below 0.5 LSB over your conversion window and temperature range. As Iq falls into the sub-microamp region, bandwidth shrinks and 1/f noise rises, so millivolt-level variation becomes easier. Below that point, longer averaging or higher Iq is usually safer.
How do I budget startup time for a nano-power reference in a duty-cycled IoT node?
Start from the datasheet startup time with recommended COUT, then multiply by two or three to cover cold temperature, large capacitors and board parasitics. Place this delay between enabling the reference and taking “good” ADC conversions, and discard early samples. Finally, verify worst-case behaviour by sweeping VIN, temperature and load on the bench.
How do I estimate the average Iq of a reference that only wakes for brief ADC conversions?
Model the cycle with an active interval and a sleep interval, then use
Iq_avg = (Iq_active × t_active + Iq_sleep × t_sleep) / T_cycle.
Iqactive must include startup peaks and sampling bursts, not only the static table value.
Measuring real current waveforms on the bench lets you refine the model and confirm duty-cycle savings.
What output capacitor and load are safe for ultra-low Iq references without causing instability?
Stay within the datasheet’s COUT range and ESR window, because nano-power references have carefully tuned compensation. Oversized low-ESR capacitors or heavy dynamic loads can slow startup and trigger ringing or long tails. If you must drive larger capacitors or multiple channels, consider adding a low-power buffer op amp after the reference node.
How do PCB leakage and humidity affect a nano-power reference accuracy?
At nanoamp bias levels, a one-gigaohm leakage path from the reference node to ground or another rail already steals nanoamps of current. That is comparable to the bias available to hold VREF, so humidity, flux residues and surface films easily create millivolt offsets. Guard rings, keep-out regions and clean boards are essential to keep leakage under control.
When should I add a buffer op amp in front of the reference node?
Add a buffer when the reference must drive several ADC channels, long traces or comparator thresholds, or when dynamic load steps are large. A suitable low-bias op amp isolates the delicate reference core from sampling capacitors and cable capacitance, trading a little extra Iq for better settling, lower droop and simpler layout constraints.
How do I translate ppm/°C into mV drift on a 1.8 V nano-power reference?
Multiply the TC in ppm/°C by the temperature swing and by VREF. For example, a 50 ppm/°C reference over a 60 °C span at 1.8 V gives 50 × 60 × 1.8 / 106 ≈ 5.4 mV of drift. Compare that number to your full-scale and LSB size to decide whether the part is suitable without additional calibration.
What is the best way to validate long-term drift for wearables and medical patches?
Use accelerated stress: store devices at elevated temperature and humidity, and run temperature cycling that mimics body or ambient conditions. Measure VREF at regular intervals, always using the same calibrated meter and fixturing. Combine this with long duty-cycle endurance tests so you see both aging drift and any change in startup behaviour over time.
How do I share one nano-power reference across multiple ADC channels without droop?
Keep the VREF trace short and routed as a low-impedance star point near the ADC pins, with only the recommended COUT. Group ADC conversions into a burst while the reference is on, and avoid large external capacitors or extra dividers on the node. If droop is still visible, add a low-power buffer stage.
What are typical failure modes when enabling and disabling the reference at high duty cycles?
Common issues include incomplete startup at cold corners, references latching in a meta-stable state, and slow VREF decay that corrupts the next wake interval. Frequent cycling can also stress internal bias structures, slightly shifting drift. Robust designs characterise these behaviours on the bench and add extra margin to delays and warm-up conversions in firmware.
How do I compare different vendors’ Iq specs when they use different test conditions?
First normalise the conditions: VIN, temperature, load and enable state. Many datasheets quote Iq at VINnom, 25 °C and no load, while your node might run colder, lower or duty-cycled. Where numbers are not directly comparable, measure Iq on identical boards under your real operating profile before final vendor selection.
What information should I provide in a BOM to help sourcing a nano/micro-power reference?
Include the required VREF, accuracy, TC limit, average Iq target and duty-cycle pattern, plus temperature grade and package constraints. Add any startup or drift requirements, preferred brands and whether automotive qualification is needed. This lets sourcing match you to the right family and plan viable alternates without compromising lifetime or measurement accuracy.
