IWR6843 integrates 60–64 GHz FMCW transceiver/processing resources with three transmit and four receive channels. Bands/channel counts do not establish end-product detection range, angular accuracy, or human recognition. Define targets/scenarios first, then antennas, supply, chirps, sampling, and algorithms. Retain data that reproduce false detections and misses.
Task objectives must be specific to location and action
Expand “presence detection” into target classes, ranges, angles, poses, speeds, occlusion, mounting heights, and allowed response times. Stationary presence, moving targets, and range/velocity measurement are different tasks requiring different observation times/features. A hand-wave demonstration does not establish stationary or multitarget compliance.
| Scenario Input | Fixed Requirement | Output Definition |
|---|---|---|
| Target | Dimensions, material, orientation, and motion | Detectable conditions and difficult samples |
| Space | Distance, azimuth, pitch, and occlusion | Coverage Area and Blind Spots |
| Installation | Height, angle, enclosure, and mounting method | Structure drawing and calibration conditions |
| Business | Latency, false positive, and false negative definitions | Event determination and statistical unit |
First distinguish between package, antenna, and power modes
This page addresses IWR6843, not directly AOP antenna-in-package variants or other mmWave families. Conventional RF-interface designs require joint antenna-array, feedline, material, and fabricator-stackup design. Antenna structures affect channel-phase calibration; recheck after thickness, radome, or mounting-metal changes.
The device has digital, RF, baseband, I/O, and other rails, not one 3.3 V supply. Build mode-specific power tables and avoid mixing normal-LDO and bypass arrangements. Ripple limits have frequency/test conditions; one value is not a universal full-band threshold. Measure near corresponding supply pins and retain probe connections.
Keep high-speed clocks, peripherals, and switching-power nodes away from antennas/sensitive RF areas. Follow package-specific decoupling/returns. Preserve debug access to startup/download interfaces; confirm startup, readable firmware, and configuration before scene algorithms. Adjusting detection thresholds cannot fix intermittent data caused by startup failures.
Use calculations to check whether chirps support the task
Ideal range resolution is approximately ΔR＝c/(2B), where B is effectively sampled configured sweep bandwidth, not chip band coverage. At an assumed 3 GHz effective bandwidth, ideal resolution is about 0.05 m. Actual resolution depends on windows, SNR, targets, and processing; 5 cm is not universal ranging accuracy.
For a simplified single target, beat frequency fb≈2SR/c, with sweep slope S. Assuming S＝30 MHz/µs and R＝10 m gives about 2 MHz beat frequency. Check ADC sample rate, usable IF bandwidth, and configured range afterward. More software FFT points cannot recover signals lost beyond the receive chain.
Chirps per frame, repetition periods, and transmit scheduling affect velocity/angle processing and consume processing time/memory. Do not add chirps for stable displays while ignoring latency. Retain profile, chirp, and frame configurations; check actual output cadence against firmware processing capacity.
Data volume determines interface and debugging plan
Assume four receive channels, 256 complex samples per chirp, 128 chirps per frame, and 16 bits each for I/Q. Raw frame size is about 4×256×128×4＝524288 bytes; at 20 frames/s, about 10.49 MB/s before protocol overhead. These are format assumptions; verify actual sample packing for the selected mode.
Point clouds are much smaller than raw data but omit information useful for algorithm diagnosis. Decide whether raw ADC, intermediate spectra, or final detections answer the current question. Logs also need configuration, firmware, timestamps, and scene IDs to prevent mixing data from different settings.
Accept echoes, ranging, and application detection separately
| Validation Task | Reference Conditions | Recorded Output |
|---|---|---|
| Basic Link | Known target and fixed installation | Startup, noise floor, peaks, and stable data frames |
| Distance and Angle | Multiple locations, poses, and targets | Error, resolution, and unavailable regions |
| Housing Impact | Bare PCB and final enclosure for same scenario | Noise, peak, and blind spot changes |
| Business Determination | Including difficult samples and no-target scenarios | Missed detection, false detection, latency, and event logs |
Scene statistics need counts, durations, and criteria. No false alarms in an empty room does not establish stability with fans, moving curtains, or crossing people; one target location does not cover the whole space. Reproduce abnormalities with fixed hardware/configuration, then change one parameter at a time to separate antenna, threshold, and tracking effects.
Runtime and temperature are also configuration conditions
Transmit duty, processing load, and power modes change consumption/temperature. Record supply, chip-temperature observations, and application stability in actual enclosures/orientations, comparing hot behavior with cold startup. Device self-calibration supports operation but cannot replace external-structure, installation, and system-error verification.
Retain denominators and difficult samples for algorithm results
Define event starts, ends, and allowed response times before counting successes, misses, and false detections. Mixing frame-based and event-based statistics yields different rates from identical data. State total events, duration, and no-target observation time; accuracy alone without a denominator is insufficient.
Separate development/tuning scenes from validation scenes unused in tuning. Repeatedly tuning one room does not establish performance after height, wall, or furniture changes. Retain raw/intermediate data for difficult samples to identify insufficient SNR, background suppression, or tracking-related misses.
Enclosure verification should record material, thickness, finish, antenna spacing, and mounting angle, not just presence/absence. Metallic coatings, curved surfaces, water accumulation, and structural parts alter propagation/near fields. Testing a flat material sheet does not validate the final housing. Retest affected areas after mechanical changes.
For jumping range outputs, check raw peak changes before peak selection, static filtering, and tracking. Stable raw spectra with unstable application outputs suggest algorithms; noise floors changing with supply activity suggest hardware coupling. Linking data levels narrows causes better than display smoothing alone.
Delivery should define target-space boundaries and currently undetectable scenarios. “60–64 GHz” or integrated processors are not application-performance commitments. Traceable configuration/data permit one-factor changes in the next iteration.
Do more antennas or all transmit channels automatically improve results?
No. Transmit combinations depend on modes, supply, and algorithms; array geometry determines angular information. Select and verify configurations by task. Multiplied channel counts are not angular resolution. Deliver scene matrices, installation drawings, configurations, and data revisions with verified boundaries.
Check installation before migrating scenarios and adjusting thresholds
Changing height, angle, or walls under the same algorithm changes target echoes/multipath. Check positions against assembly drawings first, retain old/new intermediate spectra and tracks, then decide configuration changes. Raising/lowering thresholds can change misses and false detections together.
Combine range, bearing, and occlusion into executable scene IDs for reproducible tests. Each release should record hardware, radome, firmware, and configuration, with samples linked to revisions. Demo videos illustrate behavior but cannot replace verification data with inputs, criteria, and denominators.
Source material and applicable scope
- Texas Instruments, “IWR6843, IWR6443 Single-Chip 60 to 64GHz mmWave Sensor”; version Rev. F, 2025-04; checked 2026-10-17; relevant sections: Features, Pin Functions, Power Supply Specifications, RF Specifications, Application and Implementation. Supports IWR6843 bands, transceiver channels, and power interfaces; system detection depends on configuration/scenario verification.
