Skip to content

[PARENT] [ALIGNMENT]: Sensor Telemetry UI/UX #53

Open
7 of 8 issues completed
Open

Description

@jamesarich

[PARENT] [ALIGNMENT]: Sensor Telemetry UI/UX

Decision

A given sensor reading shows with the same category, unit, precision, chart scale and color band on every client. The rules below are what Android and Apple shipped through Meshtastic-Android#6099, #6092 and Meshtastic-Apple#2046, #2040, with the places they disagree settled one way. Where neither shipped what this thread asked for, the rule says so and the gap is tracked per platform.

Context

Meshtastic supports a wide range of sensors, and the clients presented them inconsistently. @oscgonfer's breakdown at the top of this thread is the basis for the rules: categories, a fixed-scale pressure axis, separate temperature and humidity scales, zero-based particulate axes, and CO2 bands.

Air quality display and calculation was settled in #54, raw particulate data in #51 and wind speed units in #87. Units follow design#99.

The categories stay as Device, Environment, Air Quality, Power and Host rather than splitting Environment into Weather and Soil. The firmware packs weather, soil and in-box readings into one EnvironmentMetrics message without saying which sensor took them, so a split would be guesswork until protobufs name the sensor.

#84 asked for SI prefixes on large values. Neither client uses them, and the values Meshtastic reports today (lux, ppm, µg/m³) read fine as whole numbers, so this spec doesn't require them.

An audit of both clients on 2026-11-09 found the gaps listed under Platform Tracking.

Required Client Behavior

  1. Group telemetry by message: Device, Environment, Air Quality, Power and Host. Soil temperature, soil moisture, wind, rain and radiation belong to Environment. A client may also show the Soil & Water message as its own group.

  2. Pick units from the locale, per design#99: temperature in °C or °F, wind in km/h or mph, rain in mm or inches, weight in kg or lb. Pressure is always hPa. Particulates are µg/m³, CO2 is ppm, radiation is µR/h, voltage is V and current is mA. Convert the chart the same way as the text, so the axis and the reading agree.

  3. Use the same precision everywhere a value appears. In logs and chart markers: temperature, wind, rain, radiation and pressure to 1 decimal, voltage to 2, current to 1. On summary tiles and cards: temperature and pressure as whole numbers. Humidity, soil moisture, particulates, CO2, AQI, IAQ and lux are always whole numbers.

  4. Give barometric pressure its own axis, a fixed 100 hPa window that defaults to 950–1050 hPa and slides down or up so no reading is cut off. Widen it to the data only when the readings span more than 100 hPa. Never start it at zero.

  5. Don't let metrics with different units share one axis in a way that flattens them. Temperature and humidity get separate scales. Humidity runs 0–100 %. Temperature is padded around its data and doesn't start at zero. Lux, radiation and wind don't share the temperature axis.

  6. Draw particulate charts on an axis that starts at zero with a small offset, so a zero reading isn't on the bottom edge. Put CO2 on its own axis, or show it in its own chart, so it doesn't flatten the particulate lines. Each series can be shown or hidden.

  7. Compute AQI with US EPA NowCast from standard PM2.5, using the 2024 breakpoints and EPA's minimum-data rule. Show the category name with the number: Good, Moderate, Unhealthy for Sensitive Groups, Unhealthy, Very Unhealthy, Hazardous. Show nothing until there's enough history to compute it.

  8. Color CO2 by these bands: Good under 1000 ppm, Stuffy under 2000, Poor under 5000, Unsafe under 30000, Evacuate at 30000 and above.

  9. Color IAQ by the Bosch scale: Excellent 0–50, Good 51–100, Lightly Polluted 101–150, Moderately Polluted 151–200, Heavily Polluted 201–250, Severely Polluted 251–350, Extremely Polluted above 350.

  10. Let the user label each power channel with free text such as Solar, Battery or Load. Labels live on the phone only, per node, and are never written to the radio. A blank label shows as Channel N. Use the label everywhere the channel appears, including node detail.

  11. Hide a metric the device doesn't report. A reported zero is a real reading and shows; a missing value never shows as zero. Chart series and legend entries appear only for metrics with data.

Acceptance Criteria

  • The same reading shows the same category, unit, precision and color band on Android and Apple, in node details, logs and charts.
  • A node at altitude shows its pressure trend without clipping, and a pressure change of a given size draws the same height on both clients.
  • A device without a sensor shows no empty tile, zero value or empty chart series for it.
  • Each client's tracker lists its gaps against these rules, and is closed with screenshots in light and dark.

Platform Tracking

Activity

  1. changed the title [-][ALIGNMENT]: [ALIGNMENT]: Sensor Telemetry UI/UX[/-] [+][ALIGNMENT]: Sensor Telemetry UI/UX[/+] on Mar 6, 2026
  2. oscgonfer commented on Mar 14, 2026

    @oscgonfer

    This issue requires a lot of unpacking. I would try to categorise the metrics that meshtastic currently collects, which I believe are somewhat reflected on the firmware, with some nuances:

    1. Power: data coming from power sources such as current/voltage meters.
    2. Weather (note that I don't use environment): data from typical weather units, namely air temperature, humidity, barometric pressure, wind speed, wind direction, rain, radiation. This is every parameter measurable in the air, that is not related to AQ below. This is partially implemented in the EnvironmentTelemetry module.
    3. Air Quality: Pollutant data, both from particulate matter and chemicals.
    4. Device-related thermal parameters: metrics that are sadly now being collected "as if they were weather" but that are basically related to the device itself (a temperature sensor inside the box, or poorly placed).
    5. Soil: soil humidity, soil temperature, etc. Niche, but supported in the firmware.
    6. (maybe in the future) Water: currently not measured, but pH, water temperature, Conductivity, etc.

    Warning

    Note that, with the exception to AQ, all the other ones: Weather, Device, Soil are now packed into EnvironmentTelemetry, with the issue being that only one of the metrics is being sent over (for multiple temperature sensors, you only get one temperature). This last issue will be solved by changing how protobufs are described (explicitely mention which sensor collected the metric).

    Power

    This is largely done in PowerTelemetry, although there is little to show for people that don't know what is connected where. I believe there is a representation gap in how this data makes it to the client and how it's shown at the moment, largely because most current/voltage meters provide several channels, but at the moment there is no way to say "this is the channel that actually means something". I am not sure how the power metrics are used, but it'd make sense to have the metrics coming from the PV panel, and load at least.

    My suggestion here would be to:

    1. Have a way to choose a channel and describe what it is.
    2. If the power metrics are coming from a solar panel (most likely the case), and the load.

    Overview (Battery, Panel power, Load power):

    • This is the best possible representation, which may not be possible with the current data, but overall, a representation of the battery %, how much power the panel is providing, and how much we are consuming, is the best way to know if a system is balanced.
    Image

    Panel power metrics (power = voltage * current) and load power metrics:
    Image

    Image

    Note

    Note: This could be combined with temperature data (if it existed) inside the device.

    Weather

    Caution

    A lot regarding weather was unpacked in meshtastic/Meshtastic-Apple#308

    Similar to PowerTelemetry, for EnvironmentTelemetry temperature, RH and barometric pressure sensors in this category we don't have a way to know if they are measuring the weather, or the box in which the device is, or something else. For that, I would simply represent those metrics as timeseries.. Note that:

    • Barometric pressure is mostly static, except when there are weather events (storms). I'd represent it on a timeseries on it's own, with fixed axis (not autoscaling), ranging from 70kPa to 120kPa for a device in sealevel. This lets you see if there is something going to happen (a storm) or not. If we autoscale this, it will be very difficult to see trends. An important note is that we can derive the expected barometric pressure if we have elevation data (see formula)
    • Temperature and rh: These two metrics require different scales (not represented on the same vertical scale) on a timeseries. Autoscaling here would be possible, although probably interesting to have quite a bit of margin on the scale to see trends.
      • For the particular case of temperature, a heatmap may be an interesting representation of data to look for "hot days".
    Image
    • For wind, the most typical (scientific) representation is a wind-rose but it's probably too much effort to represent it. Simply a table/timeseries with indications of speed and direction with little arrows is the most typical I've seen:
    Image
    • For precipitation, in addition to the above tabular representation alongside with the wind (like most weather apps), a bar chart is probably one of the most common representations, although the aggregation level is important (similar to many other metrics). Probably this is a minimally used metric though.
    • Radiation: a timeseries would do, with fixed lower value of 0. I don't know how this is measured in meshtastic, but typically, it's best to use a UV index. However, if the UV index is not provided by the sensor, we can't estimate it based on the irradiance (uW/cm2) because UV index is dependent on the associated wavelength. Then:
      • If we only get the uW/cm2 readings from the sensor: a timeseries with no "judgement". Autoscaled, fixed value at 0uW/cm2.
      • If we get UV index from the sensor: then we can apply a "recommendation" - see index usage

    Note

    One comment for all timeseries: IMHO, timeseries charts with "big dots" are a problem if they become too cluttered. I'd aim to either have thin lines, and maybe a dot where the cursor is (see an example below).

    Image

    Adding here an example which probably overuses the color scales.

    Image

    Air quality

    AQI

    I'll unpack this in #54.

    CO2

    We now support many CO2 sensors. There is no guidance on how to report it in a general sense. There are some regulations for work environments, but nothing really on households. However, some basic recommendations are:

    • Good: 400-1000ppm
    • Stuffy: 1000-2000ppm
    • Poor: 2000 (and above)
      If we want to add more colors:
    • Unsafe to work for 8h: Above 5000ppm.
    • Evacuate: Above 30000-40000ppm

    Device metrics

    Again, the issue here is to be able to know if these metrics are something or not. In any case, the current timeseries do the job.

    Soil & water

    I don't know if this is really an interesting topic to unpack. For now, simple timeseries can be used as probably this is not really a big use case.

  3. b8b8 commented on Apr 3, 2026

    @b8b8

    My comment is on air quality. I think it should be renamed "Hazards".

    Hazards should include:

    • VOC/IAQ
    • Particulate/AQI
    • CO2
    • CO
    • Radiation (the dosimeter supported in meshtastic is the radsens)

    https://climateguard.info/radsens/

    UV readings can in a separate section.

    Image
  4. oscgonfer commented on Apr 3, 2026

    @oscgonfer

    I think Air Quality is the accepted term for it in most disciplines. Hazards feels too extreme for me, especially since Good Air Quality is no hazard.

  5. changed the title [-][ALIGNMENT]: Sensor Telemetry UI/UX[/-] [+][PARENT] [ALIGNMENT]: Sensor Telemetry UI/UX[/+] on May 17, 2026
  6. garthvh commented on Jul 5, 2026

    @garthvh
    Member

    Current state

    Spot-checked the current cross-platform rendering of the metric families this issue raises (Power, Weather/Environment, Air Quality, Device thermal), since the discussion above is a design proposal rather than a status report.

    iOS (Meshtastic-Apple) — ◐ basic timeseries only, none of the proposed distinctions

    • Environment (Views/Nodes/EnvironmentMetricsLog.swift:35-46) draws temperature, humidity, dew point, barometric pressure, IAQ, and lux all on one shared, autoscaled Y axis (chartYScale(domain: chartRange)), not the separate/fixed scales suggested above (EnvironmentDefaultSeries.swift:116-139 — pressure has no fixed 70–120 kPa domain).
    • Power (Views/Nodes/PowerMetricsLog.swift:57-60,119,137,155) only labels sources "Channel 1/2/3" — no way to mark a channel as panel/load/battery, exactly the gap oscgonfer describes.
    • Air Quality isn't ingested at all yet — no pm25/pm10/co2 fields in Model/TelemetryEntity.swift; tracked separately as open Meshtastic-Apple#2040 ("Ingest PM2.5/PM10 telemetry and compute NowCast AQI").
    • Soil/device-thermal fields (radiation, soilMoisture, soilTemperature, windSpeed/Direction — TelemetryEntity.swift:44-61) exist as plain timeseries with no sensor-identity distinction, matching the "which sensor measured this" ambiguity called out above.

    Android (Meshtastic-Android) — ◐ furthest along, still missing thresholds/labeling

    • Environment (feature/node/.../metrics/EnvironmentMetricsState.kt:57-58,118-119,206-212) already puts barometric pressure on its own left axis vs. a combined right axis for temp/humidity/IAQ — partial alignment with the "separate scales" ask — but the pressure axis is autoscaled from data min/max, not a fixed sea-level range.
    • Power (feature/node/.../metrics/PowerMetrics.kt:82-90,109-144) has the same generic-channel gap as iOS, just with up to 8 channels instead of 3 — still no panel/load/battery semantics.
    • Air Quality (feature/node/.../metrics/AirQualityMetrics.kt:79-96) plots raw PM1.0/PM2.5/PM10/CO2 values with no color bands or AQI category — tracked as open Meshtastic-Android#6092 ("Compute NowCast AQI from PM2.5 history and show an AQI category display"), consistent with the CO2-threshold suggestion above not being applied yet.

    Web (meshtastic/web) — ❌ no charts for any of these metric families

    NodeDetailsDialog.tsx:171-200,447-475 only surfaces 4 raw deviceMetrics fields (airUtilTx, channelUtilization, batteryLevel, voltage) as a plain table. Environment, Power, and Air Quality metrics have no dedicated UI at all — the only place they're visible is a JSON.stringify(node, null, 2) dump inside an "All Raw Metrics" accordion (NodeDetailsDialog.tsx:478-493). The useTelemetry hook in packages/sdk-react/src/hooks/useTelemetry.ts doesn't appear to be consumed by any app component (no references found outside its own file).

    Checklist

    The three "Create Android/iOS/Web alignment issue" sub-tasks are unchecked and, as far as I can find, none have been opened yet for the general topic — that's accurate, not stale. The AQI-specific offshoot did already happen though: this thread spun into #54, which in turn produced concrete per-platform work (Meshtastic-Apple#2040, Meshtastic-Android#6092). Might be worth using that AQI slice as the template for splitting off Power and Weather/Environment alignment issues too, since those have the same "no shared vocabulary for channels/sensors" problem across all three clients today.

  7. garthvh commented on Jul 5, 2026

    @garthvh
    Member

    Implementation issues opened for the remaining gaps (Power channel semantics + Environment/pressure axis scaling — AQI is already tracked separately via #54): meshtastic/Meshtastic-Apple#2046 (iOS) · meshtastic/Meshtastic-Android#6099 (Android) · meshtastic/web#1244 (Web, which needs the underlying chart UI built from scratch).

  8. jamesarich commented on Jul 6, 2026

    @jamesarich
    CollaboratorAuthor

    Android implementation notes (meshtastic/Meshtastic-Android#6099 → PR #6111)

    Documenting the concrete approach Android took for the two non-AQI gaps, so iOS (Meshtastic-Apple#2046) and Web (meshtastic/web#1244) can align on vocabulary and behavior rather than each inventing its own.

    1. Power channel identity → per-channel user label

    • What: each power channel (Android supports up to 8; iOS/firmware expose 3) gets an optional, free-text, user-editable label (e.g. "Solar", "Battery", "Load"). Shown in the channel selector chip and in the telemetry cards; falls back to "Channel N" when unset.
    • Storage: local only, per node — not written to the device or any proto (the firmware PowerMetrics message has no label field). Persisted in the app's node DB and preserved across device syncs (an incoming telemetry/NodeInfo packet, which never carries labels, must not wipe a user's labels).
    • Deliberately started with a plain text label rather than a fixed enum (panel/load/battery) or per-role icon/color. A free-text label resolves the core "which channel is what" ambiguity immediately and doesn't require the three clients to agree on a closed vocabulary first. A structured role enum can layer on later if we want shared iconography — but the label is the migration-free, ship-now primitive.
    • Alignment ask: if iOS/Web adopt the same "optional local per-channel label" concept (same semantics: user-settable, local, not synced), the feature means the same thing everywhere even before any richer role system exists.

    2. Barometric pressure axis → fixed-width window that slides to elevation

    Android already had pressure on its own axis (separate from the temp/humidity/IAQ axis). The remaining issue was that the pressure axis autoscaled from data min/max, so a few-hPa weather swing looked huge or invisible depending on the sampled range — not comparable across sessions/nodes.

    Rather than a hard clamp (e.g. a static 950–1050 hPa), Android uses a fixed-width window (100 hPa) that slides to the node's elevation:

    • Near sea level (readings within 950–1050): the window is exactly 950–1050 — full absolute comparability for the common case.
    • At altitude (station pressure reads low, e.g. ~840 hPa): the same 100 hPa-wide window slides down so readings aren't clipped.
    • Only if a single node's readings genuinely span more than the window (e.g. a node that moved thousands of metres) does it fall back to showing the full data range.

    The key invariant is fixed width → consistent scale: a given pressure change always renders the same height, which is what this issue asked for, while never clipping high-altitude station-pressure readings. The trade-off is that absolute position isn't identical across nodes at different elevations (1013 isn't always at the same pixel) — but scale comparability, the meaningful one for spotting trends, is preserved.

    Note: Meshtastic's barometric_pressure is raw station pressure (not sea-level-corrected), which is exactly why a static sea-level clamp would hide altitude readings and the sliding window is needed.

    Happy to adjust naming/shape if iOS/Web want to converge on something different before this becomes the de-facto reference.

  9. oscgonfer commented on Jul 30, 2026

    @oscgonfer

    Small feedback.

    Weather metrics

    Image
    • Currently we have an x-axis starting on 0, making it a bit difficult to see. From the original feedback, I'd suggest:

    Temperature and rh: These two metrics require different scales (not represented on the same vertical scale) on a timeseries. Autoscaling here would be possible, although probably interesting to have quite a bit of margin on the scale to see trends.

    AQ Telemetry

    • On AQ telemetry, a similar grid to the one on environment telemetry, would be great to have. If we are only showing PM, axis can be base on 0ug/m3 (with a bit of vertical offset to not have 0 right at the bottom of the graph).
    • If showing other values, since there is so much disparity in the ranges (for instance, CO2 is 450ppm or up), having a way to make axis decoupled (or to hide-show metrics as in Env telemetry, it'd be great).
  10. garthvh commented on Aug 11, 2026

    @garthvh
    Member

    Status update (August 2026): the iOS and Web alignment sub-issues still haven't been opened (Android's exists). The chart-scale feedback above (separate scales for temperature/RH, zero-based PM axes with offset, an AQ grid mirroring environment telemetry) is exactly the kind of decision the eventual spec should ratify — suggest capturing those as the first concrete standards in this parent, then opening the per-client issues against them.

  11. moved this from Suggestions to In Progress in Cross Platform Design Discussionon Nov 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions