Features

Everything PhantomSDR-Plus does

The complete feature list of PhantomSDR-Plus, from the receiver page to the server behind it.

The core

What PhantomSDR-Plus is

  • WebSDR which can handle multiple hundreds of users depending on the Hardware.
  • Common demodulation modes.
  • Can handle high sample rate SDRs (70MSPS real, 35MSPS IQ).
  • Support for both IQ and real data.
  • Full decoder support.
  • Lightweight mobile view at /mobile for phones on limited data plans.
  • Finer audio resolution: the audio is sent with 8 times finer resolution, which removes a faint hiss of the stream's own on strong stations (see docs/README.md).
Listening

On the receiver page

The audio, the meters and the look of the page.

  • A modern, futuristic user-interface design.
  • Optimized FLAC and Opus encoders with balanced latency. Both operate at 16-bit resolution, providing a theoretical dynamic range of approximately 96 dB. ⭐
  • Integrated C-QUAM AM Stereo decoding. C-QUAM audio uses Opus compression by default, while the remaining audio modes use FLAC.
  • Automatic C-QUAM status indication. When a C-QUAM transmission is detected, the button label turns green. While operating in AM mode, pressing the button again enables synchronous AM detection on the carrier. The label then turns yellow and displays SAM. Pressing the button once more returns the receiver to standard AM mode.
  • SAM shows its carrier: while synchronous AM runs, a line such as SAM +3.2 Hz · 97% gives the station's carrier offset from the dial and how well the detector is locked.
  • IQ for listeners, as on a KiwiSDR or UberSDR: an IQ button sends 12 kHz of the band as raw I/Q — or 24 and 48 kHz where the sysop turns on the wider IQ (iq_wide, with a limit on how many listeners have it at once; both are set in the config or on the admin panel's Receivers page) — lossless, I left and Q right, with band noise kept quiet. Recording in IQ saves a WAV that HDSDR, SDR#, SDR++ and SDRuno open already tuned. On by default; iq = false under [server] turns it off. It also feeds Dream or another DRM program through a virtual cable.
  • IQ demodulators: in IQ, up to six receivers run inside those 12 kHz in the listener's browser, each with its own mode (USB, LSB, CW, AM, DSB with carrier lock, FM), frequency, width, volume and pan — both sides of a split at once, one in each ear.
  • An S-meter that reads VHF S-units above 30 MHz (IARU Region 1: S9 is −93 dBm there, −73 dBm below), with an optional rest-at-zero on an empty VHF/UHF channel ("siteSMeterGateDb").
  • A volume slider that is silent at 0 % and fades out smoothly with fine steps at low levels; it remembers where each listener left it, and the sysop sets the volume a first-time listener starts at for each receiver (AT VOLUME in the admin panel, or volume under [input.defaults]).
  • Picking a digital mode sets both the sideband and the passband.
  • Integrated audio Compressor and Equalizer, with manual configuration and reset controls.
  • An audio analyser (the Spectrogram section's 📊 Show): the demodulated audio as a spectrum with an axis that follows the mode, the recommended bandwidths and your current filter; an RF view of the station in AM and FM; the station's occupied width with a splatter verdict; and, in USB/LSB, an SSB audio report (tonal balance, compression, mains hum, residual carrier, opposite sideband) with a listening report to download; in AM/FM an AM/FM report (modulation depth, sideband balance, deviation, CTCSS tone). It works at any zoom; choosing the ESSB reference opens the IF filter to 4.0 kHz. Adapted from PhantomSDR-ORSAT by F1NSK.
  • A “Magic Eye” signal-strength indicator that emulates the behaviour of EM84 tuning-indicator tubes used in vintage radio receivers.
  • A signal strength plot (Plotter under Recording Options): the S-meter's dBm reading as a running graph over the last 10, 30 or 60 seconds, with S-unit and dBm scales, the peak, a mark at every retune, pause and a PNG save.
  • A band propagation page for the receiver's own location (🌐 Propagation in Additional Info): every band from 160 m to 6 m graded Good, Fair, Poor or Closed for five distances, from the nearest ionosondes' live data, the sun's position and the solar and geomagnetic indices, on a 3D globe you can spin and tilt (or a flat map), with outlines showing, direction by direction, how far every open band reaches and where its skip zone ends, the MUF and foF2 contours, the greyline and every ionosonde reporting now, worldwide.
  • An optimized mobile GUI, with a taller spectrum and waterfall in the extended view.
  • Client-side CPU and memory usage have been reduced as much as possible.
Audio & tuning

Redesigned and optimised

Noise controls, AGC, squelch, tuning aids and the waterfall tools.

  • Reduced end-to-end latency: the browser's own FLAC decoder plays each packet the moment it arrives, and the playback buffer sizes itself to the connection (about 140 ms on a phone over Wi-Fi; add ?diag=1 to the address to see the figures). On https pages the buffer also trims itself back after a burst, by playing 0.25 % faster for a while, which cannot be heard (?drain=0 turns it off). See Audio latency for a measured comparison with other web receivers.
  • Noise Reduction (spectral, tone-preserving).
  • Noise Cancellation.
  • Noise Blanker (impulse).
  • Background Noise Suppression.
  • Auto Notch for heterodynes and carriers.
  • AI Noise Reduction (RNNoise, in the listener's browser, voice modes only).
  • All of the above bypass the decoders, which read raw audio.
  • Synchronous AM enabled by default.
  • Transceiver-style AGC: Fast, Medium and Slow timed in the server for each listener, and Auto, the receiver's own; a running decoder holds it on Slow and says so.
  • Automatic squelch.
  • CTCSS tone squelch (FM), with a pop-up picker for the subtone.
  • Configurable buffer settings.
  • Auto Adj.: automatic waterfall brightness and contrast from the measured noise floor, the same with every colour map.
  • 3D spectrum: the last seconds of the band in relief, in the waterfall's colours and levels (WebGL2, in the listener's browser).
  • Mouse-wheel frequency tuning, with a 12.5 kHz step for the VHF channel grid.
  • A number pad on the frequency box; a typed frequency frames its band as the band button would.
  • Band buttons remember your last frequency on each band; a second click returns to the band's default.
  • A cursor line with its MHz readout on both the spectrum and the waterfall.
  • Hover text on every button, keyboard shortcuts, and screen-reader support (NVDA, JAWS, Narrator, VoiceOver, TalkBack, Orca).
  • Zoom slider with 20 even steps, from the whole receiver to a few kHz.
  • Enhanced mobile GUI.
  • Selectable Noise Gate with presets.
  • Precise mouse-based tuning with a frequency readout directly on the waterfall.
Display & tools

Waterfall, bookmarks, recording

  • A configurable band-plan overlay on the waterfall, with independent brightness, frequency-range, and visibility settings for each band; the 2 m modes follow the IARU Region 1 band plan.
  • An improved custom colour-map system. A reversed spectrum and waterfall layout is used by default across all interface variants, although this behaviour can be changed manually.
  • Four selectable GUI templates, accessible through a floating pop-up menu.
  • Redesigned bookmarks with on-screen labels and import/export functionality.
  • Audio recording and combined video/audio recording of either the complete waterfall (with the 3D spectrum when it is open) or a user-selected region; the video marks the tuned frequency and passband with the tuner's yellow lines.
  • A channel scanner on the tuning row: it sweeps the band — or exactly what the waterfall is showing — and stops on the first channel rising a chosen number of dB above the band noise floor, resumes by itself when the channel goes quiet, keeps the step on the mode's own channel grid, and can skip empty channels by reading the spectrum. Locked-out channels and settings live in the listener's browser.
Decoders

Decoders and spotting

Every decoder runs in the listener's browser except DRM and RADE, which the server decodes; the autorun engine reports spots by itself. See Decoders for each one.

  • Integrated decoders for FT8, FT4, FT2, JS8, CW (with CW Browser), QRSS Grabber, WSPR, HF FAX, SSTV, NAVTEX, an FLDIGI window with ALE 2G HAM, RTTY with RTTY Browser, PSK31/63/125 (BPSK/QPSK) and PSK Browser, Olivia and Contestia with Olivia Browser, FST4W, MFSK16/32/64 with MFSK Browser, MSK144, DominoEX, THOR, THROB, Hellschreiber, MT63, Packet (AX.25) and APRS (with a live map of the stations heard), SERVICES (HFDL aircraft messages with a live aircraft map, Maritime FSK / SITOR, Weather RTTY, Naval RTTY (PBB CARB), DSC, Time Signals, HF Selcall, ALE 2G, AQC-ALE, SITOR-A / ARQ-E / ARQ-E3, STANAG 4285, MIL-STD-188-110A and 3G ALE), DRM (digital broadcasting, decoded on the server) and FreeDV RADE V1 and V2 (beta). RTTY, PSK31, FT8, FT4 and FT2 show a real SNR (the FT modes agree with WSJT-X), and RTTY has a squelch. The FLDIGI modes are ports of fldigi's own receivers, checked against real fldigi recordings, with no fldigi installed anywhere; Hellschreiber is shown as a picture that can be saved as a PNG. A frequency marker whose mode is a digital mode starts its decoder in one click. See the decoder documentation.
  • RADE V2 (beta), the FreeDV 3.0 waveform, on the RADE-v2U button through FreeDV's rade_c on the server; V2 keeps playing down to about −6 dB SNR on a real V2 signal, and when a station is one carrier off the panel says which way to tune, with a button that moves the dial. See the RADE manual.
  • Decoder buttons in two rows of six, the drop-down in the same order; HFDL, DSC, TIME, SELCALL, ALE, FST4W, MSK144, DRM and MORSE (the CW decoder) frequency markers start their decoders.
  • A Decoder ID tool that identifies an unknown digital signal and offers the matching decoder in one click. Off by default.
  • JS8 is shown as a conversation panel, with multi-frame messages reassembled.
  • All decoders have been moved off the main browser thread. Each decoder now runs in its own Web Worker, preventing signal decoding from competing with audio playback and GUI processing.
  • Automatic spot reporting to PSK Reporter and WSPRnet. A built-in autorun engine locally decodes off-air signals and uploads reception reports without requiring additional software, on an hourly schedule per band and mode set in the admin panel, and can start by itself after a reboot:
    • FT8, FT4 and JS8 → PSK Reporter using the native IPFIX/UDP protocol. Spots are batched and transmitted at the recommended interval of at least five minutes. JS8 is spotted at Normal speed, from heartbeats, compound frames and directed messages; group destinations such as @ALLCALL and unresolved callsigns are never reported.
    • WSPR → WSPRnet using the standard WSPRnet upload endpoint. WSPR reports are sent only to WSPRnet to prevent duplicate submissions.
  • Integrated FreeDV Reporter and DX Cluster tools.
  • An EiBi window (button next to SW) that searches the worldwide EiBi schedule of long wave, medium wave and shortwave broadcasts and utility stations — the database shortwave.live is built on — by station, language, target area, band, country, transmitter site, frequency and UTC time, with the stations on the air in green. One click tunes a station in the mode EiBi gives it (FAX, HFDL, DRM, NAVTEX, RTTY and CW rows start their decoder), zooms the waterfall to ±50 kHz and marks it in red; the results can be shown on the waterfall and the HFCC markers hidden. frequencylist/update-eibi.sh keeps the list current, checking once a day.
Rig control

Your transceiver, Kiwi clients and a second site

  • CATsync support through the CATsync Tool for WebSDRs application, TCI-CAT on the receiver page itself (ExpertSDR, AetherSDR, Thetis, or any Hamlib rig through the included bridge), and two-way rig control — frequency, mode and filter — in Desktop PhantomSDR+ (video). See docs/RIG_CONTROL.md.
  • TCI-CAT on the receiver page. A new TCI-CAT button beside IF Filters lets the page follow your own transceiver by itself, with no desktop application: the browser talks straight to the TCI server of ExpertSDR, AetherSDR or Thetis on your computer. Frequency, mode and filter width follow in both directions — turn the rig's dial or press its FIL button and the receiver follows, click the waterfall or pick an IF filter and the rig follows — and the page falls silent while you transmit.
  • Direct USB on the receiver page. The same window can reach the rig itself over its USB cable, with no TCI software, Hamlib or bridge: 54 rigs from Icom and Xiegu (CI-V, the IC-7300MK2 among them), Yaesu, Kenwood, Elecraft, FlexRadio SmartSDR CAT, QRP Labs, (tr)uSDX and Lab599, in Chrome, Edge, Opera or Firefox 151+ on an https:// page. Icom filters are selected (FIL1/FIL2/FIL3), never rewritten; ⏻ Power on and ⏻ Power off switch the rig on and off over CAT, and Rig scope shows an Icom's own scope (IC-7300, IC-7300MK2, IC-705, IC-9700, IC-905, IC-R8600, IC-7610, IC-7760) as a waterfall strip under the receiver's, each sweep under the same frequencies. See Direct USB.
  • CAT Sync, Split and Mute on TX on one line of the CAT window: switching CAT Sync on brings the page to the rig at once, Split leaves the rig where it is while you tune the page, and Mute on TX (on by default) silences the page while you transmit.
  • Transceiver control from the desktop app. Desktop PhantomSDR+ (video). A 💾 Desktop applications button under Open Additional Info, beside System Resources, opens the download folder of Desktop PhantomSDR+.
  • KiwiSDR client emulation — an optional bridge that lets Kiwi clients such as AetherSDR and kiwiclient connect to the receiver directly, on the same host and port, with real retuning, waterfall, S-meter and I/Q (kiwirecorder -m iq). Off until [kiwi_emulation] enabled = true is added to your config. See docs/Aether_config.md.
  • Receive diversity — pair the receiver with a second site (another PhantomSDR-Plus, a KiwiSDR, an UberSDR or a WebSDR) and hear whichever of the two currently has the better signal. It runs in the listener's browser; only a WebSDR as the partner needs a small relay on your own server. See docs/RECEIVE_DIVERSITY.md. It switches even while the two streams are not yet aligned, so a station heard only by the remote comes through.
Several receivers

One station, several receivers

  • Several receivers on one computer — for example an RX-888 for HF and an RTL-SDR for 2 m, each a complete server of its own with its own waterfall, chat, markers and listeners. ./add-receiver.sh (also offered at the end of every installer) adds one, with its driver; start-all.sh starts them all and ./stop-websdr.sh <name> stops one. On the page a *Receivers:* line, on /mobile too, switches between them, and proxy.py lets them share one public port. See docs/MULTI_RECEIVER.md.
  • On a station with several receivers, Show All puts each other receiver under the waterfall as a live strip — click it and that receiver opens there; each receiver's page remembers where you were, and a frequency only another receiver covers (typed, from a link, a spot, the chat or a label) opens that receiver by itself.
  • Optional server settings for multi-receiver stations: [server] host, [server] html_fallback_root, [websdr] public_port, and per-receiver siteReceiverURL / siteAntennaURL links.
Listeners

Chat, users and directories

  • Chat with a ↪ Reply button on every message; answers appear indented under the message they answer, on the desktop page and on /mobile. Messages are measured in characters, so Greek and Cyrillic get the full 200.
  • A real-time connected-user list with geolocation data, map integration, and user statistics covering configurable time periods. Your own session is marked you in the list, whether it is opened from the interface or as a standalone users.html page. With several receivers open side by side, each page's Connected Users and User Statistics windows show its own receiver.
  • The page carries a Website button beside the GitHub link, and the version line reads "original from github.com/sv1btl/PhantomSDR-Plus"; the installers, update.sh and the admin panel say so when a copy comes from a fork.
  • PhantomSDR is registered with the following WebSDR directories:
Installation

Installing and updating

More in For sysops.

  • A new install.sh installation procedure that simplifies the initial server setup, starting with a station wizard (configure-station.sh): the receiver, band, station and ports are asked once and kept in station.conf, every config is written from it, and new stations publish a single public port (9000, proxy.py).
  • https:// alongside http:// — setup-https.sh puts Caddy with a self-renewing Let's Encrypt certificate on port 443, while the http address keeps serving websdr.org, Kiwi clients and http-only partners; a station behind the sysop's own reverse proxy (nginx, Nginx Proxy Manager, Apache, Traefik) gets its guide and checks instead.
  • A recompile.sh script for quickly rebuilding the backend, the frontend, or both.
  • An update.sh tool that updates the receiver with one command, keeping your configs, markers, logs and edited scripts untouched; each update's backup is one dated archive in update-backups/, restorable with ./update.sh --restore LAST; ./update.sh --from takes the new version from a folder, .tar.gz or .zip for a station without GitHub, and each release carries the fingerprint of every published file, so only real local edits are asked about.
  • Installers that stop a running receiver first, frame every step with its own verdict, and spell out the default of each question.
  • Support for running the GUI through localhost.
  • Supported receivers currently include the RX-888, RTL-SDR Blog V4, Airspy HF+ Discovery, HackRF, SDRplay RSP1A and RigExpert Fobos SDR, including compatible clones through SoapySDR. The installers set up each one's complete driver chain, and setup-rsp1a.sh, setup-fobos.sh, setup-airspyhf.sh and setup-hackrf.sh add one to a station already installed. Support for additional receivers is planned.
Administration

Admin panel, limits and stability

  • A full-featured, password-protected Admin Panel for remote server management without requiring direct SSH access. It provides access to server logs, chat moderation, user messaging, chat-message deletion without restarting the server, user disconnection, command execution, file editing, and other administrative functions; its Users and Chat pages cover every receiver of the station, with an 8, 16 or 24-hour ban next to the kick. It also sends alerts by email or Telegram when a receiver goes off the air, comes back or restarts, the thermal guard acts or crash.log gets a new entry; makes a one-zip backup of the station's settings and restores it, keeping every file it replaces; shows private visitor statistics (visitors, sessions, busiest hours, countries, and the addresses with the most sessions at once) and a System Health card; and has six themes — Phosphor green, Amber display, Metal panel, Modern dark, Light and Auto, light from sunrise to sunset at the station's locator — picked with THEME in the top bar.
  • A Thermal Protection function in Admin Panel, to avoid thermal stress of the CPU and take the proper actions.
  • The sysop's kick and the X-Forwarded-For header are trusted only from the same machine, and a request that a web server on that machine relays counts as the visitor it carries, so nobody can kick listeners or slip past the per-IP limits through the proxy — or through a reverse proxy of your own in front of it.
  • Connection limits, and a flood guard. The receiver can now cap simultaneous listeners and the rate of new connections per address, and bound the silent half-open connections that no other limit can see. Two keys switch it on — per_ip and per_ip_rate in [limits] — and everything defaults to off, so an existing config behaves as before. A refused listener is told why instead of watching a page that never loads. Below it sits setup-firewall.sh, an optional nftables layer that sheds volume before the server reads a byte and cannot lock you out of your own machine. Full manual, in all seven languages and in the offline PDFs: docs/CONNECTION_LIMITS.md — including how to measure your own traffic before choosing the numbers. A dropped connection now ends the session. The audio socket no longer reconnects by itself. On a receiver that was wrong: the waterfall and events sockets never came back with it, so a retried session was live audio attached to a frozen waterfall, and stopping the server stopped clearing it.
  • Improved server stability and fault reporting. If spectrumserver crashes or malfunctions, diagnostic information is written to crash.log.
  • The crash.log file can be accessed both from the server’s root directory and through the Admin Panel.
  • An optional monitoring tool that adds real-time server-status information to the PhantomSDR application.
Performance

Benchmarks

  • Ryzen 5* 2600 - All Cores - RX888 MKii with 64MHZ Sample Rate, 32MHZ IQ takes up 38-40% - per User it takes about nothing, 50 Users don't even take 1% of the CPU.
  • RX 580 - RX888 MKII with 64MHZ Sample Rate, 32MHZ IQ takes up 28-35% - same as the Ryzen per User it takes about nothing (should handle many).
  • Intel i5-6500T - RX888 MKII - 60MHz Sample Rate, 30MHz IQ, OpenCL installed and enabled, about 10-12%. 100 users is no problem with OpenCL as the GPU does the heavy lifting.

Audio latency

How far behind the air does a listener hear the band? We measured the delay from the antenna to the listener's speaker on four web receivers at the same moment, each through its own web page. All four were tuned to the DCF77 time signal on 77.5 kHz, and every second mark heard was compared with an internet-synchronised clock.

ReceiverAudio delayRange over 3 minutesLocation
PhantomSDR-Plus (SV1BTL)0.30 s0.29–0.35 sAthens
UberSDR0.30 s0.21–0.31 sAthens
PA3FWM WebSDR (University of Twente)0.35 s0.31–0.44 sEnschede, Netherlands
KiwiSDR (Web-888)0.96 s0.96 sAthens

Measured on 15 October 2026 at 23:10 UTC over 180 seconds, all four at once from the same computer. The delay includes the network and the listener's sound output, and moves by a few hundredths of a second from one session to the next.

Until mid-October 2026 PhantomSDR-Plus measured 0.39 s in the same test. Three changes brought it down: the server's AGC now looks 25 ms ahead instead of 100 ms (enough for the gain to drop before a strong signal arrives); the browser's playback buffer starts with a safe 80 ms and then trims itself slowly to what the connection needs, instead of keeping whatever backlog a burst left (which also cut dropouts from 4–6 to about 1 per session); and the chat's emoji picker is prepared only when a listener first clicks into the chat box, instead of stalling the audio five seconds into every session.