Superdish Mosaic Engine

Superdish Mosaic Engine — Short Technical Manual

Version 1.1.0

What it is

Superdish Mosaic Engine is a simultaneous multi-channel viewing module: it shows several channels from the same transponder in a single grid window, without having to tune them one by one.

It is not a signal technical-analysis tool. For that, Signal Analyzer Engine, Direct Tune and Blind Scan are still the tools to use. The Mosaic has one single purpose: comfortably watching several channels together.

How it integrates

• It's part of the TBS Universal Analyzer Engine suite by PantaSat, but lives inside Signal Analyzer Engine, which in turn can either run integrated within TBS Universal (hardware already powered on and shared) or on its own, in standalone mode.

• It's launched from the Signal Analyzer Engine menu, after powering on the hardware and locking a signal.

• It doesn't touch tuning: it simply reads what's already locked. On closing, Signal Analyzer Engine returns exactly as it was before.

Basic usage

• Double-click on a channel → fullscreen. Double-click again, or press ESC, → back to the grid.

• Right-click on a channel → menu to choose audio track, subtitles, aspect ratio.

• The filter at the top selects what to show: clear channels, radio, encrypted, or all together.

• The rest is intuitive: feel free to explore on your own.

Difficulties encountered and how we addressed them

Every satellite, every transponder, has its own quirks. The most recurring ones:

• Encrypted channels disturbing clear ones on the same transponder — explicitly isolated, channel by channel, so encrypted streams don't interfere with opening the clear ones.

• "Dirty" or unstable UDP streams at the source — handled by widening tolerance margins (more time, more data before giving up); not always fully solvable when the received signal is genuinely at its limit.

• Channels that open inconsistently, or that freeze after a while — addressed almost channel by channel, patiently, looking for the combination that makes them open and stay stable.

We tried to squeeze as much as possible out of ffmpeg and mpv, the two engines under the hood. Not everything is solved: some streams remain more stubborn than others.

Current status

Works well on most channels and transponders tested so far. There's still quite a bit of work left to increase overall stability, especially on the more marginal streams. Important note: multistream and T2-MI streams have not been analyzed or handled yet in this 1.1.0 version.

What we're asking testers to do

• Test on different satellites and frequencies.

• Report channels that don't open, that freeze, or that have audio/video issues — ideally with the session log attached.

• Tell us what's missing or what you'd change.

Credits

Heartfelt thanks to PantaSat, for the work, patience and availability put into the whole TBS Universal Analyzer Engine suite and into this module — without his foundation there would have been nothing to build on.

Download Superdish_Mosaic_Engine_V.1.1.0

See here for installation

Some Fixes on 1.1.0 adopted for Turksat 42E

• TRT HD (11054 V): residual audio/video disturbances, confirmed upstream too — raw UDP stream doesn't even open in VLC, so it's not a software limitation but a signal/multiplex issue.
• Turkmeneli (12009 V): fixed — TV and Radio now split, both working (the radio was the second audio track of the same channel, previously inaccessible).
• Freqs. 11012 H and 11727 V (Russian cloned package) — raw UDP feed affected by heavy interference at the source. UMKA channel especially affected, not always clear. Confirmed with the VLC UDP test too — not a software-side issue.

Download update here

Fix 1.1.2 – Superdish Mosaic Engine (Turksat 42.0E)

• Unlocked FTA channels hidden inside encrypted Digiturk bouquets (11265V, 11305V, 11345V) - Anyway some hard spikes on CGTN Channels
• Fixed TRT audio across multiple frequencies (11053H, 11054V, 11507V, 11096H, 12034V)
• Split TRT radio from TV where it shared the same audio track (12009V)
• Data/interactive channels (e.g. TKGS DATA) now recognized and flagged as such, no longer showing wrong content
• Residual issues with some frequencies/channels not opening: signal limitation upstream, not software-related as understood
• To improve: some channels not always opening on first launch — if one doesn't come up right away, try opening it a second time

Download Fix 1.1.2 – Superdish Mosaic Engine

Pictures of Freq. 12265 V Turksat 42E

Quote from stephan94

I just tried version 1.1.1, and multistream on 9° East works very well.

...

Thanks stephan94 , I haven't gone into the specifics yet for T2-MI and Multistream, I'm glad to know about it

Fix 1.1.3 – Superdish Mosaic Engine

Turksat 42.0E:

• TRT radio completely muted (12034V) and same crash on TRT 11096H: fixed with explicit audio handling
• Data/interactive channels (TKGS DATA, Intercom) now excluded upstream instead of stealing another channel's content

First Eutelsat 5W support:

• Satellite detection added; fixed a routing bug that sent a 5W tuning into the wrong pipeline (built only for UK/Astra 2)
• French radio muted (11480H) and Italian regional multistream (12540H): same audio crash, same fix

T2-MI/HEVC RAI multistream (12606V) – three distinct causes on the same mux, isolated one by one:

1. Rai 4K: two audio tracks, the second in AC-4 (a codec with no available decoder) — any attempt to touch it took down the entire shared mosaic. Fixed by mapping only the first track (E-AC3), excluding AC-4 entirely.
2. Rest of the mux (e.g. Rai 1 HD): same "sample rate not set" crash seen elsewhere, heavier here due to 17-18 channels sharing one process. Same fix applied across the frequency.
3. OTA-B: a service-only channel (EPG) not recognized as such, causing the same content-stealing bug already fixed for TKGS DATA. Added to the exclusion list.

Known limits, not software-fixable: a few channels (Ada TV/TV2020 on Turksat, some on the French 5W multistream) occasionally lose their turn in the shared process — if one doesn't open on the first try, a retry usually works.

Download Fix 1.1.3 – Superdish Mosaic Engine

In the photo EutelSat 5W - 12606 V MULTISTREAM T2-MI MIS 5 (RAI)

Fix 1.1.4 – Superdish Mosaic Engine

New "VUOTO" (EMPTY) feature: for clear channels that aren't transmitting anything real at a given moment, a clear label now appears instead of the usual unexplained black tile. This covers both genuinely dead channels and — importantly — cases like Ada TV/TV2020 where the channel is fine but that particular launch just doesn't come up within 55 seconds due to the usual shared-process contention: previously a silent gap, now clearly labeled. Double, well-spaced check so a channel that's merely slow to start (e.g. Europe 1 on Astra 19.2E, which sometimes doesn' appear)) isn't mistaken for a truly empty one.

Eutelsat 5W: fixed a last case of degradation on French bouquets with many encrypted channels on the same mux (11096V) — choppy video and crackling audio after a few seconds (some spikes remain).

Astra 1 19.2E: first systematic, certified check of every single frequency variable in the file — found 24 with no protection against numeric coincidences with other satellites, all fixed. Also resolved a fatal audio crash on 12633H that took down the entire 14-channel mosaic together, and a missing route on 12226H to the dedicated UK pipeline.

Download Fix 1.1.4 – Superdish Mosaic Engine

Example of "VUOTO" application

Quote from superdish66

Fix 1.1.4 – Superdish Mosaic Engine

New "VUOTO" (EMPTY) feature: for clear channels that aren't transmitting anything real at a given moment, a clear label now appears instead of the usual unexplained black tile. This covers both genuinely dead channels and — importantly — cases like Ada TV/TV2020 where the channel is fine but that particular launch just doesn't come up within 55 seconds due to the usual shared-process contention: previously a silent gap, now clearly labeled. Double, well-spaced check so a channel that's merely slow to start (e.g. Europe 1 on Astra 19.2E, which sometimes doesn' appear)) isn't mistaken for a truly empty one.

Eutelsat 5W: fixed a last case of degradation on French bouquets with many encrypted channels on the same mux (11096V) — choppy video and crackling audio after a few seconds (some spikes remain).

Astra 1 19.2E: first systematic, certified check of every single frequency variable in the file — found 24 with no protection against numeric coincidences with other satellites, all fixed. Also resolved a fatal audio crash on 12633H that took down the entire 14-channel mosaic together, and a missing route on 12226H to the dedicated UK pipeline.

Download Fix 1.1.4 – Superdish Mosaic Engine

Fix 1.1.4a – Superdish Mosaic Engine
Recommended update, please download the new version.

How "VUOTO" (EMPTY) works, step by step:

• Each channel goes through a precise order: first checked for encrypted (🔒), then for data/interactive (📊) — if either applies, it stops there, the matching label appears right away
• Only the remaining channels — the genuinely clear ones — can ever reach VUOTO
• For those, the system waits patiently (up to 55 seconds) before judging — a channel that's merely slow to start gets full time to prove it works
• If nothing real has arrived after that, the "VUOTO" label appears and the useless player is closed

Fixed today: radio channels were being wrongly marked "VUOTO" after 55 seconds, even while playing perfectly — the check only looked at video, which a radio never has. Radios are now checked on audio instead.

Astra 1 (19.2E): fixed a crash that was taking down all 7 channels on 11377V together.

Download Fix 1.1.4a – Superdish Mosaic Engine

SUPERDISH MOSAIC ENGINE — V. 2.0.0

STRUCTURE
• Hardware profile cascade: satellite certainty from the active device, fixed/DiSEqC/motorised
• Ten terminal states: every channel gets its own situation within 55 seconds, no silent cell
• One single label per cell: the first diagnosis wins
• Mosaic exit reordered: no leftovers on screen

T2-MI AND MULTISTREAM
• Five family rules tied to the stream, not to frequencies: valid on all satellites
• Closed the gap that let +genpts through on unnamed T2-MI muxes
• Channel cache keyed on MIS + PID + PLP
• Dedicated audio cushion for TSDuck bursts

EXCEPTIONS
• Manual of the six cases at the top of the file: symptom, measured cause, cure
• HEVC stutter, audio/video drift, "scratched record" audio: closed and documented
• No cell can be left without a verdict any more

PERFORMANCE
• Player memory −80% (from 180 to 36 MB per cell)
• Incremental log reading, buffers and caches resized
• Shutdowns and cleanups verified: no processes or references left hanging

OPEN
• HEVC codec not reported by the scan on T2-MI muxes: detected and written to the log, awaiting field proof

Download of PantaSat's TBS Universal Analyzer V. 8.0.0 is recommended — the module is verified on it.

Download Superdish Mosaic Engine 2.0.0

In the pictures: T2-MI Multistream and Multistream on 12606 V Eutelsat 5W — French Multistreams on 12738 V on Eutelsat 5W

Superdish Mosaic Engine V. 2.1.0

NO CELL IS LEFT WITHOUT A VERDICT · NESSUNA CELLA RESTA SENZA VERDETTO

The eleven verdicts

• 📺 regular video
• 📊 DATA — data stream
• 🔒 ENCRYPTED — encrypted per scan
• 🔒 MOMENTARILY ENCRYPTED — encrypting right now
• EMPTY — declared but not transmitting
• 📭 NO STREAM — no stream in the program
• 📄 SAME STREAM — duplicate, with the donor's name
• 🔊 AUDIO ONLY — audio only
• 📭 NO PICTURE — ninety seconds without a frame, the last net; the caption withdraws itself if the picture turns up later
• ⚠️ OUTPUT DIED — dead output, with SID and name
• ⚠️ PLAYER NOT STARTED / NO REPLY FROM PLAYER

One label per cell: the first diagnosis wins. The verdict is written at the centre of the cell, in the same place for every cell, and repeated twice so no burst can swallow it. No verdict is ever given on a cell whose picture we switched off ourselves.

T2-MI and multistream

• HD cells freezing on multistream: larger UDP read queue on the players, overrun made non-fatal — 12590 H, LA7 and France 24, 13 HD cells out of 13
• The channel whose key frames are never marked — 12590 H MIS 11, GALLURA TV: the audio plays, the video decoder never receives a single packet, there is no error line anywhere, and VLC opens the channel. Copying started without waiting for a key frame, and the decoder told to show everything: declared for that one channel, 17 cells out of 17
• Clones with no stream recognised: a cell titled FRANCE 24 was showing LA7 HD
• Duplicates named: the cell says whose picture it is borrowing
• "Scratched record" audio on the French muxes and on T2-MI: guaranteed audio buffer floor
• Programs with no stream on ordinary muxes — 12633 H Astra 19.2E, three cases at once

Keys and use

• Double click on the cell title bar = full screen, ESC to leave
• Ctrl + wheel = full screen and back, and this one works over the picture as well: up to enter, down to leave
• Wheel = volume of the cell under the pointer, ten points per notch, value and bar on screen — over the picture, over the title bar and over the filter dropdown alike
• Middle button = pause, press again = play, sound and picture realigned
• Speaker icon = audio on that cell, every other cell muted
• Right click on the title bar = cell menu
• Filter dropdown = clear TV / clear radio / TV and radio / encrypted / all channels; choosing the filter already in use rebuilds nothing
• The buttons the player keeps for itself are switched off: no menu of its own, no zooming, no swallowed clicks

Right-click menu

• Aspect ratio: AUTO / 16:9 / 4:3
• Deinterlacing: AUTO / ON / OFF — per cell, on request, never on by default: it costs GPU on every cell
• Audio track: list with language and codec, tick on the active one
• Subtitles: list and disable
• One live menu at a time, destroyed on shutdown

Full screen

• Hidden cells stop decoding, and resume one at a time on the way out
• A cell with a verdict stays off: NO STREAM never comes back over a picture
• A hidden cell is never judged, and its check is re-armed when full screen ends: nobody is declared EMPTY for having been watched
• Transition made lighter: from 73 to 35 command-channel openings on 32 cells

Windows

• The mosaic is an ordinary window: open a program from the taskbar and the mosaic steps behind, as it should. Fifty native video windows forced on top were what made the Start menu struggle
• On the way out, Signal Analyzer comes back restored and centred on the screen

Stability and memory

• A command that got lost in the burst: now one cell, one knock
• Menus piling up at every right click: now only one
• Closing the mosaic, or changing filter, while it is still opening no longer leaves an ffmpeg running with no window to serve
• ffmpeg logs: the six most recent are kept, the older ones are cleared at every launch
• On shutdown, cells, commands, log, menu, dropdown and window are all released
• Satellite read from the live line of the configuration

Open

• HEVC codec not reported by the scan on T2-MI muxes: detected and written to the log, awaiting field proof

Download Superdish Mosaic Engine V. 2.1.0

In the Photos: Cairo Communication Multistrem - La 7 - from 5W

Quote from stephan94

Good evening, superdish66

I’m having a bit of a problem with two frequencies on the PaksatMMI satellite at 38° East (C-band linear).

On the frequency 4029 H 3600, it takes a very long time before ‘Empty’ appears, and 15 seconds later the picture appears but ‘Empty’ remains displayed in the middle. (With the TSreader Pro mosaic, amongst other things, there’s no problem.)

On the frequency 4103 H 1600, no picture appears (shows ‘Empty’) (TSreader Pro works fine).

The signals aren’t very strong, but that shouldn’t be a problem.

Otherwise, your programme is very good – well done!

Display More

Hi Stephan94, thanks for the report — and especially for the screenshot, it shows the problem perfectly.

The cell **is** receiving: the news bulletin is running, and EMPTY is written on top of it. That label is wrong, and the fault is ours — not your signal and not the channel.

Two things I need from you:

1. **Does the picture keep running while EMPTY stays on screen**, or does it freeze? From the screenshot it looks like it keeps running, but I want to be sure.
2. **Send me the ffmpeg log** of that run. You'll find it in `signal_analyzer\logs`, named `ffmpeg_mosaic_<date>_<time>.log`. Post it here or send it to me by private message, whichever you prefer — the log alone is enough.

Meanwhile, here is what happens, because it is already clear. The empty-check asks the player three times whether it has a picture width to show. On a signal this close to the edge — 8.7 dB SNR on C band, 7.83 Mbps — the player can answer "nothing yet" three times in a row and then start decoding a moment later. By then the caption has already been written, and it stayed there for good.

The fix is written. From now on **every label that does not switch the video off — EMPTY included — withdraws itself as soon as the player reports a width.** The verdict stays in the log, where it is useful to understand what happened, but the cell is cleared. Labels that genuinely remove the picture (NO STREAM, SAME STREAM) keep it, because there the video was removed on purpose.

**It ships tonight in the update.** If you send me the log I still have time to confirm this is exactly your case and not a second fault that looks like it.

Quote from stephan94

1_ Yes, the image continues to scroll with ‘Empty’

2_Here is the folder:

Thanks — that's the confirmation I needed. The picture running while EMPTY stays on screen means the label was simply wrong.

It's fixed: from now on EMPTY (and any label that doesn't switch the video off) withdraws itself as soon as the player reports a picture.

You'll be able to try it later tonight, when the update is released. Let me know about it, thank you in advance

SUPERDISH MOSAIC ENGINE — V. 2.2.0

STRUCTURE AND WINDOWS • The window now opens already filled: grid and cells are built before it appears, instead of an empty command bar turning into a mosaic while you watch • Minimum width measured on the real command bar, progress bar included: no more buttons cut off on the right, not even on a single-channel mosaic • That minimum is now a permanent constraint: resizing the window by hand no longer makes the buttons disappear • Guard added before the cells are built: closing the mosaic while it opens is a normal action, not an error • Eleven terminal states, one single label per cell: the first diagnosis wins • If the player queue fails to start, every cell says so: never a black and silent grid

CHANNEL OUTPUT — THE RULE CHAIN NOW GOES FROM MOST TO LEAST SPECIFIC • Rules are applied in order of precision: channel → frequency → mux → family → generic • Family rules (matched by name) used to sit above frequency rules, and in a chain the first match wins: four purpose-written cures were never executed • Recovered: CGTN 11265 V and 11305 V, TRT 11054 V, 11053 H, 11507 V • Nothing was taken away from anyone: wherever the broad rule provided deinterlacing or damaged-frame display, those lines were added to the dedicated rule • The German-mux rule, which switches on from a single channel name and then applies to the whole transponder, now excludes the frequencies that already have a cure of their own

TWO LOG LINES THAT CLOSE TWO WHOLE CLASSES OF PROBLEMS • For every cell, the options the player actually received are now written down: if a cure is not on that line, it was not applied — nothing left to guess • At every launch: satellite, frequency, polarisation and which frequency rules switched on. If a cure vanishes because the satellite being read is the wrong one, you see it at a glance instead of after half a day

CHANNEL RECOGNITION • Names are no longer matched by resemblance: "id" searched inside the name also matched KIDS and VIDEO, which lost their protection against a black cell. It is now a whole word • Same trap already closed for "l tv" (which caught LALEGUL TV and with it an entire transponder) and for "ntv" (which pulled the Russian package into the German-mux rules) • Türkmeneli is now recognised by channel and not only by frequency: its radio — which has no identifier of its own and is the second audio track of the TV service — comes out even when the selected satellite has not yet been written to the configuration • UMKA on both twin frequencies now runs video-only: its audio stream carries nothing, and it used to drag the picture down with it

TRT — 11054 V TURKSAT • On that package every channel is called TRT, and "trt" is one of the words in the Turkish name list: all of them fell into the generic family rule • The rule written specifically for that frequency — buffer margin and audio clock, against the "some arrive late but they arrive" reported from the field — sat below it, and was never executed • Frequency now comes before family, and that rule finally reaches the channel • On the same frequency the original timestamps are declared: field checks show zero flattened timestamps

EMPTY OVER A RUNNING PICTURE • The verdict is no longer final: if the picture arrives after the caption has been written, the caption withdraws itself • The check repeats at 25, 55 and 120 seconds and covers EMPTY, AUDIO ONLY and NO PICTURE • The verdict stays in the log: we clean the glass, we do not erase the diagnosis

OPEN AND DECLARED • The satellite read from the configuration can lag behind after changing card or dish: the mosaic reads that value and does not write it, but now it reports it • 10994 H Astra: measured a 1.036 s jump on the video timeline after 175 seconds, self-absorbing within one second. How often it repeats must be known before curing it • Hotbird, mux carrying HotBird 4K1: audio transcoding produces 8 milliseconds of correction per minute — inaudible, but recorded. HotBird 4K1 runs on a timebase 3.48 hours ahead of the rest of its own transponder

Download Superdish Mosaic Engine V. 2.2.0

In the Photos: Some old Hard Full mosaic viewing Turksat Packages now solved: 12685 V - 12265 V -11727 V

Quote from stephan94

Thank you very much – now ‘empty’ disappears after about ten seconds once the image appears.

As for the 4103 H frequency, unlike the other mosaics, it still can’t display an image.

Another minor issue has cropped up with this latest version: horizontal scrolling no longer works (video).

https://workupload.com/file/MdjwC3srBej

I think I’ve tried everything – vertical works, but horizontal doesn’t??

Thank you for report stephan94 . I worked on the windows last night, probably something left, well done! As per 4103 please send me log. Keep one and a half minute on, that's one case not found till now. Unfortunately I cannot try C Band in this period, only KU, I'm waiting for hard works to my system, it will need more time as prevented. You can send private, thank you. In these days I'm on holiday and I can work patiently on this fixes. Thank you again, kind regards.