Repository navigation
Releases
CRT Toolset v1.6.0
CRT Toolset v1.6.0
Includes everything in v1.6.0-dev4 (HDMI custom resolutions: 540p, 1024x768,
1280x768, 1280x1024, 1920x1200) plus the SD timing fix below.
Fix: 240p/288p test on single-SD-standard consoles (issue #1)
The SD timing test (Test 240p / Test 288p) now reprograms the display in place
instead of switching to the other SD standard. It clears the programmed AV
head's profile flag and re-requests the same SD mode with the framebuffer
format toggled, so the AV mode-setter reloads the patched timing. No
resolution change, no 50/60 Hz flash, about half a second. Region-locked
consoles that expose only 480i (NTSC) or 576i (PAL) now work. If the candidate
is not live afterwards, the old flip and then an HD bounce are used as
fallbacks. Details: docs/SD_SINGLE_FAMILY_240P.md.
- PROVEN (owner PAL console): activates and auto-reverts with no flip.
- PROVEN (issue #1 reporter, NTSC CRT, component and composite): "It works!"
Checked with Sonic CD and PSOne Classics MegaMan Legends
(#1 (comment)).
Build
python -B tools/build_ps3_app.py --sd-timing-fix -> CRT-Toolset-1.6.0.pkg,
APP_VER 01.81, in-app 1.6.0. Same code as v1.6.0-dev10.
SHA-256: a090c068241467e99dfb8c69a9625cae6ef4fdd34b91d58892b18cb1bb757172
Test
- AV Multi Out at your SD mode (480i NTSC / 576i PAL).
- Launch CRT Toolset, select Test 240p (NTSC) or Test 288p.
- Cross keeps it; otherwise it reverts after 15 s.
If it fails, the flow reverts to stock. Log: /dev_hdd0/tmp/ps3_240p.log.
CRT Toolset v1.6.0-dev10 (in-place SD reprogram)
Pre-release. The 240p/288p test now activates the patched timing without any mode change: after patching the timing table, the app re-requests the same SD mode with the framebuffer format briefly switched (XRGB->XBGR->XRGB), which makes the AV manager reprogram from the patched table. No 480i/576i flip, no black flash, about half a second. Used on every console; the old flip and an HD bounce remain as fallbacks. Fixes region-locked (single-SD) consoles that refuse the other SD family and expose no HD modes (issue #1). Proven on a PAL console (activation and auto-revert in place); NTSC 480i single-SD still to be confirmed. APP_VER 01.80. See docs/RELEASE_v1.6.0-dev10.md.
CRT Toolset v1.6.0-dev4 - HDMI custom resolutions (540p, 1024x768, 1280x768, 1280x1024, 1920x1200)
CRT Toolset v1.6.0-dev4: HDMI custom resolutions
This release adds an HDMI custom resolutions menu with five modes:
| Mode | For | Notes |
|---|---|---|
| 540p (1920x540 progressive, 60 Hz) | HD CRTs | start from XMB 1080i or 1080p |
| 1024x768 | 4:3 PC monitors / VGA-style CRT setups | start from XMB 1080p |
| 1280x768 | wide PC monitors | start from XMB 1080p |
| 1280x1024 | 5:4 PC monitors | start from XMB 1080p |
| 1920x1200 | 16:10 PC monitors | start from XMB 1080p |
Each choice opens a 15-second preview inside the app. X keeps it for the
session; Circle, or no input, restores the previous output. The helper
restores stock output by itself even if the app is closed or the PC
disconnects. It is the exact package tested on the owner's console.
PC resolutions
- PROVEN (capture + receiver, in-app): 1280x768 showed a readable
confirmation screen, the receiver reported 1280x768, and the timeout
restored 1920x1080. 1024x768 bound and drew the app at 1024x768 with a
readable confirmation screen and checked restoration. See
HDMI custom resolutions. - PROVEN (earlier XMB-path tests): 1280x1024 showed XMB with the receiver
reporting 1280x1024. 1920x1200 showed XMB with a live receiver report of
1920x1200. Both restored stock 1080p. See
PC presets
and XGA. - UNKNOWN: these four modes were not retested with this exact package,
exact refresh is unmeasured (nominal 60 Hz requested), and 1024x768 uses
non-standard clock/totals. Game compatibility is untested.
How to use 540p
- Set XMB to HDMI 1080i (recommended for HD CRTs, which usually lack
1080p) or 1080p. Other XMB resolutions are refused for 540p. - CRT Toolset → HDMI custom resolutions → 540p / 16:9.
- Wait about 25 s while the helper switches 1080i → 720p → 1080i, then
check the 540p picture and press X to keep it or Circle to restore.
Evidence
- PROVEN (owner, Xiaomi HDMI monitor): the monitor reported 1920x540
at 60 Hz with a full, correct picture. See
540p resolved. - PROVEN (capture frames, this package): the in-app preview shows the
complete menu with correct proportions. Hardware fetch pitch was 0x4000,
from both a 1080i start and a 1080p start, and stock output was restored. - PROVEN (cause of the earlier failures): forcing HDMI AVI VIC 0 broke
monitor lock. The in-app crop came from the doubled pitch writer being
skipped. Both are fixed in this build.
Limits
- UNKNOWN: real HD CRT behavior, component output, games (especially
games that keep a fixed display pitch), other models/firmware and
measured refresh. 540p samples every other row of the 1080-line source;
this is not filtered scaling. - UNKNOWN: whether some displays reject the stock AVI description (VIC 5)
for a progressive 1920x540 signal. - The 16:9 picture fills each resolution; 4:3 and other aspect handling are
planned for a later release. - Requires CFW 4.93 (Evilnat + Cobra tested), HDMI only. Never install or
update CRT Toolset while it is running.
Files
| File | SHA-256 |
|---|---|
CRT-Toolset-1.6.0-dev4.pkg (344784 bytes) |
ce14d2b178c9a6eac0489a931c96775e0b81628585bed7c3154b86d643c311e4 |
helper crt_hdmi_custom.sprx |
556ad75bb7d445628c7694a214376bf9fa2ea32fca9c19951e0aee168fe8c2a6 |
CRT Toolset v1.6.0-dev8 (single-SD HD bounce)
Pre-release. Single-SD (region-locked) consoles: the single-SD timing fallback now forces a real mode change by bouncing through 720p (else 1080i) before returning to the SD target, so the AV mode-setter re-runs and loads the patched 240p table. dev7 picked the right head but a same-mode re-request never reaches the mode-setter (issue #1 log: Post-switch match=0). Expect a brief black flash (the HD bounce) before the test picture and on revert; fully reversible, no reboot. Dual-SD consoles unchanged. APP_VER 01.77. Host tests 44/44. See docs/RELEASE_v1.6.0-dev8.md and docs/SD_SINGLE_FAMILY_240P.md.
v1.6.0-dev7 — single-SD timing targets the active AV head
Pre-release. Supersedes v1.6.0-dev6. Full dev4 (HDMI custom resolutions) feature set.
Single-SD-family timing now targets the active AV head.
dev6's single-SD fallback chose which AV head to reprogram from the SD target (480i → head 0, 576i → head 1). GitHub issue #1 (NTSC console, 480i over component) shows that assumption is wrong: that console drives its live 480i output on head 1, with head 0 idle. dev6 cleared idle head 0 and left the live head at stock timing, so the candidate never activated and the test reverted with "The active analog timing did not match the candidate."
dev7 reads the head that is actually programmed (AV cache state==2) and reprograms that one. switch_revert does the same. The old target-derived default is kept only as a fallback when zero or both heads are active.
patch_active_head()— the programmed head, or -1.switch_begin_target/switch_revertuse it;patch_invalidate_headnow logs the no-op case too.- Host tests: 44/44, including an NTSC head-1-active case.
For NTSC / component (issue #1): AV Multi Out at 480i → launch → Test 240p. Expected: the 240p picture appears with a 15-second countdown. A failed activation reverts to stock cleanly (no reboot). If it still fails, pull /dev_hdd0/tmp/ps3_240p.log — it now logs Single-SD fallback: invalidate head=....
CRT-Toolset-1.6.0-dev7.pkg — SHA-256 5de272bcc5210ef8e4500eb89ef4d28af4c0d5a4757239bf150f85d5779b89b4
v1.6.0-dev6 — single-SD-family timing (cache invalidation)
CRT Toolset v1.6.0-dev6 — single-SD-family timing via head cache invalidation
Pre-release. Supersedes v1.6.0-dev5. Full dev4 (HDMI custom resolutions)
feature set.
What changes from v1.6.0-dev5
dev5 let region-locked consoles past the mode-detection gate, but the timing
test still needed to configure the other SD family (480i↔576i) to make the
AV manager reprogram the patched table. A console that exposes only its own SD
standard (NTSC → 480i, PAL → 576i) cannot configure the other family, so the
test failed with:
RAM restored; display recovery failed. Reboot Console
dev6 keeps that flip as the primary path (no change on consoles that can
configure both families) and adds a fallback: when configure(alternate) fails,
it invalidates the AV manager's per-head cached profile index and re-requests
the current mode, which reprograms the RSX CRTC from the patched table without
any SD-family change. switch_revert does the same on recovery, so a failed
flip no longer demands a reboot.
- New:
patch_invalidate_head()(app/source/lv1_patch.c). - Changed:
switch_begin_target/switch_revert(app/source/switch_flow.c). - Mechanism and addresses:
docs/SD_SINGLE_FAMILY_240P.md. - Host tests: 44/44 (
python -B tools/test_app.py).
Build
python -B tools/build_ps3_app.py --sd-timing-fix
Output: CRT-Toolset-1.6.0-dev6.pkg, APP_VER 01.75, in-app 1.6.0-dev6.
Test (NTSC + component / single-SD console)
- AV Multi Out, start at 480i (NTSC). Confirm XMB shows 480i.
- Launch CRT Toolset v1.6.0-dev6.
- Select Test 240p.
- dev5: "RAM restored; display recovery failed. Reboot Console".
- dev6 expected: proceeds without switching to 576i; the 240p test picture
appears with a 15-second countdown.
- Press Cross within 15 s to keep, or wait for the automatic revert.
If it still cannot activate, the flow reverts to stock cleanly (no reboot
needed). Pull /dev_hdd0/tmp/ps3_240p.log for the AV head ... cache invalidated and videoConfigure lines.
Validation
- PROVEN: primary (flip) path unchanged; owner's PAL console unaffected.
- UNKNOWN until a live single-SD console test: whether the repeated same-mode
request reaches the AV mode setter so the invalidated flag triggers the
reprogram. Reversible: a failed activation reverts to stock.
v1.6.0-dev5 — single-SD-family timing gate
CRT Toolset v1.6.0-dev5 — single-SD-family timing gate
Pre-release. Carries the full v1.6.0-dev4 (HDMI custom resolutions) feature
set plus one fix to the SD timing test.
What changes from v1.6.0-dev4
PROVEN (implementation): fixes the SD timing test failing on consoles that
expose only one SD standard:
Firmware does not expose both SD modes; timing test stopped
cellVideoOutGetResolutionAvailability reports only the console's region SD
standard — a PAL unit advertises 576i alone, an NTSC unit 480i alone — so the
old gate (which required both 480i and 576i available) stopped every
region-locked console. The encoder drives either interlaced family and the app
configures either in practice, so the gate now requires only that one SD
family be exposed. The transition still drives both families (480i↔576i, both
15 kHz-safe); configure() and the automatic 15-second revert recover a
console that genuinely refuses the second family.
- Changed:
app/source/switch_flow.cselect_modes()— the availability
check becameavailable(target)<=0 && available(alternate)<=0(was||).
New message on genuine refusal:Firmware exposes no SD mode for AV Multi; timing test stopped. - Host tests: 44/44 pass (
python -B tools/test_app.py).
Build
python -B tools/build_ps3_app.py --sd-timing-fix
Output: build/hdmi-custom/deliverables/CRT-Toolset-1.6.0-dev5.pkg
APP_VER 01.74, in-app version 1.6.0-dev5.
SHA-256 5e1d519531d3451f9d6efbf0dbcd37385a930e79c2bc44b9af8c19034703f541,
344,848 bytes.
Validation
PROVEN (owner live test, 2026-11-15): PAL console, AV Multi Out, RGB SCART,
starting at 576i. Test 288p proceeded past the gate and the picture held;
previously this console hit the "does not expose both SD modes" stop. Installed
through webMAN; all four installed files matched the package byte-for-byte.
CRT Toolset v1.5.0-dev19 - Experimental HDMI 540p test
CRT Toolset v1.5.0-dev19 — experimental HDMI 540p test
This prerelease adds a timed 1920×540 progressive HDMI experiment for HD
CRT testing. It is the exact dev19 package tested on the owner's console.
It does not add permanent 540p output or 540p background enforcement.
Compatibility
- PROVEN (implementation): the HD experiment accepts HDMI only,
starting from stock 1080i at 59.94/60 Hz on the supported 4.93 CFW
layout. It refuses component, RGB SCART, clones, 50 Hz, unknown records and
active SD enforcement. Sources:app/source/main.c:hd_runand
plugin/vsh_hd540.c:prepare. - PROVEN (capture/readback): the signal recipe has produced a full XMB
picture and receiver-reported 1920×540/progressive on the owner's HDMI
capture setup. The source scanout samples alternate rows of the unchanged
1080-line framebuffer; this is not filtered downscaling. Sources:
dev18 signal result
and dev19 app-menu fix. - CLAIMED (owner): the final dev19 manual test worked: “540p works!”
Console: Slim CECH-25xx, Evilnat 4.93 CEX + Cobra, HDMI monitor/capture. - UNKNOWN: real HD CRT compatibility, component 540p, other models or
firmware, games, long-duration use and measured refresh rate. This build
targets nominal 60 Hz operation; exact refresh has not been measured. - INFERRED: ordinary HEN remains unsupported by this LV1 timing backend.
HEN support notes.
Install
- Exit CRT Toolset and install CRT-Toolset-1.5.0-dev19.pkg through XMB
Package Manager. It updates title IDPS240P001; its package version is
01.68 and its menu says v1.5.0-dev19. - Reboot once after upgrading to clear the previous resident helper.
- Use HDMI and select 1080i in the PS3's video output settings. Start
from 59.94/60 Hz, with SD 240p/288p enforcement disabled. If old SD timing
edits are active, disable enforcement and reboot before this HD test. - Keep a way to return to stock 1080i. The timer runs in the resident helper
and depends on a functioning console/OS; it cannot guarantee recovery
from a whole-console freeze.
Test sequence after reboot
Open HD CRT / experimental 540p. Press Cross on the following in order,
reopening CRT Toolset and this submenu after each return to XMB:
- Test stock 720p / 1080i round trip (exit to XMB).
- Test AVI packet only (stock 720p; 15 seconds).
- Test 540p with custom AVI packet (15 seconds).
Skip “Test AVI packet only (stock 1080i; 15 seconds)” for normal testing;
that separate diagnostic produced a black picture on the tested receiver.
The first two checks are required once per resident-helper session. After
they pass, repeat only the 540p option until the next reboot/helper restart.
An error invalidates those prerequisites. This prerelease does not combine
all three steps into one button.
Each timing action exits the app normally before changing the display.
Leave the controller untouched during the test and return. The candidate
holds for 15 seconds after native readiness, then automatically restores
stock timing and returns through 720p to the saved 1080i mode. Mode changes
and receiver resynchronization add time before and after the hold; allow
roughly 30–40 seconds for the entire candidate sequence. A few seconds of
black while the receiver reacquires sync were observed in the app-menu test,
so 15 seconds of active timing is not a promise of 15 seconds of visible lock.
Check that the TV locks to the candidate, the entire XMB fits, and the picture
returns normally to 1080i. The test always restores; Cross does not keep
540p permanently.
Feedback requested from HD CRT testers
Please report the TV model, HDMI connection chain, starting resolution,
whether the candidate locks, whether the full picture fits, and whether
automatic 1080i recovery succeeds. A photo/video or the TV's signal-status
display is useful. Attach these logs if possible:
/dev_hdd0/tmp/vsh_240p.log/dev_hdd0/tmp/ps3_240p.log
Component users should report their connection first: this HD experiment
currently refuses component, so this package is not a component 540p test.
Verification and changes
PROVEN (live regression): dev19 fixes an app-to-XMB row-pitch mismatch
that made dev18 abort early and falsely report failed cleanup. The app's
7680-byte pitch could be replaced by XMB's 8192-byte pitch during switching.
The check now requires current VSH and AV-source pitch to agree, then compares
that with the hardware pitch. The waveform and 15-second hold are unchanged.
Actual app-menu test sequence 280557604 held for 15.219765 seconds and ended
with DONE/detail 0. The owner's repeat, sequence 458662207, held for
15.453232 seconds and also completed with error 0 and stock recovery.
Evidence: docs/HD540_DEV19.md and ignored owner-test logs.
PROVEN (offline verification): host regression/failure-recovery tests,
private firmware fingerprints, compiled helper loader contract, self-authored
wrapper emulation and package-member checks passed. Normal stable release
artifacts were preserved. The normal build still excludes the experimental
HD menu by default; source builds use --experimental-hd540 for both helper
and app. This prerelease is not marked as GitHub's latest stable release.
Download integrity
CRT-Toolset-1.5.0-dev19.pkg — 308560 bytes
SHA-256:
50ca756c9dbbc1fe74a97e2d5fb120259163ab9b6b49a62860cb4489c5dde4d2
PROVEN (artifact identity): this is the existing tested package, not a
newly rebuilt binary. SHA256SUMS.txt and package_manifest.json accompany
the download. Stable v1.4.3 remains available separately.
CRT Toolset v1.4.4-rc1 - Official SD release
CRT Toolset v1.4.4-rc1 — official SD release
The owner approved promoting this tested build to the official release on
2026-11-14. The existing tag, download names and in-app version retain
rc1; the package is unchanged. This promotion does not add the rc2 audio
logger or the separate experimental HDMI features.
What changes from v1.4.3
PROVEN (implementation): fixes the fixed-PID lookup behind this error:
PATCH ERROR: 4.93 sysmgr PID 10 not found; check CFW syscalls
Timing patches and gamma now discover the supported sysmgr across all
32 process slots. Discovery verifies immutable service identity, rejects
ambiguous matches, and rechecks process/page mappings. Existing checked
writes and rollback remain in place.
Implementation and validation.
PROVEN (supplied external log): the affected CECH-2004A setup uses PID 9
and this build resolves all 54 timing targets. CLAIMED (tester): timed
240p recovery and background enforcement work in XMB and the app.
CLAIMED (owner): the same package also works on the owner's CRT setup.
External result.
PROVEN (package scope): includes SD 240p/288p, background enforcement,
stock 480i/576i selection and analog gamma. Title ID is PS240P001, package
version 01.44, menu version v1.4.4-rc1. Source: the checked package
manifest and tools/build_ps3_app.py --pid-compatibility.
The supported timing backend remains 4.93 CFW + Cobra. Ordinary HEN and
HDMI 240p/288p are not supported by this release; experimental HDMI 540p/PC
presets remain separate development builds.
Installation
- Exit CRT Toolset, install CRT-Toolset-1.4.4-rc1.pkg through XMB Package
Manager, then reboot normally once to clear an older resident helper. - Use your analog SD/CRT connection with the usual CFW/Cobra syscalls
enabled. Confirm v1.4.4-rc1 in the menu. - Start with Recheck timing memory, then the timed Test 240p or
Test PAL 288p before enabling background enforcement.
If discovery fails, stop and retain /dev_hdd0/tmp/ps3_240p.log.
For enforcement issues, also retain /dev_hdd0/tmp/vsh_240p.log.
Gamma and TV calibration guide.
PS1 compatibility note
CLAIMED (tester): ps1_emu produced lower-pitched/crackling audio and a
zoomed/cropped picture on their setup. Switching the launcher to
ps1_netemu resolved both, including Toy Story 2, Battle Hunter and
Spider-Man 2000. Try that alternative if you encounter these symptoms.
This is a reported workaround; ps1_emu is not established as universally
incompatible. Exact cause, integer scaling and latency remain UNKNOWN.
Report, emulator evidence and startup-path caveat.
Validation and download integrity
PROVEN (prior checks): all 32 relocated PIDs, SD apply/exact undo,
gamma with a relocated service, decoys, duplicate identities, denied reads,
malformed lists and changing mappings passed regression checks. Private
4.93 identities, helper loader behavior and decrypted package members were
also checked. Sources: the linked discovery evidence and
tools/test_app.py --firmware.
PROVEN (promotion check): all three existing release assets were
downloaded and verified against their GitHub SHA-256 digests. The tested
PKG remains 255568 bytes, SHA-256:
e35062e49bdb6a3cb785442c49d5e8e855389d44d92590fd2bfa3aba5614165e
SHA256SUMS.txt and package_manifest.json accompany it. Local publication
evidence: ignored local_only/releases/promote-rc1-20261114/.
v1.4.4-rc2: PS1 audio diagnostics
Diagnostic build for the reported PS1 lower-pitch/crackling issue at 240p. No audio fix is claimed. The owner reports normal 240p audio in Toy Story 2 and Battle Hunter on their CRT; the affected CECH-2004A needs game-time logs.
Install the PKG and reboot normally once to remove the old resident helper. Open CRT Toolset 1.4.4-rc2, enable 240p background enforcement, exit and play the same affected NTSC game for 30–60 seconds within the first ten minutes. Exit the game normally.
Before rebooting or reopening CRT Toolset, copy these FTP files:
/dev_hdd0/tmp/ps1_audio_diagnostic.log/dev_hdd0/tmp/vsh_240p.log/dev_hdd0/tmp/ps3_240p.log
Return those three text logs with the game title/release and launch method. NTSC region, white-RCA mono-TV audio and normal 480i playback are already known.
The additional observer is read-only: at most 120 samples, five seconds apart, no PCM/audio recording or audio/video-setting changes. It identifies both PS1 emulators, but reads the audio layout only for the fingerprinted 4.93 ps1_emu. Cached buffer state and cursor indices do not measure the analog sample clock or pitch. The original 240p timing/correction policy is unchanged.
SD-only build, same CRT Toolset title ID: replaces the installed app and excludes the experimental HDMI menus. Stable v1.4.3 remains unchanged.
Validation: full host/private-firmware suite, helper loader/signing checks and all packaged members pass. The observer was also exercised during Battle Hunter alongside the owner's original SD helper. The complete diagnostic PKG on the affected console remains untested.
Detailed instructions and evidence
PKG SHA-256: 9ec4bfb7325f68f57c653cb6ae5249754b3d4468fa24eb7e52921a85941b7779