History / Dongles

GitHub already prints the page name, so pages drop their own title The wiki renders the file name as an h1 above every page, so a leading "# Title" showed twice on all 41 pages. Removed it everywhere. That makes the file name the title, and it has to be chosen as one: every ASCII hyphen in it renders as a space, so six pages were titled wrong in the page list and search results as well as at the top. They are renamed, using U+2010 for a hyphen that has to survive and U+2215 for the slash in TI-99/4A. A slash that was only shorthand is dropped instead, which is how the HDMI dongle lost the one in A/V. Renamed pages are linked with markdown links rather than [[...]], so the exotic character stays in the slug and out of the prose. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

visrealm committed Aug 25, 2026

Dongles says the dongle is detected, not how The overview spelled out the detect mechanism twice, once in the intro and again in the FFC pin table, and Display-Output already owns that explanation in full: the GPIO read, the 1k bridge between HSYNC and VSYNC, and why a VGA dongle reads as no dongle. Dongles now states the outcome and links there. What survives is what a reader choosing a dongle needs: SCART is detected automatically with nothing to set, VGA and HDMI need nothing configured, and the probe runs once at boot so the host has to be powered off before swapping. Pin 9 still says it is used for detection, without saying how. Trimmed my own +5V note on pin 12 to the same depth, since it had the same problem. It listed where the cable spends the rail, down to the 180 ohm; now it says what the rail is for, which is switching the display to its AV input and to RGB. The resistor is on the SCART page with the rest of the pinout. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014FKyZHSa2tzJ58B1zNQRbx

visrealm committed Aug 24, 2026

the SCART dongle does need +5V The comparison table said it was unpowered and the FFC pin table said a passive dongle leaves pin 12 unconnected. Both are wrong for SCART. Passive is not the same as unpowered. The dongle has no active parts, its whole bill of materials being two connectors, the audio header and the 1k detect resistor, but it does take +5V off FFC pin 12 and pass it out on Mini-DIN pin 4. The cable spends it on two things: SCART pin 8, which switches the display to its AV input, and SCART pin 16 through 180 ohm, which tells it to show RGB rather than the composite signal on pin 20. Without that rail there is no RGB picture, so it is a requirement, not a pass-through convenience. The SCART page already had this right down to the resistor. Only the summary tables on Dongles contradicted it, which is the worse place to be wrong: they are what someone reads when choosing a dongle. Only the VGA dongle actually leaves pin 12 unconnected, and its schematic has no +5V net at all. Checked against pcb/scart/pico9918_scart_v1_3.pdf, where the +5V0 net does run from the FFC to Mini-DIN pin 4, and both dongles' BOMs. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014FKyZHSa2tzJ58B1zNQRbx

visrealm committed Aug 24, 2026

all seven board-art figures now follow the reader's theme Rolls out what the VGA figure proved: a light and dark bake per figure, chosen in the page with <picture> and a prefers-color-scheme source. GitHub's wiki sanitizer keeps both the picture and source elements with the media query intact, checked by fetching the rendered page rather than assuming. This was the alternative to a single theme-neutral PNG, which is lossy in a way worth recording. The best contrast any one ink can reach against both a white page and GitHub's dark canvas is 4.35 to 1 at #797979, short of the 4.5 WCAG wants for text, and it squeezes every colour into a narrow mid-grey band: the ink, muted and faint tiers collapse into each other and the teal accent falls to 3.25. A transparent background is the same problem wearing a different hat, since the ink then sits directly on the page. Baking both themes keeps the full palette instead of compromising it. The standalone view still works, which was the point of going to PNG: whichever bake is active is what right-click opens, and a PNG has no sub-resources for the raw host's default-src 'none' to block. The img src is the light bake, so anything that ignores picture still shows something sane. 2.8 MB for fourteen files. The light bakes were already committed, so this adds seven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014FKyZHSa2tzJ58B1zNQRbx

visrealm committed Aug 23, 2026

the seven board-art diagrams ship as PNG, so they open on their own These figures embed their board renders as data: URIs, which is what makes them work in a wiki page at all. It is also why they could not be opened on their own: raw.githubusercontent serves every file with Content-Security-Policy: default-src 'none'; style-src 'unsafe-inline'; sandbox and with no img-src override that blocks every image a document loads, a self-contained SVG's own embedded art included. Nothing was wrong with the file. Hashing the served copy against the one on disk gives the same sha256, which is exactly why viewing source told you nothing: CSP travels as a response header, and view-source shows the body. In the page the SVG arrives inside an img tag, so it renders as an image rather than a document and the header never applies, hence identical bytes looking fine in the page and broken in a tab. A PNG is the resource rather than a document that fetches resources, so there is nothing left to block. Right-clicking one of these figures now opens something that works. Verified first that a PNG really does render from that host under the same header, rather than assuming it. Baked light, since these pages already carry board photographs shot on white. Two costs worth stating: a PNG cannot follow the reader's theme the way the SVG did, and with the SVG gone the vector form lives nowhere in the repo, so these seven are regenerable only from the scratch generators plus the untracked board masters. Worth tracking that pipeline separately. 1.4 MB for the seven, which is slightly less than the SVGs they replace, since base64 inflates embedded art by a third. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014FKyZHSa2tzJ58B1zNQRbx

visrealm committed Aug 23, 2026

FFC pin 1 is BOOT, wider 22-way ribbons, VGA and JST figures Corrects a real error. Pin 1 is a BOOT line, not a ground: on PRO v2.0 an unmarked solder jumper beside the socket brings it out, most dongles leave it unconnected, and a dongle with BOOT button pads lets the board be reflashed over a USB-C extension without opening the machine. The grounds are pins 2, 4, 6, 8 and 10, still one immediately ahead of each signal. The bad claim came from inferring the grounds by exhaustion off the VGA dongle's six-net list. That argument cannot see an NC pin at all, so a pin the dongle simply does not connect looked like a spare ground. Noted in the generator so the next person does not repeat it. Ribbon widths now come from the socket each cable plugs into rather than being picked by eye, so the 22-way is the 1.7x wider cable it really is: 22 ways at 0.5 mm pitch is 11 mm of contacts against 6 mm. The adapter render is drawn far larger than the other boards for its physical size, measured through the FFC parts they share, so it takes a scale factor to bring its sockets into line with the F18A's. The FFC pinout figure was a bare strip floating on the page, which told you what pin 3 carries but not which contact pin 3 is. It now shows the socket enlarged off the board render, with the pin-1 triangle the board silkscreens and a ring round the unmarked jumper, since there is nothing printed on the board to find that jumper by. New: a VGA dongle wiring figure, deliberately parallel to the HDMI and SCART ones so the three compare, where what is missing is the point: no audio lead, because the board has no audio header, and no use made of pin 12. New: the 6-pin JST pinout for v0.4 to v1.1. The order is from the v1.1 schematic, where connector H5 has its pin numbers and net labels on matching rows: 1 VSYNC, 2 HSYNC, 3 BLUE, 4 GREEN, 5 RED, 6 GND, with 47 ohms in series on both syncs. It cross-checks against the board silkscreen, which prints a ground symbol then R G B H V, the same sequence read from pin 6 back to pin 1. Also recorded: these boards only ever had a VGA dongle and its cable was hard-wired into it, so there was no cable to swap and no HDMI or SCART option. Index: the adapter and the no-cut mod now sit under Dongles rather than beside it, in both the sidebar and Home. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014FKyZHSa2tzJ58B1zNQRbx

visrealm committed Aug 22, 2026

more dongles

visrealm committed Aug 22, 2026