// blog
How a PDF Schematic Gets Linked to a Boardview, in the Browser
A schematic PDF usually holds its text along with the drawing. The viewer reads each word and its position with PDF.js, then matches words against the board’s part and net names. Matching words become click targets. Case counts, and a name that could mean two nets stays unlinked rather than risk showing the wrong one.
The best thing about working with a boardview and a schematic together is not switching between them by hand. Click a part in one and it’s found in the other. Our free board viewer does this in the browser, and getting it right took a few decisions worth sharing.
Reading the words
A schematic PDF drawn by a CAD program normally holds real text, even where the lettering is drawn as lines. PDF.js gives us each run of text with its position. We split the runs into words, keep a box for each, and turn the boxes with the text, because schematics are full of vertical labels.
Then every word is checked against the board’s part and net names. On our sample board, the Glasgow Interface Explorer, 257 of its 266 parts link. The other nine are things like fiducials and mounting marks, which never appear on a schematic. On a real MacBook board with a 75-page schematic, 1,668 of 1,672 parts and 1,578 of 1,579 nets linked.
Why u1 isn’t U1
Our first version ignored upper and lower case. The test board’s U1 then showed up in 142 places. The real number is five. The rest were capacitor values: u1 means 0.1 µF. Upper and lower case now count, which also stops the word “reset” in a note from linking to a net called RESET.
Names need some give too. A board may call a net /SDA where the sheet says SDA, +3V3 where the sheet says 3V3, and multi-unit parts appear as U3A and U3B. Those are handled.
Some names stay unlinked on purpose
On hierarchical schematics the same label can appear on two sheets for two different nets. If X3 could mean either, we don’t guess. When you’re probing a board, showing the wrong net is worse than showing nothing.
A scanned PDF has no text at all, so it can’t link. The viewer says so instead of failing quietly.
On a second monitor
The schematic can move to a window of its own, so you can drag it to another screen. The two windows talk to each other inside your browser, with no server involved. Search covers all the text on the schematic, not only linked names, and invert gives you a dark sheet.
Nothing is uploaded. The PDF is read on your own computer.
When you want the schematic read for you
Finding names is one thing. Reading a power sequence out of the drawings is another, and it needs an AI that can see the page. We didn’t want your schematics on our server, so the viewer gives you a prompt to take to the AI you already use, with this board’s rail names in it. Attach the PDF, paste the answer back, and the viewer steps through the sequence on the board. It keeps only names that are really on the board. AI can misread a schematic, so check the result against it.
