diy-layout-creator/diylc/docs/plans/tube-socket-datasheets.md

Tube Socket Datasheets

Naming every electrode of a TubeSocket from the tube that plugs into it.

1. Goal

Follows DIP IC Pin Names and Transistor Pinouts. A tube socket already carries a free-text "Electrode Labels" property, and every consumer reads it through getControlPointNodeName: the netlist, hover tooltips, the AI project context and the schematic view. In practice it is almost never filled in. Of 95 tube sockets in the sample corpus, the user typed a designation on about half (12AY7 24 times, 6L6GT, 5Y3, KT88, 12AX7) but real electrode names on only four; the rest carry the default 1,2,...,9. A datasheet of the tubes DIYers use (guitar amps, hi-fi, radio) sets the base, the electrode names and the designation in one step, from the toolbox or from the AI.

What is wrong today

Pins of one component that share a name are one net (NetlistBuilder, Node.railName). The six factory variants that carry electrode names break that rule:

Variant	Labels	Effect in the netlist
Chassis Noval 9A (12A?7)	A,G,C,H,H,A,G,C,CT	both anodes, both grids, both cathodes and both heater ends shorted together
Chassis Noval 9CV (EL84)	n/c,G,C,H,H,n/c,P,n/c,S	heater ends shorted; the three n/c pins joined (only NC and RSV are exempt)
Chassis Octal 5T	n/c,H,n/c,P,n/c,P,n/c,H	both rectifier plates shorted, filament ends shorted, n/c pins joined
Chassis Octal 7AC (6?6)	n/c,H,P,S,G,n/c,H,C	heater ends shorted, n/c pins joined
Chassis 7-pin 7BK	G,IS,H,H,P,S,C	heater ends shorted
Chassis Octal 8ET (EL34)	IS,H,P,S,G,n/c,H,C	heater ends shorted

Heater ends are separate nets: a 12AX7 runs on 12.6 V across pins 4-5 or on 6.3 V with 4 and 5 joined against 9, and that choice is made by the wiring, not the tube.

2. Design decisions

#	Decision	Choice	Rationale
V1	Datasheet row	group, tube, base, basing, name1, ..., nameN, e.g. Dual Triode,12AX7,B9A,9A,2P,2G,2K,H4,H5,1P,1G,1K,HCT. base is a TubeSocket.Base constant; basing is the RETMA basing code, or - for a tube that has none (most Soviet types); the name count must equal the base's pin count	The trailing-names shape of the IC datasheets. The basing code is what tube data books print, so rows can be checked against them, and it lets a test assert that tubes sharing a basing share their names.
V2	Model identity	enableDatasheet = true, datasheetCreationStepCount = 2, datasheetModelColumn = 1	The toolbox groups by family, the AI names the tube.
V3	applyModel	Sets base, electrode labels and designation. Leaves mount, angle and colours alone	The datasheet says what the tube is, not how the user mounts the socket.
V4	One row per tube	Equivalents get their own rows (12AX7, ECC83, 7025, 5751; 6L6GC, 5881, KT66), as the ICs list TL072 and NE5532 separately	Users and the AI look a tube up by the number printed on it.
V5	Electrode names	Role letters as in American data books: P plate, G grid of a triode, G1/G2/G3 control, screen and suppressor grids, K cathode, BP beam plates, IS internal shield, SH the metal shell or base sleeve (pin 1 of metal and GT octals), HCT heater centre tap. A tube with two or more sections puts the section number first, as the ICs do: 1P, 1G, 1K, 2P (and T/P prefixes for triode-pentodes: TP, TG, PG1)	Section-first keeps G1 meaning control grid, never "grid of section 1".
V6	Heater and filament ends	Named after their pin: H4, H5 on a noval, H2, H7 on an octal. A directly heated filament, which is also the cathode, is F2/F8. The _n suffix is never used for them	Each end is its own net (§1). Naming by pin follows VCC+1/VCC+5 on the LM3886, the precedent for repeated roles that the package does not join.
V7	Joined pins	A repeated name with the _n suffix, numbered by occurrence as on the ICs, only where the data book says the pins are connected inside the tube (e.g. the EF80 cathode on pins 1 and 3, K_1/K_2)	Same rule as the ICs, so NetlistBuilder joins exactly what the tube joins.
V8	Unused pins	NC for no connection. Pins the data book marks "IC" (internal connection - do not use as a tie point) are IC, and IC joins NC and RSV in Node.NON_NET_NAMES	Several IC pins on one tube must not become one net. One-line core change.
V9	Factory variants	The six named variants of §1 are retired (deletedOn), as the DIP8/DIP16 and resistor variants were; the four plain ones (PCB/Chassis Noval/Octal) stay as mount presets	The datasheet replaces the named ones with correct names; leaving them would keep handing out shorted netlists. The plain ones preset the mount, which a datasheet model deliberately leaves alone (V3), and their numbered labels short nothing, so a plain preset plus a model is the intended way to place a tube.
V11	Drawn labels	Drawn through AbstractMakerBoard.getDisplayPinLabel, so G1_1 shows as G1 and parenthesised notes are dropped, as on the ICs; node names keep the suffix	Joined pins carry one role; printing the suffix crowded small sockets (6S19P).
V12	Equivalents on one basing	A tube given a RETMA basing takes that basing's names, including its section numbers, even where its own data sheet numbers the units the other way (6N8S and 6N9S use the 6SN7's); tubes with no RETMA code keep their sheet's numbering	Swapping a 6N8S for a 6SN7 must not relabel the socket; section numbers are labels, the pin grouping is identical.
V13	Property name	The electrodeLabels field and its accessors keep their names; the property is shown as "Pin Names", as on the ICs	One name for the same idea across ICs and tubes, for users and the AI. Variants store values by property name with no alias mechanism, so the two factory presets are re-keyed; a user's own saved tube variants lose their labels, as user variants lost "Display Pin #s" when the ICs renamed it.
V14	What a pin shows	New pinLabels field with the shared PinLabels enum, shown as "Pin Labels": None, Names (default), Numbers. Numbers are the base's pin numbers, drawn where the names go	Matches SIL_IC and MultiwattIC. Old files deserialize with null, which reads as Names - exactly what every socket drew before.
V10	Existing projects	Unchanged. A project placed from a retired variant keeps its labels and its shorted nets; nothing rewrites user data	Labels are user data. The release note can say how to re-apply a model.

Explicitly out of scope

• Jumbo4PinTubeSocket (fixed G,F1,P,F2 for 211/805/845): a separate class with its own ceramic body and terminals. The UX4 base of 300B and 2A3 was added later (§7, phase 5).
• Magnoval (B9D: EL509, PL519), loctal and the B12C compactrons: no datasheet rows until a user asks; B12C exists as a base but few DIY tubes use it.
• SubminiTube and the tube schematic symbols.
• Translating group names, as for the ICs and transistors.

3. Changes

diylc-core

• Node.NON_NET_NAMES: add IC (V8).
• import-defaults/variants.xml: retire the six named TubeSocket variants (V9).

diylc-library

• TubeSocket: implements IDatasheetSupport, descriptor attributes (V2), applyModel (V3), a TubeSocket(String[] model) constructor. @AiNote extended with the naming rules of V5-V8 and the datasheet model.
• datasheets/TubeSocket.datasheet: rows per V1, each read from a manufacturer data sheet or a tube data book (RCA, GE, Sylvania, Tung-Sol, Philips/Mullard, JJ, Svetlana, Electro-Harmonix), or two independent agreeing sources.

Generated

• catalog_full.json regenerated; pinout-drivers.txt updated if the analyzer starts reporting "Electrode Labels" differently.

4. Datasheet inventory

TubeSocket.datasheet small-signal rows, 46 tubes:

Group	Basing	Tubes
Dual Triode	9A	12AX7, ECC83, 7025, 5751, 12AT7, ECC81, 12AU7, ECC82, 12AY7, 6072, 12BH7, 12DW7, 12AZ7, 5965, ECC99
Dual Triode	9AJ	6922, E88CC, 6DJ8, ECC88, 7308, 6CG7, ECC85
Dual Triode	9LP / 9H / -	6FQ7 / 5687 / 6N1P, 6N2P, 6H30P
Dual Triode	8BD	6SN7, 6SL7, 6080, 6AS7, 6BX7, 6BL7
Triode	9AG / 6Q	12B4A / 6J5, 6C5
Pentode	9CQ / 9AD / 9AQ	EF86, 6267 / 5879 / EF80, EF184
Pentode	7BK / 7BD / 8N	6AU6, 6BA6 / 6AK5 / 6SJ7

Small-signal sourcing notes:

• Read from RCA, Tung-Sol, GE, Sylvania, Philips, Brimar and JJ sheets (frank.pocnet.net mirror, tubebooks.org) cross-checked against TDSL basing codes. 6SL7, 6080, 6BL7, 6BA6 and 6C5 rest on their basing code without the sheet opened; 6H30P on a blurry lead drawing - the weakest rows.
• Lookalikes that differ: 6FQ7 has pin 9 NC, not a shield, so it is 9LP, not 6CG7's 9AJ; 5687 is not a 12AU7 (plate 1 on pin 9, centre tap on pin 8); 6N2P has a shield on pin 9 and no centre tap, so it is not a 12AX7; 6N1P has the 9AJ functions but its sheet numbers the units the other way.
• 12DW7: unit 1 (pins 6-8) is the 12AX7-like section, unit 2 (pins 1-3) the 12AU7-like one.
• Left out: 6C45P (no pin list found), 6N8S (said to equal 6SN7, unit order unconfirmed), 6SF5.

TubeSocket.datasheet radio rows, 54 tubes:

Group	Tubes
Converter	12BE6, 6BE6, 6SA7, 12SA7, ECH81, ECH83, UCH81
Pentode (RF/IF)	12BA6, 6BD6, 6SK7, 12SK7, 6SG7, EF85
Diode-Triode / Diode-Pentode	12AV6, 12AT6, 6AV6, 6AT6, 6SQ7, 12SQ7, EABC80 / EBF89, UBF89
Triode-Pentode	UCL82
Beam Tetrode (radio output)	50C5, 50L6GT
Rectifier	35W4, 35Z5GT
Tuning Indicator	EM84, EM80, EM81, 6AF6, 1629
Space-Charge	12K5, 12AD6, 12EG6, 12AF6, 12CX6, 12DZ6, 12EK6, 12AC6, 12BL6, 12AE6, 12FM6, 12FK6, 12EL6, 12FX8, 12DS7, 12EM6, 12J8, 12DL8

TubeSocket.datasheet power, rectifier and reference rows, 63 tubes:

Group	Tubes
Power Pentode	EL84, 6BQ5, 7189, 7189A, 6CW5, EL86, EL34, 6CA7, 7591, 6F6G, 6K6GT, 6EH5, 6DL5, EL95, 6AK6, 6AR5, 6CL6
Beam Tetrode	6V6GT, 6V6, 6L6GC, 6L6, 5881, 7581A, KT66, KT88, 6550, KT120, KT150, 6Y6G, 6AQ5, 6BF5, 50B5, 6973, 12AB5, 6AS5, 6CU5, 35C5, 6CZ5, 6EM5, 6BW6, 6CM6
Triode-Pentode	6BM8, ECL82, ECL86, 6GW8
Rectifier	5AR4, GZ34, 5V4G, GZ32, 5Y3GT, 5U4GB, EZ81, EZ80, 6CA4, 6X4, 6X5GT, 6X5, 6AX5GT
Voltage Reference	0A2, 0B2, 0D3, 0C3, 0A3

Radio and power sourcing notes:

• Read from RCA, GE, Tung-Sol, Sylvania, Philips, Amperex, GEC/M-OV, JJ, Brimar and Svetlana sheets (frank.pocnet.net), with TDSL as the second source. Where TDSL's basing code disagrees with the printed data book the book wins: EL34 8ET (TDSL 8EP), 6X4 7CF (5BS), 6F6 7S (7AC), 7591 8KG (Sylvania; TDSL 8KQ).
• A pin carrying several electrodes takes the main one's name: K for cathode + g3/beam plates/shield, G3 for g3 + shield on 7BK, G5 for shell + g5 on 6SA7, H5 for heater end + triode cathode on 12FX8.
• Space-charge tubes follow each data book's grid numbering: on 12K5, 12DS7 and 12DL8 G1 is the space-charge grid and G2 the control grid; on 12J8 and 12EM6 G1 is the control grid.
• Other names introduced: HT the panel-lamp heater tap on 35W4/35Z5GT (not a centre tap); HK8 the heater-cathode pin of 5AR4-type rectifiers; J_1/J_2 the base jumper of octal regulators (0A3/0C3/0D3), which opens the supply when the tube is pulled; TA target, DE deflection electrode, 1RC/2RC ray-control electrodes of tuning indicators; DK a separate diode cathode.
• Rectifier plate numbers follow each book: 5L (Amperex) has plate 1 on pin 4, 5T (RCA) plate 2.
• KT88 conflict: the row follows GEC (pin 1 base shell); JJ brings g3 out on pin 1.
• The same-basing test treats NC, IC and SH as interchangeable, because the unused pin 1 of a basing is a shell, a no-connect or an internal connection depending on envelope and maker.
• Left out: EF89 (basing code and joined shield pins unconfirmed), 12DK6 (no sheet), metal 6F6 (shell pin assumed), 6V4 and 6005 (TDSL only), 6AV5 (octal TV sweep tube); 6E5 and 6U5 are on the unsupported UX6 base - 1629 is their octal counterpart.

TubeSocket.datasheet Soviet rows and EL91, 30 tubes (6N1P and 6N2P are counted with the small-signal rows):

Group	Tubes
Dual Triode	6N1P-EV, 6N2P-EV, 6N6P, 6N23P, 6N23P-EV, 6N24P, 6N3P, 6N8S, 6N9S, 6N13S, 6N5S
Triode / Power Triode	6S45P / 6S19P, 6S4S
Pentode	6Zh32P, 6Zh1P
Power Pentode	6P14P, 6P14P-EV, 6P18P, 6P15P, EL91, 6AM5
Beam Tetrode	6P1P, 6P3S, 6P3S-E, 6P6S, 6P27S
Rectifier	5C3S, 5C4S, 6C4P

Soviet sourcing notes:

• Ranked by use in guitar and hi-fi DIY (6N2P, 6P14P, 6N23P, 6N1P, 6P3S, 6N8S lead); read from Soviet reference sheets (magictubes.ru, r-type.org handbook scans, chipdip, radiolamps.ru) and Voskhod/Svetlana/Sovtek sheets.
• Compatibility, never assumed: 6P14P and 6P18P are EL84 pinout; 6P15P is not (g3 + shield on pins 1 and 6); 6P3S/6P6S are 6L6/6V6 pinout; 6P27S fits an EL34 socket but is a 7AC beam tetrode, not 8ET; 6N8S/6N9S/6N13S/6N5S are 6SN7/6SL7/6AS7 pinout; 6N23P is 6922 pinout; 6N1P and 6N6P have the 9AJ functions but no RETMA code; 6Zh32P is EF86; 6Zh1P is close to 6AK5 but carries k + g3 + shield on both pins 2 and 7; 6C4P is not a 6X4.
• 5C4S is indirectly heated with its cathode on pin 8, so it takes 5AR4's names. EL91/6AM5 is listed as 7BZ by its source but does not match the 6AQ5, which is 7BZ, so its basing is left as -.
• Weakest rows: 6S19P and 6P27S (one sheet each), 6S45P (Sovtek sheet only), 5C3S and 6C4P (pinout pictures without labels), 6P3S-E and the -EV grades (assumed identical to the base type).
• Left out: 6N30P (lead drawing unlabelled - the existing 6H30P row covers it), 6P43P, 6S41S, 6Zh52P, 6P9, 6Zh4 (not sourced or conflicting); 6S33S/6C33C, GU-50, GU-19, 6D22S and 6P45S sit on bases TubeSocket does not have.

Rules for every row: names as in V5-V8; a part is added only after its basing diagram has been read from a data book or manufacturer sheet, or from two independent sources; equivalents are listed only where the basing is confirmed identical (the 6L6 family is not uniformly 7AC, and Russian "equivalents" such as 6N1P are not pin-compatible with 6DJ8).

5. Tests

• TubeSocketDatasheetTest, after SIL_ICDatasheetTest: every row parses and applies; the name count equals the base's pin count; designations are unique; within a row no two names share a net unless both carry the _n suffix; rows sharing a basing code have identical names; findSeries finds a tube and not a group.
• TubeSocketTest: applyModel sets base, labels and designation and leaves mount and angle; IC and NC pins stay apart in the netlist.

6. Verification

• mvn -q test over the whole reactor (core and library change).
• Done: whole reactor passes (core 227, library 508, swing 29).
• Done: 12AX7, EL84, EL34, 5AR4, 6AQ5, 12K5, 0A2 and 6S19P rendered headlessly from the datasheet; labels sit by their pins and fit once the suffix is dropped (V11).
• Regression suite: no rendering or netlist change is expected for existing projects (V10); the run confirms it.

7. Phases

1. Plan: this document.
2. Code: done - V2, V3, V8 (IC in Node.NON_NET_NAMES, covered by NodeNameTest), V9 (six variants retired 2026-11-02T14:00), TubeSocketTest, TubeSocketDatasheetTest, @AiNotes. The datasheet test fails until the datasheet exists.
3. Library: done - 188 tubes (46 small-signal, 54 radio and space-charge, 58 power, rectifier and reference, 30 Soviet and EL91). catalog_full.json regenerated; release notes in 6.10 (datasheet, and the shorted-variant fix).
4. Pin labels: done - V13 ("Pin Names"), V14 ("Pin Labels"), factory presets re-keyed, TubeSocketTest covers the old-file default; rendered in all three modes.
5. UX4: done - Base.UX4, four pins 90 degrees apart on a 16.3 mm pin circle (Radiomuseum base table; pins 1 and 4 are 4.0 mm, 2 and 3 are 3.2 mm), drawn on a 30 mm body like the octal. Pin 1 follows the convention of the other bases: the gap between the thick filament pins 1 and 4 points right at 'Angle' 0, numbered clockwise in the bottom view. Rows for 2A3 and 45 (RCA, basing 4D), 6A3 (Tung-Sol, 4D) and 300B (JJ; same layout as 4D), named F1,P,G,F4 per V6, and the 5Z3 rectifier (RCA 4C: plate 2 on pin 2, plate 1 on pin 3; Tung-Sol agrees but does not number the plates), F1,2P,1P,F4. The 50 and 80 are also UX4 and left out until a sheet is read.
6. Later, on request: the jumbo 4-pin base (211, 845) in the datasheet, magnoval.