This document describes the machine as it stands after the concept review. Dimensions are the ones used in the concept model and are good to a centimetre; the CAD phase fixes them to the millimetre. The coordinate system is the machine’s: X runs along the bench from the carousel towards the deck, Y runs across the bench (positive towards the front), Z is up. The origin is at the bench top under the flip station.
The bench is about 190 × 100 cm. From left to right it carries the carousel magazine (Ø 86 cm, standing free on the bench), then the frame (121 × 88 × 83 cm over the panels) that spans the flip station and the turntable. The gantry rides on top of the frame; its Z column rises to about 135 cm when fully raised, so the machine wants a metre and a half of headroom.
The frame’s left end is open and its X beams overhang it by 8 cm so that the gantry can reach the carousel’s pick slot, which is the only point where the two halves of the machine meet. The carousel can be lifted off the bench for loading without touching the frame.
The frame is a glued box of 18 mm birch plywood: two side panels at Y = ±43 cm and one end panel at the deck end, each cut as a frame around a large window so the interior stays open and reachable. The panels give the racking stiffness; there are no diagonals and no corner posts. A 2040 V-slot extrusion is bolted along the top edge of each side panel and is the X axis: the carriages run on Delrin V-wheels in the slot, as on a hobby printer. One 2020 tie across the open end keeps the two beams parallel. Brackets for the cue-lever servo, the deck camera and the LED bar screw directly to the panels’ inner faces.
The electronics are not boxed: the controller board and its supply mount on the outside of the rear panel under the X beam, the pump and valve on the same panel near the open end, and the Pi with its USB devices on the outside of the deck-end panel. The flip station’s post stands on a printed foot on the bench.
The gantry is a Cartesian XYZ with one rotary axis:
| Axis | Mechanism | Travel | Notes |
|---|---|---|---|
| X | GT2 belt, NEMA17, Delrin wheels on 2040 V-slot beams | ≈ 80 cm | From the carousel pick slot to the tonearm rest |
| Y | GT2 belt, NEMA17, Delrin wheels on a 2040 V-slot cross beam | ≈ 63 cm | The concept assumed 45 cm; the planner’s re-grip poses reach from −34 cm to +28.5 cm. The 94 cm cross beam gives about 70 cm; the frame is 86 cm wide so the fork clears the rear panel at the re-grip pose |
| Z | T8 lead screw, NEMA17, moving column on an MGN12 rail | ≈ 60 cm | Column slides through a guide on the cross beam; self-locking; the one axis that keeps a linear rail |
| Wrist | NEMA17 with ≈ 5:1 planetary gearbox, Hall home sensor | −90° … 180° | Axis parallel to X |
The Z axis is a moving column rather than a fixed tower: the 76 cm rail is part of the carriage and slides up through a guide block on the cross beam, so when the carriage is raised nothing hangs below it. Every X move happens at travel height and clears the deck and the station by construction. The lead screw is self-locking, so a power loss leaves the record where it is.
The wrist sits on a 9 cm outrigger beside the column so the arm can swing without meeting it. It carries three things: the vacuum arm (22 cm, with the cup on its end pointing perpendicular to the arm), the wrist camera looking along the cup’s axis, mounted on the outer (+X) side of the arm 11 cm up from the cup, and the finger-lift fork on the opposite side of the hub, 5.5 cm long, which points down when the arm points up. The cup bracket does not reach below the cup, so the spindle tip, which enters the hollow bellows cup when a record is released over the spindle, meets nothing.
The gripper is a single 40 mm bellows vacuum cup on the label area, fed by a 12 V diaphragm pump through a solenoid release valve, with a pressure sensor on the line that confirms a seal before any move and aborts a move if the seal is lost. The label area is the one thing every record has in common, so the cup handles 12”, 10” and 7” records and shaped discs alike; 7” singles with the large jukebox hole are out of scope. What the cup can hold and what the carousel can present are not the same set: the magazine locates a record by its round edge, so a shaped picture disc, or one whose outline is not centred on its hole, is placed on the platter by hand and the machine does the rest (see the operating cycle).
The cup’s bellows give about 5 mm of compliance, which is what lets the cup land on a spoke standing slightly off its nominal position. At the pick position the cup stops 8 mm short of the record’s face and makes the last 8 mm as a slow approach; the gap between spoke faces at the arm’s radius is 65 mm and the cup-to-arm-back assembly is 48 mm, so the approach and a 9 mm margin use the rest.
The magazine is a horizontal disc with 24 radial slots at 15° pitch. Records stand vertically in the slots like spokes, each in a 5 mm slot on top of a printed comb. The floor of that slot is a shallow V, sloping 5° each way from an apex 26 cm from the axis, so a record of any diameter rolls to the apex and stands with its centre on the same vertical line: 12”, 10” and 7” records all sit at the same radius, and only the height of the centre differs, by the record’s radius. The pick is therefore the same X and Y for every size, and the machine takes the height from the size in the batch manifest. At that radius the gap between neighbouring spokes is about 6.5 cm face to face, enough for the 5 cm gripper assembly to descend beside a record. A 10” is picked 25 mm lower between the same neighbours, through the same gap. A 7” is picked 62 mm lower, and there the gap runs out: the wrist hub, the outrigger and the foot of the column sit 22 cm above the cup and are wider than the gap at the radius where they descend, so they have to stay above the top edge of the neighbouring spoke on the front side, the side the wrist hangs on. Beside a 12” neighbour a 7” pick puts the hub 18 mm and the column’s foot 7 mm below that edge. So the slot on a 7”’s front side holds a 7”, a 10” or nothing, a rule the manifest editor enforces; a run of 7”s costs one empty slot at its front end. There is no rim: the V does the locating, and a wall tall enough to touch every size at its own centre would stand in the way of a record being carried past at travel height.
The disc is 18 mm plywood cut roughly round and is nothing but a floor. It rolls on ten identical printed brackets, each holding a 608 bearing, arranged on a 72 cm circle, and is located by a 6005 bearing on a stub shaft at the centre. A printed GT2 tooth ring in eight dovetailed segments on the disc’s underside is driven by a NEMA17 with a 20-tooth pulley, about 34:1 at the ring and about 170:1 with the motor’s 5:1 gearbox, and an optical sensor reads a home mark; one slot index is therefore repeatable to a fraction of a millimetre at the rim. The 24 slot combs plug into a printed centre hub, which carries the angular positions, and are screwed to the disc at their outer ends, so the wood never has to be cut precisely.
Each record returns to the slot it came from, so the magazine is both infeed and outfeed, records never touch each other, and pick and return happen at one fixed position: the slot that points along X towards the deck.
A ring rest on a post 12 cm behind the machine’s centreline, at X = 20 cm, Z = 30 cm: a printed ring of 100 mm mean diameter and 6 mm section, so it contacts a record between 44 and 56 mm from its centre — the edge of the label, where the record is stiffest — with a 3 mm rubber O-ring in a groove on top that the record actually rests on. It holds a 7” with 32 mm of overhang, leaves 16 mm between the cup and its inner edge, and its 60° opening towards the front, 44 mm wide at the inner edge, passes the 24 mm arm with 10 mm a side. That the ring touches a 12” just outside the label is accepted: the O-ring is soft, and stability matters more than the last millimetre of dead wax. The station is deliberately off the centreline so that a vertical record travelling along X at travel height passes beside it.
The deck is an Omnitronic DD 3120: a Technics-type direct drive, 45 × 36 × 9 cm, with a manual cue lever, a remote start/stop connector, a pitch output, 33/45/78 rpm with quartz lock and ±10/±20 % pitch range, and no auto-return. Its layout, measured on a top-down photo (images/dd-3120-top.jpg) and to be confirmed with a ruler: the spindle sits 18.4 cm from the left edge and 18.6 cm from the rear edge; the arm pivot is 19.4 cm to the right of and 8.4 cm behind the spindle (pivot-to-spindle 21.1 cm, consistent with a 23 cm effective length); the arm rest is 19 cm right of and 13 cm in front of the spindle; the cue lever is at the front-right of the arm base, 3 cm inside the deck’s right edge and about 5 cm behind the spindle line; start/stop and the speed buttons are at the front left, the power knob at the left, the pitch fader at the right in front of the arm, and the deck’s own red strobe LED shines on the platter rim at the front left. The platter rim carries four rows of strobe dots.
The felt DJ slipmat in the photo should be replaced by a rubber or cork mat for digitising: a slipmat is made to let the record slip, and it also holds static and dust. (A Stanton T.92 USB is also available but has no cue lever, which would bring back a separate arm-lift mechanism; it stays the spare.) The cartridge is an Ortofon DJ S, a spherical DJ stylus tracking at about 3 g, robust against the kind of handling this machine does; an elliptical stylus can be swapped in for a final archive pass once the machine is trusted. The machine touches the deck in four places, none of them modifications:
Start and stop go through the deck’s own remote start/stop connector, a 6.3 mm jack made for the fader-start feature of older mixers and hi-fi systems, which started the deck when the fader came up: it expects a contact closure and does nothing but start and stop the platter (speed is chosen with the buttons). One channel of the Conrad 393905 USB relay card (already on hand, driven from the Pi) on a 6.3 mm plug does it, with no wire soldered inside the deck. The 33 and 45 buttons are momentary switches and are pressed by two more channels of the same card wired across the switch contacts inside the deck; a relay is a dry contact, so polarity and the switch’s voltage do not matter.
The cue lever is worked by an MG996R servo through a pusher rod, on a bracket on the deck-end panel of the frame; the rod comes in along X at lever height; the lever is only 3 cm from the deck’s right edge, so the rod is short and never crosses the arm’s swing. A bar from the same bracket runs in front of the arm base, under the arm, and carries a soft-sleeved end-stop pin standing in the arm tube’s path 45 mm from the pivot, so the arm cannot swing past a stylus radius of roughly 53 mm: the stylus physically cannot reach the label. At 45 mm the pin stands outside a 12” record’s outline; at the 60 mm first drawn it stood under the record’s edge.
The tonearm is moved only while the cue lever holds it up, by the fork straddling the headshell’s finger lift with prongs fore and aft of it. The fork never carries the arm’s weight.
The arm rest clip stays open.
There is no cleaning module. Records are brushed by hand before they go into the carousel; a brush pass takes a minute per record and is the right moment to look each one over anyway. A frame-mounted brush with a dust edge was designed and dropped as a simplification; the decision log has the details, and the rear panel has room for it if it ever comes back.
The wrist camera (a USB webcam, a Logitech C270) sits on the outer side of the wrist arm, 11 cm up from the cup, looking along the cup axis. With the cup pointing down it looks straight down, and that is where it works: hovering over the platter after a record is placed it photographs the label that is facing up, finds the centre (the spindle is in the hole) and measures the lead-in and run-out radii; over the ring rest it photographs the other label before the re-grip. Both label photographs come from it, each taken face-on with nothing in the way. At the carousel it does not look at the record: inside the 65 mm gap between spokes there is no room for a camera at a usable distance, and none is needed, because a record that has rolled to the apex of its comb’s V has its centre within a millimetre of a known position, which the cup’s bellows and the spindle tip absorb.
The deck camera (Pi Camera Module 3) is mounted low on the deck-end panel at platter height, facing the headshell across the platter from 20 to 40 cm. It sees the stylus in profile, the groove bands edge-on and the strobe dots on the platter rim. It tracks the headshell to locate the arm wherever a side ended, confirms the stylus landed on the lead-in, watches for the run-out and for skips, and reads the platter speed.
An LED bar beside the deck camera has two modes: pulsed at 50.000 Hz from the controller’s crystal it is a strobe that makes the rim dots (or a printed strobe ring on a deck without dots) show the speed error as a drift; run continuously it is a low-angle raking light that makes the grooves and the stylus contact visible.
Printed parts are PETG for structure and TPU wherever a record is touched: the cup lip if not bought, the fork lining, the end-stop sleeve; the ring rest carries a bought rubber O-ring instead. Bought parts are everything long, stiff or precise: extrusion, the Z rail, screws, bearings, motors, electronics; a used 3D printer is the intended source for most of the motion parts. Plywood is used for the frame panels, the bench top, the carousel disc and its base plate, and never for anything whose position matters.