Sega CD and 32X Repair Reference

This page covers the two add-ons that bolt onto a Sega Genesis: the Sega CD / Mega-CD, in both the Model 1 tray-loading and Model 2 top-loading forms, and the 32X. The console underneath them lives on my Genesis and Mega Drive reference. Its board revisions, its own power trap, its video chain, its recap lists and the general bench method are all documented there. Read that page first if the fault might be the host, because a surprising share of “my Sega CD is dead” and “my 32X locks up” turn out to be exactly that.

The organising idea for this page is one sentence: these are three separate computers that only look like one machine. A full tower is a Genesis, plus a Sega CD carrying its own 68000, its own RAM, its own BIOS and a complete CD servo front end, plus a 32X carrying two SH-2 cores and its own video encoder. Each one has its own power input. Each one boots on its own terms. Almost every diagnosis on this platform starts with working out which of the three machines is actually broken, and the good news is that the stack is unusually cooperative about telling you.

One thing up front. Unlike the Game Boy and Game Gear pages, this one is written from documentation, not from a stack of my own repairs. What sits behind it is four Sega OEM service and maintenance manuals (the Model 1 export maintenance manual, the Model 2 service manual with its A3 foldout schematics and its Funai-versus-Sony supplement, and the 32X service manual with its parts specifications), three Sega developer hardware manuals, the 32X factory service bulletins, and the usual technical wikis. Where a claim is somebody else’s prescriptive judgement I attribute it, and where a number rests on a single source I say so rather than rounding it off. There is no bench confidence tag anywhere in the data below, and that is deliberate.

Safety first: three power inputs, two opposite polarities

Read this before you plug anything in. A full Sega CD plus 32X plus Genesis tower wants three separate wall adaptors, and they are not all the same polarity. There is no input bridge rectifier on any of these units, so getting it wrong does not simply fail to run: it drives the rail backwards into the board.

Three things to burn in.

Both Sega CD models are centre-negative, and they take the Genesis Model 1 supply. This one runs against a widely repeated belief, so here is the evidence rather than an assertion. The Console Power Supply Bible files both MegaCD 1 and MegaCD 2 under the centre-negative Mega Drive 1 adaptor. ConsoleMods states that both models take the same tip-negative 9 V adaptor as a Model 1 Genesis. The SEGA Hardware Database records the North American MK-4102A Model 2’s own rating label as “DC 9 V 1.2 A, minus inside” and names the Genesis 1602-family adaptor by part number. And the OEM Model 2 service manual settles the related myth that a Model 2 “draws its power from the Genesis”: the rear panel has its own AC Adaptor Jack, the accessories page lists its own bundled adaptor, and the specification page gives the unit its own 18 W wall input. What a Sega CD takes from the console is a power-enable signal, not operating current.

The 32X is the opposite polarity on a different barrel. Centre-positive, EIAJ-03, in the Genesis Model 2 family. The two barrel sizes are different, and that mechanical difference is doing real safety work in a stack that has three bricks lying behind it. It stops protecting you the moment an adaptor or a forced plug enters the picture, and the 5.5 x 2.1 mm barrel the Sega CD uses is physically identical to an NES jack, where nearly every generic supply of that size is centre-positive.

Do not reason from the shape of the machine. The CDX and Multi-Mega combos play CDs and are centre-positive. “It’s a Sega CD, so it must be centre-negative” gets a combo unit killed.

So the bench rule, same as the one on the Genesis page: label your supplies by polarity, not by plug, and meter any unknown adaptor for voltage and polarity before the first power-on, every time. After a suspected reverse-polarity event, check the fuse and the input protection before recommissioning anything.

What my USB-C adaptors will and will not power here

I build USB-C power adaptors, so it is worth being exact about which parts of this stack they cover, because the wrong answer here damages hardware.

Both the standard USB-C 9 V adaptor (an onboard boost converter, so it runs off any 5 V USB port with no PD negotiation) and the cheaper Lite version (a PD trigger, so it needs a real 9 V-capable USB-C PD source) output 9 V on a centre-positive EIAJ-03 barrel. That is the Genesis Model 2 family. So:

• The 32X: yes. Its OEM rating is 10 V at 850 mA centre-positive on exactly that barrel, the PAL-G and PAL-I units are specified at 9 V outright, and 850 mA sits inside the standard adaptor’s 9 W budget. This is also the right adaptor for the Genesis Model 2 underneath it, which is the other centre-positive unit in a typical tower.
• The Sega CD: no, not either model, on my reading of the sources. Every source I trust on this platform puts both Sega CD models on the centre-negative 5.5 x 2.1 mm barrel, which is the opposite polarity and a barrel these adaptors do not fit. Separately from polarity, the North American Model 2’s own rating label calls for 1.2 A, which is more current than a 9 W adaptor has to give. I want to be straight that my product listings currently name a Sega CD Model 2 in their compatibility list and I have not been able to reconcile that with the research above, so I am flagging the conflict rather than papering over it. Meter the jack on your own unit before you connect anything, and if it reads centre-negative, do not use a centre-positive adaptor on it.
• The Genesis Model 1: no. Centre-negative, wrong barrel, and the listings say so already.

If your unit’s jack does not match, that is what a correct adaptor is for, not an adapter plug. I would rather lose a sale than replace somebody’s mainboard.

Know your unit first: which Sega CD, which 32X

Three machines stack up here (a Genesis underneath, a Sega CD, a 32X), and two of them ship as a “Model 1” and a “Model 2”. Everything in this section is about the add-ons, so read Model 1 and Model 2 below as Sega CD Model 1 and Model 2 unless a line says Genesis. Within the Sega CD, two different manufacturers laid the boards out independently, and a reference designator on one of them means nothing on another. Identify what you are holding before you order a single part.

Sega CD Model 1 versus Sega CD Model 2, from across the room

• Model 1 is the front-loading motorised tray that sits under a Genesis Model 1. No controls on the add-on itself: the tray opens from the console’s system menu or a reset. Front bezel carries a green “Ready” LED and a red “Access” LED. Japan got it December 1991, North America late 1992, PAL and Korea in spring 1993.
• Model 2 is the cheap top-loading lid on a thin base that the Genesis sits on top of, with the drive off to the side. Single access LED: the green “Ready” LED was dropped, and that is the fastest tell. Rear connectors are identical to the Model 1. Japan April 1993, North America November 1993.

Model numbers, because listings get these wrong constantly: the Model 2 is MK-4102 (Funai board) or MK-4102A (Sega board) in North America, MK-4102-50 in Europe, HAA-2912 / HAA-2912A in Japan. There is no MK-4103: that belief traces to MK-2103, which is the Genesis 2 adaptor part number. The Model 1 is HAA-2910 in Japan and the 1690 family in Europe; the North American number is genuinely poorly documented, with one wiki listing MK-1601, which collides with the Genesis Model 1 number. Do not stamp a North American Model 1 number into a listing without reading the unit’s own bottom label.

The six Sega CD Model 2 boards

Five Funai-designed boards and one Sega-designed board, and the split decides the cap kit, the laser, and BIOS compatibility.

The one-glance triage: a Funai board integrates the drive controller and the activity LED onto the motherboard. The Sega board (171-6528C) puts the drive controller on a separate PCB under the laser and the LED on its own little board, and the motherboard is smaller. The Sega board is what earns a model number the A suffix.

• BE2200G02A01: Funai, Japan only, earliest. The only Model 2 carrying the old export-Model-1 ASIC (315-5548). Spotted by a small bodge PCB glued on top of the PCM chip and its RAM.
• BE2200G02A02: Funai, first North American board. Moves to the 315-5632 (MCE3) and still carries the bodge PCB.
• BE2200G02B02: Funai, first PAL board. An A02 with minor layout changes; bodge PCB present.
• BE2200G02C02: Funai, serial T30282001 and later. The bodge contents are folded into the board, so no glued daughter.
• 171-6528C: the Sega board. Smaller, split drive controller, and it uses a “universal” BIOS. An earlier Funai-board BIOS is not compatible with it, which matters the moment you transplant a board or fit a region-free image.
• BE2200G02C22: Funai, final revision. The audio DACs and op-amps merge into a single 315-5619; drive circuitry and LED integrated.

Lasers key off the board, not the model. Funai boards take a Sony KSS-210A or a Samsung SOH-O4T; the Sega 171-6528C takes a JVC Optima-6S or a Sony KSS-240A, and there the spindle colour tells you which: silver is Sony, black is JVC. On the Model 1 it keys off the drive vendor instead: Japanese and most North American units carry a Sony transport with a KSS-240A, most PAL units carry a JVC transport with an Optima-5S.

32X VA0 versus VA1

Silkscreened on the board, and the OEM manual carries separate circuit-board diagrams for each. What actually differs is compatibility, not repair:

• VA0 (MK-84000) shipped with the Extension Unit, a terminator that plugs into the Genesis side-expansion slot and pulls down data bits A4 to A11 and A13 to A20. On a VA0 Genesis Model 2, a VA0 32X will refuse to run many games or lock up within minutes unless something is in that slot: an attached Sega CD does the job just as well as the Extension Unit.
• VA1 (MK-84000A) built those pull-downs onto the 32X board, so the Extension Unit is no longer needed. Its cartridge slot is also slightly larger and accepts Mega Drive carts, though that last detail comes from a single reader tip and I would treat it as a lead rather than a spec.
• Neither revision fixes 32X behaviour on the VA2.3 or VA4 Genesis Model 2 boards. There is no documented fix there beyond loading from a flash cart.

Common problems and fixes

Ordered roughly by how often each lands on a mixed intake, per the sources. Model 1 dominates because it is the oldest, the most capacitor-plagued and the only one with a real mechanism; the Model 2’s headline fault is a ten-cent fuse; the 32X’s headline fault is cosmetic video noise rather than death.

Start here on any no-boot Sega CD: run the boot ladder

This is the highest-value thing on the whole page, and it costs nothing. The Sega CD boot is a progressive self-test, so how far it gets localises the fault. Watch the bezel LEDs while you do it: the Model 1 has the green “Ready” plus the red “Access”, the Model 2 has the access LED only.

Three shortcuts that fall out of it. Black screen with the LEDs blinking is not a power fault at all: that is the bus translator. Stuck at the first stage is worth attacking with software before you pull chips: the 240p Test Suite has a Sega CD section that exercises the BIOS CRC, Program RAM and Word RAM in situ, and it runs from a flash cart in the attached Genesis. My Genesis flash cart will boot it for that purpose. To be clear about what it is and is not: that cart runs the Genesis and Master System libraries and cannot itself run Sega CD or 32X software: its value here is as a known-good Genesis test cart and a way to get a diagnostic ROM in front of the stack. And reaches the logo but spins forever is a laser job, not a logic job, so start with the lens.

One more thing not to chase: a delayed boot that hangs at the first stage for several seconds and then springs to life on its own is a specific Model 1 fault, covered below. Every other first-stage hang hangs forever.

Leaking electrolytics on a Model 1 (the marquee job)

Any or all of: no boot, stuck at the first stage, distorted or absent CD audio, a disc that will not read, visible green crust, lifted cans. The classic framing is a unit that “worked last year” and now does not.

Both the surface-mount and through-hole aluminium electrolytics leak, and the electrolyte corrodes the pads, vias and fine traces underneath and between the caps. ConsoleMods is blunt that the Model 1 main-board capacitors must be replaced even with no visible leakage: that is their prescriptive call, not mine, but every other source I have agrees on the underlying failure mode.

The procedure that actually matters is the cleanup, not the soldering. Shell it, inspect both the main and sub boards under magnification for crust and discoloration around the cap skirts, and clean the electrolyte off before anything else: leaked electrolyte and old flux alone can cause a no-boot, so cleaning sometimes is the repair. I handle the neutralise-and-rinse step the same way I describe in my general restoration and testing writeup: if you use distilled vinegar to neutralise, wash it off afterwards with isopropyl, because vinegar left on a board is itself corrosive.

Then bell out the traces in the corroded zone. A cap that “came off with almost no force” means the joint or pad was already eaten. One recap writeup flags the traces at the upper-left corner of the Mega CD Engine ASIC as the classic spot where a recap silently lifts a run, because they pass right beside the leaking caps: that location comes from a single user comment and nothing corroborates it, so check the corner but confirm the designator against the manual before you go probing a specific pin.

Two things that get skipped and then bring the unit back: the drive board and the laser assembly’s own filter cap. A Sony transport has a 100 uF cap sitting on the laser assembly itself that is not in any vendor kit, and JVC-drive units add leaking electrolytics on the drive controller board that Sony-drive units do not have. “Recapped but still will not read” is usually one of those two.

And read the cans before you order. On the Model 1 the same board position ships as either a leaded radial or a surface-mount can depending on the production run: Console5 sells a distinct leaded kit and SMD kit for the identical board, which tells you how real the split is.

Model 1 tray will not open or close

The rubber loading belt has hardened, stretched or gone gummy. Watch the loading motor pulley: if it spins while the tray stalls or tracks weakly, the belt is slipping. A shiny, cracked or sticky belt confirms it. This is the number one Model 1 mechanical fault and it is a five-minute job with the right part.

Sizing disagrees between sources (one guide specs a square 25 mm by 1 mm belt and retail replacements list 24 mm by 1.2 mm), so measure the old belt, or buy the matched Console5 Sega CD / Neo Geo CD tray belt SKU rather than a generic assortment.

Model 1 “phantom tray”: opens then immediately recalls

Distinct from the belt, and the intermittency is the tell. The mechanical limiter switch that tells the logic the tray has reached its end-stop is oxidised or out of adjustment, so the open position never registers and the mechanism keeps trying to home. Continuity-test the switch through its travel and exercise the mechanism by hand, watching whether it actuates cleanly at the stop. Clean the contacts, re-seat or adjust the actuator, and only replace the microswitch if the contacts are pitted.

Spins but will not read

Reaches the logo, the disc spins forever, no “Press Start”. Or it reads pressed discs but not CD-Rs, which is the signature of a marginal laser near end of life.

Work it in this order, and do not skip steps.

1. Clean the objective lens with isopropyl on a swab.
2. Confirm the diode actually emits. A phone camera sees it glow red during a read attempt with no disc in. This one check separates a dead laser from a servo problem and costs nothing.
3. If it emits and the lens is clean, it is a servo tune: focus, tracking, RF. That is a scope-and-test-disc job, and the authority is the OEM CD Drive Adjustment procedure with the TOC Read-in Flowchart alongside it. Do not turn pots blind. The Funai Model 2 pot map that circulates (VR401 for E/F balance, VR402 tracking gain, VR403 for the VCO/PLL, with an 800 Hz / 0 dB tracking-gain target on the SOH-O4T) comes from exactly one source and the pots differ by drive, so confirm against the OEM procedure before you move anything.
4. Replace the pickup only if tuning cannot recover it. Age alone is not the replace signal; failure to calibrate is. One guide gives a numeric threshold for the Model 1 (swap the KSS-240A if RF falls below about 35 mV after cleaning and adjustment), and that figure is single-sourced, so treat it as a guide.

Before all of that on an intermittent, try a fresh CD ribbon. The flat flex between the drive and the main board is a common intermittent and it is the cheapest test on the list.

Model 1 Sony drive: the sled will not traverse

A different no-read, and worth separating because the fix is mechanical. The Sony mechanism’s helix or “Archimedes” worm gear wears, or the drive gear strips, so the pickup cannot be driven along its rail: the diode is fine but nothing can move it. It may read the very start of a disc and then fail, or never focus at all.

Power off and gently walk the sled along its rail by turning the feed gear, feeling for stripped teeth, a worn worm or binding. There is no distributor part for this, so it is a clean-and-re-grease job or a transplant from a donor drive. Old grease on the rails also goes gummy and stalls the sled; clean it off and re-lube with a plastic-safe synthetic, never anything conductive or graphite anywhere near the optics.

Black screen, no LEDs at all on the Sega CD

The unit has its own adaptor plugged in, the Genesis is powered, and there is nothing: no LEDs, no drive activity. This is the classic side connector or power-enable signature, and it looks far worse than it is.

The Sega CD does not self-power even with its own 9 V present. The +5 V that the Genesis presents on side-connector pins A26 and B26 is what enables the CD’s own 5 V regulator. So with the console on, walk the chain: 9 V at the DC jack, through the noise choke and the fuse, to the pass transistor’s emitter; the Genesis +5 V arriving across the power ribbon into a divider that turns on a second transistor; that transistor sinking the pass transistor’s base; and finally +5.0 V out of the regulator and back up the ribbon to the rest of the unit. The first discontinuity you find is the fault.

The topology is solid, and the +5 V identity of A26 and B26 is independently confirmed by the GameSX Genesis expansion-port pinout. The component designators and the divider values are single-sourced, so use them to orient yourself and confirm against the OEM power schematic before you condemn a part by its reference number.

Fixes, in order of likelihood: clean and re-tension the side connector and reflow its joints; replace a blown fuse; if a noise choke has shorted the 9 V to ground, remove it and jumper with solid wire; replace the regulator if it has 9 V in and nothing out.

Completely dead Model 2: check the fuse first

The single most common Model 2 fault is a blown input fuse. On the Funai boards it is F301, and it takes anything from 1 to 2 A of any type. It blows from a wrong-polarity adaptor, or from attaching and removing the Genesis while the CD is still powered.

Continuity-check it before anything else on a dead Model 2. Then work out why it blew (confirm the adaptor is the correct centre-negative unit) or the replacement blows too. This is a ten-cent part that restores the unit, and it is worth checking on every dead Model 2 that comes in regardless of what else you suspect.

A note on Model 2 capacitors, because the sources genuinely disagree and I would rather show you the disagreement than pick a side. ConsoleMods says the Model 2 does not suffer the Model 1 leaky-cap plague. RetroSix’s Model 2 cap tables document a mostly-surface-mount electrolytic population, which is exactly the type that does leak. My reading is lower-urgency but not zero: inspect the surface-mount positions on any Model 2 with audio or read faults, and treat the four large through-hole power caps (the 1000 uF pair, the 470 uF and the 330 uF positions) as the ones that actually vent and corrode near the regulator.

Delayed boot that recovers on its own (Model 1)

A very specific fault, easily misdiagnosed, and worth knowing because it looks expensive and is not. The unit hangs at the first stage (blank BIOS background, no logo, no music) for several seconds, and then springs to life on its own and runs perfectly. Power cycling behaves the same way.

The recovery after a delay is the entire tell. Every other first-stage hang (bus translator, drive, ribbon, Program RAM, BIOS, 68000) hangs forever. The attributed cause is a drifted MB3790 power supervisor next to the save battery, still asserting reset for seconds after the 5 V rail is already stable and holding the CPU and the Mega CD Engine off until it finally releases. It is a motherboard fault, not a power-board fault, which is what misdirects people into swapping power boards.

Two flags on this one. The whole symptom-to-cause-to-fix chain comes from a single writeup: the part’s reset-generator role is independently corroborated, but nothing else confirms that replacing it cures this. And the reference designator is unreliable: that writeup calls it IC9, while Console5 lists the MB3790 as IC6 on the Model 1 and IC114 on the Model 2. Read the silkscreen before you desolder anything.

Audio that stutters about twice a second

Boots and plays, but the audio cuts out roughly every half-second. The symptom is diagnostic on its own: aged aluminium electrolytics on the CD power board, commonly the 100 uF positions C25, C26 and C27. ESR-check them or just replace the power-board electrolytics as a set, and on a Model 1 that is in for service anyway, do it as part of the main recap.

Intermittent snow over the boot screen

Snow or static that rolls over the picture after the Sega CD boots, coming and going at random, with an otherwise clean image. This is a power-delivery problem and it is usually upstream on the Genesis side: a weak or aged supply that cannot hold rated current, a dirty or worn DC jack, or dry joints on the side and power connectors starving the CD’s supply.

It is emphatically not a CD video fault, because the Sega CD has no video encoder at all. Try a known-good supply first, since that rules the source in or out in seconds, then clean and reflow the DC jack and the connector joints.

No CD sound through the TV, but the CD’s own RCA jacks work

A useful split, and the answer is not in the CD’s audio circuit. The CD’s mix returns to the Genesis over the expansion-connector audio-in pins SL2 and SR2 (B29 and A29), where it joins the Genesis FM and PSG and leaves through the normal Genesis A/V. That is an internal path across the side connector. So this symptom points at those pins and their joints.

While we are here: the rear 3.5 mm jack on a Sega CD is an INPUT, not an output. The OEM manual labels it the Mixing Jack, fed by a separate Stereo Mixing Cable accessory, and its job is to carry a Model 1 Genesis’s headphone audio into the Sega CD so the CD’s RCA line-out can put out a combined mix. It does not carry CD audio out and it does not connect to the expansion connector. That gets mixed up with the internal SL2/SR2 path constantly.

32X video noise and horizontal sync instability

The 32X’s famous flaw, and it is a design consequence rather than a fault. There is no 7805 in a 32X. The 5 V rail comes from a switching buck converter (IC13, an IR3M03A or an MC34063A depending on production), and that switcher’s output oscillation and ripple inject straight into the video path. Console5 notes that Sharp-branded parts are anecdotally worse than the Motorola/ON variant, which is an anecdote and I would treat it as one.

Confirm the noise is actually on the 32X output (Genesis-only video through the same chain should be clean), and scope the 5 V rail for ripple if you are instrumented for it. The canonical cure is the kevtris cap swap: bulk up CE17 from 470 uF to 2200 uF / 10 V low-ESR, and change C30 on the vertical board from its small ceramic to a 100 uF / 6.3 V ceramic. Two practical warnings. Height on CE17 matters: anything over about 14 mm fouls the shell. And the designator is contested: RetroRGB and the schematic call it C17, the silkscreen and Console5 call it CE17. Identify it on the actual board before you desolder.

For stubborn cases the escalation is the regulator itself, or the big filter cap in the original Sega brick if you are still using one.

32X coil whine

An audible whine from the 32X that varies with load, often at its worst idling in a menu or on a plain Genesis game. Same root cause as the video noise: the buck converter drops into discontinuous conduction and the L7 inductor vibrates. It can also be nothing more than a dry joint on the coil legs, which is why reflowing and inspecting L7 comes first: a dry joint alone causes whine with no other symptom.

Past that the published fixes escalate: a timing and feed-forward tweak (cheap, but the inductor still whines at light load), a NCP3063 / NCV3063 regulator retrofit (the best DIY result, reported silent at any load), or a drop-in replacement power board. The component values for all of those are single-sourced from ConsoleMods, so verify against the board before ordering. Because it shares a root cause with the video noise, a proper regulator job cures both.

Dead or won’t-boot 32X

Almost always mechanical. Inside, the main and sub boards are joined by two 40-pin flat cables and a 64-pin connector, and a lifted flex is a dead 32X. Outside, the cartridge contacts get dirty and the through-cart connector and the A/V ports develop cold solder joints, because the 32X runs hot in a tightly packed shell and thermal-cycles every session.

So: remove, clean and reseat all three internal interconnects; clean the cartridge connection thoroughly; inspect the passthrough connector’s solder under magnification and reflow what looks tired. You can isolate a cart-port fault by running Sega CD or 32X-exclusive software, which does not depend on the same path.

32X locks up or games won’t run, and it isn’t the 32X

Two distinct patterns, both diagnosed on the Genesis, not on the add-on.

Jitter then lockup after roughly an hour of warm-up, on Asian and PAL Genesis Model 1 VA4 boards. The Genesis EDCLK (external dot clock) and VCLK signals on those boards are not clean enough for the 32X and drift as the board warms. Sega’s own service bulletins fix it on the host motherboard: Bulletin No.008 of December 1994 specifies substituting C92 from 51 pF to 33 pF next to IC4, shorting or removing FB7, removing C112, and jumpering C10 (which is IC4 pin 31) to ground with wire heavier than AWG28. The community transcription adds removing C78 and C100, and for VA5, VA6, VA6.5 and VA6.8 gives a different fix: remove C78 and jumper its top pad to 68000 pin 15. Test for at least an hour warm: this fault is thermal, and a ten-minute test proves nothing.

Refuses to run or locks up within minutes on a VA0 Genesis Model 2. That is the expansion-terminator issue covered above: attach a Sega CD, or use the Extension Unit, or fit a VA1 32X. Note the diagnostic question that settles it in one move: does it stabilise when a Sega CD is attached?

Which Genesis board you have is documented on the Genesis page, and this is one of the cases where identifying it is the whole job.

32X and Genesis destroyed after a cable or a mod

Not a wear fault at all, but it belongs here because it is the most expensive mistake available on a 32X. Any A/V patch cable that plugs INTO the 32X must have the +5 V pin unconnected.

Sega deliberately kept the Genesis cartridge-slot +5 V and the 32X’s own internally generated +5 V separate, so the 32X rail can only come up once the Genesis rail is present. The 32X’s A/V IN jack routes its +5 V pin (pin 2 of CN6) straight to that internal VCC net with no isolation diode. A cable that carries the Genesis +5 V down that pin back-feeds the 32X’s regulator from the wrong side and can destroy the 32X, the Genesis, or both. This is confirmed on the OEM schematic and independently by RetroRGB, and note that the ConsoleMods page on the same subject is a verbatim copy of the RetroRGB text, so it is not a third source.

The exception, which is where people get confused: a female-input adaptor that does not itself plug into the 32X may carry all pins. It is the 32X-side connector that must not carry +5 V.

“Won’t play imports”: not a hardware fault

There is no CIC-style continuous lockout chip anywhere on the Sega CD. Region control is software, inside the BIOS ROM: the BIOS checks the disc’s region and refuses out-of-region discs, and the Genesis-side TMSS still gates the whole stack. So region-free is a BIOS swap, a mod rather than a repair. Very early Model 1 units with serial numbers starting A10 socket the BIOS EPROM, which makes that swap solderless. The 32X’s own region check is a Genesis-side cartridge-bus handshake, not a lockout chip of its own.

Inside the stack: what differs from a bare Genesis

A short tour of what the add-ons actually do, which is what makes the faults above make sense. Anything shared with the console is on the Genesis page.

• The Sega CD is a second computer, not a peripheral. Its own MC68HC000 at 12.5 MHz, its own 512 KB Program RAM, its own BIOS, a custom “Mega CD Engine” ASIC, a Ricoh RF5C164 PCM chip and a complete CD servo and decoder front end. The CD 68000 runs the game code; the Genesis 68000 handles video timing and controller I/O. They rendezvous through Word RAM: 2 Mbit to one CPU at a time in 2M mode, or two swappable 1 Mbit banks in 1M mode. That shared window is exactly why bad Word RAM gives you “boots all the way but never loads the game” instead of a black screen.
• The Sega CD has its own clock domain. A ~50 MHz crystal feeds the Mega CD Engine, which divides it down. The OEM Model 2 manual’s pin table for the 315-5632 labels pin 84 ICK50 and pin 1 OCK25, which is what pins the architecture down: 50 MHz in, a 25 MHz derived clock out, and the 68000 at 50 divided by 4 = 12.5 MHz. Practical consequence: a dead crystal takes out the CPU and the engine at once, so on a CD that has a healthy 5 V rail and never draws its BIOS background, scope the crystal early.
• The Sega CD has no video encoder. None. It produces graphics data and an overlay plane that the Genesis VDP rasterises, and all final video leaves the Genesis A/V port. So every Sega CD “video problem” is diagnosed on the Genesis video chain. This is the single most useful architectural fact on the page.
• The 32X does have a real video path, and it genlocks. Its own VDP and DAC (315-5781) feed its own RGB encoder (315-5788) out to its own 9-pin mini-DIN, mechanically the same connector as a Genesis Model 2. The Genesis picture arrives over the link cable, gets combined with the 32X layer, and the merged image leaves the 32X. The 32X locks its encoder phase to the Genesis composite colour-subcarrier reference carried on that link, which is why an under-wired cable gives you phase misalignment and dot crawl even when it damages nothing.
• The two audio paths are entirely separate. The Sega CD mixes 8-channel stereo PCM from the RF5C164 with Red Book CD-DA off the disc, through a 16-bit DAC with an 8x oversampling filter, and presents that mix in two places: its own rear RCA line-out, and back into the Genesis over the expansion audio-in pins. The 32X generates PWM stereo through a completely different chain and mixes into the Genesis over the A/V link. A full tower is therefore mixing three sources through a chain of cables, which is worth remembering when one of them goes quiet.
• The 32X passes the cartridge bus through. It sits in the Genesis cart slot, taps the 68000 bus for its own SH-2 pair, and passes the bus to a game-cart slot on top. That shared slot +5 V is also part of its power sequencing, which is the whole reason the A/V +5 V trap exists.
• The optical front end is a self-contained CD-player chipset, and it differs by drive vendor. Disc to data runs pickup, RF amplifier, CD DSP, CD-ROM decoder, CD 68000, with a servo processor closing the focus, tracking, sled and spindle loops off the DSP’s error signals through a bridge driver. Stated once because it sends people to the wrong chip: LC8951 / LC89515 is the CD-ROM decoder and 315-5476 / RF5C164 is the PCM sound source. One widely-read page gets that backwards. Do not chase a sound fault on the decoder.

Mods worth knowing

I do not reproduce anyone’s install guide here. This is an orientation to what is worth doing and the traps that cost real money on a resale build. Mods that are really Genesis mods (the host RGB bypass, the host region switch, TMSS defeat) live on the Genesis page.

The preventive recap is the baseline, not an enhancement. On a Sega CD it is the expected condition of a sold unit, because a leaking board that works today fails on the buyer’s bench and generates a return. The 32X gets sold un-recapped constantly because it “looks fine”. Recap it anyway; its board is all through-hole and the job is quick.

The 32X video-noise cap swap is the best value on this platform. Two caps, low risk, and it cures the 32X’s single best-known flaw. Do it on every 32X you resell. The coil-whine fix is a step further and silences an annoyance rather than a fault; the timing tweak is a quick add-on and the regulator retrofit is for premium listings.

Region-free BIOS and MultiBIOS. All Sega CD games are region-locked and most units only run same-region discs, so a hacked BIOS is a real value-add for import buyers. A MultiBIOS goes further, carrying several region images bank-selected by two address lines: pins 38 and 39 of the 40-pin BIOS, which RetroSix and ConsolesUnleashed name independently. Wire pin 38 to the host’s video-rate line and pin 39 to its language line and the whole thing becomes switchless. Prefer a socketed install so the buyer can revert, and because a socket protects the desolder work you already risked. One caution on the published switchless passthrough wiring: it routes the host’s region signals across interconnect pins B25 and B27, described as otherwise unused, but the expansion pinout I have on the Genesis page assigns B25 to VA18 and B27 to SL3. That is a real conflict between sources and both are single-sourced. Buzz those pins out on your own board before you wire anything to them.

For the surface-mount combo units (CDX, Multi-Mega, X’Eye, Wondermega) the BIOS is an SOP-40 with no modern DIP equivalent, so it is hot-air removal plus an adapter board. There is an open-source KiCad adapter (0xca01/scd-bios-adapter) for a modern flash chip, single BIOS with no banking, and commercial MultiBIOS boards exist too.

FRAM battery-free saves. Replaces the battery-backed 8K SRAM with a pin-compatible FRAM module so saves persist with no coin cell, which kills both the dead-battery failure and the leak risk of an aged one. It drops into the same 28-pin footprint except at pin 1 and pin 26, which get lifted and either tied off for a fixed bank or switched for four banks. The Model 2 differs only in the format procedure: unlike the Model 1 it will not format internal RAM with no disc present, so boot to the CD player menu, insert a game to unlock the Memory options, and format there. “No battery to die” is a genuine longevity selling point and it pairs naturally with the recap.

The 32X 50/60 Hz switch. Because the 32X has its own video output, its field rate is set independently of the host. It is one pin: pin 32 of the 315-5788 encoder, high for 60 Hz and low for 50 Hz, set by whichever of R42 and R43 the factory populated. Desolder the populated resistor and wire an SPDT between the two pads with the wiper on pin 32. Worth knowing that the OEM manual corrects the community here: those are 1 kohm chip resistors, not zero-ohm jumpers, even though the board does use true zero-ohm jumpers elsewhere. It makes no difference to the mod. This only means anything paired with a host region mod, since the game’s region is still gated by the console.

Optical drive emulators, stated accurately. The Terraonion Mega SD is a cartridge-slot FPGA re-implementation of the Sega CD, not a drive replacement: it runs from the Genesis cart slot and the physical Sega CD unit is not required at all. That makes it genuinely useful on the bench, because it boots Sega CD software with zero optical risk and isolates “is the fault the drive or the mainboard?” in seconds. It also lets you sell a recapped Genesis plus Mega SD to a buyer who wants the CD library without a laser to maintain. Two corrections to the marketing that follows it around: the handful of games that use the Sega CD and the 32X together do not work on it, because the attached 32X blocks the CD data path in the cart slot; and it is not the only Sega CD ODE: the Phode ODE is a genuine optical-drive replacement supporting Sega CD Models 1 and 2, the CDX and the X’Eye. Fenrir, MODE and Rhea are Saturn-only and are correctly excluded, but that does not make the Mega SD the only option.

The A/V link cable is a build item, not a mod, and it is the damage-critical one. Covered in full above. Ship a 32X with a correct 5 V-safe cable and say so in the listing; it pre-empts the buyer’s most common complaint and a possible board-damage return.

Getting a 32X onto a modern TV. Take video off the 32X, not off the Genesis: the merged picture only exists at the 32X output. From there my SCART to HDMI converter handles the RGB path with a Genesis Model 2 SCART cable, and my composite AV to HDMI converter handles the composite path if you are not running an RGB chain. Both are the general-purpose converters I put in bundles rather than low-latency competition boxes, and I say so on their own pages.

Recap and parts

The one rule to carry away: identify the exact board before you order, and on a Model 1, read the actual capacitors first. Nothing on this platform carries across between the Model 1, the six Model 2 boards and the 32X.

A few practical notes on top of the tables below.

• Leaded versus surface-mount is a production-run question on the Model 1. The same position ships either way, which is why the vendor sells two different kits for one board. Eyeball the cans and order the kit that matches what is in front of you. The Model 2 is predominantly surface-mount with a few large through-hole power caps; the 32X main board is all through-hole radial, which makes it the friendliest of the three.
• Buy the revision-matched kit for one-off work. Right count, right values, name-brand parts, one order, and the vendor has already solved the leaded / SMD and Sony / JVC / Funai splits. For a-la-carte, price DigiKey and Mouser first; after the current US import tariffs, LCSC usually loses on landed cost for small distributor-stocked parts and only wins on bulk reels.
• Uprate voltage freely, never downrate. A 25 V or 50 V part covers every 6.3, 10 and 16 V position on these boards, which is the easy way to consolidate stock across all of them. Two exceptions worth respecting: CE17 on the 32X must be a low-ESR part because it is the switching rail’s bulk, and CE18 is a factory OS-CON solid polymer that does not dry out like a wet electrolytic and usually should be left alone: swapping it for a wet cap is a downgrade.
• Match the pickup to the transport, not to the model. The KSS-240A is the safest order because it is a common Sony part shared with many CD players, but reproduction quality varies badly, so buy from a reputable retro supplier and expect to calibrate. The KSS-210A and the Samsung SOH-O4T are scarcer, and on a Funai board they are not drop-in interchangeable with each other: C412 is populated to match the fitted pickup and the servo tune differs.
• There is no 7805 in a 32X. Its 5 V is a switching buck converter, so do not go looking for a linear regulator to check. The Sega CD’s +5 V is linear (an NEC uPC2405, a 78M05-class part) and it sits right where the leaking caps are, so it is a common secondary casualty. Verify the pinout of any substitute against the board before fitting it.
• Confirm the backup cell’s chemistry before you fit one. Some units use a rechargeable and some a primary; a primary cell on a charging circuit vents.

Bundle a resale-grade Sega CD refurb as: the correct cap kit, a trace inspection after cleanup, the drive-board caps, belt and mechanism service on a Model 1, a laser clean and RF check, and a backup-battery decision. Unlike a Genesis, the recap here is not optional: the corrosion is a board-killer rather than a cosmetic issue.

If you would rather buy hardware that has already had this work done, everything I restore is in the shop, and my general bench method lives in the guides.

Everything below this point is the same material as structured data: the per-revision cap maps, the parts I would order, the pinouts worth metering against, and the one-line numbers. Each row carries its source and a confidence tag, so you can see exactly where a figure came from and how hard it is. Long notes fold away: expand any row for the detail.

Capacitor lists

Per board, per revision, because nothing carries across. Two things to read before ordering. On the Model 1, package is production-run dependent, so check whether your board has leaded or surface-mount cans in those positions. And the drive boards are separate entries for a reason: the Sony transport’s laser filter cap and the JVC drive-board caps are the ones people skip and then wonder why a recapped unit still will not read.

Replacement parts

The non-capacitor parts a Sega CD or 32X job actually consumes, plus the ones that are service-or-donor only.

Chip and connector pinouts

The component library: each IC and connector defined once, with the revision badge on the card showing which of the three machines and which board it sits on.

Two things to expect here. The Sega custom 315-xxxx gate arrays have no public datasheet (not the Mega CD Engine, not the PCM chip, not the 32X MARS parts) so those cards carry the part number, the position and what the part does, and no pin table. And several of the cards that do have pins carry only one or two, because that is what an OEM manual page actually names. Where a card is empty it says what would verify it.

The three cards worth opening before anything else are the two DC jacks, for the reason the whole top of this page exists, and the 32X A/V IN jack, whose pin 2 is the most destructive pin on the platform.

Schematic facts

Reference confidence key

How to read the confidence tags and source citations on the data above. Note that bench does not appear anywhere in this device’s data, because this page is built from documentation rather than from my own measurements.

Sources and further reading

These are the outside references I trust for the Sega CD and the 32X. I link them rather than copy them; go read the originals. The OEM Sega service and maintenance manuals behind most of the hard numbers here are scans that circulate freely: search for the Mega-CD Maintenance Manual (Export, Aug 1992), the SEGA CD II / MEGA CD II Service Manual No.002 with its A3 foldouts and its Additional Information supplement, and the GENESIS 32X / MEGA DRIVE 32X Service Manual No.012 with the 1994-95 Sega service bulletins.

• ConsoleMods: Sega CD model differences (the six Model 2 boards, lasers, and the service notes on belts, limiter switches and the Model 2 fuse)
• ConsoleMods: 32X model differences
• ConsoleMods: 32X service bulletin fixes
• ConsoleMods: 32X video output (A/V cabling and the +5 V-pin warning)
• ConsoleMods: 32X noisy inductor fix
• ConsoleMods: 32X region mod
• ConsoleMods: Sega CD BIOS mod
• Console5 TechWiki: Sega CD v1, Sega CD v2 Funai, Sega CD v2 Sony and 32X (IC lists and per-board cap maps)
• RetroSix wiki (the boot-stage diagnostics, the power-enable circuit walk-through, per-board capacitor tables, and the FRAM and MultiBIOS installs; a vendor, so read the product comparisons as advertising)
• RetroRGB: 32X and the kevtris video-noise fix
• The Console Power Supply Bible (the cross-console voltage and polarity table this page’s safety section leans on)
• SEGA Hardware Database (per-model rating labels and regional part numbers)
• RetroTechCollection: Sega CD Model 1 (tray belt, laser RF threshold, backup-battery options; a single-source guide, so its specific numbers are flagged as such above)
• 8bitplus: Mega CD recap walkthrough
• PowerUpGaming: console laser repair cross-reference
• Terraonion Mega SD and RetroRGB’s writeup of it
• scd-bios-adapter (open-source SOP-40 BIOS adapter)

And the companion page: my Genesis and Mega Drive reference covers the console these all bolt onto.