Sega Genesis / Mega Drive Repair Reference

The Sega Genesis, sold as the Mega Drive outside North America, is one of the more satisfying 16-bit machines to work on, because most of what fails on it is understood and fixable, and because the platform rewards knowing exactly which board is in front of you before you touch anything. That last part is the whole game here. Sega re-laid this board out roughly once a year for a decade, almost always to fold more discrete logic into a custom gate array, and those revisions change the audio circuit, the video encoder, the mod paths, and even the number of voltage regulators. A fix that is right for one revision is wrong or impossible on another. So this page leads with board identification and comes back to it constantly.

This is a reference for the standalone consoles: Model 1 revisions VA0 through VA7, and Model 2 revisions VA0 through VA2.3, NTSC and PAL. The Model 2 short boards (VA3, VA4 with its all-in-one GOAC die) and the Genesis 3 show up for lineage and the occasional bench visit, but they are not my core range and I flag them as context when they appear. The Sega CD and the 32X are separate devices with their own repair story, and they get their own page later. They intrude here in exactly one place that matters: the power connector, because the 32X shares the Model 2 supply and polarity, which is the opposite of the Model 1 — while the Sega CD add-ons, confusingly, take the Model 1 type instead. More on that below, because it is the single most damaging mistake you can make on this platform.

Where the source material is thin or the community disagrees, I say so rather than papering over it. A few specific numbers on this platform rest on a single source, and I have called those out in plain language instead of pretending they are settled.

Know your board first

Because a bare Genesis PCB swaps between shells trivially, the case only narrows it down. The board confirms it. Here is the shape of the two families and then the per-revision detail.

The line splits into two case families with materially different A/V, audio, and power hardware:

Model 1 revisions (VA0 through VA7)

Revisions are silkscreened VAn on the board, with a few exceptions: “VA0” was never marked and is a retro-name, and “VA6.5” and “VA6.8” come from Sega service docs (VA6.8 is marked “V6.8”). The trend across the whole line is consolidation of discrete logic into Sega gate arrays.

• VA0 (837-6656, 1988, Japan only). Uses an EDCLK-fix daughterboard plus discrete 74LS00 glue. Overdriven preamp, the worst rainbow banding of any revision.
• VA1 (837-6832, 1989, Japan). The daughterboard and 74LS00 fold into the 315-5339 gate array.
• VA2 (837-6955 JP / 837-6992 NA, 1989, first North American board). 315-5339 becomes 315-5345, fixing a DMA-refresh glitch. The NA layout adds an internal RF modulator.
• VA3 (837-7071, 1989 to 1991, NA and first Brazil). The bus arbiter and clock generator merge into the 315-5364. Sega fixes the overdriven preamp and adds a gentle low-pass filter. This is the audio baseline enthusiasts prize. Some 1991 boards get the die-shrunk 315-5313A VDP.
• VA4 (837-7138 JP/KR / 171-5872-20 Asia/PAL, first Asia/PAL/Korea). Roughly a VA3 with the video encoder flipped and the circuit rearranged. Most PAL “High Definition Graphics, Stereo Sound” units are VA4.
• VA5 (171-5963, 1990 to 1991, Japan). The I/O and arbiter merge into the big 315-5402 gate array. Carries factory JP5/6/7 jumper wires, which are not a mod. Some 315-5402 boards show TMSS only in export mode.
• VA6 (171-5963-xx, 1991 to 1993, most regions). Same as VA5 but the I/O chip is the 315-5433, which brings the first TMSS (the “Produced by or under license” license screen). Last revision with the EXT port in most markets.
• VA6.5 (171-5963-10/-30, 1992 to 1993). A VA6 with the EXT port and line filters deleted, leaving a blank rear panel. Many use the 315-5313A VDP. The most common PAL Model 1.
• VA6.8 (171-5963-40, 1993, PAL only). Like VA6.5 but an SMD Fujitsu MB3514 encoder replaces the DIP Sony CXA1145P. Community documentation (ConsoleMods) reports that the factory omitted C30 on this board, so there is no mono audio in stock form and you get sound only from the headphone jack. I have not been able to corroborate the C30 detail against a second independent source, so treat it as a lead to verify on the board rather than a settled fact. Final PAL Model 1, and often confused with VA6.5.
• VA7 (171-6217-xx, 1992 to 1993, NA/JP/Brazil/Taiwan). A total redesign. The first standalone unit built on the Yamaha FC1004 ASIC (315-5487), which puts the VDP, the I/O, and a modified YM3438 FM core on one die. Single 7805 instead of two, SMD RAM, SMD CXA1145 encoder, PLCC 68000. On non-Japanese units the rear ports shift toward center, which is the fastest visual tell. The 315-5487 has a broken 50 Hz mode; some Brazilian VA7s use the fixed 315-5660.

A couple of oddities: VA7 is essentially a Genesis 2 board in a Genesis 1 case, and it even shares the PC PCB prefix. A Korean Samsung variant (SPC-200/201R Rev.A) is a VA4 derivative with no EXT port, making it the only board with neither TMSS nor EXT. Brazilian VA3 units output PAL-M via a 53.634 MHz oscillator and encoder pin 7 tied to ground.

Model 2 revisions (VA0 through VA2.3, plus VA3/VA4 for context)

• VA0 (1993, all regions). A condensed VA7 with the Z80 in a QFP. The worst-sounding Model 2. The encoder varies (Samsung KA2195D is blurry, Sony CXA1145M and Fujitsu MB3514 are cleaner).
• VA1 (171-6534A, 1993 to 1994, last Japanese board). Uses the 315-5660 ASIC (YM3438 core). Single main-RAM and single VRAM chip. Audio resistors tweaked to fight distortion.
• VA1.8 (171-6534A/B, 1994 to 1996). No real electrical change from VA1; some add power-jack reinforcement wires.
• VA2 (171-6535F, 1994, NA only). Reverts to a discrete YM2612 (the 315-5786 ASIC has no FM core), and FM comes out too loud versus PSG. Carries a bodge PCB with a 74HC14 that fixes /ASEL for Virtua Racing and the 32X. Bad RGB jailbars baked into the trace routing.
• VA2.3 (171-7039x, 1995 to 1996, NA only). Discrete YM2612 again, Schmitt trigger integrated so no bodge board, cleaner RGB. Watch the model number: an MK-1631A label means this board carries the flawed 315-5685 VDP (broken shadow/highlight, glitched rasters); a plain MK-1631 gets the good 315-5786.

Beyond my core scope but worth recognizing: VA3 (first short board, RF shield stops partway) and VA4, which uses the 315-5960 GOAC die (Yamaha FJ3002). GOAC stands for Genesis-on-a-chip: the 68000, Z80, VDP, YM3438, I/O, and Z80 RAM all live on one die, leaving only main RAM, VRAM, the encoder, and the amp as discrete parts. VA4 short boards have two square RF-shield cutouts and glitch on 32X titles. The Genesis 3 takes this even further with a different GOAC and merged RAM.

How to identify it, and why it decides the repair

On a Model 1 without opening the case:

• FCC ID on the bottom label. FJ846EUSASEGA is an early non-TMSS board (VA2/VA3); FJ8USASEGA is a TMSS board (VA6, VA6.5, VA7). TMSS units also carry a patent-number block.
• Silkscreen slogan. “High Definition Graphics” is NA 1989 to 1992; “High Definition Graphics, Stereo Sound” is almost always a PAL VA4.
• Rear ports. A power jack that is not hard right, or an A/V connector shifted toward center with a gap on the right, means VA7. Ports in stock positions with an EXT-port blank means VA6.5.
• Cart-port trace tell. VA2 has thick traces near the slot; VA3 and PAL VA4 have a small circle in the same spot.

On a Model 2 without opening:

• Board width through the bottom vents. A full-width board is VA0 through VA2.3; a 3/4-width short board is VA3 or VA4. This is the quickest “will this sound good” check.
• Short board with rectangular RF-shield cutouts is VA4; short board with no cutouts is VA3.
• Model number. MK-1631A means VA2.3 with the buggy 315-5685; plain MK-1631 means the good 315-5786. MK-1451 with a white label is a Majesco VA4.

Three things ride on getting this right:

• Audio quality. The FM chip and the amp circuit both change across the line, and the difference is real and audible (next section).
• Mod compatibility. The subcarrier pin for a jailbar fix, whether a 50/60 Hz switch will even work, and which RGB-bypass tap points apply are all board-specific. Selling a “50/60 Hz switch” on a 315-5487 board is selling a mod that does not function.
• Region and lockout. TMSS arrives at VA6, and the region jumper method differs sharply by revision.

The audio boundary, stated precisely

People talk about “the good early Genesis audio” and usually get the reason wrong, so here is the actual delta. It splits along the chip-integration boundary, not neatly at any one revision:

• Model 1 VA0 through VA6.8, and Model 2 VA2/VA2.3, use a discrete YM2612. Its 9-bit DAC has an undocumented quantization flaw, the “ladder effect,” where the top and bottom halves of the waveform stick. It is most audible on quiet passages and fades, and composers exploited it (Ecco soft-panning, After Burner II drums). This grit on quiet sounds is original and expected. Do not recap or mod it away as if it were a fault.
• Model 1 VA7 and Model 2 VA0 through VA1.8 use a YM3438 core inside the FC1004 ASIC. The core itself is fine, but Sega paired it with a cheap LM324-class op-amp and a too-high-Q Sallen-Key low-pass filter. These are the muffled, distorted, worst-sounding Genesis units, and the fault is the circuit, not the chip.
• “Louder early audio” is mostly the amp, not the chip. VA0 through VA2 run a slightly overdriven preamp: louder, but it distorts on some games. VA3 fixed the overdrive and added mild low-pass, which is the enthusiast baseline. So a healthy VA6 should sound like a clean VA3, not overdriven. A loud or distorted VA6 is a cap fault, not “how it is supposed to sound.”

The practical upshot: a bad-sounding VA7 or early Model 2 that survives a recap is a design limit you fix with a mod, while a bad-sounding VA6 is a repair.

The power trap: polarity runs opposite between models

Read this before you plug anything in. This is the most damaging bench mistake on the platform, and it is not intuitive.

The Model 1 is center-negative and the Model 2 is center-positive. Two independent sources agree on the polarity flip and the connector sizes, so the polarity is the hard fact to trust. The sources differ slightly on the Model 2 nominal voltage (one says 10 V / 850 mA, the other calls it the same voltage as the Model 1 with lower current), so I treat 9 to 10 V as the healthy window and do not hang anything on the exact figure. The PAL Model 1 OEM block diagram labels its adapter 10 V / 1.2 A, so a “9 V” and a “10 V” center-negative barrel are both correct depending on region.

Why it bites:

• Wrong polarity damages the console. This input is not the polarity-agnostic bridge-rectifier front end an American NES has. Feeding reverse polarity drives current the wrong way into the input protection and downstream.
• The Model 1 barrel is physically identical to the NES barrel but wants the opposite polarity from most center-positive adapters. That is an easy and destructive mismatch.
• The Model 2 shares its supply and polarity with the 32X and the Nomad — but NOT with the Sega CD. This one catches people, including me when I first wrote this page. Both Sega CD add-ons, Model 1 and Model 2 alike, have their own supply and it is centre-negative on the larger 5.5 x 2.1 mm barrel — the Genesis Model 1 type, drawing up to 1.2 A. A Sega CD Model 2 sitting under a Genesis Model 2 does not take the Genesis Model 2’s brick. The all-in-one CDX / Multi-Mega is the exception that sows the confusion: being a combo machine it really is centre-positive, 9.5 V on the EIAJ-03. Sources: the OEM Mega CD II service manual, junkerhq’s power-supply bible, ConsoleMods, and the SEGA Hardware DB reading of the MK-4102A’s own label. Meter your own jack before you trust any of us.
• The Model 1 tolerates a bit over 9 V (10 to 12 V works but runs hotter; avoid 12 V). Voltage headroom is not the hazard here. Polarity is.

The bench rule, the same one I use on the Game Gear with its own region-dependent polarity trap: meter the barrel of any unknown adapter for voltage AND polarity before the first power-on, every time. Do not trust a label, a bin, or my table alone. After any reverse-polarity event, check the input protection diodes and the regulators before recommissioning.

Common problems and fixes

Ordered by how often it actually shows up on a refurb bench, not by severity. The healthy signal baselines every one of these diverges from (clock frequencies, rail voltages, pin numbers) are summarized in the last section.

Glitchy graphics, crashes, or a dead slot (most common)

Garbled or checkerboard graphics, freezes, colored static, a black screen, or a game that runs only with sideways pressure on the cart. This reads as a dead console but it is almost always the connector.

The 64-pin (2x32) card-edge slot exposes the raw 68000 bus: full address, 16-bit data, control, +5 V, and ground. Any intermittent contact on an address or data line corrupts fetches into garbage or a lockup; a bad +5 V or control contact gives no boot at all. The cause is tarnished contacts, splayed fingers, or a cracked solder joint at the slot. This affects every revision.

Prove the contact before suspecting a chip. Try a known-good, freshly cleaned cart and a second title. Wiggle-test: if the fault tracks pressure, it is the connector. On the bench you can probe +5 V at cart pins A2/A31 and the 68000 clock (~7.67 MHz NTSC) at cart pin B19 to confirm the bus is live at the slot. Fix by cleaning both the cart edge and the slot with high-purity isopropyl alcohol, re-tensioning splayed fingers, and reflowing the slot’s through-hole pins if a joint is cracked. This is the highest-yield first move on any “won’t play” Genesis.

Buzzing, hum, weak or distorted audio

The classic aging Genesis. Buzz, hum, or crackle in the audio, volume that fades or distorts, and sometimes worsened jailbars or dim video. It is progressive, not sudden.

The cause is aging through-hole electrolytics losing capacitance, going high-ESR, or leaking. Unlike the Game Gear, the Genesis uses through-hole radial electrolytics, not surface-mount, so leakage-under-cap board rot is far less of an issue and a recap here is usually preventative rather than a board-saving emergency. The load-bearing offenders are the audio-coupling caps, the main power filter (a 220 uF on most boards), and on VA3 a 220 uF / 6.3 V VRAM VCC cap that sits on the board underside where it is easy to miss on a top-side recap.

A buzz that changes with the Model 1 volume slider points at the slider or the coupling path; a buzz present at every slider position points upstream. Scope ripple on +5 V and on the amp output, and ESR-test the electrolytics in circuit. Because the Model 1 rear A/V audio is mono and stereo lives only on the front headphone jack, a “no rear audio, fine on headphones” split is the mono coupling path, not a chip. Fix with a full recap from the correct per-revision value list. It does not fix the design-limited muffled audio on VA7 and early Model 2 (below) or jailbars, which are circuit topology, not tired caps.

Completely dead: no LED, no video

Work outward from the DC input, and start with the polarity trap above. Meter the barrel for correct voltage and polarity first. Then walk the path: is voltage present at the jack, is there continuity through the input fuse or ferrite, and is +5.0 V present on the 7805 output? A dead output with good input is a bad regulator or a downstream short pulling it down.

The specific suspects: a wrong or reverse-polarity supply; a cracked-solder or bent-pin DC jack (very common on the board-mounted Model 1 jack); a blown input fuse or protection diode, often from a reverse-polarity event; and the 7805 itself. Remember the regulator count is revision-dependent: Model 1 VA0 through VA6.8 use two 7805s; VA7 and most Model 2 use one. One more gotcha worth knowing: a shorted RGB or SCART cable can crowbar the +5 V pin on the A/V connector, so test the “boots without cable, dies with cable” case by unplugging the A/V.

Powers but won’t boot, or won’t release reset (the VA6 trap)

LED on, +5 V present, but a black screen and no boot, sometimes with no video at all and the CPU held. On a VA6 this is a known trap.

The Genesis reset is not a plain RC on a CPU pin. It is a comparator-based supervisor, which is exactly why a marginal cap or reference here holds the whole board off. A power-on RC (about 47 kohm and 10 uF) generates MRES#, and an LM358 comparator produces the CPU reset by comparing that charging cap against a reference borrowed from the CXA1145 video encoder’s Vref pin. The comparator sits at 0 V, holding reset, until the cap charges past the reference. The exact reference voltage (about 2.0 V) and the exact power-on timing come from a single deep forum source and I treat those specific numbers as unconfirmed, though the topology itself is corroborated by the schematic.

This bites on VA6 restores because the reset now depends on three things being healthy: the 10 uF timing cap, the LM358, and the CXA1145’s Vref being present. A leaky reset cap, a sick LM358, or a missing encoder Vref all produce a board that has 5 V but never releases reset. So on this symptom, confirm +5 V and then check reset before chasing the CPU or ASIC: verify the 10 uF reset cap is in spec and measure the comparator reference. Scoping the controller SELECT line (pin 7) helps: if it never leaves reset, chase the supervisor first. Only then move on to the master clock and CPU. A caution that matters on VA7 and early Model 2: one section of the audio quad op-amp is reused as the reset comparator on those boards, so a botched audio op-amp swap can also kill boot.

Non-reset causes of the same symptom are a dead master clock, which takes out the 68000, Z80, VDP, and PSG at once, or bad main/Z80 RAM or a dead CPU.

Stuck, non-latching, or intermittent reset (mechanical)

Distinct from the trap above: this is the physical switch, not the supervisor. A worn or oxidized reset switch either asserts reset continuously (won’t run) or does it intermittently (random resets, often tracking case flex). Continuity-test the switch off-board and scope MRES# while pressing. Clean or replace the switch.

Controller: stuck direction, missing buttons, dead port

A direction permanently held, buttons dead or wrong, one port dead while the other works, or 6-button pads misreading. The cause is a worn or cracked DE-9 port, or a blown input buffer in the I/O path. The console drives pin 7 (SELECT) to multiplex which buttons appear on the data pins, and a shorted pin or stuck SELECT collapses the mux (a permanent “down plus left” is the classic short signature). Six-button pads need clean, rapid SELECT toggles, so a marginal port breaks the ID handshake and the pad misreads.

Swap in a known-good 3-button pad first; if a direction stays stuck, it is the port, not the pad. Reflow the DE-9 pins, verify +5 V on pin 5 and ground on pin 8, and scope SELECT pulsing when a game polls. Verify the fix with both a 3-button and a 6-button pad.

Scratchy volume or cutting-out audio (Model 1 only)

The Model 1 front slide-potentiometer wears and oxidizes, giving scratchy or intermittent audio that changes as you move the slider. Model 2 has no volume control, so this is a Model 1 fault. If the noise tracks slider movement it is the pot; if the same buzz is present at every position, look upstream at the coupling caps. Clean the slide-pot with contact cleaner and exercise it, replace if worn, and test the headphone jack independently.

Vertical “jailbars” over the picture (inherent, a mod not a repair)

Fine vertical bars, worst on blue, most visible once you take clean RGB. This is crosstalk from the VDP’s color-subcarrier clock coupling into the RGB or sync traces on the way to the encoder. It is a board-layout artifact, not a failed part. A board can be electrically perfect and still show jailbars, so do not “repair” toward removing them.

Confirm it is the fine, regular jailbar pattern (layout) rather than irregular hum bars (a power-filter cap) or snow (the video chain). Severity runs by revision: Model 1 VA0 through VA6.8 show composite rainbow banding plus blue jailbars, VA7 and early Model 2 clean the composite up with minor blue bars, and VA2 Model 2 has severe jailbars its trace routing makes unremovable even by a Triple Bypass. Removing them is a mod (subcarrier lift or a full RGB bypass), covered below.

No video with working audio, garbage video, or wrong colors

Audio playing while the screen is wrong is a useful split: it proves the CPU, cart contact, and 5 V rail, and corners the fault into the video path. Candidates: a marginal master clock (usually a fully dead board, but a sick oscillator can give unstable video), the 220 uF VRAM VCC cap undervolting VRAM into sparkle and corrupted tiles, a bad video encoder or its output caps (no composite while RGB may still be fine), or a dead VDP (garbage tiles with audio alive). Check RGB at the VDP and composite sync, and if RGB is good at the VDP but composite is dead, the encoder or its caps are the fault. Recap the VRAM and encoder caps; a dead encoder or VDP is a donor-board job.

Muffled or low audio that survives a recap (design, not fault)

Specifically on VA7 Model 1 and Model 2 VA0, VA1, and VA1.8. This is not a fault. Sega paired an adequate YM3438 core with an inadequate circuit: LM324-class op-amps and a too-high-Q Sallen-Key filter, plus under-valued resistors that overload the op-amp inputs on a single +5 V supply. Confirm the revision, and if a VA7 or early Model 2 sounds bad after a good recap, stop chasing electrolytics. The fix is a mod: either the stock-circuit component swap or a full audio bypass (below).

One channel dead, or FM without PSG (or the reverse)

Two independent sources are mixed: FM from the YM2612 (analog MOL/MOR out, about 1 Vpp) and PSG from inside the VDP. FM requires the full 68000-to-Z80 path alive; PSG rides the VDP. So the split localizes: FM dead with PSG OK points at the YM2612 or the Z80/arbiter/IO path feeding it; PSG dead with FM OK points at the VDP PSG output or its mix resistor; one channel only points at that channel’s coupling cap or one op-amp section. You can probe the YM2612 analog-out pins directly and quietly hear whether FM is being generated. Do not “fix” the ladder-effect grit on a discrete-YM2612 board, since that is original behavior.

License screen then hang, or some carts won’t run (TMSS and region)

The “Produced by or under license by Sega” screen appears and the game hangs, or one import or unlicensed cart won’t run on a console that plays everything else. This is a software lockout, not a hardware fault. TMSS, introduced on Model 1 VA6 via the 315-5433 I/O chip, checks for the ASCII string “SEGA” at two cartridge locations at boot and shows the license screen; if neither matches, the game does not run. Pre-VA6 boards go straight to the game with no license screen, and that absence is not a fault. A working license screen is actually a positive boot milestone: it proves the 68000, VDP, video, and reset are all healthy enough to render text. Reproduce with multiple carts; if licensed carts play and only an import hangs, it is TMSS or region, and region-free handling is a mod, not a repair.

Crashy or heat-related lockups after warmup

Boots and plays, then freezes or resets after minutes, sometimes recovering when cool. The cause is a marginal +5 V under thermal load: a high-ESR 7805, ripple from a tired input filter cap, or a cracked joint that opens as the board heats. Note that the 7805 dropout from a 9 to 10 V input is large, so the regulators run genuinely hot as normal. A single-regulator VA7 or Model 2 has less thermal margin than a dual-7805 early Model 1. Scope +5 V for droop as the board warms, freeze-spray the regulator, RAM, and ASIC to see if the fault clears, and reflow suspect joints. Distinguish “runs hot but stable,” which is normal, from “hot then crashes.”

Inside the Genesis

A short tour of what the hardware is doing, which makes the faults above make sense.

• It is a dual-CPU machine. A Motorola 68000 runs the game; a Zilog Z80 (from the Master System) runs as the sound co-processor, provides Master System backward compatibility, and helps with controller I/O. Around them sit Sega custom ASICs whose integration level is the main thing that changed revision to revision.
• The integration trend predicts the board. VA0 and VA1 start with an EDCLK daughterboard; VA3 merges the arbiter and clock generator into the 315-5364; VA5 and VA6 merge again into the 315-5402 and 315-5433 (the 5433 adds the TMSS ROM); VA7 folds the VDP, I/O, and a YM3438 into the FC1004 ASIC; Model 2 continues that line and then VA2/VA2.3 drop back to a discrete YM2612; Model 2 VA4 goes fully GOAC. That one sentence predicts most of the per-revision audio and video behavior.
• One master oscillator runs everything. There is no separate CPU crystal. The master clock is 53.693175 MHz on NTSC and 53.203424 MHz on PAL; the 68000 is that divided by 7 (~7.67 MHz NTSC, exposed conveniently on cart pin B19), and the Z80 is divided by 15 (~3.58 MHz). A dead master clock kills the 68000, Z80, VDP, and PSG at once, so scope the oscillator first on a fully dead board. The odd exact frequencies are the NTSC and PAL color subcarriers multiplied up, which is why a true composite-accurate region conversion (not just a 50/60 Hz flag) wants the matching crystal.
• Power is two split +5 V rails on the dual-regulator boards. On Model 1 up through VA6.8, one 7805 feeds the main logic and the other feeds secondary loads, each with its own filter cap. There is no second voltage level; both are 5.0 V. VA7 and Model 2 consolidate to a single rail.
• The video chain is VDP to encoder to A/V. The VDP outputs analog RGB and composite sync directly, the encoder (CXA1145 or a later substitute) turns that into composite plus buffered RGB, and later boards swap encoders (CXA1645, KA2195D, MB3514, BH7236), which is why composite quality varies by revision even when the digital side is identical.
• The A/V connector differs by model, and the Model 1 pin numbering is genuinely contested. The Model 1 8-pin DIN carries the same signal set everyone agrees on (composite, ground, mono audio, RGB, +5 V, sync), but the two community text sources disagree on the pin numbers and the OEM schematic uses a third numbering entirely. The Model 2 9-pin mini-DIN pinout is triple-confirmed and solid. So before you wire a Model 1 RGB cable, verify by continuity on the actual board rather than trusting a published Model 1 pin table.

Mods worth knowing

I do not reproduce anyone’s install guide here. This is an orientation to what is worth doing and where to get the real instructions. Two facts frame every A/V decision: the Genesis already outputs RGB natively, so no mod is needed to add RGB, and a full A/V bypass almost always sacrifices composite video and is irreversible on many revisions. For a preservation-grade unit, a recap plus the correct cable plus an external scaler often beats cutting the board.

• RGB is a cable, not a mod. Every model already puts analog RGB on the multi-out; you only need the right cable. Model 1 is 8-pin DIN with mono on the port, so its RGB cables carry a 3.5 mm tail to grab stereo from the headphone jack. Model 2, 3, Nomad, and CDX are 9-pin mini-DIN with stereo on the port. A common no-cut value-add is fitting a Model 2 9-pin multi-out into a Model 1 so it takes cheaper, more available Gen 2 cables. Buyers pay for “RGB modded” units, but a stock console plus the right cable already does RGB.
• Subcarrier-pin lift. The cheap partial jailbar fix: lift and trace-cut the VDP chroma-subcarrier pin so it stops crosstalking into the RGB and sync lines. Big RGB cleanup for near-zero cost, but it disables composite. The target pin depends on the ASIC (315-5313/5313A is pin 50, the FC1004 family is pin 131, and the later families differ again). Two independent wikis and the OEM schematics agree on the first two families; meter against the board before lifting.
• Full RGB bypass (discrete). Tap raw RGB straight off the VDP, isolate the multi-out from the stock amp, and drive it through a better external amp. High effort, irreversible, kills composite. One hard rule: on Model 1 you solder to the VDP sync pin but do not lift it, because some revisions need it for the Z80. Most people now just install a bypass board instead of doing this by hand. The Model 2 die-pin tap numbers for the 315-5660 are single-sourced, so confirm on your board before cutting.
• Triple Bypass (3BP). The standard open-source full A/V bypass board (audio plus video plus the Model 1 DIN). One board fits every revision via jumpers. It outputs 75-ohm CSYNC on the composite pin, so true composite cables will not work but any sync-on-composite cable will. On Model 1 the win is mostly video (audio is already good); on Model 2 and Genesis 3 it is the big fix for the notoriously bad audio. Current KiCad files live on GitHub (tianfeng33’s v2, Zaxour’s remix, db-electronics’ original); assembled boards come from Insurrection Industries, RetroGamerStuff, and Zaxour.
• Mega-AV. The current-best full bypass and the 3BP successor. Marginally better RGB and audio, but its headline is the best composite video seen from a Genesis, better than stock. Choose it over a 3BP when composite matters. Open-source files are on GitHub (RIP-Felix/MEGA-AV-main).
• Mega Amp 2.0 / 2.0 Pro. An audio-only bypass (no video) that restores Model-1-grade sound on any model and gives every model real headphone drive. Choose it when you want the audio cleanup but to keep stock composite and RGB untouched, which makes it a natural fit for a Genesis 3 (mono to stereo) or a bad Model 2 where you do not want to bypass video. Boards are orderable on OSHPark.
• Model 2 VA0/1/1.8 and Model 1 VA7 audio component swap. Instead of a bypass board, fix the stock circuit on the worst-sounding revisions by swapping the LM324-class op-amps for better single-supply quads and correcting the Sallen-Key filter’s passive values. It cleans up the muffled sound without cutting traces, which some buyers prefer as more original-looking. The critical caution: one of the four op-amp sections is the reset comparator, so do not fit an op-amp with output phase reversal (for example a TL074) or you will break boot. Good choices are the TL974I, MC34074, MC33204, TS924, TLV9104, and the pricier OPA4171.
• Region and 50/60 Hz. Most Model 1 boards expose four jumper pads: JP1/JP2 for language and JP3/JP4 for TV system. The region half is corroborated by two sources; the 50/60 Hz-jumper half is single-sourced, so meter against your board’s schematic before committing a switch. A hard warning: VA7 Model 1 and early VA0 Model 2 have no working 50 Hz mode. The 315-5487 ASIC does not implement 50 Hz properly (games lose sound or crash), and the Japanese 315-5487-10 has no 50 Hz mode at all. Only the revised 315-5660 runs both cleanly. Do not sell a “50/60 Hz switch” on a 315-5487 board. Model 2 modding difficulty is board-specific: VA0/VA1/VA1.8 hardwire the ASIC pins (hard), VA2/VA2.3 put jumpers on the underside (easy), VA3 exposes a pad for a single cut, and VA4 uses 0-ohm jumpers.
• Overclock. Raises the 68000 clock to reduce in-game slowdown. It does not speed up gameplay or raise audio pitch; it only cuts lag in games that have it. Stock is master-clock-divided-by-7; divided-by-5 gives about 10.74 MHz NTSC, which most games tolerate with occasional mild glitching. It is out of spec for the 8 MHz-rated 68000, so per-game glitches are expected. A keeper mod, not resale prep.
• Open Heart. A cheap Raspberry Pi Pico board that bundles TMSS bypass, a dual-frequency oscillator (region and 50/60 Hz without a crystal swap), in-game reset, and overclock into one module driven by controller combos. Much easier than doing region plus overclock discretely, and attractive as a single “region-free plus 60 Hz plus no license screen” package on an import-oriented listing.
• HDMI. The MegaSwitchHD is an internal FPGA mod that taps the VDP and generates 1080p HDMI from the console’s own chips while leaving analog video working, compatible with Model 1 revisions using the 315-5313 VDP (VA0 through VA6.5). It is install-service territory. The RAD2x is an external, no-solder, plug-and-play line-doubler on the multi-out and the sensible path for most setups. FPGA clones (Analogue Mega Sg) and MiSTer are the no-mod digital route.
• Genesis 3 compatibility restore. Out of core scope but it still comes up. The Genesis 3 drops pins that break the 32X, Sega CD, SMS, Game Genie, and Virtua Racing, and outputs mono. You can restore much of it by wiring the missing GOAC pins back to the cart port (the pin lists are per-source and board-swap risky, so verify against the cart pinout before wiring). Neither Genesis 3 revision can talk to a Sega CD.
• Flash carts. Not a console mod, but a Mega EverDrive or Terraonion Mega SD is invaluable as a known-good test cart during repair (rules out a dirty game cart in one move), and a flash cart bundled with a refurbished console is a strong listing.

Cartridge save batteries

Not a console fault, but it lands on the same bench. Any Genesis cartridge with battery-backed saves carries a coin cell that is now three decades past its shelf life, and when it dies the game boots normally and simply forgets everything. Check how the cell is actually held before you reach for an iron. On some carts it is soldered; on others a metal tab retains it by friction even though the assembly does not look like a socket, and those come out without heat.

Recap and parts

The one rule to carry away: every in-scope Genesis and Mega Drive electrolytic is a through-hole radial, not surface-mount. Through-hole caps age far more gracefully than the surface-mount cans on a Game Gear, they rarely vent onto the board, and there is no SMD-corrosion trap on either console here. So a Genesis recap is usually preventative or symptom-driven, not the board-saving emergency it is on a Game Gear. Recap when you see audio buzz or weak sound, flickering power, or dim or rolling video, or when you are warrantying a resale unit. Do not sell a recap as a jailbar cure, since jailbars are a layout issue.

The one thing that will bite you: the caps are board-specific. Sega re-laid the board out every revision, so the designators and value splits do not carry across. Match the cap map or schematic for the exact board in hand.

• Buy the revision-matched kit for one-off work. Console5 sells a pre-matched kit for the exact board: an “IC series” kit covers Model 1 VA0 through VA6, and a “PC series” kit covers VA7 and Model 2 VA0/VA1/VA1.8. Right count, right values, name-brand parts, one SKU. A couple of per-revision gotchas: VA3 has the easy-to-miss 220 uF / 6.3 V VRAM cap on the underside; Model 2 VA1/VA1.8 have an encoder-dependent cap (CE14 is 220 uF on MB3514 boards but 10 uF on CXA1145/KA2195D boards, so check the encoder before ordering); and VA2/VA2.3 add two small odd values (2.2 uF and 22 uF at 50 V) that a VA0/VA1 kit will be missing.
• For à-la-carte, use Panasonic FR or Nichicon UPW. Both are low-ESR 105 C parts stocked at DigiKey and Mouser and a clear upgrade over the OEM caps. You can safely uprate voltage to standardize a kit (higher working voltage is always fine); never downrate. The two Model 1 audio-coupling caps (C61/C62, 1 uF / 50 V) are best as bipolar/non-polar types, though I will note that recommendation is single-sourced and the OEM schematic draws them polarized, so a polarized part there also works.
• 7805 regulators. Model 1 VA0 through VA6.8 carry two; VA7 and most Model 2 carry one. They run genuinely hot from the 9 to 10 V input, which is normal. A direct replacement is the STMicro L7805CV (keep the heatsink). A cool-running upgrade is the Recom R-78E5.0-1.0 switching module in the same TO-220 footprint, good on single-regulator boards and inside tight Model 2 shells, but keep it away from the analog audio and video section on an audiophile build (it is a switcher), and verify at least 1 A of headroom.
• A/V socket. The Model 1 8-pin DIN and Model 2 9-pin mini-DIN are non-standard Sega footprints, so DigiKey and Mouser do not stock a drop-in. Failures are almost always cold or cracked joints, so reflow first; replace only with a salvage socket or a specialty repro.
• Controller ports. The Sega port is a female DE-9. A generic right-angle PCB-mount DE-9 is electrically correct but the mounting footprint differs from Sega’s, so dry-fit before committing or pull a salvage port from a donor. Most “dead controller” faults are joints, so reflow the existing port before replacing.
• Cart slot, reset switch, volume slider. The cart slot is a cleaning job, not a replacement (IPA and light burnishing; no distributor equivalent). The Model 1 reset tact switch and volume slide-pot are common mechanical wear items with standard substitutes.

A sourcing note for a US bench: price DigiKey and Mouser first. After the current US import tariffs, LCSC usually loses on landed cost for small distributor-stocked parts (caps, 7805s, DE-9s) even where the unit price looks lower; it only wins on bulk reels you would import anyway.

If you would rather buy a console that has already had this work done, everything I restore is in the shop, and my general restoration and bench-testing method lives in the guides. I will link specific Genesis and Mega Drive service and mod listings here as those pages firm up.

Everything below this point is the same material as structured data: the per-revision cap maps, the parts I actually order, the pinouts I meter against, and the one-line numbers I look up. Each row carries its source and a confidence tag, so you can see where a figure came from and how hard it is. Expand any row for the detail.

Capacitor lists

The electrolytic map for each board revision, with the substitutes I fit. Two things to carry in: the designators do not survive a revision change, and a few revisions have no published list of their own. VA5, VA6.5 and VA6.8 are in that group. VA6.5 is a VA6 with the EXT port and line filters deleted, so the VA6 list is the right starting point minus the parts that went with those sections, but confirm against the board rather than trusting the count.