NEC TurboExpress / PC Engine GT Repair Reference

This page covers the NEC TurboExpress (US, 1990) and the PC Engine GT, its Japanese twin. They are one machine: same board, same capacitors, same faults, same chipset. The only electrical difference between them is a single 0 ohm resistor that sets the region, so everything below applies to both unless I say otherwise.

For transparency, this page is different from most of the references here. The part of this machine I have real hands-on with is the power side: I design, build and bench-test the two USB-C adapters that let one of these run without its long-dead original brick, and the polarity trap below is the exact hazard those products exist to eliminate. The rest of this page is built from documentation rather than from my own repair count. So I have said which documentation, and how strongly each claim is attested, every time. Where the sources disagree with each other, or where something rests on one person’s bench rather than two, that is called out rather than smoothed over.

That attribution matters more here than on most machines, for one reason: no NEC service manual for the TurboExpress has ever surfaced publicly. The best electrical references that exist are a public-domain reverse-engineered schematic set traced from copper-layer scans of a donor GT board, and RetroSix’s bench-traced wiki pages. The genuine NEC TurboGrafx-16 console manual does exist, and it shares the chipset and the cartridge interface, but its power section describes a completely different machine. Do not carry the console’s 10.5 V across.

The organising idea for this handheld is simpler than most: the capacitors are the machine. Console5 puts it as bluntly as anyone puts anything in this hobby, saying that without total capacitor replacement a Turbo Express is “guaranteed to fail,” and that you should not replace only the audio or display caps, change them all. Almost every symptom below traces back to those capacitors, so the correct first move on nearly any fault is a full recap, and only then diagnosis.

Safety first: the polarity trap

Read this before you plug anything into one. It is the fastest way to destroy a TurboExpress, and it is nastier than most power traps because the wrong plug fits perfectly.

Same barrel. Opposite polarity. The two supplies physically interchange, nothing mechanical stops the mistake, and each one reverse-feeds the other. I cover the same trap from the other side on the Game Boy page, because it is equally lethal in that direction.

The TurboExpress figure is well attested: the GameSX power-supply table, a photograph of the OEM adapter’s own label reading OUTPUT: DC7V 700mA with a centre-positive diagram printed next to it, and the reverse-engineered schematic itself, where the jack’s tip contact feeds the battery-positive rail and the sleeve is the return. Wikipedia’s “6 volt AC adapter” line is simply wrong, and worth knowing about because it is the figure most people will hit first.

Three more things about this jack that are worth burning in.

Do not count on reverse-polarity protection. The reverse-engineered schematic draws a series diode, D500, between the jack node and the battery rail, but it draws it cathode-facing-the-jack, an orientation that would block the adapter rather than pass it. Either the schematic has the part flipped or the topology is something I have not worked out, and its authors mark the whole set provisional. Until somebody meters a real board, treat this machine as having no protection at all and get the polarity right at the plug.

The jack switches ground, not power. Battery negative reaches system ground through inductor L501 and the jack’s own normally-closed switch contact. Inserting a plug breaks that path and hands the ground return to the adapter. That is why a worn jack does not give you a clean dead machine, it gives you random on-and-off cycling on wall power, which is exactly the symptom Console5 lists against a damaged jack or plug. It is also why any adapter or dongle has to make a solid ground contact, not just a solid centre contact.

The safe input window is 7 to 9 V, centre-positive. The OEM adapter is 7 V and a fresh six-cell AA pack is about 9 V nominal, so the machine’s own design already spans that range. RetroSix’s bench procedure confirms the internal 5 V rail still regulates with as little as 6 V injected, which sets a comfortable floor. What is not correct is the TurboGrafx-16 console’s 10.5 V, and it is not a fixed “9 to 10 V” either.

Where my two adapters fit

This is the trap the TurboExpress adapters I build exist to close, so here is exactly what each one does and what it needs. Both use a right-angle 3.5 x 1.35 mm barrel wired centre-positive, and both are labelled on the enclosure with their voltage and polarity, which is the whole point.

• USB-C 7.5 V adapter, model CA-DTE1. An enclosed boost converter that takes plain 5 V in and regulates 7.5 V out, which sits just above the OEM adapter’s 7 V and comfortably inside the machine’s own 7 to 9 V window. Because it boosts from 5 V rather than negotiating, it needs no USB Power Delivery at all and runs off any USB port that can source about 2 A, including dumb USB-A bricks, laptops, power banks and car ports.
• USB-C PD 9 V adapter, model CA-DTEL1, the Lite. The budget build. A PD trigger asks the charger for 9 V and passes it straight through, with a status LED that lights once the contract holds. 9 V is the top of the window and exactly what a fresh set of six AA cells delivers. The condition, stated plainly: it requires a USB-C PD source that can actually do 9 V. A 5 V port or a dumb USB-A cable will not work, and a few unusually strict chargers will refuse to hold the contract. If you want something that works off literally any USB source, take the boost version above.

One reconciliation worth flagging because my own listings say it differently: those product pages describe the original NEC brick as 7.5 V at 1.25 A. The repair sources I trust most here (the GameSX power table and that photograph of the adapter label) both say 7 V at 700 mA, and that is the figure I use on this page. It does not change what either adapter is safe to do, since 7.5 V and 9 V both land inside the machine’s design window, but the listing copy is the number I would correct, not this one.

Unlike a Game Gear, the TurboExpress will not shock you through its main rails, but it does carry one genuinely dangerous node: the cold-cathode backlight inverter runs roughly 900 to 1600 V, and it is live on batteries alone because the inverter boosts up from the system rail. Do not put an ordinary scope probe on the backlight connector or the transformer secondary.

Know your board first

There is nothing to identify, and that is the useful finding.

There is no published mainboard revision history for this handheld. ConsoleMods states outright for the whole TurboGrafx-16 and PC Engine family that models are arranged chronologically and there are no revisions, and no source I have enumerates distinct GT or TE mainboard revs. One Console5 board-position capacitor map is explicitly marked as covering both “USA -A-” and “JPN” boards, which is about as direct a statement as you get that the layout does not differ by region.

So the axis that actually matters is region, and you can read it off the outside of the machine:

• US TurboExpress. Black shell, TurboExpress branding, product code HES-EXP-01 (a single-source code, so do not lean on it). The card slot takes US TurboChips.
• Japanese PC Engine GT. PC Engine GT branding. Takes Japanese HuCards.

Inside, the one documented electrical difference between them is a single 0 ohm resistor: HuC6280A pin 29 tied to ground selects the US region. It is populated on a US TurboExpress and left open on a Japanese GT. That is half the region lockout; the other half is that the card-edge data bus runs in reverse between the two markets, so D0 on one is D7 on the other. There is no lockout chip anywhere on this machine, and a unit that refuses imports is behaving exactly as designed.

Two board numbers are worth recording if you want to build this section out over time. The reverse-engineered schematic names its donor mainboard 7610 060, and the front control board carries 7610-0641-A. If you see a different silkscreen on a real unit, that is genuinely new information.

Common problems and fixes

Ordered by how often the sources say each one comes up, not by how interesting it is. Nearly all of them route back to the same place, so start there.

Recap first: all 18 capacitors, every unit

This machine is the poster child for surface-mount electrolytic leakage. The electrolyte creeps out under the can, corrodes the pads and eats the traces underneath, and the same failure wears half a dozen different disguises depending on which capacitor let go first.

The community position here is unusually unanimous, and I will attribute it rather than assert it. Console5 says a TurboExpress with original capacitors is “guaranteed to fail,” and follows it with “do not replace only audio or display caps! Change them all!!” RetroSix calls capacitor replacement mandatory maintenance rather than optional for this platform. ConsoleMods says any unserviced GT or TurboExpress must have its capacitors replaced, plus a full clean of leaked electrolyte and repair of corroded traces, on both boards.

The count is 18: seventeen on the main board and one on the front control board, and three independent lists agree designator for designator. Two of them are through-hole, which trips up anyone working from an “it’s all SMD” mental model:

• CC702, 470 uF, the main power reservoir. It lays over at 90 degrees to clear the RF shield, and its body must not exceed about 6.3 mm diameter or the shell will not close. Do not upsize it for longevity.
• CC800, 47 uF, on the front control board behind the button membrane. This is the single most-forgotten capacitor on the machine, because it is on a separate board that a main-board recap never touches. If you have a “recapped” unit with flaky buttons, this is what the last person missed.

Whichever you fit, clean and inspect every former cap site under magnification before the new parts go on. The whole point of doing this is that the electrolyte has already been working on the copper, and a beautiful recap over an eaten trace fixes nothing. There is a documented case of exactly that: a completely dead unit recapped with on-hand parts, and the recap “did nothing,” because the problem was corrosion and wrong values, not the capacitors.

Full values, voltages, body dimensions and substitutes are tabled at the bottom of this page. For a resale-refurb operation, Console5 sells a matched kit that covers all 18 plus the two jailbar-fix ceramics, and that is the fastest correct path. Their leaded kit carries one caveat straight from Console5: the factory RF shielding may need altering or removing to clear the taller radial parts, so on a unit you are selling as original-condition, take the surface-mount kit.

Backlight is on but the picture is dim or absent

The signature TurboExpress fault after the recap itself, and it has a named cause. pcengine-fx’s teardown identifies a single 4.7 uF 35 V electrolytic on the line that feeds the LCD’s power as “THE culprit” of the backlight-but-no-picture fault: when it dries or bleeds, the panel no longer gets enough to run. That part is CC501, and RetroSix independently lists CC501 against dim or absent display. Its 50 V sibling CC502 is the other flagged position in the same circuit.

The first thing to establish is that this is not a dead backlight. If the screen glows at all, the whole inverter is doing its job and the fault is downstream.

Then confirm the panel’s bias supply is present. On the LCD ribbon CN201, pin 3 should carry roughly -25 V DC on a healthy machine. That rail is generated by transformer T500, which turns a small AC pulse into about 35 V AC and then rectifies and regulates it down to an adjustable -10 to -30 V. The exact number is one bench source and forum reports of around -29 V also exist, so do not condemn a rail on its precise value; condemn it for being absent or small. What you need to see is a large negative voltage.

The useful structural fact here, and it saves a lot of time: all eight signals at CN201 are present with nothing else connected to the mainboard. So if the drive signals are good at the ribbon and the panel is still dark, the fault is the LCD panel or its daughterboard, not the video source. The full CN201 pinout with expected levels is in the pinout section below.

If the picture is washed out or cannot be dialled in rather than absent, check the contrast path. The thumbwheel is not a divider in the usual sense, it is a 0 to 4.7 kilohm variable resistance across ribbon pins 2 and 8, with 0 ohms giving an over-bright picture and 4.7 k giving black. Verify it actually sweeps that range. The real contrast drive voltages are generated on the panel’s own board around its OKI M6370 driver using that -25 V as a reference, so a contrast fault can live on either side of the ribbon.

Dead, distorted or low audio, the speaker especially

Leaking audio-path capacitors, first and second and third. The two 33 uF coupling capacitors CC101 and CC102 are the ones the sources name for low or distorted audio, and the bank of 100 uF power and audio decoupling parts is what they name for no audio at all or a machine that will not power up.

The signal path, so you can bisect it: audio leaves the CPU as analog left on HuC6280A pin 17 and right on pin 18, passes through the coupling capacitors and resistors to a dual 5 kilohm volume thumbwheel, and lands on IC100, an NJM2073M dual power amp fed from the raw battery rail through two 68 ohm resistors. A game card can also inject its own mono audio into that mix.

The output routing is where a very common misdiagnosis lives. With headphones plugged in you get true stereo. With headphones out, the jack’s switch mixes left and right into the amp’s left channel only, and that single channel drives the speaker through front-ribbon pin 11. So the speaker is mono by design, and a speaker that is silent while headphones play perfectly is very often the jack’s switch contacts rather than the amp or the speaker.

RetroSix’s isolation trick is worth knowing: jumper the amp’s left output directly to front-ribbon pin 11 and listen on the speaker. That proves the amp independently of the headphone jack. Solder a speaker across pins 11 and 12 to test the speaker element itself.

One documented gotcha: low speaker output can persist through a correct recap. When that happens, suspect the amplifier, the volume pot, or corroded traces under the old cap sites, not more capacitors.

No backlight at all, sometimes with a blown fuse

Before anything else, confirm the backlight is even enabled, because two separate things gate it and either will look like a dead inverter.

1. The front board gates it. The circuit only runs when front-ribbon pin 8 is pulled to ground, and the front board is what does the pulling. Bench testing without the front board connected means grounding that pin by hand.
2. The BLH solder jumper. There is a pad labelled BLH at the top right of the board. With its jumper removed the backlight is intentionally off. That is a documented power-saving modification and also the standard step when fitting a modern screen, so on a used machine it may well have been done on purpose.

Past that, the backlight is a self-oscillating Royer inverter built around transformer T900 and a matched transistor pair, Q900 and Q901, stepping the roughly 9 V system rail up to the 900 to 1600 V that strikes the fluorescent tube. Console5’s common-problems table sends “display backlight does not turn on, fuse may or may not also blow” straight to a shorted Q900 or Q901, so diode-test them. Check fuse F500 for continuity. And check L900, the inductor under epoxy above T900, along with the T900 centre-tap trace that should sit at about 9 V; RetroSix reports that inductor as a frequent open.

Q900 and Q901 must be replaced as a matched pair: RetroSix is explicit that a mismatch burns them both out. And do not confuse them with Q502: the backlight pair are 2SD1898, Q502 is a B1182. Ordering the wrong one is an easy mistake and my own working notes contained exactly that mix-up at one point.

If the tube itself is gone or hopelessly dim, the pragmatic answer in 2026 is usually a modern screen, which deletes the whole high-voltage circuit. See the mods section.

No power, or random on and off

Split this cleanly in two before you start: is there anything on the system rail at all, or is the system rail present and the 5 V dead?

Nothing at all. Walk the input chain. From the battery positive pin the rail runs through inductor L502 (with a 330 ohm resistor in parallel across it), then fuse F500, then the power switch. Battery negative returns through L501 and the DC jack’s ground switch. Any one of those open is a dead machine, and both inductors are usually hidden under epoxy, so buzz them rather than looking at them. Then the jack: the rivet-style centre pin loosens and cracks its joint with age, and Console5’s field fix is to reflow solder onto that centre conductor before condemning the jack. Random on-and-off cycling on wall power specifically is the jack’s signature, for the switched-ground reason in the safety section above.

System rail present, 5 V dead. The +5 V logic rail here is a discrete adjustable series regulator, not a 7805. RetroSix identifies the pass device as Q502, a PNP, with Q503, R507 and C507 forming the loop, and the reverse-engineered schematic gives Q502 the value B1182. Treat Q502’s collector, the large centre tab, as the +5 V test point. RetroSix’s bench procedure is to inject 6 V at the battery terminals first, confirm the collector regulates, and only then raise to 9 V. That procedure is single-source, but it costs nothing and it is the right instinct.

Here is the reading that trips people up: a healthy 5 V rail on this machine can measure anywhere from about 4.5 to 5.0 V. Console5 puts the factory setting of trimmer VR501, the Vcc adjust, at around 4.5 V and notes that these units are not restricted to TTL voltages, while RetroSix’s bench procedure expects something nearer 5 V. Both are inside a healthy range, so treat the whole window as normal. Do not chase a 5.00 V reading and do not “fix” a correct rail by winding VR501. The narrower 4.5 to 4.75 V figure this page used to give came from one source that hedged it itself.

Finally, remember the capacitors. The 100 uF decoupling group is what the sources name for a machine that will not power on, and CC702 is the main reservoir. A shorted power electrolytic drags the rail down and looks like a regulator fault.

Brightness pulses in time with the sound

Weak cells or a sagging supply. Console5 lists it directly: on batteries, if the image is bright or the brightness changes in time with the sound, replace the batteries or test with an adapter. If a stiff supply clears it, that was it. If it persists, the power and decoupling capacitors are no longer holding the rail steady under the combined audio and backlight load, and that is a recap.

This is the one fault where a known-good adapter is a genuine diagnostic instrument rather than a convenience, which is a large part of why I make the adapters at all.

Jailbars and vertical lines

Two different things get conflated here.

Capacitor leakage in the video path produces jailbar artifacts and horizontal lines, and a recap clears those. Separately, there is a distinct jailbar fix that targets fixed-pattern vertical banding by adding two 4.7 uF ceramic capacitors on the video encoder’s decoupling rails. Both Console5 cap kits bundle those ceramics, which is the sensible way to buy them since they arrive with the recap.

So: recap first, then if fixed vertical bars survive on an otherwise good picture, fit the ceramics. Console5 places them at C709 and C715. That designator pair is Console5-only for this handheld (the general fix is documented across the PC Engine family but not those specific positions here), so read the board before you solder.

Genuinely scrambled graphics, meaning wrong tiles and sprites rather than bars, are a different animal. That points at the digital core, the video SRAMs, or corroded address and data traces under a leaked capacitor, and no amount of recapping will help.

Buttons dead or flaky

The front control board carries the BU5782K ASIC and the leaded capacitor CC800 on its 5 V feed. Leaked CC800 or a damaged ribbon path is the usual cause, and CC800 is photographed failed in Console5’s own writeup.

If you want to prove the interface rather than guess, the protocol is documented. The ASIC polls the buttons roughly every 33 ms; a 1 microsecond 5 V pulse on ribbon pin 7 clocks the read, pin 6 carries a reference pulse map, and pins 2 through 5 each encode two buttons by pulse polarity: idle 5 V, one button gives low pulses, the other high pulses, both together pull the line to 0 V. Scope pin 7 for the trigger, then the data pins for their response.

One thing that is not a fault: the I and II sliders are three-position turbo switches, off then roughly 4 Hz then roughly 7 to 8 Hz. A machine that “presses buttons by itself” usually has a slider left in a turbo position.

Replace the BU5782K only if you have proven it dead. It is a scarce NEC ASIC with no drop-in.

A card will not boot

Almost always the contacts or the region, not the board. Clean the card edge and the slot, and try a known-good card of the correct region. RetroSix measured no change in current draw with or without a card inserted, so a bench meter will not see a card problem at all.

If it is region, that is not a fault. See the board section above for the mechanism, and the mods section for the modern fix.

Inside the family: what differs

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

• It is a TurboGrafx-16 shrunk into a handheld. Same HuC6280A CPU with its integrated 6-channel PSG, same HuC6270 video display controller, same HuC6260 colour encoder, same 8 KB of work RAM and 64 KB of VRAM, same media. What the home console does not have, and what therefore has no OEM documentation anywhere, is the battery front end, the LCD, the negative bias rail and the cold-cathode inverter. That is precisely the part that breaks.
• One video display controller, not two. Wikipedia’s TurboExpress infobox lists two HuC6270A chips. That is the SuperGrafx console, not this. The schematic shows exactly one, at IC701.
• The video is natively RGB. The HuC6260 produces analog RGB and drives the LCD with it directly, with no composite round-trip anywhere inside the machine. That is unusual for the era and it is the entire reason RGB screen driver boards give such a good result here.
• Four separate power rails come off one board. The schematic derives +5V, +5VA analog, +5VL for the LCD and +5VP, plus the negative LCD bias from transformer T500 and the backlight high voltage from T900. The split analog and LCD rails are why the jailbar fix targets specific decoupling positions rather than the logic rail.
• The front board does three unrelated jobs. It reads the buttons, it gates the backlight, and it carries the speaker. So a disconnected or capacitor-damaged front board presents as “no backlight and no speaker” at the same time while the picture and headphone audio are perfectly fine. That combination is a strong tell.
• There is no save memory at all. Nothing on this machine is battery-backed; it plays the same save-less HuCards the console does. A community project exists to add battery-backed saves using a harvested HuC6201, but it is shelved with unresolved problems and is not something to plan around.
• Wikipedia contradicts itself on the panel resolution, giving 336x221 in the infobox and 400x270 in the specifications section. The panel is a roughly 2.6 inch active-matrix colour TFT; I have no panel datasheet, so I am leaving the native grid unsettled rather than picking one.

Mods worth knowing

I do not reproduce anyone’s install guide. This is an orientation to what is worth doing and where the traps are.

• The recap is not a mod, it is the baseline. A machine that has not had it is a project unit regardless of whether it currently boots, because the leakage is progressive. Nobody pays a premium for it; they price down without it.
• RGB screen replacement is the marquee upgrade. Because the machine is natively RGB, an RGB-input driver board (LCDDRV, and the newer RGBDRV) gives a sharp direct image with no lossy conversion. The electrical install taps RGB at the LCD connector CN201 but takes sync from HuC6260 pin 44, because the composite sync present at CN201 is inverted. That pin identity is corroborated twice, by RetroSix’s install page and independently by the schematic’s own labelling, so it is one of the more solid facts on this page. The install also disables the original contrast circuit and desolders the BLH jumper to kill the high-voltage inverter, and the shell window has to be widened because the aftermarket panel is larger than the OEM one. Two warnings worth carrying: Console5 states that the good RGB driver boards are not sold through US distributors and that the majority of LCDDRV kits in circulation are counterfeit, and glue-free bracket kits exist specifically so an install does not wreck future serviceability. A hot-glued screen is a mark against a unit.
• Composite is the budget screen path. Cheaper controller boards want composite, which means first generating a composite signal off the HuC6260 with a small encoder board. It works and it is softer than RGB. Prefer RGB when boards are available.
• Region-free HuCard. The Kytor connector board transplants your factory slot onto a board that can present the card edge either way round, and you hold SELECT at power-on to switch region, power-cycling to go back. No shell cutting and fully reversible, which is exactly what you want on a machine you might resell. The soldering is expert-level, around 0.2 to 0.3 mm work.
• USB-C power is the accessory almost every one of these needs, because working original bricks barely exist any more. That is the two adapters in the safety section above.
• A replacement mainboard is the rescue path, not a value-add. When the electrolyte has eaten more traces than are worth rebuilding, RetroSix’s ReBoard is a full replica you transplant every part onto, and the Board-Folk project publishes public-domain gerbers you can simply have fabbed. Either way it is an expert-level transplant that restores a dead machine rather than making a better one.
• Small mechanical wins. The slide-switch actuator breaks or goes missing constantly and there is a printed replacement, and printed replacement shells exist for cracked or yellowed cases.

Recap and parts

Order the kit. For this machine specifically, the buy-the-kit argument is stronger than usual: the 18 positions span five capacitance values across four voltage ratings and three body sizes, and two of them are through-hole with mechanical constraints on the can diameter. A vendor kit has already solved the footprint problem, comes with the jailbar ceramics, and costs less than the time you would spend assembling distributor part numbers.

If you do source loose, hold to three rules. Match capacitance and use 6.3 V as the minimum wherever the map says 6 V, because the original 6 V rating simply became 6.3 V in modern parts and the two are the same position. Hold to the original body dimensions so everything clears the RF shield and the shell closes, and in particular keep CC702 at or under 6.3 mm diameter. And keep CC501 and CC502 distinct: they are both 4.7 uF but at different voltages and different sizes, and CC501 is the one behind the video fault.

The parts a TurboExpress repair actually consumes beyond capacitors are short: the DC jack, the Q900/Q901 backlight pair, Q502, the NJM2073M amp, and F500. Everything else is either not a wear part or has no source. All of it is tabled below with what each source actually says.

The hard data behind all of the above (the per-board capacitor maps, the non-capacitor consumables, the chip and connector pinouts, and the schematic facts) follows. Long per-item notes fold away to keep it readable; expand any of them for the full detail. It is organised by region rather than by board revision, for the reason in the board section: there are no documented revisions, and region is where the one real electrical difference lives.

Capacitor lists

All 18, per board. The two tables are deliberately identical between the US and Japanese machines, because one Console5 capacitor map covers both and I would rather show that explicitly than make you infer it. The values are the consolidated agreement of three independent lists; the body dimensions come from the one of them that measured them.

Replacement parts

The non-capacitor parts a TurboExpress repair actually turns on, plus the two escalation paths (a replacement mainboard and a region-free card connector) that are worth knowing exist before you start a job you cannot finish.

Chip & connector pinouts

The component library: each chip and connector defined once, with an interactive pin diagram and a folded pin table. The revision badge on each card shows which machines it applies to.

Read the confidence tags on this section carefully, because they vary more than usual here. The HuCard card-edge pinout is from the genuine NEC service manual, so it is as solid as anything on this site, but read its region note, because the data bus runs the other way on a Japanese machine. The LCD ribbon and the front-board ribbon are corroborated twice each, by a bench trace and independently by the schematic. The 40-way HuCard flex, the internal AV header and the backlight connector come from that reverse-engineered schematic alone, and its own authors mark it provisional, so if your meter disagrees with those tables, believe your meter.

Where I could not verify a pinout against a datasheet or manual I actually have, the card says so and the table is empty. The three big Hudson chips are the obvious case: I know exactly which parts they are, and for the CPU and the colour encoder I have listed only the handful of pins that two independent sources confirm.

Schematic facts

Reference confidence key

How to read the confidence tags and source citations on the data above. Note that nothing on this device carries a bench tag, for the reason in the introduction: this data is documentation-derived, not measured by me. Everything here is either two-source verified or explicitly flagged as resting on one.

Sources and further reading

These are the outside references this page is built on. I link them rather than copy them, and I would encourage anyone working on one of these to go read the originals, because the documentation for this machine is thin enough that every source matters.

• Console5 TechWiki: Turbo Express (the chip list, the capacitor list, the common-problems table, the VR500 and VR501 adjustments, the BLH backlight-disable jumper and the DC-jack fix)
• Board-Folk: PC Engine GT, the public-domain reverse-engineered KiCad schematic and gerbers, traced from copper-layer scans. The single most valuable electrical document that exists for this handheld, and there is a companion repository for the front control board
• RetroSix wiki, the bench-traced circuit pages for the PC Engine GT: the 5 V circuit, powering from the bench, the backlight driver, the LCD circuit, the audio circuit and the front button board. A vendor, so read the product comparisons as advertising, but the circuit work is real and it is the only place a lot of this exists
• ConsoleMods: TG16 model differences and TG16 region information
• GameSX: PC Engine power supplies (the adapter voltage and polarity table)
• pcengine-fx forums: the repair-guide sticky set is the definitive community corpus for this machine, and the video fix and A/C adapter threads in particular are where the CC501 identification and the adapter-label photograph come from
• Copetti: PC Engine / TurboGrafx-16 architecture (what the chipset is actually doing)
• Kytor: region-free HuCard board

If you would rather buy a working power solution than build one, the TurboExpress adapters are the two described above, and everything else I make is in the shop.