AA3000+ PCB Explorers

Notes

Practical takeaways from Chucky’s three build posts, all five pages of the EAB project thread, and the assembly notes supplied with these boards. Check PCB revision, component markings, and matching firmware before applying an instruction.

Source coverage: the three blog posts and the text of all 90 posts across EAB pages 1–5 were reviewed. Forum posts were read through a live browser session. Designer confirmations, individual build reports, and unresolved questions are distinguished below; historical claims are not current compatibility guarantees. Reviewed 25 October 2026.

1. Match the daughterboard revision first

• Do not mix the blog’s Altera programming procedure with the supplied rev 1.1 board instructions. The blog uses Altera MAX CPLDs, a USB Blaster, and Quartus II (the author used 13.0 SP1). Your rev 1.1 notes specify two Xilinx XC95144XL devices, a Xilinx Platform Cable USB, and ISE iMPACT 14.6. Use the firmware and tools for the actual devices fitted. [Blog part 3] [Rev 1.1 notes]
• PCI needs the daughterboard’s PSU connection, including during programming. The rev 1.1 notes identify power as CN500 and JTAG as CN400; the blog’s connector descriptions differ. Follow your board’s labels and pinout rather than copying a photograph. Program and verify both CPLDs with the matching U400/U401 files. [Rev 1.1 notes] [Blog part 3]
• The blog fits the daughterboard’s 100 µF electrolytics before the Zorro connectors for access, and removes two pins from one video-slot connector. Check the illustrated connector and your PCB before making that mechanical change. [Blog part 3]

2. Bring up the motherboard in stages

• Start with passives and inspect the small resistor/capacitor networks for bridges before connectors obstruct access. Chucky repeatedly warns that poorly installed chipset sockets are a common source of faults. [Part 1] [Part 2]
• Follow the guide’s checkpoints: power LED → reset and RGB sync → CPU running DiagROM → serial diagnostics → Chip RAM → video → Fast RAM → remaining interfaces. This makes it easier to isolate a fault than fully populating the board before its first test. Initial sync is checked at the 23-pin RGB port, not VGA. [Part 1]
• In the guide’s onboard-CPU setup, initial jumpers are J102: 2–3, J104: 1–2, and J100: 1–2. These are not universal accelerator settings: the post explicitly warns that its photos may show incorrect jumpers, and leaves its later full settings list unfinished. Check the CPU-card documentation when changing configurations. [Part 1]
• The guide reaches a normal boot with A3000 Kickstart ROMs. Use DiagROM and serial output while debugging rather than relying solely on a blank display. [Part 1]

3. Battery, memory, and SCSI cautions

• Battery type determines the charging components. For the blog’s CR2032/lithium setup, fit D192 and omit R192. The NiCad setup uses R192 and omits D192. Do not copy the rechargeable-battery arrangement for a non-rechargeable coin cell. [Part 2] [Motherboard notes]
• Do not assume the extended 64 MB Fast RAM feature works with the original build files. In the December 2020 post, Chucky reports no GAL code for U893 and says a redesign would be needed. The supplied notes list U893/C893 as optional. This is a dated limitation, not proof that no later solution exists. [Part 1] [Motherboard notes]
• Plan SCSI termination for the actual internal/external bus layout; the guide says to retain board termination when no external SCSI device is attached. The supplied board notes specify three 220/330 Ω, 8-pin SIP resistor packs at RP810–RP812 and show their orientation. [Part 1] [Motherboard notes]

4. Flicker fixer and video routing

• The flicker fixer is optional. If populated, its PLL needs adjustment with VC470; the blog uses an oscilloscope at the probe point shown in its photos to help obtain a stable waveform/picture and recommends an insulated adjustment tool. See the original photos before probing. [Part 2] [Motherboard notes]
• Without an external source on CN490, link pins 1–2, 3–4, 5–6, 9–10, 11–12 to route native Amiga video through the switch. Do not treat every pair as a jumper position. J490 selects automatic (1–2), forced Amiga (2–3), or forced GFX (3–4). [Motherboard notes and pinout]
• SW481 up bypasses the scandoubler (15 kHz); down enables it (31 kHz). A VGA connector alone does not make a 15 kHz signal compatible with every monitor. [Motherboard notes]

5. DSP: optional hardware, patch, and firmware

• Get the basic machine working before adding the optional DSP. The documented hardware fix is a wire from U122 pin 5 to U124 pin 27. The patch is also recorded in the supplied motherboard notes. [Part 1] [Motherboard notes]
• Part 2 uses updated JED files for U122, U124, U701 from the DSP project, and U123 from the original AA3000+ archive; the author used 5 ns GALs. Fit the appropriate delay line at U702 or U703 for the chosen device, and test with the project’s DSP software before fitting the obstructing connector. Treat this as the author’s build recipe, not a guarantee for every revision. [Part 2] [DSP project referenced by the blog]

6. Rev 1.1 PCI software and clockport

• The supplied rev 1.1 software sequence is MMU libraries → Picasso96 with Prometheus selected → matching Prometheus Firestorm drivers → reboot. The bundled Picasso96 driver files target 2.1c; other versions may require different files. [Rev 1.1 installation notes]
• The default clockport firmware is U401.jed; RTC clockport devices need U401_rtc.jed, which changes the jumper/address mapping. Clockport caching must be disabled for its selected 64 KB range (default 0xEF0000). Use the exact firmware/jumper table and MMU configuration in the notes rather than guessing. [Rev 1.1 clockport notes]

7. Useful troubleshooting reports

• No boot / no serial: a Part 1 commenter ultimately traced this to two bad or counterfeit 7407 chips; the custom chips were intact. This is one builder’s diagnosis, not a universal cause. Inspect soldering and verify ordinary logic parts before concluding an expensive custom chip is dead. [Part 1 comments]
• Chip RAM detected but DiagROM crashes when using it: a Part 2 commenter fixed a dry joint on a Chip RAM resistor pack. Detection alone does not prove the RAM path is sound. [Part 2 comments]

8. EAB: hardware compatibility and power limits

• Designer’s baseline: onboard 25 MHz 68030, 68881/68882 FPU, 2 MB Chip RAM, and 16 MB Fast RAM using two SIMMs. Hese explicitly describes extended Fast RAM as unfinished and needing another motherboard revision. The original Firebird specification is three PCI slots, four Zorro III slots, and one AGA video slot. [Hese, post 1]
• A3000 desktop fit is confirmed; universal compatibility is not. Hese describes it as an A3000 case drop-in and reports that the A3640 and SCSI work. He had not tested an A3660. Do not turn the CPU-slot specification into an assurance that every accelerator/SCSI combination works. [Hese’s replies, page 3]
• Firebird PCI is not a drop-in upgrade for a stock A3000 motherboard. Hese says it needs fifth-Zorro-slot signals that the original motherboard does not route to the daughterboard. The thread also asks about A4000 operation without establishing a confirmed result. [Designer confirmation] [Unanswered compatibility question]
• Budget the 3.3 V rail, not just PCI slot count. Hese quotes a 3 A converter for the early daughterboard. Another participant corrects an earlier “10 A” reference to 10 W and warns that demanding PCI cards can exceed a 3 A supply. Check the regulator actually fitted and the combined card load; these early figures are not a rating for every Firebird revision. [Early design rating] [Correction and load warning]
• 3.3 V / 5 V card support does not mean universal drivers. The introduction identifies a Prometheus-based bridge. A later question asks how to get a Radeon 92xx working, but the reviewed thread provides no confirmed recipe. Check specific card, bridge firmware, and driver support before buying. [Project specification] [Unanswered Radeon question]

9. EAB: sourcing and SCSI trouble spots

• Identify the rare chips before committing to a build. AGA chips and CIAs can come from an A1200; Super DMAC is A3000-specific, while Ramsey/Buster also appear in A3000/A4000 systems. A contributor highlights the difficulty of obtaining Super DMAC-04 for a Ramsey-07 combination. Later builders still identify Super DMAC as a major sourcing bottleneck. Forum price lists are historical, not present-day quotes, and their mixed part lists are not a replacement for your revision’s BOM. [Donor-chip guidance] [Historical cost breakdown] [Later sourcing report]
• “Newer DMAC/Ramsey fixes SCSI” is not established. One owner reports different results with DMAC-02/Ramsey-04 and DMAC-04/Ramsey-07 across two original A3000s using A3660 cards. That is useful evidence of configuration sensitivity, not a direct AA3000+ test or a guarantee for either pairing. Hese’s AA3000+ A3640 success is a separate report. [Original-A3000 owner report] [AA3000+ designer report]
• CyberStorm PPC SCSI remains an unresolved report in this thread. A user reports DMA / phase-mismatch errors with a cable/drive/termination setup previously working in an A4000. The suggestion that DSP DMA interferes is the poster’s hypothesis; no diagnosis or fix follows in the reviewed thread. Do not present it as a confirmed DSP fault. [Problem report]
• Extended ROM mapping needs care. A contributor warns that $A00000–$AFFFFF is AutoConfig I/O space and should not casually be repurposed for an extended Kickstart image. Treat this as an address-map warning to verify against your chosen ROM solution, not a tested recipe supplied by the thread. [ROM mapping warning]

10. EAB: DSP software exists, but is not automatic acceleration

• The later posts supersede early doubts about available DSP tools. Wrangler announces Asm3210 v2.2, GCC 3.2 for DSP3210, and elf2exem68k: write C, cross-compile, assemble, link with M68k code, and construct an executable on a modern computer. GCC 4/6 attempts hit internal compiler errors, so the announced working toolchain was GCC 3.2. These are versions reported in the thread, not a claim about today’s latest releases. [Toolchain announcement] [Asm3210 on Aminet]
• Check Aminet rather than assuming the GitHub tool binaries are current. Wrangler explicitly says the assembler/compiler copies in his repository were outdated. His Mandel_dsp example is a port of sorts of ARTAbrot; adapting other Mac DSP software also depends on source availability and the missing ARTA/VCOS integration layer on Amiga. [Version warning] [Porting limitations] [Wrangler’s DSP repository]
• MP3/JPEG decoding and other DSP uses are proposed in discussion, not demonstrated there as ready-to-install features. Installing the DSP does not make unmodified applications use it. [Suggested uses] [Author’s response]