Replacement chip: Difference between revisions

From NeoGeo Development Wiki
Jump to navigation Jump to search
(Added CPLD replacement proposition)
(Updated options, reasoning, done/todo)
 
Line 1: Line 1:
Dirty page for SNK chips replacements development.
Notes about modern SNK chips replacements. Some info can be generalized to replacement of digital custom chips in general.


Reading various [[http://newlifegames.com/nlg/index.php?topic=1859.0| repair logs]], it appears that dead or partially dead SNK chips are a common issue. It forces cannibalism and forcibly reduces the worldwide stock.
Reading various [[http://newlifegames.com/nlg/index.php?topic=1859.0 repair logs]], it appears that dead or partially faulty SNK chips are a common issue. Original replacements aren't available outside of some limited new-old-stock occasionally found on Aliexpress and eBay, which might exhibit the same problems if faults are a material or environmental issue. Another common solution is to pull chips from donor boards which are deemed unfixable. Both rely on available stocks that obviously aren't replenished anymore.


For preservation, it would be desirable to produce drop-in replacement chips to save original boards.
For preservation, it can be desirable to produce drop-in replacement chips that are reliable, faithful to the originals, and economically sensible.


=Considered solutions=
=Considered solutions=


Nobody likes rewiring to accomodate different pinouts, especially 80+ pins chips. Producing new ASICs is out of the question because of the price at low quantities, and because finding pin-compatible models todays would be difficult if not impossible.
Nobody likes rewiring to accomodate different pinouts, especially 80+ pins chips. Producing new ASICs is out of the question because of the price at low quantities, and because many packages used back in the day aren't available anymore.


The only solution would be adapter boards the size of the original chips, with castellated vias for surface mounting ?
* Dedicated ASICs ? No: not worth it for such low quantities, development and production of bare dies in the $30-40k range. Doesn't solve the packaging problem.
[[User:Furrtek|Furrtek]] ([[User talk:Furrtek|talk]]) 18:41, 15 January 2016 (CET)
* Shuttle service ASICs ? Possible but development still quite expensive if several iterations needed. Doesn't solve the packaging problem.
* Reconstruct the chip's logic with regular logic chips ? Possible if logic is very simple and footprint large enough. TSSOP and QFN parts.
* Use 2 boards on top of each other with a mezzanine connector ? Possible but more expensive and fragile.
* Use a CPLD ? Possible if logic simple enough and no RAM blocks needed. Microchip ATF series, Lattice LC4000ZE, and Greenpaks are 5V-tolerant, still made and affordable. Altera MAX3000 and Xilinx XC9500 are phased out but stock is still available.
* Use a FPGA ? No 5V-tolerant parts available anymore, requires voltage shifting (space issue). Flash-based parts like Gowin GW1N, Lattice iCE40 and Altera MAX10 are of interest.


* Reconstruct the chip's logic with regular logic chips ? No: they're all too big. BGA parts aren't 5V compatible.
The only solution would be adapter boards the size of the original chips, with castellated holes for surface mounting ?
* Use 2 boards on top of each other with a mezzanine connector ? No, no, no.
* Use 1- or 2-gate SOT-23-5/6 chips ? Maybe in some cases, but would be expensive (long BOM, lots of parts to assemble).
* Use a CPLD ? Probably the best solution. Altera's MAX3000 and Xilinx's XC9500 are 5V-tolerant and are still available at reasonable prices.


=CPLD proposals=
=List=


*[[NEO-257]]: 53 I/Os, EPM3064 in TQFP100 ?
*[[NEO-257]]: QFP64R, 5 standard logic chips, done
*[[NEO-273]]: Only used in cartridges, rarely dies. A clone wouldn't be that useful ?
*[[NEO-273]]: QFP64R, 5 standard logic chips, done
*[[NEO-BUF]]: Really a pain, too small for QFP CPLDs. 74LVCH16245 in BGA56 ?
*[[NEO-BUF]]: QFP48, 2 standard logic chips, done
*[[PRO-C0]]: Need pinout, quite big so EPM3128 in TQFP144 ?
*[[PRO-C0]]: Possible but doesn't seem to fail often and only used on 1st gen systems and games, see ZMC2
*[[NEO-C1]]: EPM3128 has enough I/Os, but too big in TQFP144.
*[[NEO-C1]]: QFP100R, fits in 128LEs, all options too big to fit in footprint, flex adapter being attempted
*[[NEO-D0]]: 55 I/Os, EPM3064 in TQFP100 ?
*[[NEO-D0]]: QFP64R, CPLD, done
*[[NEO-E0]]: 54 I/Os, EPM3064 in TQFP100 ?
*[[NEO-E0]]: QFP64R, not enough IOs on QFP100 CPLD, standard logic chips and Greenpaks, done
*[[NEO-F0]]: 60 I/Os, EPM3064 in TQFP100 ?
*[[NEO-F0]]: QFP64R, possible with CPLD ?
*[[NEO-G0]]: 56 I/Os, EPM3064 in TQFP100 ?
*[[NEO-G0]]: QFP64R, 5 standard logic chips, done
*[[NEO-I0]]: 56 I/Os, EPM3064 in TQFP100 ?
*[[NEO-I0]]: QFP64R, possible with CPLD ?
*[[NEO-ZMC]]: Very small, can be found on AliExpress (January 2016). BGA ?
*[[NEO-ZMC]]: Very small, can be found on AliExpress (January 2026). Possible with Greenpak ?
*[[NEO-ZMC2]]: 71 I/Os, EPM3128 in TQFP100 ?
*[[NEO-ZMC2]]: QFP80R, CPLD, done
*[[PCM]]: Only used in cartridges, rarely dies. 65 I/Os, EPM3064 in TQFP100 ? Or AFT1508AS. -7ns minimum ?
*[[PCM]]: QFP80R, CPLD, done
*[[NEO-DCR-T]], [[NEO-SDR-T]]: Need pinouts, probably same problem as for NEO-C1.
*[[NEO-DCR-T]], [[NEO-SDR-T]]: QFP100R, same problem as for NEO-C1.
*Line buffers ([[PRO-B0]], [[NEO-B1]]) could be replaced with MAX10's and level shifting ?. Probably not worth it.


*LSPCs could be replaced but the I/O requirements directs to >$60 CPLDs.
Most other custom chips are too complex to be worth replacing. LSPC requires FPGA, level shifters, very fine pitch adapter, ...
*Line buffers ([[PRO-B0]], [[NEO-B1]]) could be replaced with MAX10's and level shifting ?. Might be too expensive.
 
==Altera==
 
MAX3000 I/Os.
 
*EPM3032: 34. $1.61
*EPM3064: 34 or 66. $2.93
*EPM3128: 80 (or 96 but doesn't fit). $9.89
 
==Xilinx==
 
XC9500/XL I/Os.
 
*XC9536: 34
*XC9536XL: 36. $1.46
*XC9572: 34, 69 or 72
*XC9572XL: 34, 38, 52 or 72. $2.86
*XC95108: 81. Hard to find.
*XC95144/XL: 81. $7.25
 
=Board proposals=
 
Which board houses are castellation friendly ? Is classic 1.6mm good ?
 
==QFP64R==
==QFP80R==
==QFP100R==


=Logic definitions=
=Logic definitions=

Latest revision as of 20:21, 25 January 2026

Notes about modern SNK chips replacements. Some info can be generalized to replacement of digital custom chips in general.

Reading various [repair logs], it appears that dead or partially faulty SNK chips are a common issue. Original replacements aren't available outside of some limited new-old-stock occasionally found on Aliexpress and eBay, which might exhibit the same problems if faults are a material or environmental issue. Another common solution is to pull chips from donor boards which are deemed unfixable. Both rely on available stocks that obviously aren't replenished anymore.

For preservation, it can be desirable to produce drop-in replacement chips that are reliable, faithful to the originals, and economically sensible.

Considered solutions

Nobody likes rewiring to accomodate different pinouts, especially 80+ pins chips. Producing new ASICs is out of the question because of the price at low quantities, and because many packages used back in the day aren't available anymore.

  • Dedicated ASICs ? No: not worth it for such low quantities, development and production of bare dies in the $30-40k range. Doesn't solve the packaging problem.
  • Shuttle service ASICs ? Possible but development still quite expensive if several iterations needed. Doesn't solve the packaging problem.
  • Reconstruct the chip's logic with regular logic chips ? Possible if logic is very simple and footprint large enough. TSSOP and QFN parts.
  • Use 2 boards on top of each other with a mezzanine connector ? Possible but more expensive and fragile.
  • Use a CPLD ? Possible if logic simple enough and no RAM blocks needed. Microchip ATF series, Lattice LC4000ZE, and Greenpaks are 5V-tolerant, still made and affordable. Altera MAX3000 and Xilinx XC9500 are phased out but stock is still available.
  • Use a FPGA ? No 5V-tolerant parts available anymore, requires voltage shifting (space issue). Flash-based parts like Gowin GW1N, Lattice iCE40 and Altera MAX10 are of interest.

The only solution would be adapter boards the size of the original chips, with castellated holes for surface mounting ?

List

  • NEO-257: QFP64R, 5 standard logic chips, done
  • NEO-273: QFP64R, 5 standard logic chips, done
  • NEO-BUF: QFP48, 2 standard logic chips, done
  • PRO-C0: Possible but doesn't seem to fail often and only used on 1st gen systems and games, see ZMC2
  • NEO-C1: QFP100R, fits in 128LEs, all options too big to fit in footprint, flex adapter being attempted
  • NEO-D0: QFP64R, CPLD, done
  • NEO-E0: QFP64R, not enough IOs on QFP100 CPLD, standard logic chips and Greenpaks, done
  • NEO-F0: QFP64R, possible with CPLD ?
  • NEO-G0: QFP64R, 5 standard logic chips, done
  • NEO-I0: QFP64R, possible with CPLD ?
  • NEO-ZMC: Very small, can be found on AliExpress (January 2026). Possible with Greenpak ?
  • NEO-ZMC2: QFP80R, CPLD, done
  • PCM: QFP80R, CPLD, done
  • NEO-DCR-T, NEO-SDR-T: QFP100R, same problem as for NEO-C1.
  • Line buffers (PRO-B0, NEO-B1) could be replaced with MAX10's and level shifting ?. Probably not worth it.

Most other custom chips are too complex to be worth replacing. LSPC requires FPGA, level shifters, very fine pitch adapter, ...

Logic definitions

See https://github.com/neogeodev/SNKVerilog