I’m fully focused on Replay2 currently, but just a quick update on the Namco CUS34 replacement. I really wanted to complete the alternative mode used on ToyPop and Libble Rabble.
Using my 20 year old bit of vero-board, I can compare the replacement and real chip directly. The logic analyzer connects to both parts and dip switches let me isolate outputs from the replacement while still feeding it all the inputs. I worked out which pins where inputs/outputs and bidirectional first by isolating each pin and applying pull ups/downs.
I’ve got it all running now, and the advantage of this direct compare is that the analyzer can tell me if there are any differences at all between the chips, but also I can compare and tweak the timing. The CPLD used is way faster than the original and sometimes delay cells need to be used to add some hold timing. The disadvantage compared to reverse engineering the die is that I can only test functionality used in the game.
Now I understand what the chip does, I can put it back on the Zynq tester shown in the previous post and apply some unusual signals just to make sure it works correctly.
I’ve seen posts from a Japanese chap who has reverse engineered the die. Great job, but no source code as far as I can see yet.The code for this, along with all the other reverse engineer chips will be in the public R2 git repository, and used in our latest FPGA cores.
I’ve been shipping replacement modules for many years for the Namco28 pin chips (and other parts) so message me if you are in need. I have quite a few in stock currently.
I’ve been shipping Namco CUS34 replacements for a while and the majority of the engineering work was done over 10 years ago. Reverse engineering the silicon by decapsulation of the die, taking pictures and manually tracing them is best way to understand what is going on, but it’s very time consuming. Most of these chips are quite simple and have little internal state – but to make sure all paths are covered I’ve started to retest the chips using a little jig designed by my friend Wolfgang. This sits on top of an AMD Zynq board, and allows each pin to be driven and measured.
The test framework is written in Python, and then wavedrom (https://wavedrom.com/) is used to draw the waveforms.
Sometimes you need a bit more timing information, so a logic analyser is also hooked up. This lets me see which edges are used.
A month or so ago I thought all was good – and then somebody tries it in a ToyPop board… ToyPop and Libble Rabble use the Namco System 16 Universal hardware.
Unlike the other boards which use this chip – Dragon Buster for example – pin 13 is tied low rather than high. This seems to put the chip in an entirely different mode. As far as I know, no schematics exist for these boards, so a brute force reverse engineering effort is underway.
I dug out my old adapter board from 10 years ago. This let’s me isolate each pin and with the logic analyzer – here the big Agilent 1680 – I can work out which pin is an input, which is an output etc.
I’ve got to fit this around Replay2 design work – more on that shortly, but I hope to be done soon – and I’ll document it publicly.