The UM348x family of sound generating devices is currently unemulated mainly due to a lack of understanding of the chip internal rom structure and data.
The sound of these devices should be familiar to many of you as they have been broadly used in many 80s and 90s low end gadgets, doorbells, etc. Bootleg arcade games have used these chips as a way to integrate music at a minimum cost.
Each chip contains a fixed number of melodies hardcoded in the silicon during manufacturing.
Sean Riddle and ClawGrip have made extensive efforts to document these devices, full decap images, sound recordings, and main mask rom dump available for UM3481A and UM3482A here: http://www.seanriddle.com/um348x/
The following notes taken during my analysis of the mask ROM array structure:
Overall view of the UM3481A chips internals with metal removed. Mask ROMs highlighted.
Zoomed structure view of the main internal mask rom, metal removed.
To obtain usable data, individual bits from each bank are grouped together to form 7-bit words.
The resulting data consist of 8 banks of 7 bit words containing the melody information. The pattern 100011 marks the beginning of the data space, this pattern is also used at the very end in any unused remaining space. Perhaps it resets the oscillators to not generate sound.
The word pattern 000011 signals the end of every melody. The UM3481A chip contains 8 songs, therefore we find 8x 000011 words in the data.
The eight melodies occupy the corresponding space according to their length, melody #6 is the shorter of them all by far.
These lengths are also observable in the WAV, the recording matches all observed melodies in the mask rom.
Most notes seem to take just one word, here's an example from the first UM3481A melody "Jingle Bells", in the song three notes repeat almost at the very beginning, this is clearly visible in the data itself.
Longer duration notes seem to take two words, in the following example also from the beginning of "Jingle Bells", two longer duration notes are played after the initial repeating single notes.
That's all from my observation notes, I hope the better understanding of the data structure helps the emulation of this family of sound generating devices.
The FD1089 module variants from Hitachi / SEGA were fabricated in a plastic case, on the back of the module a epoxy layer is visible together with two rows of pins arranged as DIP64. This arrangement mimics a standard 68000 CPU as intended by SEGA.
The sample shown below features the SEGA code 317-0013, this indicates the module was used as the CPU in Enduro Racer.
As is, the module measures 8.72cm by 2.85cm, not a small piece for a cpu.
The plastic case features a top cover providing access to the battery container, as shown below as many as three batteries could be fitted at once, perhaps allowing for longer data retention configurations. All modules I've seen myself just make use of one battery slot only.
Hitachi seemed to keep most things home by employing Hitachi Maxell CR-2032 3 volt batteries, this one was dated 1986 week 06, that is long enough!
For some reason a white looking dust was found inside all over the battery container, perhaps a battery byproduct over the years?
Lastly, there's a big letter B stamped inside, this corresponds with the specific module type under review: FD1089B. Modules A stamped inside correspond to FD1089A variants.
Time to wear our x-ray glasses so we can see what it looks like inside the module, this is usually one of the most fun parts of a project. For the first time you have a look inside and start making sense of the internals, this one is busy inside and is no standard IC on a package. It's time to start planning an attack.
Easy options first, in order to try gain clean access inside we test the epoxy with a strong paint remover for a couple of days.
Unfortunately this didn't produce any significant results, the epoxy in the FD1089 seems to be well formulated and is resistant to this type of attack.
One step forward, by employing a combination of heat and patience it is possible to separate the plastic case from the epoxy block.
A weak spot on the epoxy curing caused by trapped air is discovered, this allowed for a unique early view of the shinny internal pcb as shown below.
Time to keep going deeper and figure out how to attack the epoxy, first we take a few extra measurements to understand dimensions, we will need them later on. The epoxy block on its own does 8.49cm by 2.60cm.
With the help of higher resolution x-rays we start to get full control of what's going on in there, eg enumeration and type of devices, guessing the purpose of each device, producing early diagrams, etc...
As shown below, the FD1089 uses a total of four different chips in its operation, left to right: A custom IC (this is where the security magic happens), a 68000 CPU, a 6264 SRAM chip, and a MB3771 voltage monitor (when needed, this takes care of switching power from VCC to VBATT and vice-versa).
The first three chips are HITACHI bare dies directly glued onto the top of the pcb, the MB3771 is in full form as a surface mounted device soldered to the back of the pcb.
Next, we need to have a fully validated understanding of the internal interconnect (aka, how things connect to each other and the outside)
The tool of preference for this project was a precision CNC, nothing fancy, just a standard 1610 model kit from Aliexpress. The goal here is precise enough milling so we reveal the circuit while avoiding fatal damage to the interconnect and chips, especially the custom IC.
Hours later, a significant part of the the copper surface is fully exposed and the custom IC preserved in place, though this didn't come without surprises and a couple of drill bits broken in the process. The reason: just below the IC you can see another unexposed rectangle area, this happened to be a ceramic insert placed there during fabrication to protect access to a sensitive area of the module
At this point a further donor module got dremel down bare in order to understand how many more ceramic inserts were there, in total the FD1089 module has four ceramic inserts, two on the front and two on the back shown below in white, they protect key SRAM signals such as the data bus to prevent extraction of the encryption key through direct drilling.
By now I was convinced these modules were probably very expensive to produce back in the day.
For illustration purposes lets take the following example shown below: Four data bit vias from the SRAM are covered with ceramic, both at the origin near the SRAM as well as at the destination close to the custom IC.
Beyond being cool and almost a hand craft, I'm not sure why this was done, if an attacker could drill from above with the intention of making contact, it could do so at other part of the circuit by exposing the copper, you don't require a via specifically for that.
More hours of careful drilling allow us to gain access to further key areas of the pcb, when this process is fully completed it will allow us to work on the next part of the reverse engineering.
Diagraming the interconnect and how chips relate to each other is a critical step before we proceed to examine any deep down chip logic.
Once we are ready to explore the custom chip logic, a further donor unit is put under the knife, the goal here is to cut down the minimum possible sample to allow extraction of the custom IC die housed inside.
Bonus: a cross section of the pcb reveals its four layers, the inner two are dedicated to VCC / GND distribution only.
The smallest possible bite size sandwich is produced, inside not visible is the custom IC, on the left you can see a couple of the white ceramic inserts guarding access.
The sample was then put in a nitric acid beauty spa for several rounds.
At each round the sample was inspected for progress and cleaning. IC dies are very fragile so patience is key, this one waited 30+ years, so it may as well wait another day if necessary.
Finally getting closer, both ceramic bits are released, we can already see the back of the IC die now exposed.
A final round allowed for the chip die to be fully exposed clean. Pharaonic honors please.
Stay tuned for the next article, we will explore this chip under the microscope. Happy reversing.
