Donkey Kong by Nintendo PCB #1

Updated 19/11/24

I've been asked to test a couple of Nintendo Donkey Kong Arcade PCB sets with a view to repairing any faults and confirming which of the Donkey Kong ROMsets are fitted, possibly changing to a preferred software revision. Both have been stored for some time but were reportedly running prior to that, one apparently had an issue with the positioning of some of the sprite details on screen.

The one on the bench at the moment is destined for an original cocktail machine and is the TKG3 type, 4 board PCB set. I'm no expert on this particular hardware which has been extensively documented elsewhere so will refer to the existing knowledge base at times but hopefully add my own observations and interpretation of their workings as I progress.

The first step as always is to set the PCB up on the test bench with suitable connections, here I can once again adapt my Nintendo Space Fever test setup which has already been updated to include variations for Space Firebird and Space Demon PCB sets. Although some of the connections are compatible with Space Firebird pinouts I'll need to make an additional cable with 10 pin and 7 pin power connections for the CLK and Sound PCBs respectively.

Having done that I can make all the connections on the bench for power to the PCBs, Video out (via an RGB inverting amplifier to cope with Nintendo's 'negative' video polarity) and low level audio to my test bench RGB video monitor. To 'coin up' the game I can just use a small wire link to ground the input pin on the CPU board coin connector and for control I have an adapter lead to connect my Quickshot (Atari / C64 style) joystick to either the main or sub control input connector on the CPU PCB.

Donkey Kong by Nintendo PCB #1

Powering on, there is a brief sound and the screen comes to life with a jumble of characters then the attract mode commences. Initially apparent are 3 issues, firstly the screen is predominantly blue with an elevated blue background level as well as washed out, overly intense blue characters on screen.

Additionally, some of the colours within the character sprites appear to be missing such as the middle of the Kong body, Mario face / skin tone details and the part of the hammer graphic. After a little warm up some of the objects begin to appear glitchy and intermittent most noticeably the barrels and flame sprites.

Donkey Kong by Nintendo PCB #1

Looking into the most obvious symptom first, the increased blue levels turn out to be the result of a deliberate design revision rather than a fault condition. Referring to the video output section of the circuit diagram the three colour channels appeared to be the same originally but the blue channel has a number of components which have been omitted or bypassed.

I haven't seen any other instance where a single colour channel has been altered to produce unequal signal levels but assume it was to give the intended Sanyo monitors a bit more adjustment range in the blue channel. In an arcade monitor the three colours are pretty independant, each with separate gain and lowlight adjustment so could cope with signals having unequal levels.

Donkey Kong by Nintendo PCB #1

For my test bench TV monitor adjusting the colour balance would involve using the service menu to change factory settings so instead I'm temporarily adding a 330 Ohm series resistor to the 68 Ohm blue output resistor R24 just to balance up the 3 colour signal levels - close enough for the test bench, anyway.

That done, the colours do look better but some of the colour areas within the sprites are still missing or incorrect and some of the sprite details still intermittent. Checking the circuit diagram, graphics information arrives on the CPU PCB from the video PCB, is latched into 2 x 74LS174 ICs in positions 2G and 2H and output via a pair of 256 x 4bit palette PROMs which determine the final output colours.

A problem with the colour palette could cause specific colours to be missing or incorrect so that would seem a good place to start. I don't have the means to program a replacement PROM but can read and verify the existing PROMS against known files using my old EPROM programmer with an adapter which I made to convert the PROM pinout to read as a 2716.

Donkey Kong by Nintendo PCB #1

The adapter which I originally made for this purpose was on a small piece of veroboard, using a standard 16 pin IC socket for the PROM. I've recently made a PCB version with a ZIF socket which is preferable as the legs of the PROM ICs are notoriously fragile. Anyway, after reading the contents of the two palette PROMs in positions 2J and 2K on the CPU PCB both verify correctly against known files from the Donkey Kong ROMsets.

Before going too much further I'd also like to read and verify the program EPROMs on the CPU PCB and check the program and video RAM for errors. So the next hurdle will be reading the 4 x 2532 EPROMs on the CPU PCB which are not pin compatible with the more common 2732 IC.

Checking the data sheets for the 2532 EPROMs in use and the common 2732 type the majority of pins have the same functions and the programming sequence is similar for both, the only differences being the assignments of pins 18, 20 and 21. Those comprise the highest order address line A11, Vpp programming Voltage and the IC enable / programming pulse input (in different order for the 2 IC types).

So it should be possible to reprogram the 2532 EPROMs if required by making up another adapter to reorder those 3 signals. One other difference I noted between the 2532 and 2732 in read mode; the 2732 datasheet advises Vpp should be at logic Low i.e. 0Volts but Vpp on the 2532 should be held at +5V in read mode. So to use the same adapter to read and verify the 2532 EPROMs an additional switch will be required to select either read or programming mode.

Donkey Kong by Nintendo PCB #1

To try it out I've made up the adapter, using veroboard again and successfully read the 4 CPU EPROMs which verify correctly as Donkey Kong, US set 1. I haven't tried programming any 2532 EPROMs as yet but that may be a task for later. The remaining EPROMs on the other PCBs within the set all appear to be more common 2716 types which my older EPROM programmer already supports.

16/11/24

I have a Nintendo Donkey Kong PCB set on the bench which runs with some minor graphics issues. So far I've checked the CPU program EPROMs as well as the palette PROM ICs and verified all are working, with data matching known ROMsets. I can verify the remaining EPROMs using my programmer to read them but as none of the RAM ICs are socketed there is no easy way to remove them and test them out of circuit.

Donkey Kong by Nintendo PCB #1

To assist with a number of previous repairs I wrote some 'homebrew' test programs to run on various game PCBs, checking the RAM in circuit from the CPUs point of view. I've since ported the original programs to run on other game PCBs which employ similar hardware and think the program which I originally wrote to test some Exerion PCBs could be adapted to work on the Donkey Kong PCB set.

Both are Z80 based with separate RAM banks for background video / text and sprites as well as program RAM so with minor changes to the memory locations the program should run on the Donkey Kong Hardware. The video RAM is checked first which causes the jumble of characters on screen then once cleared the test results are placed into screen memory. For Exerion the text characters are Ascii based but the characters for Donkey Kong use different encoding so I've nicknamed it Donscii.

Unlike Ascii, codes 00 to 09 represent numeral characters 0 to 9, a blank space is represented by decimal 16 (10 hex) and upper case letters A to Z begin at code 11 hex. I haven't checked the remaining codes as other characters are not required for the test display, many seem to be segments of the background 'scenery' in the game screens. After assembling the adapted program and debugging in MAME it's ready to run on the actual Donkey Kong PCB set.

I don't have any spare 2532 EPROMs to load the test program onto so, comparing the pinout of the 2532 to the 2716 EPROM I realise as the test program is less than 2k Bytes a 2716 will work in place of the original 4k Byte 2532. The 2716 has 2 select inputs, /CE on pin 18 and /OE on pin 20 whereas the 2532 is selected when the PD/PGM pin 20 is low. Pin 18 on the 2532 is address input A11 which will also be low for the first 2k Bytes so the 2716 will work without any modification required.

Donkey Kong by Nintendo PCB #1

And with the test program loaded onto a 2716 EPROM then placed in the boot ROM position 5F, powering on the tests run and each result is a pass. Even though that doesn't solve the missing colour problem it is good news as the memory components are becoming scarce and expensive. So the missing colour issues are more likely the result of a fault in the associated logic components or even possibly a bad interconnection between video and CPU PCBs etc.

19/11/24

I've been testing a Nintendo Donkey Kong PCB set on the bench which appears to be working apart from some incorrect colours and intermittent dropouts within the sprite graphics. So far I have verified the colour palette PROMs and CPU board EPROMs as well as running a RAM test on the main, video and sprite RAM banks with no errors found. The next step will be to rearrange the board set on the test bench to allow access to the video and clock PCBs and do some signal tracing within.

This TKG3 board set is evolved from earlier Nintendo games in particular Radarscope and being a 4 PCB stack can be difficult to access all areas for fault tracing. The CPU and sound PCBs are affixed to one side of a central mounting / shield plate with the video and clock PCBs on the opposite side and interconnection by ribbon cables as well as separate power connections to each PCB.

Donkey Kong by Nintendo PCB #1

I've deliberately made my bench test cabling quite long to allow the PCBs to be removed from the mounting plate and spread out on the bench, by pivoting the video and clock PCBs through 180 degrees these can be positioned next to the CPU and sound PCBs. At this point, a short power interconnect cable between the video and clock PCBs prevents these two boards from being spread apart further.

To access the video PCB I could either make up a longer interconnect cable or just pivot the clock PCB a full 360 degrees so that it sits below the video PCB then reconnect the short link cable, also placing some non conductive bubble wrap between the two PCBs to prevent shorts. But first let's take a closer look at the clock PCB.

Donkey Kong by Nintendo PCB #1

The clock PCB contains the main crystal oscillator and divider chains which provide the video sync and other timing signals, there is also a bank of 4 x 2716 EPROMs on board. These contain the 'gfx2' data which I suspect may provide some of the 'cut scene' animations while the 'gfx1' EPROMs on the video PCB may handle more interactive details such as in-game, player character etc.

Removing each EPROM in turn to check their contents, all verify correctly against the known Donkey Kong ROM file checksums.

An unusual feature of this PCB is the very high oscillator frequency 61.44 MHz and use of ECL logic components in the initial divider stages. I haven't come across this component type in other arcade PCBs to date and imagine by 1981 they would have already been largely obsolete apart from applications which demanded very fast response times.

There is also a pair of ECL RAM ICs on board, type 10422 each having only 256 x 4 bit capacity but extremely fast access time of 10 nanoseconds or less whereas the common 2114 static RAM used on the CPU PCB would have access times in the hundreds of nanoseconds. The ECL RAM does not appear on the CPU memory map so is not directly accessable by the CPU, it appears to be addressed using a pair of binary counters so I'm guessing it may be used for fast scrolling of sprites.

An added complication when using both ECL and TTL logic within a circuit is the need to interface between the two types, each having different logic levels. A number of 10124 and 10125 ICs perform this function, interfacing from TTL to ECL and vice versa. The ECL components also require a -5V power supply, their Vcc pins being at 0V, circuit ground level.

Donkey Kong by Nintendo PCB #1

Looking more closely at the PCB I noticed one of the ceramic packaged, ECL ICs in position 5L was missing the top of its case. It had been covered with a label so had been that way for some time but I carefully removed the label to inspect the damage. Apart from the 'lid' being missing entirely there appeared to be some cracking of the adhesive cement which held the ceramic top as well as the IC legs in place.

Amazingly the IC was still working as none of the incredibly fine internal wires which join the wafer to the pins had broken. Nonetheless I think the best action is to replace the component as some old stocks of these ICs are still available from sellers overseas. This particular IC is a type 10136, high speed hexadecimal counter used in the initial clock frequency division.

To be continued...

Web Resources (External Links) -

2732 Datasheet - ALLDATASHEET.COM

HN462532 Datasheet - ALLDATASHEET.COM

Donkey Kong Memory Map - arcaderestoration.com

TTL/Bipolar PROM / RAM Reference Guide V2.f3 - retrotechnology.com

Pinout: Nintendo Radarscope/Donkey Kong/Donkey Kong Jr. - mikesarcade.com

Top of Page


All images and text on this website are Copyright.

Contact: jbtech at telstra dot com

Home