Cube64-DX - Ameliored GC controller to N64 adapter.

[Nully]

Oh cool, thanks.

[DarthCloud]

Btw I did another small update.

When doing the button mapping layout reset combo keys (AL+AR+Z), you now only erase the currently active button layout instead of erasing everything.

[DarthCloud]

First official release!! Build r37

Precompiled .hex available in download section and pdf schematic too:
http://code.google.com/p/cube64-dx/downloads/list

Also this is the last build that will support PIC16F84A, 12F629 and 12F675. Precompiled .hex will stay available in the download section however. I flaged then "Deprecated" so you need to search in the download section in "All downloads" to find then.

Their is no reason to use those PICs anyway, maybe only if you have some of those chip already and want to try out while waiting to receive your 12F683 ;)

I also made a tutorial like promised to help people building it:

http://code.google.com/p/cube64-dx/wiki/BuildingAdapter

This is targeted to beginner.

If you have good skill already you can skip this and build a nice pcb with SMD component :)

[APE]

A few months ago I grabbed the interface for the microphone used in Hey You Pikachu! With no microphone and no idea what I was going to use it for.

Now I have the PERFECT box for this.

[Annoying_one]

that would make a nice case to house it in. and look rather pro

[Nully]

Wow you built the whole adapter on top of a DIP socket. That's genius and cost effective. Awesome.

[Nully]

I finally ordered all my parts to build this. This is my first venture into microcontrollers and I want to thank you for finally motivating me to do it. I'll report back with results in a week.

[derekb]

wow this is some straight up McGuyver adapter

[DarthCloud]

wow this is some straight up McGuyver adapter

At some point I had a better looking adapter but I dismantled it since I needed the GC/N64 cable to build some sniffing interface for my LA.

Using a dip socket make development easier and quick.

What it could look like with pcb and smd component (This included the charge pump circuit to provide +5v to the rumble motor):

[derekb]

the charge pump would certainly be nice, is it complicated to add?

[DarthCloud]

the charge pump would certainly be nice, is it complicated to add?

Not really if you are ok with SMD soldering, the chip used is only available in SMD package. You need 3 additional capacitors and one more resistor for the charge pump circuit.

[Nully]

So the 5v is just for the rumble motor? It's not needed at all for a basic controller?

[DarthCloud]

So the 5v is just for the rumble motor? It's not needed at all for a basic controller?

Correct.

[link83]

Thanks for your continued work DarthCloud :-)

Another update :D, I improved the gc to n64 axis conversion function. But I'm still unsure if I'm going to keep it or not, I guess I will wait to get some feedback once some peoples build their adapter to make my mind about it.

Well i'm glad the axis conversion is still there :nod: If you will be upgrading to a 64MHz PIC at some point I guess the CPU and flash usage issue will become redundant?

Also, maybe this is just me, but I have been looking at the schematic and it doesnt seem clear to me where capacitors C3, C4 and C5 fit in? Especially since C3 and C4 both connect to +3.3V.

Has anyone tried making an eagle board/schematic for the Cube64-DX? I am tempted to have a go, but theres no point if someones already done it.

Oh and slightly OT but could I just ask something which has been bugging me for a while - what is the maximum possible size for the N64 controller pak's SRAM? I previously thought it was 32 kilobytes, since that is the size of the SRAM chip used in official controller pak's, and I read in the 'n64-observations' doc from the original Cube64 here:-
http://svn.navi.cx/misc/trunk/wasabi...4-observations

Note 1:
With 15 address lines, that gives standard memory paks a 32k capacity.
The Nyko Hyperpak tested has 128k of SRAM, but includes a switch that
selects between four 32k banks.

However, in the same doc it also says there are 16 address lines labelled A0-A15? :confused:

[DarthCloud]

Thanks for your continued work DarthCloud :-)

Well im glad the axis conversion is still there :nod: If you will be upgrading to a 64MHz PIC at some point I guess the CPU and flash usage issue will become redundant?

I found some problem regarding Wavebird controller. The joystick look to be rotated to the right more than regular controller. Didn't made any precise measurement but if I look a both pcb side by side, The joystick indeed look more rotated. I did some measurement with my LA however for example the center idle horizontal position is always near ~0x80 but if you push the stick straight up in the "alignment hole" (Don't know how to call it) I read a value -8 below the center, but I should be in the center in the X axis!!

This create the problem that in adventure game when you want to run straight up and put the joystick at what you believe the center you will turn a little bit to the left.

I played with the Wavebird with some VC N64 game on my Wii and I noticed a little dead-zone around the center, you push the joystick for about 3/4 of millimeter nothing happen. I believe this dead-zone is to avoid problem with misaligned controller.

I implemented a dead-zone in the adapter and this fixed that problem.

Doing so I modified the possible range of the Gamecube controller from [-0x64, 0x64] to [-0x5A, 0x5A] this still an higher range than n64 range [-0x54, 0x54] but it's damn close. I modified the axis range function so it fit to new spec but it don't really make any difference now that the range is so close to the real one. So I removed that. The code exist so I can put it back at some point if I receive some feadback about that.

But honestly I think it's better without that function. It will fill more sensitive but it's more caused by the fact the GC joystick is better than because the range is a bit higher.

Also, maybe this is just me, but I have been looking at the schematic and it doesnt seem clear to me where capacitors C3, C4 and C5 fit in? Especially since C3 and C4 both connect to +3.3V.

Capacitors in middle of nowhere is a schematic are always decoupling caps. You need to put then very close to the power and ground pin of ICs. So C3 is a regular 0.1µF caps so it go between the power pin of the micro-controller.

If you check the MAX682 datasheet you can see that C4 go between +3.3v input and gnd of the IC, and C5 go on the +5v output and gnd of the charge pump.

Has anyone tried making an eagle board/schematic for the Cube64-DX? I am tempted to have a go, but theres no point if someones already done it.

I'm planning to do one once I implement the N64 controller input for peripheral (Got my sample SPDT switch for that today btw)

Oh and slightly OT but could I just ask something which has been bugging me for a while - what is the maximum possible size for the N64 controller pak's SRAM? I previously thought it was 32 kilobytes, since that is the size of the SRAM chip used in official controller pak's, and I read in the 'n64-observations' doc from the original Cube64 here:-
http://svn.navi.cx/misc/trunk/wasabi...s/cube64/notes

However, in the same doc it also says there are 16 address lines labelled A0-A15? :confused:

The mazimum adressable memory is 32k. Bigger memory card use the old trick of switching between memory bank with a switch. So It's like if you had 4 independant memory card into one single package. You select between then with a physical switch.

[link83]

I found some problem regarding Wavebird controller. The joystick look to be rotated to the right more than regular controller. Didn't made any precise measurement but if I look a both pcb side by side, The joystick indeed look more rotated. I did some measurement with my LA however for example the center idle horizontal position is always near ~0x80 but if you push the stick straight up in the "alignment hole" (Don't know how to call it) I read a value -8 below the center, but I should be in the center in the X axis!!

How strange...I wonder if this applies to all Wavebird's? Might this just be your Wavebird? Just seems like an odd thing for Nintendo's designers not to notice.

180 steps is quite close to 168 steps, but doesnt this mean that the GC analog stick area mapped to N64 analog stick is now 10% smaller? So when you press full in one direction it has already reached the maximum value earlier.

Capacitors in middle of nowhere is a schematic are always decoupling caps.

I see, thank you for the info - I learn something new every day!

If you check the MAX682 datasheet you can see that C4 go between +3.3v input and gnd of the IC, and C5 go on the +5v output and gnd of the charge pump.

I am curious how much extra drain does MAX682 and wired GC Controller with rumble on max setting put on N64's +3.3V power rail?

I'm planning to do one once I implement the N64 controller input for peripheral (Got my sample SPDT switch for that today btw)

Thats great, I look forward to seeing it.

The mazimum adressable memory is 32k. Bigger memory card use the old trick of switching between memory bank with a switch. So It's like if you had 4 independant memory card into one single package. You select between then with a physical switch.

Hmm, maybe I am missing something, but if the biggest addressable memory is 32k then why are there 16 address lines which allows up to 64k?

[DarthCloud]

How strange...I wonder if this applies to all Wavebird's? Might this just be your Wavebird? Just seems like an odd thing for Nintendo's designers not to notice.

The fact they used a deadzone suggest that they indeed noticed! I have two Wavebird like this, one gray which is probably quite old, and my personal platinium one which a buy brand new some month after the Wii release, and I didn't used it much.

I'm currently watching an auction on ebay for 3 Wavebird without receiver, so they are cheap. I will try to get them so I have a better sampling for testing.

180 steps is quite close to 168 steps, but doesnt this mean that the GC analog stick area mapped to N64 analog stick is now 10% smaller? So when you press full in one direction it has already reached the maximum value earlier.

No it reach the maximum value later in comparison of the direct value mapping (But a bit sooner in comparison of the adapted value). The first 10 value in either direction are always mapped to 0. and value higher than 10 are subtract by 10. So you have 0 for the first 10 value them the joystick value start to increase from 1 to 0x5A

Hmm, maybe I am missing something, but if the biggest addressable memory is 32k then why are there 16 address lines which allows up to 64k?

Good point! I must admit I have no idea why it's only 32k if their is 16 address line. Micah in is doc talk about 15 address line but it's controller pak pinout suggest their is 16. :O

[DarthCloud]

Checked again both controller pcb and maybe their are not misaligned, maybe it just en illusion created by the mark on pcb which are present on the wavebird and not printed on my regular controller. Very hard to tell.

But still from my reading the idle center don't match the value you got at either the top or bottom position.

Wavebird:


Regular:

[link83]

No it reach the maximum value later in comparison of the direct value mapping (But a bit sooner in comparison of the adapted value). The first 10 value in either direction are always mapped to 0. and value higher than 10 are subtract by 10. So you have 0 for the first 10 value them the joystick value start to increase from 1 to 0x5A

I see, so the GameCube analog stick gives readings for 208 steps, and you added a dead-zone of 10 steps on either side bringing the total to 188 steps, and then you minus off another 10 from the reading each side to give 168 steps - is that right?

Good point! I must admit I have no idea why it's only 32k if their is 16 address line. Micah in is doc talk about 15 address line but it's controller pak pinout suggest their is 16. :O

I decided to look into this further, and I think A15 might only be for the Rumble Pak? I found some info in the Controller Pak patent:-
http://www.google.com/patents?id=GW0...bstract&zoom=4
Page 23

Page 24

Also the patent seems to suggest that Nintendo had planned to make a combined Controller Pak/Rumble Pak at some point?

[DarthCloud]

I see, so the GameCube analog stick gives readings for 208 steps, and you added a dead-zone of 10 steps on either side bringing the total to 188 steps, and then you minus off another 10 from the reading each side to give 168 steps - is that right?

The GC range is more close to 200 actually after making an average of all my controller value. I only remove the 10 on either side of the middle so it reduce to range to 180. I leave it like that. Nothing is done to adapt it to 168.

Anyway while N64 range is -0x54 to 0x54, from my iQue controller (which use very same hardware) the range I get is -0x58 to 0x58. I believe it's because the joystick hole is not high enough so the stick can go a bit further.

I played a lot last night to test the latest r44 build (Now available in the project download page) and I'm very happy by how it's performing.

I decided to look into this further, and I think A15 might only be for the Rumble Pak? I found some info in the Controller Pak patent:-
http://www.google.com/patents?id=GW0...bstract&zoom=4
Page 23

Page 24

Also the patent seems to suggest that Nintendo had planned to make a combined Controller Pak/Rumble Pak at some point?

N64 issue a write to address 0x8001 to identify what the controller slot contain. So A0-A14 is for controller pak memory, and A15 is to access a virtual memory area for device "management".

- Memory map

0x0000 - 0x7FFF : SRAM, on the memory pak

0x8000 : Identification/initialization.
Write all 0xFEs to it for initialization, then
read back 0x00 on the memory pak or 0x80 on the rumble pak.

0xC000 : Motor control latch (low bit) for the rumble pak