Back in February, Nintendo of America decided to release the *New* Nintendo 3DS following its earlier releases in Europe and Japan. The *New* 3DS is a refresh of the 3DS system, with upgrades such as a quad-core processor, C-stick nub, dual rear triggers (on each side), and face tracking 3D. As with the original 3DS, there are two variants of this system: a pocketable version and a slightly-less pocketable version with larger screens. The smaller model also provides new opportunities in terms of customizability, with removable front and back cover plates that make it possible to change the top and bottom appearance of the console in seconds. There are 59 official cover plates as of writing, and they are pretty neat. The *New* XL, unfortunately, is the same as the old consoles in terms of personalization– there is not much you can do non-destructively besides putting on decals.
Being a fan of good-looking video game consoles, I naturally wanted one of the smaller *New* 3DSes.
The problem? Nintendo of America decided to release only the XL version of the new console, with reports citing low demand for the original small model as the reason. The difficulty of setting up the supply chain for all the different cover plates was probably another cost adder with very low potential returns. There is certainly a demand for consoles with a bit of personality– there was the drama that unfolded over the limited stock of Majora’s Mask *New* 3DSes, where many people were left frustrated with the flawed pre-order system (why bother pre-ordering when the stock that the retailer acquires is determined beforehand?). Nintendo seemed very adamant that the smaller *New* 3DS was not coming to America, and I was not interested in having another 3DS that could not be personalized. I couldn’t just purchase one from Japan or Europe to play my current selection of 3DS games either. 3DS consoles are region locked and cannot play games that identify with other regions.
After some thought, I decided that I could probably just assemble one myself. It certainly didn’t make much sense to assemble one since I already had a perfectly fine 3DS XL, but It was a challenge I felt comfortable taking on.
For the sake of cost-saving and 100% cross-compatibility between the XL and non-XL model, Nintendo must have maintained many constants between the two models. Of course form factor would change, but it didn’t make sense to use different schematics or software codebases for each model. Let’s take a look at both motherboards, side by side:
We can see they are very similar in terms of the general layout of the components. Assuming that the only difference between the boards is the software loaded onto them, we narrow down our parts of interest to the eMMC (embedded MultiMediaCard) chip and the Nintendo processor. Doing some research on the 3DS architecture over at 3dbrew.org, we find that these two chips are tied to each other by the AES (Advanced Encryption Standard) Engine in the Nintendo processor. The eMMC is encrypted with a key that is stored in the processor, and can only be decrypted by a processor with that specific key.
So if my assumptions are correct, the creation of a US Non-XL *New* Nintendo 3DS can be done by taking the processor and eMMC from a US *New* Nintendo 3DS XL (sheesh I hate typing that) and installing it into a Non-XL unit from another region. Those parts can then be used to create a Non-US XL in return.
I flipped some coins, drank some beer, thought about whether I really wanted to do this risky project, and then proceeded to order a Black Japanese *New* 3DS and a pulled US *New* 3DS XL motherboard from eBay. I could have then proceeded to give both system motherboards to a professional rework service to perform the operation. In hindsight that may have been the right thing to do, to save some sanity over a bit of money. But hey, I like to try everything. I solder small surface mount components all the time and QFNs are a cinch when I have the right tools. How hard could two BGAs be?
It turns out is is HARD. Very HARD. Do not try this if you are not a professional or do not have a proper rework station. Let’s begin.
Step 1: Remove chips from both boards. I’ll start with the eMMC.
Ugh, What is this yellow crap around the chip? I hate it. It’s underneath the damn thing too. It’s not quite the silicone I am used to but also not quite epoxy. Let’s call it glue. Luckily, the eMMC on the US version doesn’t have it. Something also different about the eMMC on the US version is that it is a Toshiba chip instead of a Samsung. The package shape of the chip is different as well. I confirm that both chips have the same pinout (FBGA-153) and proceed. Nothing surprising about having alternate parts from different vendors. From iFixit’s teardown of the XL, I expected a Samsung chip. Maybe they tore down a Japanese model.
I start off by using a hobby knife to cut at the glue on the Japan board. The final removal of the chip depends on the weakening of the glue underneath the chip by heat. I decided to remove the glue around the edges to make that final liftoff removal step as easy as possible. Not sure if it was completely necessary, but I wanted as much freedom as possible for that chip when it was ready to come off.
Using a hobby knife for this job sucks. It works, but it sucks. I damaged the board by pushing down a little too hard in a few places, but it was mostly cosmetic solder mask damage. I switched to a hot soldering iron and got way better results. I was able to essentially plow the glue away from the board using the iron. I made sure to work slowly and carefully to avoid desoldering the tiny resistors and capacitors that are hidden under some areas of glue.
After doing a satisfactory job, it was time to desolder the chip. Easy enough with the right tools. Don’t try to do this with just a heat gun– you’re at high risk of damaging the chip or the PWB (printed wiring board) if you do so. A copper plane spreads across the entire board and absorbs a lot of heat. A lot more heat on the chip is thus required before the solder underneath it melts. Even if you do get it hot enough, you have a small window of time to remove the chip before the solder cools and becomes solid again. A board pre-heater is definitely required for this job. I used two in the process of reworking these boards, a Zephytronics ZT-1 and an Aoyue 866. For heat guns I used a Metcal HCT-900 and the Aoyue 866. I bounced back and forth between the lab at work (after hours) and my home lab. I have the Aoyue at home and like it a lot more that the Zephytronics and Metcal setup.
Preheater to ~100C, Heat gun to ~300 C. Preheat the board for a minute before proceeding to attack the chip. To be extra careful, you can place foil tape over the neighboring components of the board to make sure they don’t get too hot. The eMMC should come off in less than 30 seconds with light nudging. If it doesn’t come off, let the chip cool, increase preheater or heat-gun temperature, then retry. I won’t try to give exact temperatures because everyone’s setup will be different.
After the chip was off, I proceeded to clean the area. The glue was removed with the soldering iron method, and the lead-free solder on the pads was removed with a wick. Make sure the board is being heated from below when removing solder from the pads. The ground plane will otherwise absorb all the heat and make it hard to drag the wick along without ripping off some pads.
A little bit of acetone, and we have a beautiful clean landing pad. I removed the eMMC chip from the US board and then proceeded to prepare that to be soldered. You can see how nice the procedure can be when there isn’t glue involved.
At this point, I decided to see what would happen if I tried to turn on the system with no eMMC installed. I reassembled the whole unit, powered it on, and was presented with the following error screen.
Cryptic. Certainly looks like it’s reporting that it’s reading nothing. The important things that this error screen does tell me are that there is some software loaded on to the processor that will provide information about whether or not the eMMC is installed correctly, and that if the processor is installed correctly, the system will boot and display video output onto the screens. I can assume that if the data, clock, power, and ground connections of the eMMC are soldered properly, I will get something other than all 0’s or all F’s on this error screen. This assumption convinced me that it was safe to continue.
Now we move on to the nightmare of balls. So many balls.
To prepare the US eMMC to be soldered onto the JPN board, I had to reball the BGA package. I decided to go the route of soldering preformed solder balls to the chip to maintain good uniformity in ball size. The package drawing for an FBGA-153 specifies .3 mm diameter pads, so I went and bought a tube of .3mm diameter leaded solder balls off eBay. So I can just tack the balls onto the pads with flux and then heat up the chip to attach the balls, right?
No, don’t do this without a stencil. You will hate it. It was a terrible idea for me to even try.
The problem with this method is that once you heat up and start activating the flux, stuff moves around. The airflow from the heat gun doesn’t help either. All the balls that you spent time placing one by one will shift and stick to each other and you will be sad. Some will solder onto the pad just the way you want them to. Some will not. And you will be sad trying to remove the messed up balls without disturbing the perfect ones.
I gave up and picked up a Samsung Galaxy phone stencil card that had the package I needed. I couldn’t find a singular stencil with that package on eBay for domestic delivery.
Using a stencil simplified the process greatly. I fluxed the chip and aligned the solder pads with the stencil openings. The stickiness of the flux helped make alignment easy. I then taped the chip onto the stencil using kapton tape to ensure immobility. All I had to do to align the solder balls with the BGA pads was to pour them over the stencil openings. The stencil and the flux did the rest of the work.
To align the eMMC onto the board for soldering, I used a microscope imager connected to a PC. There was no part outline silk screened onto the board, so I was on my own. I found a piece of software that allowed me to overlay a crosshair over the video feed to use as a placement aid. On the chip, I used a pen to draw an X using the corners as guides. The lines intersected on a letter ‘A’ which I used as the center reference. I first lined up the camera with the center of the landing pad and taped the circuit board into place. I then placed the chip onto the landing pad using the crosshairs to line up the chip center with the landing pad center. The rotation alignment was eyeballed. I figured I could get it close enough where the flux and the process of solder melting would auto-correct any misalignments.
A board preheat and a soldering operation later (with similar temperatures as before), I have the eMMC installed. What does the board do now when I try to boot it up? There was a chance that it could boot straight into US software.
Nope, just another error message. This error message is different though! I took the different error code and the non-FFFFFF messages as a sign that there was some communication happening between the processor and the flash.
I moved on to removing the processors on both boards. I followed the same procedures as before to remove the glue around the chip and to desolder the chip. The chips came off with a little more nudging than the eMMC required. The USA processor was unfortunately glued down like the JPN one was.
Nintendo processor markings.
Now, how do I reball this monster? It is a 512 pin BGA, .4mm pitch. Learning my lesson from before, I decided I needed the help of a stencil. With no stencil I could buy off eBay for this pin pattern, I attempted to make use of a small .4mm pitch one off of the Galaxy stencil card. Since it was so small, I had to reball the chip section by section. I placed a couple balls one-by-one, reflowed them, then repeated.
I knew this was going to be bad for the chip. All these thermal cycles were stressing the joints and could break the chip at any moment. I persevered for a while until I took a step back and realized there were 512 damn balls on this thing and I was putting these balls on at a rate of about 40 per hour. It wasn’t perfect, either. The balls had a tendency to get stuck on the stencil and would leave me with gaps in the ball pattern when I removed the stencil. Fixing these one or two balls required another thermal cycle and pushed the limits of the chip further and further.
I needed to get a proper stencil. I gave up my pride and called up a local board assembly house nearby that I had some connections with through work. I made some measurements with calipers, drafted up a drawing of the package pad layout, and sent it to them to get a quote for a stencil. They offered to solder the chip onto the board for a little more and I agreed. I had enough headache with this thing already. I asked if they could do additional x-ray inspection and they added that on for no extra cost. This is when we break away from the DIY spirit of things, but as you will see it was probably the right decision.
The additional X-ray inspection brought to light the difficulties of soldering very fine pitch BGAs. I would say that .5 mm pitch and up can be done at home when the proper care is taken. Once you get smaller than that, then stuff like this can happen (even with the help of machine placement).
The first time they performed the rework, the X-ray inspection discovered a couple bridged pins. Looking at the corresponding pads on the PWB, I could tell that the pins did not already share copper and were not meant to be shorted. I did not approve the rework. They asked if it was okay to try again, taking into account the number of reflow cycles experienced by the chip. Since I had told them to use a lower temperature leaded soldering process, I was hoping that any damage caused by these reflow cycles would be minimal on these designed-for-RoHS-process chips. Approving another attempt meant that the chip would go through another 3 reflow cycles to be removed, re-balled, and resoldered. I didn’t have any other options at this point, and approved the second attempt. It turned out to be very successful.
I took both the board and the stencil home and reassembled the 3DS. Fingers crossed.
Everything worked out as hoped. The 3DS operates as you would expect, and there does not seem to be any glitches with the different hardware. I’ve played a couple games with it and have had no problems so far. System updates work, as well as all the online features and the eShop. To provide some comfort to those who wish to perform this mod, I’ve created a video showing the unit side-by-side with a first-generation XL and a US region game. I’d like to believe this is the only US non-XL *New* 3DS in existence, but I don’t know that for sure. EDIT: It looks like it isn’t, from what I’ve seen in comments on this post. It might be the first that was created this way, but there are some software modders who’ve done some outstanding work to understand the eMMC contents and the system transfer protocol to create their own non-XL 3DS systems. From what I understand, they have been able to get everything working normally except for eShop. Not sure If I would have went with this method if I had seen those posts when I was just starting this project.
For those who want to see it work.
All in all this wasn’t a very complicated mod. It is technically difficult, but conceptually simple. It is certainly not a beginner’s soldering project, and I would only recommend those who are professionals to attempt this mod. I do believe that I have a finite chance of performing the mod in my home given that I have both of the proper stencils now, but I’d need a good incentive to try it again. Without an X-Ray machine, hitting that power button is a risk. Though I did solder on the eMMC properly, it was a 0.5mm pitch part and was more forgiving.
Feel free to use my package drawing to order your own stencils if you’d like to do the same mod. I hope this helps some other Nintendo fans out there. Unlike region unlocking modifications, I doubt this mod can be detected by or patched out by Nintendo.