For those of us who have worked on SNES and GameCube controllers, we know that these are pretty simple to get into and maintain. However, in the trend of making modern game controllers more complex and less maintainable, Nintendo’s new Switch 2 Pro controller is giving modern Xbox and PlayStation controllers a run for their money in terms of repair complexity. As shown in a teardown by [VK] on YouTube (starting at nine minutes in), the first step is a disappointing removal of the glued-on front plate. After that you are dealing with thin plastic, the typical flimsy ribbon cables and a lot of screws.
The main controller IC on the primary PCB is an ARM-based MediaTek MT3689BCA Bluetooth SoC, which is also used in the Switch 2’s Joy-Cons. The 3.87 V, 1070 mAh Li-ion battery is connected to the PCB with a connector, but getting to it during a battery replacement might be a bit of a chore.
The analog sticks are Alps-branded and do not seem to match any other sticks currently on the market. These are disappointingly also still resistive potentiometer sticks, meaning they might have to be replaced before long due to stick drift. Reassembly has a few tricky parts, especially with the two sticks being not identical, yet easy to swap by accident. Which would require a second disassembly round.
There’s also a soft-touch coating on these controllers, which have been known to get… gunky after a few years, so time will tell what the lifespan is here. As is typical, these controllers also only work with the Switch and not with a PC or other consoles. Overall, it seems like a nice, silent controller, but the repairability seems low at best.
Reading about the coating is sad. Why arent manufacturers learning we hate that type of coating? Now you can remove it with baking soda and a bit of water (making what appears to be a paste), but why?
Of course manufacturers learned that the old coatings sucked, that’s why you’ve not seen it used on anything for years.
The real question is do modern rubberised coatings actually degrade, or has the technology improved sufficiently over the 20 years or so since it was last in common usage? Whatever Nintendo is using here is guaranteed to not be the same stuff.
The rubber coatings on old thinkpads have held up pretty well. That said that’s the only rubber coating I can think of that hasn’t turned into a sticky mess. Only time well tell if Nintendo got it right.
I have no idea what Nintendo is using; but my impression is that the entire class of silicone ‘rubber’ materials that(so far) seem to be pretty well behaved in terms of both things like temperature range, limited alarming results from the toxicologists, and few “leave it in the sun and the plasticisers leach out/oops, it dissolves certain wood finishes on contact” things is a relatively new one at least for consumer products.
It looks like the history of silicone elastomers goes back a fair way(initial developments late 19th century; some applications shipping during WWII); but I really don’t remember them being all over consumer stuff, ice cube trays, cutting boards, ‘rubber’ encapsulated electronics in activity monitoring bands and such; until much more recently.
They know exactly what the material they are over-molding with will do.
It is a science.
Everything mass produced is ‘made to a price’, that price is often unrelated to retail price.
If you don’t like the plastic, wait till you see the sintered brass, perma-greased ‘bearings’.
Better than a lot of old rubber but still not great.. I have a bunch of thin pads from early 2000s and while the rubber hasn’t like melted, they’re all a bit tacky
I find it shocking how nowdays you need an entire ARM SoC with computing power comparable to Pentium 4 PC just to read some buttons and send them over bluetooth. A competent programmer could easily do this with ATmega8 and some BASCOM.
Your ATmega8 is more power hungry, less available, and most importantly more expensive and has been for more than a decade!!
U wot m8?
Even right now I can order 1000 ATMEGA8L-8PU MCUs for $2.55 a piece and have them delivered by Friday. They are perfectly available if you can use internet. It’s not my fault that you live like it’s still 2001 and parts are available mail order only.
Digikey doesn’t seem to carry much Mediatek stuff, so I don’t have a direct comparison with the part Nintendo is using; but $2.55 puts the ATmega8 in more or less exactly the range of various outfits’ “BT+MCU” products, mostly with bigger MCUs.
I don’t hate the classics or anything; and for my quantity-1 stuff I’m only going to go novel if it’s the novelty I’m interested in(or the novel involves a capability I can’t get any other way); but if I’m trying to justify a part choice for a consumer wireless device the numbers are just screaming at you to not try to DIY.
Sorry but $2.55 is truly expensive for what it can do. I could get infinitely more capable Cortex M0 (I’m not even talking about raw power but stuff like NVIC) with goodies like CAN under 1eur for volumes just over 100 in 2015 for god’s sake + that PU variant is uterly huge and even worse: nonSMD. Populating THT crap was expensive back then and is still today maybe even more so.
Yeah, that’s right around the price for an nrf52840, famously power sipping and capable with a great set of peripherals and few required supporting components.
ESP32-C3FH4 is $1.30 at 1qty. That chip has FLASH integrated, WiFi, Bluetooth and runs circles fast.
EFR32BG22C112F352GM32 -> $1.36. With support for BL(E), Zigbee, or proprietary 2.4GHz protocols.
Add a few passives and you got your whole wireless transceiver complete.
If you want real bottom of the barrel, even Mouser carries some Inplay or Telink chips that go for 56-95 cents.
ATMEGA8 is baked on an ancient process which means they are relatively expensive. Many manufacturers also bet on part lock-in long term and raise prices over time. ST does this, NXP too.. That makes this kind of an unfair comparison, but also no wonder why many are choosing for modern (32-bit) MCUs over older parts.
But it’s also very resilient againt radiation. If there was accident like Chernobyl today I think 99% electronic devices would simply stop working because air around the globe became radioactive (remember how coal miner in Sweden tripped x-ray detector which finally made Soviets admit what happened?) and it makes modern electronics glitch out.
Your Cortex-M0 projects will be scrap. My ATmega8 projects will happily work. That’s the difference between “maker” types chasing the latest, flashiest chips and hardcore engineer working with proven tech.
I’d have to say not really – some would no doubt glitch a little bit more, and in the really hot zones you’d probably be right. But most things globally would likely keep working just fine as all the RF shielding and interference tolerating required for a modern electronic product along with a reasonable quantity of error detection and correcting in the RAM, and the datapackets etc – the actual system should be pretty tough to that sort of problem really.
To some extent you can look at the computers that have been working on the ISS for a reasonable comparable situation to being around the hot zone of a radiological incident – for the most part they are using regular off the shelf hardware in regular off the shelf ways without issue, and all in an environment that is rather irradiated compared to the ground.
Also the bit that makes most of the older micro’s that have been tested against radiation perform well seems to be rather more because of the silicon production process and feature size than the design. So a modern production of that chip is likely not any more radiation hard than a modern production Arm chip (though I guess being a little simpler and smaller in die area terms does lower the odds of an energetic enough interaction). But it is those giant and much more energy intensive to operate old silicon gates having the side benefit that it takes much more significant added energy to trigger them in error more than anything.
Coexistence with other wireless devices demand this. The BLE only supports several modes which aim to provide fast communications (e.g. 1Mbps), with addressable devices, etc. A basic BLE network stack therefore will requires dozens of KB of FLASH and some CPU horsepower to run.
Sure everyone can connect an OOK/FSK RF synthesizer together with to a timer output of the ATMEGA8 and blast out some differential Manchester encoding at a few 10s of kbps, but if we want resilience and coexistence unfortunately this doesn’t scale.
In this specific case we’ve also got audio in play(controller has an analog headset jack; receives system audio from console; sends mic input to console); which is a major point in favor of BT.
Your point remains that Real EE Fu could absolutely handle it(presumably either as a quick course in DACs and ADCs as part of the digital data; or with an elegant but antiquated pure analog; mapping audio to RF and back being something that we were doing back when the semiconductor junction was involved a cat’s whisker and some manual futzing, after all); but when ‘Be a headset, along with some lightweight GPIO for status messages and stuff’ is basically what BT was born for you are going to have a hard time beating out someone’s off-the-shelf implementation for doing exactly that. And then you’ve got the actually-not-totally-useless encryption that modern BT is supposed to be good for; which is a relative pittance in reasonably contemporary fixed-function silicon; but would actually take some doing to handle properly on a really resource limited general purpose microcontroller.
I doubt that they actually need the whole thing; it’s more a question of whether there’s actually a meaningfully cheaper part(or a custom/semi-custom part where the unit savings would outweigh the upfront cost) that they could target instead.
This is especially the case with something like BT, where the ongoing changes to the standard to try to make it suck less mean that you’ll likely want a comparatively recent part for reasons beyond package size and lifecycle status; and where the spec’s requirements for what the CPU doing housekeeping for the RF needs to be capable of are sufficiently large that there’s a lot to be said for just slapping something well-known like a smallish ARM core on to do the job; allowing enough room for a lot of customers to sneak their application-specific work onto the same core or cores and cut an entire microcontroller and any supporting components out of the design; rather than trying to pare it to the bone for absolute minimum part cost while forcing every customer to add another microcontroller or small CPU to host their application.
There are still economic pressures in favor of integration; but this pressure seems less acute on the wired side; where a lot of low-bandwidth interfaces are basically solved-ish because the period of standardization where really egregious sins are getting hammered out was 30-40 years ago; so 20 year old parts are the highly mature ones; and because there’s no ongoing “if we use levels of DSP that would have been witchcraft or nation-state customer 20 years ago we can get slightly better performance in the RF hell that is the contemporary ISM band” spec development; unlike the RF interfaces.
It’s still entirely possible that The Classic is a $1.50 SOIC part whose datasheet is stamped “NOT FOR NEW DESIGNS”, competing against 50 cent Cortex M0 BGAs and almost impossibly cheap mystery 8051s with Chinese characteristics; but if you want to talk I2C or SMBUS or whatever it’ll be just fine; where a 20 year old BT device is Not good news.
I am pretty sure the chip is a lot like the nrf52 parts (lower end of the series) but costs less per unit because Nintendo order quantities.
I find it a little silly how much folks find anything but hall effect disappointing now – in practical real world terms a decent resistive stick is fine, better than just fine even. Yes hall effect has some nice features and can be a great choice but ultimately for a thumbstick the centring spring is probably more important than the sensing technology by a huge margin – People just are not sensitive or dexterous enough to really notice the difference over that 15mm of throw between the sensitivity and repeatability of hall stick and a decently made resistive sensor. The bit you’ll definitely notice is when it stops re-centring properly, and how big the deadzone it ships with is, and both of those tend to because the pivot or centring spring is bad not the sensor tech in use at all…
NB don’t get me wrong some resistive sticks are garbage and I do very much like hall effect and optical etc sensing, and sometimes it might even really really make sense to be upset they didn’t use them. But sensing tech only matters if the centring spring and pivots are good enough to actually make use of them, and in the case of resistive sticks you can usually just clean them should they start drifting for sensing reasons and end up getting just as much if not more life out of them as the hall effect stick will – as the life span limiting part ends up being the same and made much the same way for all sticks in that formfactor…
on first principles i’m inclined to agree with you, but my wife did recently take her switch controller apart and apply contact cleaner to it. and she’s happy with the result.
making me think that gunk on pots is not a red herring, and the centering spring must be holding up. but i didn’t actually ask her what she did so who knows. the one thing that is for sure is that there’s a lot of end user frustration with the way they built them in the recent past
Indeed, dirty contacts are absolutely a potential source of problems, and i did in fact say clean them if needed. The big problem with ’em in recent history seems to be just that the Big N put very cheaply made junk that has no chance to last a reasonable lifespan in their very expensive toys.
If the resistive potentiometer was more durable I’d have no issue with them. As the current manufacturing standard with a thin film that gets worn through by the wiper or gunked up. That’s not hypethical, all but like 2 thumbstick of the well over 30 thumbstick
replacements I’ve done are worn through resist material.
Most cars use the same technology for the drive by wire throttle body, the only one of heard failing in great quantities was a late 90s early 2000s Volvo with an Italian made throttle body.
Get a Swiss made potentiometer and it should be good for 10-20 years easily.
I don’t get everyone banging on about hall effect, the mechanicals still need to be tight.
As I said some of them are junk – with the joycon being perhaps the worst mass produced thumbsticks ever. Doesn’t mean you can’t have perfectly good versions though that last decades of abuse…
The whole point being that it really isn’t the sensing technology that is the problem, it is the quality of the execution – can’t judge a new stick till folks have actually tried wearing them out it might be like the 360 controller sticks that are darn nearly immortal sensing wise (though you’d still end up replacing them for other reasons) or might be Joycons…
Ah they’re not using the original 3DS battery for the pro controller this time.
It’s worth mentioning that there are a fair number of well-made third party controllers that work with the Switch 2 and are easier to repair, such as 8bitdo’s Ultimate and Pro controllers (with the latest firmware updates).
Now, if you want something that has the new dedicated “chat” button, a headphone jack, and support for reading NFC (amiibos) all in one, I don’t know of anything else that supports that yet. But I suspect it’s only a matter of time.
It’s 90 euro’s and you can’t even repair it. I really don’t understand the Switch 2. It’s a console targeting younger children. Sure, some adults buy it too but they aren’t the target audience as they go for the Playstation, Xbox or even better, the Steam Deck. Almost all the games on the Nintendo are games designed specifically for younger children. So why is a console for children so much money, why are the games so expensive. It makes no sense to me. And a controller like this made in the cheapest way is just bad.