Analog sticks have become a core part of modern video game controllers. They also routinely fail or end up drifting, consigning expensive controllers to the garbage. [sjm4306] recently did a repair job on an Oculus VR gaming controller with drifting analog sticks, and decided to do an autopsy to figure out what actually went wrong.
The video starts by taking apart the analog joystick itself by prying off the metal case. Inside, we get a look at the many tiny individual components that make up a modern thumbstick. Of most interest, though, are the components that make up the potentiometers within the stick. Investigation revealed that the metal contacts that move with the stick had worn through the resistive coating on the thin plastic membrane in the base of the joystick, creating the frustrating drift problem.
It doesn’t have to be this way. Analog sticks in modern controllers could be manufactured with higher-quality components that don’t wear so easily. After all, it’s hard to imagine a 90s video game controller wearing out as fast as this modern Oculus unit. But everything is built to a price, at the end of the day, and that’s just how it goes. Video after the break.
34 thoughts on “Autopsy Of A Drifting Thumbstick Reveals All”
Funny, I tore down my drifting switch joysticks and the contacts were barely worn. Sure there are marks but they’re faint.
Meanwhile my working 3DS circle pad had the carbon peeling off the contacts from wear.
… it only takes barely wearing the contacts to ruin the resistance gradient of the potentiometer and cause drifting. Comparing it to a much simpler button (on/off) is as apples and oranges as you can get.
The 3ds uses potentiometers too
Drift can either come from wear on the stick or wear on the pots (encoders). Just depends which is the weaker link. If you think about the old N64 sticks (yes they weren’t “analog” but it’s just bits of resolution in the end) the encoders didn’t wear at all, but the sticks did.
Even if you use an encoder that won’t wear, you’re still jamming bits of plastic around.
You can also get drift from wear on the centring springs – though in that case a recalibration will get you something close to new in performance – just whichever direction the spring biased the stick now has less movement for its travel range.
Yeah, I included that as wear on the stick. You can’t lump it into “mechanical” because of course the wear on the pots is mechanical as well.
You can recalibrate out wear on the pots as well, it’s just a change in the overall curve. And you can add trimpots to rebalance the range until you actually cut through the track itself. Physical wear on the stick is when you’re totally screwed.
Carbon tracks are all but gone. In the automotive world, replaced with magnets and some kind of solid state compass. Nothing to wear out, the tech is there perhaps joystick OEMs will adopt it or perhaps they enjoy the planned obsolescence.
That’s already a thing, but it’s probably too expensive to be a standard option.
Gulikit makes controllers with Hall effect joysticks and sells those sticks as a drop-in replacement for the steam deck.
Nintendo, microsoft, sony and co just want to sell more controllers.
I have my doubts that they’ll actually be significantly *fundamentally* better in terms of longevity – often times when the pots are gone in the controllers, the actual stick mechanism has a fair amount of wear as well.
I think the main benefit of the Gulikit sticks is that they start off better with no effort. If you spend a bunch of effort aligning and balancing a potentiometer-based one, they’ll last a while. I think *some* of the manufacturers at least deal with imbalance by calibrating it, which means they naturally won’t last as long.
That sounds like a hall effect joystick. I know there are upgrades for the Steam Deck for it and the new Playstation controller has them as well.
Unfortunately the $200 Dualsense Edge controller has the same crappy ALPS standard sticks that every other controller uses, including Nintendo’s. Unless you just meant that replacement hall sticks are available, which Gullikit will be making at some point.
The gulikit hall effect sensors have been available for MONTHS on Amazon…
Not for the Dualsense or the Dualsense edge they aren’t.
A link would be helpful, so here’s one: https://www.amazon.com/dp/B0B8GPWF37/
If it’s something without a mechanical stop like a dial and it’s sealed from ingress, there’s a huge benefit to that since it’ll basically never wear out. But not so sure it makes sense on sticks: the plastic will wear from use eventually anyway, and if the cost difference is large enough it doesn’t make sense. The real sad part is not having standardized modules easy to replace.
I mean… they *are* mostly, but customers don’t see it that way.
Dreamcast controller have hall sensors in their analog stick and triggers.
“It doesn’t have to be this way. Analog sticks in modern controllers could be manufactured with higher-quality components that don’t wear so easily. ”
Like to many things these days it seems that they are designed to fail so that the gullible public will buy more; the same applies to devices which are apparently made to be hard to repair, or for parts to be difficult or expensive to source, or simply that parts can’t even be switched between devices despite the fact that the device is the same model/type (yeah, with the latter few observations I’m thinking of certain iPhones ……).
Not really: as the article says, everything is built to a price, and it is cheaper to use resistive elements than magnets and a chip.
Of course there’s always hall effect joysticks, no carbon tracks to wear in those.
hall effect isnt all its cracked up to be. they might have wear immunity, but bad engineering can lead them to be worse than a good pot stick. i want to put some resolvers or vericaps on my joystick. i got some antique mil spec pots, made in the usa when analog was king, and they are possibly the best rotation sensors i own.
hall is actually a good choice for thumbsticks though, especially if they have on-chip adcs. because you can make those extremely cheap.
not to mention you can get single chip vector hall sensors with a dsp on board to apply software linearization. one chip, one magnet solution that can use a digital interface and greatly simplify controller design and ensure the shortest possible signal path on the analog side (entirely in-chip). its only a matter of time before a module like that becomes ubiquitous.
I hope by “mil spec”, you are not referring to the military standard of the cheapest thing to be mass produced efficiently by the lowest bidder. You would be better off buying Dollar Store controllers in that case.
Cheaply made products. They expect the consumer to roll over and buy a game pad. Yea those game pads are built into a certain price but it’s probably 10 bucks in parts and probably another ten bucks in labor. The rest is markup and tax.
If they used something better instead of resistive analog sticks and jack the price up a bit the consumer will still buy it.
Has anyone tried building an RVDT or LVDT using PCB traces?
90s controllers? Are you kidding me?
The SNES has 12 total contacts on a 1 layer board. One chip. Wired, so no power management. That’s it: 12 bits of information. And they *still* charged what, 10-20 for each?
Nowadays a controller has multiple motors, a wireless module, more buttons and multiple analogs, power management and barely costs more.
It’s a miracle the things don’t break when you touch them. They’re *incredibly* well made for the price.
They cost a good bit more nowadays, although it’s justified like you said. DS2 was $25-$35, DS4 was $60-$70 (wonderful controller), Joycons are $80 (for 2 separate controllers). I’m just miffed that some sticks work for 10yr and some last for under 2yr, it has become the point of failure rather than the cord or the battery. I’ve had to repair almost every single Joycon I own or my family owns.
That’s modern prices, though. The NES Advantage retailed for $50 when it was introduced in 1987. That’s $135 today. That’s practically an Elite 2 controller, and have you *seen* what an NES Advantage looks like in a teardown? It’s a pair of basic logic chips that were bog standard for 20+ years. It’d be like opening an Elite 2 controller now and finding the whole thing run by a ATTiny or something.
It’s not the same thing as comparing the consoles – yeah, an SNES is way less powerful than a modern console, but it wasn’t using 20+ year old technology, either. The NES is a modified 6502 – Apple II’s were introduced with those ~6 years prior at $1000/pop – with a bunch of custom chips. Consoles have been basically tracking processor technology at roughly the same pace (maybe a bit quicker now, but processor tech is slower).
Controllers were mostly cheap overpriced crap basically until Sony came out with the DualShock and Nintendo came out with the Wavebird, and since then it seems it’s been a race to push as much tech into one as possible.
Wouldn’t drifting sticks be an asset in Grand Theft Auto?
GTA 12: Joycon Drift
Hall effect sticks are used in power wheelchairs; pricey to buy fom OEMs but these chairs and parts are turning up at garage sales and as junk. I have obtained used chairs in good condition for anywhere from free to $100. These chairs in general are a thing ripe for all sorts of hacking.
There is another option. I have a MS sidewinder precision pro joystick that uses an optical sensor. Much alike the sensor in an optical mouse. It still works today (these were released in 1995).
Wikipedia has an article devoted to that joystick (https://en.wikipedia.org/wiki/Microsoft_SideWinder) and while it says it’s optical, no design details are provided (maybe there’s a patent out there?).
I too thought that optical means might be a more reliable method and my idea involves using two layers of polarizers, one stationary and the other rotates along the respective axis, so that as the angle between the two changes, so does their transmittance/luminescence on a light sensor, but there only needs to be one LED for both axes.
On of the things highlighted by the Wiki article is that these non-resistive joysticks are not drop-in replacements, but they may be better if the game is designed for them, with lower operational overhead than continuously measuring a (noisy) resistance (and maybe averaging them to make changes smoother). Seem that a sensor could be made to output on a SPI/I2C bus and maybe even a interrupt to signal changes. Retrofitting the new style to old style would require a lot of unrewarding rework.
Here’s a link to an optical unit: https://www.ctielectronics.com/OEM-Analog-Industrial-Joysticks/Optical-Industrial-Joystick.php
Please be kind and respectful to help make the comments section excellent. (Comment Policy)