Hacker News Reader

Comment by spdustin | Hacker News Reader

Seems to me like it might be a calibration issue, and third parties aren’t able to perform the calibration.

Apparently it is not. The reddit thread has an ex Apple employee who runs repair shops in Germany. He says if you just change the serial number of a iPad screen(using specialised hardware) in existing iPad this issue will appear. This seems to indicate that it’s intentionally done by Apple.

Other option is that Apple bakes in calibration details of all possible serial numbers in every iPad sold which doesn’t sound like a plausible scenario.

Why would the calibration details need to account for all iPads sold?

Just bake the one that the device is bonded with, using the display serial number as a lookup key.

Then, if you change the serial number of the display, the lookup would fail, and fall back to whatever "default" values are.

and just be shear coincident two screens, the broken one and replacement, match calibration data completely?

Actually not that unlikely, imagine the allowed range of calibration is 0-100, the factory is likely to produce parts with small variance, but having same constant offset, resulting in output of screens between 70-80. While default in software might be set exactly in the middle at 50. Knowing how a large organization works, the default set in sw is probably not even close to what comes out of production lines, it's just an assumption some dev made that happened to work good enough on his desk 4 years ago.

Even if the screens would cover the whole spectrum, there is still a high chance this coincidence could happen, and more data would be needed to validate either theory. Is there any statistics showing that this happens all the time or just some times.

Could it be something like this:

At software level iOS registers some calibration data (after some factory calibration) tied to that specific serial number (as calibration makes sense only for that specific instance of the panel anyway) so when serial number changes iOS can't find calibration data on device and continues uncalibrated instead of trying to use calibrations for another-serial screen (from the OS perspective).

It makes sense, if there actually is some sort of calibration going on in the first place of course.

Yeah I think this is most likely. Then when you change the serial you're using the calibration data for the wrong screen, but most of the time you're lucky and the two screens behave close enough that an average user doesn't notice.

Probably for touch the effects are negligible as the accuracy needed for touch is much more coarse compared to accuracy needed for Apple Pencil drawing. So if calibration is perfecting things at Apple Pencil-level accuracy, it should already be below the threshold of touch-based highest acceptable margin of error anyway.

Touch and pen would generally require completely different calibrations because touch is detected by the capacitive touch screen whereas the pen position is tracked inductively.

However, I wonder why the pen calibration data is not stored on an controller on the display panel? That would absolutely make sense.

Can it be like:

Pencil also has a small calibration table and when paired with a new iPad it checks and downloads that immediately.

If not, it might also be (more likely) sending whatever it senses to iPad and iPad processes that raw input with the calibration data to determine the final output.

It only makes sense from the perspective of screwing third-party repair and then trying to come up with something with plausible deniability. Are people swapping screens frequently enough to justify the extra complexity of keeping calibration data segregated by serial number?

Printer manufacturers have been pulling the same trick with storing (approximations of) ink levels in chips in the cartridge, claiming that it makes it easy for users to swap cartridges and continue to have accurate quantities and in that case there's a little bit more truth to the argument, but not this one.

I'm just speculating but the accuracy needed for Pencil to work properly is, by nature of the application, much higher than estimating ink in cartgridges.

I mean, if cartgridges' calibration drift a bit it might also be more acceptable for many, but if Pencil starts drawing incorrectly many artists would be extremely frustrated and move away from the ecosystem.

The fact that they showed how using the controller chip from the existing screen causes it to start working correctly shows definitively that this is not about calibration at all.

Unless the calibration data is stored on the controller chip.

But the physical screen is a different screen so any such data is totally wrong.

Could be calibration is looked up via serial number by the iPad and the third party ones don’t have any cal on file.

There was no third party display, replacement one came off another working ipad.

Serial numbers encode production-specific parameters. We don’t yet know how to produce uniform displays batch to batch. Running Toyota engine code on a Subaru will produce glitches without requiring any conspiracy.

But realistically this would mean I couldn't use an iPad with it's pen without connecting it to the internet first. I mean how would Apple bake every calibration for every display into all devices, even ones it hasn't even produced yet?

Because the screen was on the device when it was calibrated.

It's the display that needs calibration. Pen doesn't know where it is, the screen does. So each complete device or at least the screen needs to have a valid calibration table or a function for raw xy -> calibrated xy.

I have no way to verify this, but this Redditor claims it's due to the serial number being different, not calibration (also see the grandparent comment that they posted): https://www.reddit.com/r/gadgets/comments/15cxwaa/apple_penc...

That said, I see a generous explanation that I can't rule out based on what they're describing. It seems like the only test this Redditor mentioned was that you can modify the serial number and cause it to go from correct behavior to misbehavior. But I would think the reverse direction is the real test, since otherwise this could just imply that the device is looking up some embedded calibration information for that serial number and not finding it, and therefore falling back to some default behavior?

That redditor says it's just serial number matching, complains at length about how this issue is destroying their business while in the same breath saying that reprogramming the screen's serial number on a working unit will bring up the same problem.

So, if it's just a matching serial numbers problem, why aren't they programming the new screens with the serial number of the old screen, and thus the problem goes away?

It's a reasonable question; I don't know the answer. It might require damaging the device too much to return it to the consumer at that point? Or maybe they can't due to contractual reasons? It might be worth asking them on Reddit.

That's exactly what their "calibration" procedure on iPhone does, it just changes the serial number pairing, nothing else, you don't even have to touch on the screen.

They have no clue what they’re talking about.

They were a genius grunt and their penchant for drawing conclusions on incomplete information is exactly why they aren’t trusted to perform manual calibration that hinges on their manual interaction with the screen.

If you read their other comments then you see that they believe that all components used are identical and thus wouldn’t need calibration.

Never mind the fact that different vendors are used to supply screens depending on the model in question, even between “identical” screens there’s plenty of subtle variation.

The controller holds calibration information in the form of offsets for touch and color amongst other things. The controller is married via serial numbers and if there’s a mismatch it’ll default to no offset.

This also serves as a clear indicator, with other indicators, that calibration is needed in an effort to prevent the calibration from being forgotten.

Despite not every screen being identical, the accepted tolerance makes it so that offsets for one screen are often “good enough” for another screen to the point most people won’t notice it’s not perfect, provided they’re from the same vendor.

There’s no other super secret mystery to it.

> third parties aren’t able to perform the calibration.

Which is an even bigger issue caused entirely by apples anti-competitive anti-user practices.

Reply on news.ycombinator.com