How good are the nubs supposed to be?

[Nardella]

Occasionally my nubs stop working properly (one or more of then axis cannot reach its max(32 or -32) and sometimes doesn't even get past 10 or -10) and require me to restart the OS for them to begin working properly again. When in the input tester, moving the nubs around they are almost always a tiny bit jumpy, can NEVER move the icon in a perfect circle and NEVER pass the test on the first (and rarely on the second) full circle.


Sometimes it feels like the nubs get "caught" on something inside them, but there is nor correlation between this strange feeling and problems with my nubs.


It has been this way from day 1.


These problems do not make the nubs unusable as even if they can be jumpy this is not very noticeable. And I only have severe problems once every 5-10 hours (wild guess) of use or so.


Are these problems normal or are most nubs virtually perfect?


Is the nub replacement something that can be done without voiding the warranty, are they through hole soldered?

 

[craigix]

You can reset them, however when you run the tester you may need to push them up down/left right a few times for them to calibrate, then do some circles.


The nature of them being truly analogue means they can never be digital perfect in their response.

 

[Nardella]

You can reset them, however when you run the tester you may need to push them up down/left right a few times for them to calibrate, then do some circles.


The nature of them being truly analogue means they can never be digital perfect in their response.

Resetting them seems to have no effect, the system has to be rebooted.

 

[Lafazar]

I have the same problem. A nub may randomly "die" and not trigger fully anymore in one single direction (cannot reach +32 or -32 anymore). I have noticed this problem on an old Pandora (with the old nubs), a newer Pandora (new nubs, I think) and an ICP (new nubs, I think). When this happens only a power cycle helps. It happens rarely (dozens of hours of working fine in between), but it does happen from time to time. I also think doing a proper "nub dance" (slowly pushing both nubs in all four directions and keeping them at the maximum for a few seconds, then doing a few circles after that, like Craig said) after rebooting may lower the chance of it happening at all, but it's hard to say with the low number of occurrences I have for data.


Since there seem to be more and more reports of this phenomenon on the boards I think this may well be a bug in the self calibration firmware on the nubs, so maybe this can be fixed with a nub firmware upgrade?


One thing that speaks against this being a firmware problem is that from my (sketchy) memory I think it is usually always the same axis on the nub that is affected, so maybe it is an electronic tolerance issue?


We should gather more information: Could people who notice this look out if it is always the same axis that stops responding or if it is a different axis all the time? Does doing a proper "nub dance" before using the Pandora prevent this from happening or lower the number of time it ocurs?


I must say I can live with this problem, because it only happens rarely, but if this can be fixed it should be.

 

[FaeMinx]

We should gather more information: Could people who notice this look out if it is always the same axis that stops responding or if it is a different axis all the time? Does doing a proper "nub dance" before using the Pandora prevent this from happening or lower the number of time it ocurs?


I must say I can live with this problem, because it only happens rarely, but if this can be fixed it should be.

Make what you will of this. If this is the quality of the new nubs, I wonder if there's any point in doing an RMA. I certainly hope that you fellows who are experiencing similar issues to what I had in the beginning, don't have your nubs eventually deteriorate like mine did.


I agree, we need more data on this issue.... because there *is* something wrong... and it needs to be addressed. :mellow:

 

[WizardStan]

By my estimates, over 20% of the original nubs were failing almost immediately. If these few people with problems represent even 1/10th of the problem out there, that's still only about a 5% failure rate based on 1000 units since replacing the nubs, not ideal but not outside the realm of acceptable either. That's also a high estimate, it's probably more like 2% or 3% which is really good for a fledgling company.

 

[FaeMinx]

It's probably more like 2% or 3% which is really good for a fledgling company.

That's encouraging. I'm going to monitor this issue for a few more months, while keeping an eye on my nub to see if there are any further changes before I RMA.

 

[Nardella]

I believe it is always one axis but I cannot be certain.

 

[Nardella]

It seems that the problem is mostly with the left nub being unable to go even a little bit left.


Also the input test should be eight directions not four.


x -32


x +32


y -32


x +32


x +23 y -23


x +23 y +23


x -23 y +23


x -23 y -23


My Pandora has never even come close to being able to pass this test reliably even after several minutes of moving the nubs around. I am pretty confident that if this test was part of QA nub problems would be a thing of the past.

 

[MWeston]

The nub firmware itself isn't the greatest either. That whole 'calibrate on the fly' thing has a tendency to lose itself. It isn't easy to get the nub manufacturer to look at the firmware considering the team that first designed it was from the company that went bankrupt. If only they could give me the source code under NDA or something!

 

[Exophase]

The nub firmware itself isn't the greatest either. That whole 'calibrate on the fly' thing has a tendency to lose itself. It isn't easy to get the nub manufacturer to look at the firmware considering the team that first designed it was from the company that went bankrupt. If only they could give me the source code under NDA or something!

If push comes to shove you can write new firmware.. I'm sure that some people from the community would help with that or do it entirely. I know they're using some magic algorithm but how complex could it be? Or maybe its' a little too complex for its own good, hence the unreliability. They only have a few KB to spend on it so the options are pretty limited. At least that means that it probably wouldn't be very difficult to reverse engineer, assuming you have the image or a way to download it in the first place.


You're stuck with the electrical interface so it's going to be some kind of fixed transfer function to go from the ADC input (which I guess is muxed 5 ways) to the nub position. Even if it's not linear, which if it's really just a variable resistance and not a potentiomter like output, and it's connected to a voltage divider with a fixed resistance (1M?) it wouldn't be. From the voltage ranges urjaman measured before it looks like it's at least not taking up even close to the entire range. This and potential non-linearity would explain why you only get a 6-bit result despite being interfaced with a 10-bit ADC.


But I guess it's all a little moot when users won't be able to change the firmware themselves, at least not without opening their Pandoras, soldering a connector, and having the right interface equipment.


It sounds concerning that resetting the nubs doesn't have the same effect as resetting the Pandora. The Pandora controls the nub controller's reset line on a GPIO, right? This would suggest that it's a hardware problem that needs power cycling, because from the ATmega8L software's point of view a toggle to reset should be the same thing as a power cycle.