I'm currently sitting in the train, on my way back from Vienna to Ingolstadt.
It's a five hour trip, so I got the time to summarize everything that happened.
Quick meeting up with torpor and the MetaLab team
I arrived in Vienna yesterday at 9:20pm - so still plenty of time to meet up with long-time community member torpor and pay a visit to the local MetaLab (with a Pyra in my bag, of course).
It was a fun meetup and everyone was impressed by the Pyra hardware (case, keymat, PCBs), as they know how hard hacking something together like this is.
There were quite a few questions which I happily answered and I arrived back at the hotel at 1am and went to bed at about 3am.
You can see some pictures torpor took here:
https://pyra-handheld.com/boards/threads/travelling-to-vienna.78146/page-2#post-1393385
The keymat-meeting
10am, it was time for the meetup with the company taking care of the keymat.
And it REALLY was good that I paid them a visit.
It's a lot easier discussing issues and finding fixes when meeting up in real life, having samples with you and can test things.
So there was quite some weirdness going on, but I think we now found out why it's not yes as good as I hoped it should.
Here's a quick summary of what happened during the development of the keymat:
A mixture of the old and new keymat would probably be best... and as I was confused, I decided to visit the company here and discuss it with them.
Which was a fantastic idea!
After a few tries, everyone was certain something is wrong here - something in the production of the samples went wrong. (The keymats are still being produced in China, btw. There are no companies doing something like that in Europe).
But what exactly went wrong? That's what we needed to find out!
We couldn't do exact measurements as the designers didn't have the machines for that yet (but they will get the machines soon, they're already ordered).
But they had a sample keymat by the manufacturer which has multiple travel distances and softnesses to test out.
0,8mm travel distance with 80gr force needed just feels PERFECT for the normal keyboard, whereas 1mm travel distance seems to be great for the action buttons and DPad.
Hmm... but the first sample was supposed to have 0,7mm travel distance, so why did this feel like it was too far?
Comparing the sample keymat with the Pyra keymat (just by pressing it with your fingers!) made something clear:
It seems something went wrong with the production of the FIRST keymat.
The travel distance CLEARLY was more than 0,7mm (more felt like 1 - 1,2mm). So something went wrong there!
However, as I based the changes for the second keymat on the sample I got (lowering the travel distance as it still felt too far), we changed the travel distance from 0,7mm to 0,6mm.
Well... the second keymat seems to have been produced properly: So the travel distance now was 0,6mm - which is WAY to little. Which is while the tactile feedback felt worse.
Argh!
This is another story of things that happen during a production which you don't expect!
Welcome to real life
Ah, and before you worry: No, the chinese company is not incompetent, they design company in Vienna has produced keymats with them for the last 5 years without any issues, but sometimes something simply goes wrong.
(they're all humans, after all).
However, it's not hard to fix that. Just something to remember for future projects
The fix will not cost anything, as it was a manufacturers issue.
The company in Vienna is paying the manufacturer a visit next week anyways, so they're taking the keymat samples with them, do proper measurements (the chinese company has the machines to do exact measurements) and will discuss the fixes and changes.
As the stiffness and travel distance is the only thing that needs to be changed, it could very well be we've got final, fully nicely working keymats in about 4 - 6 weeks
The touchscreen
I think I already mentioned that I received a new touchscreen from Fatih a while ago (the same that's on the current Pandora models).
BOE will visit me tomorrow with another new touchscreen-sample and I'll test both of them.
However, if that STILL won't work: The keymat-design company is also specialized on touchscreens (I didn't know that...).
They can also easily do custom designs - and they're not as expensive as I feared (500 - 600 EUR tooling costs, Minimum order quantity of 500).
And yes, they DO know a lot about touchscreens.
I always thought it's the thickness of the foil that defines the sensitivity - but that's not it: The grid size is what defines that.
So yay, I learned something more about touchscreens as well here
Well, let's see how the new touchscreens tomorrow will turn out. But it's good to know there's a company that's also specialized on that not too far away from me, sitting in Austria
Stability issues with the EVM and the Pyra
As I mentioned with my last post, there are some stability issues using the latest kernels when using the EVM (the OMAP5 development board) and the Pyra.
Me, Nikolaus and some members from the community (who have an EVM at home) did some various tests.
It seems the EVMs run stable when feeding them with 6V or higher. The reason for that: The EVM has a switch included that changes the power suppply at a certain level.
And it seems that once that switch changes to the direct jack input, everything becomes unstable.
Right now it looks like that it only happens when having HDMI connected (we're still working on that).
The Pyra however uses a power circuit that's more similar to the one that runs stable on the EVM. So that probably isn't the issue.
So I ran various tests with different setups and different Pyra PCB combinations.
And found out some interesting things:
Both MIPI (the display connection) and HDMI are handled by the DSS subsystem of the OMAP5.
So it points like something goes wrong there - and it seems to be connected to the power supply.
Mabye there's a driver bug that causes a short peak which either leads to a freeze or a not-being-found display?
Maybe it's a different driver bug where the power management chip and the DSS have some issue?
As it runs stable once it's booted up, it can't really be an issue with the power supply. I can even do 200% CPU usage for a longer period of time once it has booted up, so the power supply is stable.
But something seems to go wrong whenever DSS is being initialized.
Are we on the right track? It's too early to say that, but these were just some thoughts I had.
But we'll continue!
We will squish that bug, you can be sure of that!
Phew, that was quite a long newspost.
Didn't think it would get that long!
But I hope you enjoyed it.
I'll arrive back in Ingolstadt in about 2 hours and will continue playing Dragon Quest VI now.
See you later!
It's a five hour trip, so I got the time to summarize everything that happened.
Quick meeting up with torpor and the MetaLab team
I arrived in Vienna yesterday at 9:20pm - so still plenty of time to meet up with long-time community member torpor and pay a visit to the local MetaLab (with a Pyra in my bag, of course).
It was a fun meetup and everyone was impressed by the Pyra hardware (case, keymat, PCBs), as they know how hard hacking something together like this is.
There were quite a few questions which I happily answered and I arrived back at the hotel at 1am and went to bed at about 3am.
You can see some pictures torpor took here:
https://pyra-handheld.com/boards/threads/travelling-to-vienna.78146/page-2#post-1393385
The keymat-meeting
10am, it was time for the meetup with the company taking care of the keymat.
And it REALLY was good that I paid them a visit.
It's a lot easier discussing issues and finding fixes when meeting up in real life, having samples with you and can test things.
So there was quite some weirdness going on, but I think we now found out why it's not yes as good as I hoped it should.
Here's a quick summary of what happened during the development of the keymat:
- Pandoras keymat has a travel distance of 1mm. I wanted to improve that a bit, so we changed that to 0,7mm for the new keymat.
- I also wanted to make it softer, so you need less force to press the keys.
- Then the first samples of the keymats were made (the one with the flat keys). It felt good for a first try, but the travel distance still was too high in my opinion. The flat buttons were uncomfortable (because of the sharp corners) and the buttons were too recessed in the case, so we moved them up a bit. And the keymat also felt a bit too stiff in my opinion.
- So we made the following changes: Round the keys, increase the height, decrease the travel distance by 0,1mm to 0,6mm. And try different softnesses of materials.
- The new keymats arrived. The height and feel of the key caps was now perfect - but even the keymat which was made using a lot more soft material than before felt more stiff. And the tactile feedback got a lot worse.
A mixture of the old and new keymat would probably be best... and as I was confused, I decided to visit the company here and discuss it with them.
Which was a fantastic idea!
After a few tries, everyone was certain something is wrong here - something in the production of the samples went wrong. (The keymats are still being produced in China, btw. There are no companies doing something like that in Europe).
But what exactly went wrong? That's what we needed to find out!
We couldn't do exact measurements as the designers didn't have the machines for that yet (but they will get the machines soon, they're already ordered).
But they had a sample keymat by the manufacturer which has multiple travel distances and softnesses to test out.
0,8mm travel distance with 80gr force needed just feels PERFECT for the normal keyboard, whereas 1mm travel distance seems to be great for the action buttons and DPad.
Hmm... but the first sample was supposed to have 0,7mm travel distance, so why did this feel like it was too far?
Comparing the sample keymat with the Pyra keymat (just by pressing it with your fingers!) made something clear:
It seems something went wrong with the production of the FIRST keymat.
The travel distance CLEARLY was more than 0,7mm (more felt like 1 - 1,2mm). So something went wrong there!
However, as I based the changes for the second keymat on the sample I got (lowering the travel distance as it still felt too far), we changed the travel distance from 0,7mm to 0,6mm.
Well... the second keymat seems to have been produced properly: So the travel distance now was 0,6mm - which is WAY to little. Which is while the tactile feedback felt worse.
Argh!
This is another story of things that happen during a production which you don't expect!
Welcome to real life
Ah, and before you worry: No, the chinese company is not incompetent, they design company in Vienna has produced keymats with them for the last 5 years without any issues, but sometimes something simply goes wrong.
(they're all humans, after all).
However, it's not hard to fix that. Just something to remember for future projects
The fix will not cost anything, as it was a manufacturers issue.
The company in Vienna is paying the manufacturer a visit next week anyways, so they're taking the keymat samples with them, do proper measurements (the chinese company has the machines to do exact measurements) and will discuss the fixes and changes.
As the stiffness and travel distance is the only thing that needs to be changed, it could very well be we've got final, fully nicely working keymats in about 4 - 6 weeks
The touchscreen
I think I already mentioned that I received a new touchscreen from Fatih a while ago (the same that's on the current Pandora models).
BOE will visit me tomorrow with another new touchscreen-sample and I'll test both of them.
However, if that STILL won't work: The keymat-design company is also specialized on touchscreens (I didn't know that...).
They can also easily do custom designs - and they're not as expensive as I feared (500 - 600 EUR tooling costs, Minimum order quantity of 500).
And yes, they DO know a lot about touchscreens.
I always thought it's the thickness of the foil that defines the sensitivity - but that's not it: The grid size is what defines that.
So yay, I learned something more about touchscreens as well here
Well, let's see how the new touchscreens tomorrow will turn out. But it's good to know there's a company that's also specialized on that not too far away from me, sitting in Austria
Stability issues with the EVM and the Pyra
As I mentioned with my last post, there are some stability issues using the latest kernels when using the EVM (the OMAP5 development board) and the Pyra.
Me, Nikolaus and some members from the community (who have an EVM at home) did some various tests.
It seems the EVMs run stable when feeding them with 6V or higher. The reason for that: The EVM has a switch included that changes the power suppply at a certain level.
And it seems that once that switch changes to the direct jack input, everything becomes unstable.
Right now it looks like that it only happens when having HDMI connected (we're still working on that).
The Pyra however uses a power circuit that's more similar to the one that runs stable on the EVM. So that probably isn't the issue.
So I ran various tests with different setups and different Pyra PCB combinations.
And found out some interesting things:
- Without a display board attached, they always boot without any issues and run stable.
- With a display board but only a 5V USB AC Adaptor connected, it either froze after about 21 seconds or didn't find the display. Sometimes it booted up fine.
- With a display board and only a battery connected, it didn't freeze as often, but also often simply didn't find the display. Sometimes it booted up fine.
- In both cases it worked without any issues once it booted up properly!
- With a display board AND a battery as well as AC adaptor connected, it almost ALWAYS booted into the system.
Both MIPI (the display connection) and HDMI are handled by the DSS subsystem of the OMAP5.
So it points like something goes wrong there - and it seems to be connected to the power supply.
Mabye there's a driver bug that causes a short peak which either leads to a freeze or a not-being-found display?
Maybe it's a different driver bug where the power management chip and the DSS have some issue?
As it runs stable once it's booted up, it can't really be an issue with the power supply. I can even do 200% CPU usage for a longer period of time once it has booted up, so the power supply is stable.
But something seems to go wrong whenever DSS is being initialized.
Are we on the right track? It's too early to say that, but these were just some thoughts I had.
But we'll continue!
We will squish that bug, you can be sure of that!
Phew, that was quite a long newspost.
Didn't think it would get that long!
But I hope you enjoyed it.
I'll arrive back in Ingolstadt in about 2 hours and will continue playing Dragon Quest VI now.
See you later!
