Air Manager misses clicks when turning the Knobster

[Ralph]

Can you please report this trough a support ticket? I'll ask Corjan to review this.

[sjbell]

Hi Ralph,

Sure. I'll do that.

Thanks,
Steve

[ericrolfe]

I've attached the siff file for my test instrument. The right (NAV) side has the knob set to directly increment/decrement the large and small count on the display. It works directly and does not interact with X-Plane. For me it will miss a click about 10% of the time.

It would be great if you could test it please on your set up and see if you get similar results.

....

knobster-test.siff

Your test works fine for me - not noticing any missing clicks.
I've got a DIY knobster, based on the PropWash, which will miss clicks if I twist it as fast as I can (not realistic for me in the cockpit), but if I go anywhere near slow enough to count the clicks, it's fine.
I believe the new Knobsters use a different
I've got an I7 - 7700 at 3.6 GHz and a GTX 1070.

By the way - I noticed that when I'm working on an instrument in create/edit mode, I don't seem to be able to use the Knobster. I have to put the instrument into a panel. I'm sure it's an obvious thing I'm missing....

Sorry - probably not much use other than validating that it does work with some set ups....

Cheers,
Eric

[Ralph]

The Knobster does not work in the Create Edit mode because it's not assigned, like you would do in the panels tab.

Support ticket received. Corjan is back on Friday.

[sjbell]

Hi Eric,

Thanks for testing. It is really useful to get comparisons from different set ups because that helps to identify where the problem lies.

Yes, I wouldn't expect it to keep up perfectly when going fast, but this is very much a slow problem. On the GNS 430/530 I'm talking about operations such as "I have to turn the small right knob 5 clicks (pages) to the right", 1-2-3-4-5, and it goes 3 or 4 pages rather than 5 and you have to give it an extra couple of clicks.

On the test instrument it isn't anywhere near as bad. That is more along the lines of "Oh great! It's working." "Ah, it just missed a click".

If this is a clock speed issue, the fact that the knobsters work perfectly with KnobXP on my setup means that it is at least possible to correct.

Anyway we can leave it up to Corjan to investigate when he has the time.

Thanks again,
Steve

[jph]

Hi Eric,

Thanks for testing. It is really useful to get comparisons from different set ups because that helps to identify where the problem lies.

Yes, I wouldn't expect it to keep up perfectly when going fast, but this is very much a slow problem. On the GNS 430/530 I'm talking about operations such as "I have to turn the small right knob 5 clicks (pages) to the right", 1-2-3-4-5, and it goes 3 or 4 pages rather than 5 and you have to give it an extra couple of clicks.

On the test instrument it isn't anywhere near as bad. That is more along the lines of "Oh great! It's working." "Ah, it just missed a click".

If this is a clock speed issue, the fact that the knobsters work perfectly with KnobXP on my setup means that it is at least possible to correct.

Anyway we can leave it up to Corjan to investigate when he has the time.

Thanks again,
Steve

Hi Steve, Just imho, - knobxp is a dedicated module for the knobster (which I presume must use an interrupt on the arduino otherwise it would be a bit pointless as it is a dedicated unit)
Hence, I can see no real reason why a dedicated plugin like knobxp wouldn't work ok.
HOWEVER - and of course I am sure Corjan will correct me here - the ability to interpret a pulse / press or whatever you want to call it in a reasonably fast sequence is extremely difficult in a PC looped cycle.
I presume there is a kind of ring buffer with discard ? .. which would explain the fact that the more you add (in code - in AM) the more it has the potential to miss an input. In most cases a missed pulse is a non issue providing it captures the first pulse / press / iteration etc. It is then - or does then, become - potentially an issue when capturing the second input.
There is a huge difference on reading xp or other sim data refs and OUTPUTTING as opposed to reading INPUTS and ACTIONING in this format. In the first case, it is a relaxed and polled method, in the second case it is far far more difficult to respond to an input during a polled session (if that is what is happening) ?
Also, allowing for debounce is difficult on polled inputs.. whereas a knobster may not suffer from a de-bounce issue due to interrupts, a 'normal' encoder on an arduino would and has to be allowed for.
If the same code is used, then it is again, difficult on the receive input side. Even if the hardware interrupt code was identified as different to a software interrupt on a normal encoder then it is still a timing nightmare - unless you use a dedicated module for the one specific unit.
edit - and to add, that would seem to tie up with a slower processor in a massively processor bound application such as a flight sim..................
Just imho of course.
Joe

[sjbell]

Corjan and I spent some time debugging and researching this issue.

It is not an issue with the Air Manager software itself or even the knobsters. Instead it appears to be an issue with how Windows 10 manages USB power management (Enhanced Power Management and/or Selective Suspend) in connection with USB 3.0 and USB 3.1 hubs.

We found the following relationship between the knobsters and USB hubs:

- A SINGLE knobster will work perfectly regardless of whether it goes through a USB 3.x hub. (This is why KnobXP works perfectly).

- Two knobsters will work perfectly when they are both connected directly into USB 3.0 ports on the motherboard.

- Two knobsters will work perfectly when they are both connected via a USB 1.1 hub.

- When one or both of the knobsters are connected via a USB 3.0 or USB 3.1 hub, BOTH knobsters will miss clicks. This happens even if one knobster is connected via a USB 3.x hub and the second knobster is connected directly to the motherboard.

Presumably old hubs do not support the more advanced power saving modes that are causing the issue. I didn't have any USB 2.0 hubs to test with, so that is unknown.

So for me the solution is simple; connect the two knobsters through an old USB 1.1 hub.

If anyone else experiences anything similar you should try changing how your knobsters are connected to the computer. Either run them straight to the PC, or use an old USB hub.

Thanks to Corjan for helping me to resolve this issue. Very much appreciated.

Cheers,
Steve

[Keith Baxter]

Hi Steve,

I don,t have knobster. But did you try running and external power supply to the arduino?

Not sure it will help. Just a suggestion.
Had a similar issue with siatek. many years back. Had to use a powered usb hub.

Keith

[sjbell]

Hi Keith,

Thanks for the suggestion.

These are the Sim Innovations commercial knobsters, not the DIY knobsters. They are USB powered.

The USB hubs tested above were all powered hubs.

Steve

[jph]

Hi Steve, it still sounds like a timing issue, no disrespect to Corjan.....

sometimes backward compatibility is flaky on usb 3 / 3.1.. however, if you need to use an earlier hub that is not the way to go.
It still to me sounds like a timing issue relating to the differing usb speeds.
Selling a product that needs to use a specific hub / port in what is designed as a multi function / multi port environment to support cockpit building etc is simply not appropriate imho. ! that cannot continue.
The Nano is USB 2, but should be compatible with 3 / 3.1 - which you say it is ! - unless you use more than one device on the same port / split / hub
The use of 2 ports on one pc is misleading as a pc can often use a usb 3 hub internally to drive its usb ports.

Windows aint going to change for this. By using an earlier hub (1.1 etc) all you are doing is lowing down the data rate - hence a timing issue.

It is a work around, but definitely not a solution.

the issue seems to be more in the usb hub than in either windoze OR the nano. Try different hubs.

I would say it is NOT a nano issue with only using usb 2 (very, very unlikely!) possibly a HUB issue - or software... almost certainly NOT ! a windoze issue - smoke and mirrors..........................................
It is a difficult one.. however, if the 'solution' is to use an obsolete hub etc - then the case is lost................

Turn of the power saving features if possible, although, I very much doubt this is an issue

It could well not be an issue with AM... however. - and, it could well not be an issue that AM could be 'fixed' for.............. but, having to use an old tech hub ? yeah, ok.....