Problem Aircraft shaking over the wings

[Firefly77]

Hi all, in my case the shaking problem was definetly and only linked to turbulence settings. It was fully reproduceable. Having the turbulence scale effect in active sky >25 and trigger turbulences via FMGS IOS or FS Flightcontrol let the aircraft start to shake with AP on. Below a setting of 25 in active sky I was not able to produce the shaking and it also did not start by itself in any other flight situation. I am not sure if this turbulence scale effect setting is somewhere saved in a P3D config file and therefore also valid if active sky is not started. If I remember right the default value of this setting in active sky is 70. So lets assume it sits still on 70 and P3D induce a turbulence because of weather, turbulence effect option in P3D, real turbulence Europe or A2A accu sim then this could also start the shaking for you? Maybe you could also check and try this ?
Long frames or low fps did not trigger the shaking problem in my setup.
Best regards Michael

[dkreiskott]

Hi all, in my case the shaking problem was definetly and only linked to turbulence settings. It was fully reproduceable. Having the turbulence scale effect in active sky >25 and trigger turbulences via FMGS IOS or FS Flightcontrol let the aircraft start to shake with AP on. Below a setting of 25 in active sky I was not able to produce the shaking and it also did not start by itself in any other flight situation. I am not sure if this turbulence scale effect setting is somewhere saved in a P3D config file and therefore also valid if active sky is not started. If I remember right the default value of this setting in active sky is 70. So lets assume it sits still on 70 and P3D induce a turbulence because of weather, turbulence effect option in P3D, real turbulence Europe or A2A accu sim then this could also start the shaking for you? Maybe you could also check and try this ?
Long frames or low fps did not trigger the shaking problem in my setup.
Best regards Michael

Hi Michael,
in my case Aktive Sky isn't activ. And the Prepar3d.cfg is default fresh generated.
I'm shure long frames are the problem. But I do some other tests.
Yesterday I set all values for world to minimum, the problem occurs at the same point at 120 FPS. The Aircraft shakes but only three or four times.
After this it flies at normal parameter. And the AP was always on.

[jeehell]

If it's always a t the same altitude, and if it's close to clouds (even if outside clouds) P3D introduces clear air turbulence at the clouds level. Try turning p3d turbulence effect on aircraft on weather options.

[michael1508]

SoI finally found time to do some test flights. I did around a dozen flights today, always restarting all software from scratch.

Here's what I found out:

First, I could reproduce the shaking, when flying the same flight plan as Dieter (EDDL 05R - DL013 - MISGO ...) > shaking starts exactly at DL013, where the plane had to initiate a left turn at about 3000 ft. I tried with and without ActiveSky and with and without P3D turbulence effects. The shaking occured in all scenarios.

I then tested a similar route, following the NETEX SID from EDDL 05R. This route has two waypoints (DL52 and DL53) in the same area. Same effect here. In both cases, there was a stutter (long frame) in P3d before the shaking started. DL52 was actually a flyover. Here there was a short stutter, but no shaking. The shaking started 2 miles further down at DL 53, where the plane had to initiate a left turn. I then repeated the same flight path without DL52. Now the stutter and shaking just started at DL53.

As the stutter / shaking effect always appeared close to a waypoint / turning point, I wanted to proove the causality by flying the same path without any programmed wapoints. But I was surprised to see, that P3D again stuttered in the same region. However it was not that strong this time and no shaking.

So I already guessed, that there is a causality with the programmed waypoints. However a few miles later on - I was flying a selected heading at 3000 feet downwind EDDL 05 to circle back to land on 05R, the shaking started out of nothing and with no waypoint being near. And this time there was also no P3D stutter prior to the shaking. At least, I was not aware.

After these test flights I am confident to say, that this issue appears independent of weather settings in AS and P3D. Except of this one time exception it always occured at waypoints. However, I still wonder, why this is happening only at certain waypoints and in certain situations. For example I was circling back to EDDL 05R overflying the airport and approaching the same waypoint DL013 at same height as after take off - and this time, no shaking.

So no 100% clear pattern, but at least one waypoint (DL013), where the behaviour can be reproduced. I hope, this helps to do further analysis for JL, or anyone else.

[bbruechmann]

Michael, many thanks for your investigations.
My suspicion is still, that this issue is caused by latency problem. From a logical point of view, most long frames in P3D are caused by loading scenery data. Depending on how well the specific scenery or even the transition between sceneries is programmed, the biggest single data the simulator has to load are textures and these long frames are commonly issued by searching for those textures. It can happen that these textures are between other textures that are missing. I have monitored that a while ago when I was using a specific tool which is monitoring every single microsecond of the computer doing whatever on process is in every programmes, and there I have seen that usually there are a lot of textures this simulator is searching for not existing textures. Those textures can be scenery textures but it can also be caused by AI traffic and lots of other things, but most of them are remains which are left from updated sceneries or uninstalled sceneries etc.
The reason why in the case of your scenario that the issue occurs always at the same point could be caused by a virtual scenery line, where the scenery you have passed is ending and the next scenery tiles need to be loaded. So it could be that in that specific situation, where the memory lanes are busy loading scenery data, can't process the data coming from fmgs as fast as expected, which then put the simulator and the fmgs server out of synchronisation for just a second, what ends in that shaking.
I have still the opinion that something like a buffer or whatever is worth to think about anyway, because we have to expect that long frames or stutters will persist even in the near future as we have seen from msfs, which is also not resistant against stutters even using much more modern and state-of-the-art codes using the capabilities of modern hardware.

[jeehell]

I believe some tweaking of cpu cores affinity could maybe help a bit.

[dkreiskott]

I did test flights again because this stuttering always occurs on the waypoint DL013. The fact is that the dropout is caused by ORBX GES. If you switch off GES, the stuttering is hardly noticeable and the Airbus catches up immediately. It is not necessary to switch off the AP.

I agree with Bernd where the problems come from and which agrees with Jean Luc's explanation.
I think a "buffer" is not the solution

@Jean Luc
But that gives me an idea. Of course I don't know in what form the data is sent from the AP/FBW of the FMGS server to the P3D.

Only when the AP is turned off does the plane calm down. I think because at the moment the data from the FMGS server is not being evaluated.

Shouldn't it be possible to see the rocking of the Airbus on the FMGS server when the AP is switched on? Wouldn't it then be possible to briefly interrupt the data transmission so that the P3d would not receive any new data for a short time until the Airbus was quiet again and then continue the data transmission?

No matter how good the hardware the P3D is running on, everyone will experience these dropouts.

[michael1508]

Hi Bernd, Dieter,

Thank you for your further investigations and explanations. So it has something to do with the data exchange during scenery loading.
Anyhow - and I say this with full respect and appreciation of Jean Luc's development work - it is just unfortunate, that we now have to look for a solution for this issue, while the issue was not existing before B57. Well, I assume JL has had good reasons to rework the AP function with B57.

I consider going back to B56, as this shaking is really annoying and as I don't want to degrade my scenery or move the FMGS server to the P3D PC to get it fixed.

[dwsugden]

I think we are looking at this correctly and that the shaking is not associated with turbulence. I have turned off all turbulence settings in Active Sky and P3D and still get the violent shaking every now and again. I know another active member, here on the forum, experienced the shaking problem only a couple of days ago with the latest release. It can become frustrating, particularly after nipping into the kitchen to make a coffee whilst on a long flight, only to come back to the cockpit 15 minutes later only to find the aircraft has met terra firma!

Hopefully the true cause can be found in due course.

Regards,
Darren

[jeehell]

Could you all make sure there is no issues at all in manual flight? Because disconnecting AP when it happens does not mean it won't happen in manual flight. It could just be resetting some controller inside the code at a proper timing.

[dwsugden]

Currently flying from Bristol to Glasgow at 32,000 feet. The scenery paused, maybe for a second..?? Then the shakes occurred. The autopilot did not disconnect. The shakes, unlike in most cases, occurred for 20 seconds or so and in that time the auto pilot did not disconnect. After 20 seconds or so, the shakes stopped and now I am flying perfectly once again...as I speak I am now in the descent.

It does appear, as one poster mentioned earlier, to be a scenery [loading/pause] issue....

Darren

[dkreiskott]

Could you all make sure there is no issues at all in manual flight? Because disconnecting AP when it happens does not mean it won't happen in manual flight. It could just be resetting some controller inside the code at a proper timing.

I'm a few days in holiday. After this I do some test with manual flying.
Happy eastern.

[michael1508]

Could you all make sure there is no issues at all in manual flight? Because disconnecting AP when it happens does not mean it won't happen in manual flight. It could just be resetting some controller inside the code at a proper timing.

So I was flying the same track (EDDL05R DL013 MISGO) today without any programmed route. No SID, no AP, pure manual flight. I monitored my route on my Navigraph map. And at DL013 I could turn left towards MISGO without any shaking.

Hope that helps!

[Barrykensett]

I got what I call the Wobblies after updating FMGS about three cycles ago. I have not done much about it as I have been doing some major hardware upgrades to the sim and have only just started flying again.
Did a short flight today, the Wobblies started at 5000ft and were there for the rest of the flight. Active Sky was off and I am on default P3D scenery.
Wondering about rolling back FMGS, has anyone else tried this?
Barry

[basti]

We have an i7 CPU from the 12th generation. In theory, the CPU has enough performance, but we had also from time to time frame rate drops, stuttering and aircraft shaking. I think this could have something to do with the Intel P-Cores and E-Cores and the Windows Task Scheduler. The P-Cores are the cores with high performance and the E-Cores have lower performance but high efficiency.
When the Windows Scheduler has decided that important tasks only get CPU time on slow E-Cores, then the problems might start.
To give tasks CPU time on P-Cores you can change the priority to high in the Windows Task Manager.
We gave a high priority to FBW.exe and Prepar3D.exe. Maybe this might also work for you.

[jeehell]

We have an i7 CPU from the 12th generation. In theory, the CPU has enough performance, but we had also from time to time frame rate drops, stuttering and aircraft shaking. I think this could have something to do with the Intel P-Cores and E-Cores and the Windows Task Scheduler. The P-Cores are the cores with high performance and the E-Cores have lower performance but high efficiency.
When the Windows Scheduler has decided that important tasks only get CPU time on slow E-Cores, then the problems might start.
To give tasks CPU time on P-Cores you can change the priority to high in the Windows Task Manager.
We gave a high priority to FBW.exe and Prepar3D.exe. Maybe this might also work for you.

if this is a solution on your issues, it's an interesting theory.
Please other people with that issue let me know if it helps.


@Barry: do you run FMGS erver/FSXconnect/FBW on a remote PC? or is it on P3D PC?

[Barrykensett]

Remote PC JL. What is puzzling me is that I have not had major upgrades to P3D or sceneries which could push the processors. I have been running fine for a long time, my FS computer is getting on a bit but has not given me these problems. That is why I am sceptical about some of the tuning adjustments described above. Also I have had the server on separate PC for a long time. I have been preoccupied with a major hardware upgrade to the sim but the Wobblies occurred before I started that. I thought of rolling back FMGS but I think that is a bit complicated by the change of aircraft model.
Barry

[dkreiskott]

I did some test flights today. But before that I had a few changes. I switched from P3dv4 to P3Dv5 and installed B58.0.2.

I did the same flight as Michael, but with the difference that I programmed the MCDU. In all cases I flew without autopilot.

First try without Active Sky.
At the same point as always, there is a stuttering of the scenery (GES) at this point there is no wobbling of the Airbus. it can be flown normally.

Second attempt with Active Sky.
Even without the autopilot, the Airbus starts to wobble, but can fly on again immediately by briefly operating the sidestick.

I'll try Basti's tip.

@Basti
Do you use HT on or off?

[michael1508]

That is why I am sceptical about some of the tuning adjustments described above.

I concur this. Because the issue only exists since version B57, in which the remodelling of the AP logic has been changed.

However, I will also test the tip from Basti with the CPU priorities. I have a K-pcocessor, though, but I guess, it's worth trying.

[dkreiskott]

I once tried the tip of Basti. It doesn't change anything for me. I use an i7-10700KF with 3.8 Ghz, 32 GB RAM and Nvidia GTX 3070 myself.
So definitely not a slow PC.
And also the FMGS server is an Intel I5

[jeehell]

The new AP/FBW is more prone to performance problems, it's unfortunate but this is to make it work more closely to the real one and with a better flight model. Yaw and pitch are really sensitive and can have induced roll effects.
I will try to improve it but I personally use single PC so it's not easy for me to reproduce.
There are users without any issues using a remote configuration as well. So optimizing both PCs (P3D and server) and making sure the network is not used for other purposes than the soft is a valid point.

[dkreiskott]

Hi Jean Luc,

In my case, I work as an IT administrator, so I know how hardware and networks work.
Your proposal to optimize the two PC's is okay.

In my case, only your software runs as a server on the FMGS server. Only P3D and your intercom are installed on the P3D PC. Everything else runs on separate PC's.

So if you need help or someone who can test something for you, feel free to contact me.

[michael1508]

@Jean Luc, which hardware specs does your single machine have?

Maybe someone with dedicated FMGS server machine and without shaking problems can post their hardware specs for both machines, too!

Would be interested to see, in which scenario the shaking can be avoided.

[jeehell]

I cant speak for all, but those I am aware of probably use a core i7 or core i9 on each PCs, and have a 1000bps network.
I made some tests with a core i3 as a remote FMGS server PC, over wifi, so this is really far from ideal. I do get issues as mentioned by others.
I could reduce the impact somehow, so please try this FBW: https://www.jeehell.org/FBW.exe and let me know.

[dkreiskott]

Hi JL,

thanks for you work.
I'll do a test with your FBW.EXE today or tomorrow and then I'll get back to you with the result.