PMDG 747-400 trouble with turbulence with 4.28

[Pete Dowson]

I cannot emphasise enough that the treatment of wind smoothing and turbulence is absolutely identical in 4.26 and 4.28.

Just to prove this to myself (apart from examining code changes through my Source Control system) I've flown with your saved flight with both 4.28 and 4.26 now, and the results I get are identical. I also added winds with turbulence, and the effects are the same for both wind and cloud turbulence -- as, indeed, they must be as it is the same code.

With the severe (maximum turbulence) you had set (number 4 as shown by WeatherSet), the 747 A/P does have difficulty keeping on track. this is to be expected. It does tend to reach its maximum bank before turning back towards the intended course. With light to moderate turbulence (wind or cloud, or both) it copes well. I think you'll find that in real life the pilot tends to take over manual control during severe turbulence, and requests a course or level change to get out of it, re-enabling the A/P when the effects lessen.

Incidentally, I am just about to upload version 4.282 to the "FSX downloads" Announcement above. This has a few important changes, none to do with the weather, but if you are going to test 4.28 (again?) you might as well get that version instead so we know we are definitely talking about the same code.

Regards

Pete

[ahaka]

Thank you very much Pete for investigating.

From my memory I can tell that at least the version number shown in FSUIPC user interface was indeed 4.28. As you mentioned, I also remember a similar problem in verions before 4.26. I really have no idea what could have went wrong. At least when I installed I asked the setup to "just look for existing registration", but that shouldnt have anything to do with the installation process itself, right?

I will try the version 4.282 then, when it becomes available and report back here. It will probably have to wait until tomorrow night though, but I will post here as soon as I have tested it.

[Bill Casey]

Sir Peter - this may be a red herring but last week PMDG released a major update for the 747 and since then some users have been reporting some aircraft instability flying-wise. It may be coincidence but thus far those reporting problems are all Vista users (and most of their names begin with J fwiw). It might just be worth checking if you and this user are using the same 747 version and OS when testing this problem?

While I'm here, since I'm running ASX SP3 and the latest 747 version I'll bung the latest FSUIPC on (I'm currently on 4.26) and see how things go pour moi.

[Pete Dowson]

... last week PMDG released a major update for the 747 and since then some users have been reporting some aircraft instability flying-wise.

If PMDG have messed something up I'm afraid it would have to be they who fix it, unless they can tell me precisely why it is now going wrong again, if due to my turbulence eemulation. I actually depended upon their expert 747 captain, who tests their models, to test the FSUIPC4 turbulence emulation for me and he gave it his approval.

The only example shown to me so far has the maximum turbulence set ("severe" or 4), which does indeed make life difficult for the autopilot -- so much so pilots would normally fly manually and seek a way around it. But certainly with mild to moderate levels the previous PMDG 747X is flying fine here.

And the cloud and wind turbulence is the same -- the only time it was ever different was well before 4.26, some 4.25 test release I think. That was a bug, and it was also before the turbulence was re-written to do Gaussian ramping rather than sudden changes. The problem reported here is EXACTLY what happened with the 747X with the sudden changing, even if they were mild, in the much earlier version of FSUIPC4.

Realistic turbulence, as in FSUIPC4 now, has smooth changes to random extremes governed by the severity, with those extremes chosen to fit a "normal" or Gaussian distribution. That's what changed in 4.26 and it hasn't been different since then.

Considering all this, it does seem to me that, somehow, these two fliers have been using an older version of FSUIPC4 even though they thought it was 4.28. I even downloaded 4.28 from the Schiratti site and tried that, and it was fine, exactly as expected.

Regards

Pete

[John Veldthuis]

Sir Peter - this may be a red herring but last week PMDG released a major update for the 747 and since then some users have been reporting some aircraft instability flying-wise. It may be coincidence but thus far those reporting problems are all Vista users (and most of their names begin with J fwiw). It might just be worth checking if you and this user are using the same 747 version and OS when testing this problem?

While I'm here, since I'm running ASX SP3 and the latest 747 version I'll bung the latest FSUIPC on (I'm currently on 4.26) and see how things go pour moi.

It most certainly is a red herring. A whale in fact. What they are having trouble with is something completely different which shows up as the speed tape not working and when you hit the AP button with LNAV/VNAV set results in the plane flipping vertically 360 degrees in a matter of seconds continuously. If you have stress causes damage then results in an instant breakage. I saw this via an early beta but has not happened with the latest release. It has something to do with the Simconnect.

[ahaka]

Pete,

Sorry for the late post, I was a bit busy early this week.

I have tested with 4.284 now, and as much as I'd love to say its working fine, its not. Even in moderate turbulence the plane suddenly tips and banks steeply all the way to 90 degrees and comes crashing down. I checked and double checked that I have the right version installed.

What on earth could it be... Its behaving just like before 4.26. 4.26 seems to be the only version that works perfectly.

[Pete Dowson]

What on earth could it be... Its behaving just like before 4.26. 4.26 seems to be the only version that works perfectly.

Sorry, it is working perfectly here, and I can actually prove, through my source control (which keeps track of all changes) that the code concerned is identical to that in 4.26. Sorry, I don't know what to suggest. There seems to be only two of you with this problem, and I've no idea how it can be happening.

I am away from tomorrow until June 10th. If I do think of any way to find out what is happening, I'll get back to you. There are a lot of logging facilities which might record the details, so you could try getting me a log. Edit the FSUIPC4.INI file, adding

Debug=Please

LogExtras=512

LogWeather=Yes

to the [General] section. Start FSX with a saved flight that you know runs into this trouble. Keep it short, only long enough to prove that you have a problem. Close FSX, save the FSUIPC4.Log.

Then do the same with 4.26 so I can compare them. Zip both logs up and send them to petedowson@btconnect.com. I doubt i'll be able to do anything till I get back in June so there's no rush.

Regards

Pete

[truck911]

Pete said

There seems to be only two of you with this problem, and I've no idea how it can be happening.

Pete just so you should know I also experience this issue. Folk’s like me often lurk on forums like this without participation. The hope is that we Lurkers will simply implement the recommendations without need to participate.

I can say that when using FSUIPC 4.28 with ASX SP3 and the PMDG the plane becomes uncontrollable in flight. I uninstalled ASX and reinstalled up to SP2 and do not experience the issues when only using ASX sp2 with FSUIPC 4.28.

Regards

Gary Anderson

[cshark172]

Those having this problem, are you using the new Wind stabilization (not smoothign) feature in ASX SP3? My first flight with ASX SP3, I had that enabled, and the PMDG 747x went all over the place when crossing a wind layer. The wind vector was twitching like crazy. I disable the wind satbilization feature, and have now completed 3 long flights with no problems.

[jase439]

I'm not flying the 747-400, but seeing the same behavior (violently gyrating wind directions/speeds) with ASX SP3 and Wind Stablization. I'm turning this feature off in favor of FSUIPC's wind smoothing. I'll report back how it goes.

[Pete Dowson]

Pete just so you should know I also experience this issue. Folk’s like me often lurk on forums like this without participation. The hope is that we Lurkers will simply implement the recommendations without need to participate.

But I think you mis-read. It is NOT the same issue. The reference you quoted was regarding an apparent difference in turbulence emulation between FSUIPC 4.26 and 4.28, even though the code is (and has been proven to be) identical. You are not saying that at all, you are talking about a difference between ASX SP2 and SP3.

Now I suspect that, in the end, I'll manage to prove that the 4.26/4.28 apparent different is also due to changes elsewhere, but without actually standing behind the person reporting it I can't do that -- I need comparative logs to show the actual numerical data. That's what I asked him for.

I don't want that for ASX SP2/SP3 differences. That would be more a matter for HiFi Simulation.

Regards

Pete

[ahaka]

Pete,

I have sent you the log file while the problem is occuring. Hope it helps.

[Pete Dowson]

I have sent you the log file while the problem is occuring. Hope it helps.

If you mean the one you sent on the 3rd, no, I'm afraid not. I already emailed you a reply, as follows:

Hi Antti,

> Here's the FSUIPC log captured during the cloud turbulence issue.

Unfortunately it shows me nothing useful at all because you did not do as I

asked, i.e. (to quote my message in the Forum thread sent before my

holiday):

"There are a lot of logging facilities which might record the details, so

you could try getting me a log. Edit the FSUIPC4.INI file, adding

Debug=Please

LogExtras=512

LogWeather=Yes

to the [General] section. Start FSX with a saved flight that you know runs

into this trouble. Keep it short, only long enough to prove that you have a

problem. Close FSX, save the FSUIPC4.Log.

Then do the same with 4.26 so I can compare them. Zip both logs up and send

them to petedowson@btconnect.com. "

Instead you enabled all of the logging options in the Logging options tab,

which I certainly don't want (it simply floods the log with irrelevant

matters!), there was no equivalent for 4.26, and the log was not Zipped.

I assume you've now read the later messages in the same thread which are

pointing more conclusively to a new feature in ASX SP3?

Regards

Pete

[ahaka]

Pete,

I've sent you new log files.

[ahaka]

Oh, and just to add something, in the 4.284 log file, the problem starts almost immediately after loading the turbulence test flight.

Both tests have severe turbulence, but in 4.26 it does not make the plane go to extreme banks beyond 30 degrees and the autopilot still has some control over the heading of the plane.

[Pete Dowson]

Oh, and just to add something, in the 4.284 log file, the problem starts almost immediately after loading the turbulence test flight.

Both tests have severe turbulence, but in 4.26 it does not make the plane go to extreme banks beyond 30 degrees and the autopilot still has some control over the heading of the plane.

The 4.284 log looks like it might be useful if I had something to compare it with, but something went odd with the 4.26 one -- there's no FSUIPC wind smoothing or simulated turbulence operating at all!

Are you sure you had the exact same FSUIPC.INI file in both tests? I don't understand why there's no sign at all of ANY smoothing in your 4.26 example!

The 4.284 problem seems, in the log, to be some king of initialisation problem. See this part of the log:

234219 Results: FS98 AmbientWind at PlaneAlt=5758: dir 91T, vel 7

234219 Results: FS98 CurrTemp at PlaneAlt=5758: -4C

235016 Pre- Wind Check: Dir=359.826, Vel=9.66536, X=-7.35979, Z=0.197568

235016 Ambient Winds Set: Dir=359.826, Vel=9.665, X=0.02941 (0.01513m/s), Z=-9.66531 (-4.972m/s)

235016 ==============================

235109 Pre- Wind Check: Dir=359.826, Vel=10.804, X=0.0294058, Z=-9.66531

235109 Ambient Winds Set: Dir=359.826, Vel=9.665, X=0.02941 (0.01513m/s), Z=-9.66531 (-4.972m/s)

235109 ==============================

235250 Pre- Wind Check: Dir=359.826, Vel=10.804, X=0.0294058, Z=-9.66531

235250 Smoothing now by 0.23 knots/degrees (time difference = 234 mSecs)

235250 Ambient Winds Set: Dir=91.3037, Vel=7.128, X=-7.127 (-3.666m/s), Z=0.162184 (0.08343m/s)

The ambient wind from 91T must come from some initially loaded flight (Cessna EFHK cold cockpit.FLT), which didn't get overwritten as the "weather at aircraft" from the turbulence test flt until much later --

238562 Weather Received (type 4 request, Interpolated): "????&A0 241802Z 35912KT&D0NG 100KM&B0&D4572 8CU045&CU000FSVN000N 03/-6 Q1013 "

which has just the one layer of wind with direction 359M.

It appears to FSUIPC as if the wind has just been incorrectly shifted from 91 to 359, one of the dreaded FS wind shifts, so it tries to smooth it, by keeping to the 91 and changing it gradually. By the end of your log, 60 seconds later, is has reached 26.8, but FS is still trying to set 359 throughout. I'm not sure that this isn't a cause of the problems you are seeing (with the odd 359 value creeping through), rather than the turbulence. I'll need to look and see whether my mechanism for restarting the smoothing target after a flight re-load has changed between 4.26 and 4.28.

All the turbulence logging looks okay, and to specification. Those are the lines showing like this:

297047 Randomising Dir 25.6-33.6, Spd 4.9-7.4

297078 Randomising Dir 25.6-33.6, Spd 4.9-7.4

297109 Randomising Dir 25.6-33.6, Spd 4.9-7.4

297141 Randomising Dir 25.6-33.6, Spd 4.9-7.4

With 4 of these between smoothing changes, each of those entries will represent a shift of 1/4 of the difference between the current value and a randomly selected target in that range. These are deliberately paced over 30-40 mSecs per change. The problem in 4.25 was that the changes to the andom target were instantaneous -- that DID upset the PMDG model. This new method was proven by PMDG testers to be good.

Just a brief look at the otherwise uninformative 4.26 log, the first sign of Ambient wind at the aircraft is this:

226969 Results: FS98 AmbientWind at PlaneAlt=5748: dir 0T, vel 11

even though the same two FLT files were loading. I don't understand how a simple change of FSUIPC version could do that, so I'm going to assume it is just a random timing thing. Maybe next time it would be reversed.

[LATER]

Big Ouch! I've just checked the history of my changes leading up to 4.26 release -- the smoothing logging facility was disabled by then, and reintroduced a little later. Duh!

I've found the closest version to 4.26 which has the logging enabled, so I'll email that to you. Could you repeat BOTH tests (4.284 and 4.262) please but with two important differences, one to eliminate the FLT loading difference at the start and the other so I can see the Ambient winds being seen by PMDG's code:

1. On loading FSX, before loading the test flight, go to FSUIPC Logging tab and add these offsets to the Monitor (right-hand side):

0E90 as type U16

0E92 as type U16

and check "normal log" below. Don't touch the left hand side.

2. Also disable wind smoothing until you have loaded the test flight and it is underway for a few seconds, then go to the Winds tab and switch it on.

Okay?

Sorry for all this trouble, but it still appears no one else has quite what you are getting (although they say so, so far the rest has resolved to ASX SP2/SP3 difference in options), and I'm intrigued that you could get such different results with what appears to me as identical code.

Regards

Pete

[ahaka]

Pete,

What exactly do you mean by eliminating the FLT loading difference? How? Is the "0E90 as type U16 and 0E92 as type U16" that I should add the ones you meant with this or do you mean I should add the same wind to the startup flight as in the test flight? (Btw please note that in the test flight theres only cloud turbulence not wind turbulence)

I will do the tests again but dont have time to do it tonight, I'll try to do it tomorrow though.

[Pete Dowson]

What exactly do you mean by eliminating the FLT loading difference? How?

By following step 2 of the two steps I asked you to take! I was trying to explain why the steps were there, so you'd understand. I seem to confuse you by explaining so maybe I shouldn't in future, if you prefer? Here is the step which will accomplish what I said:

2. Also disable wind smoothing until you have loaded the test flight and it is underway for a few seconds, then go to the Winds tab and switch it on.

Is the "0E90 as type U16 and 0E92 as type U16" that I should add the ones you meant with this or do you mean I should add the same wind to the startup flight as in the test flight? (Btw please note that in the test flight theres only cloud turbulence not wind turbulence)

Oooh, ouch! What's this about? Don't add any new winds!!! Where are you reading such things? Just repeat the tests with the extra two steps EXACTLY as I asked!!!

Please just follow the step as written, don't take parts of it out and make up other things with it. See it says:

1. On loading FSX, before loading the test flight,

... With me, still?

go to FSUIPC Logging tab

... This means the FSUIPC options Tab marked "Logging". I'm sure you must be able to find that. You have before

and add these offsets to the Monitor (right-hand side):

...This refers to the Monitor which fills the right-hand side of the Logging tab page. You will see columns for Offset and Type. Just fill two lines in, those two columns, offset and type (the latter is a drop-down choice, but you'll see that -- you must have seen drop-downs before) as follows:

0E90 as type U16

0E92 as type U16

... Okay so far?

and check "normal log" below. Don't touch the left hand side.

... all that means is find the checkbox below labelled "normal log" and make sure it is checked. Then press Okay. Don't press anything else.

Phew! :-(

I will do the tests again but dont have time to do it tonight, I'll try to do it tomorrow though.

Don't rush on my behalf. There's only you with this problem, so there's no pressure for me, and I'm not well at present so I'm only working part time as yet (got a tummy bug at the end of my holiday).

Oh, if and when you do send more logs, please could you include the test FLT + WX files, and your FSUIPC4.INI file. If I can see the difference it the logs I might try to repro it again.

Regards

Pete

[ahaka]

Ok Pete, understood.

I was only confused by whether you meant the "eliminate FLT loading difference" as a separate step from those you listed, but I get what you meant to say now.

Get well soon.

[jase439]

I found this thread after following a link in the PMDG forums. I too am having this problem with lateral navigation control in the presence of turbulence. I am running ASX SP3 and FSUIPC 4.28. I have observed that:

1. If I enable cloud turbulence in ASX and leave this option unchecked in FSUIPC, I get uncontrollable LNAV behavior

2. If I disable cloud turbulence in ASX but enable this option in FSUIPC, I get uncontrollable LNAV behavior

3. If I check "suppress cloud turbulence" in FSUIPC (regardless of settings in ASX), the aircraft recovers quickly and tracks the desired course

4. If I do not run ASX, but simply load a saved scenario with cloud turbulence enabled in FSUIPC, LNAV is equally uncontrollable

5. If I do not run ASX or FSUIPC but enable turbulence via the weather interface in FSX, LNAV behavior is fine

Wind stablisation in ASX is off (cases 1-3). Wind smoothing in FSUIPC is enabled in cases 1-4.

Key observation (I think). Both FSUIPC and ASX seem to model turbulence by introducing high frequency "jitter" in wind direction and speed. Even though the variance is small, the rate at which the sim vars are updated seems to be quite high. When I enable turbulence in FSX, this jitter also occurs but does so at a MUCH lower frequency. In fact, it seems to model turbulence differently than either ASX or FSUIPC. It seems to add artificial perturbations to the indicated airspeed (even though reported wind speed may be constant).

It may be that the LNAV guidance in PMDG's 747 simply cannot tolerate the high-frequency "jitter" caused by the turbulence modelling in FSUIPC and ASX. Is there anyway to "throttle" this effect (as a test)?

[Pete Dowson]

3. If I check "suppress cloud turbulence" in FSUIPC (regardless of settings in ASX), the aircraft recovers quickly and tracks the desired course

Do you have the option in FSUIPC checked to allow FSX's weather to be changed? It sounds as if you must, because otherwise it cannot inflcuence ASX's weather. This issue would be rather clearer for me if you were to first update to the current FSUIPC increment (4.287 at present, above), as this automatically suppresses FSUIPC's interference in ASX's weather when the latter is running.

Key observation (I think). Both FSUIPC and ASX seem to model turbulence by introducing high frequency "jitter" in wind direction and speed. Even though the variance is small, the rate at which the sim vars are updated seems to be quite high.

But the turbulence modelled by FSUIPC is identical for Cloud and Wind turbulence -- in fact it IS the exact same code.

The changes are not "sudden" but spread using a Gaussian ("Normal") distribution. The maximum range of change is computed first, then random targets are computed. The targets follow a Normal distribution about the mean. Then the distance to that target is divided by a value (adjustable in the INI file) to provide an increment, and the values are changed by that increment at a frequency also adjustable in the INI file until the tasrget is attained, at which point another target is computed.

This rather complex sounding system was derived from studies made by academics on how turbulence behaves, and the default parameters were tuned in cooperation with the PMDG 747 test pilot, as it was the PMDG 747 which seemed to be most sensitive to the results.

When I enable turbulence in FSX, this jitter also occurs but does so at a MUCH lower frequency. In fact, it seems to model turbulence differently than either ASX or FSUIPC. It seems to add artificial perturbations to the indicated airspeed (even though reported wind speed may be constant).

The FSX turbulence was described by the 747 test pilot as rather unrealistic. In any case, the only reason I added the emulation in FSUIPC4 is that the wind smoothing removes most if not all of FSX's own turbulence. This was the same in FS2004 -- with FSUIPC's wind smoothing enabled you never ever got any turbulence. I was trying to do better in FSX, but I am starting to wish I'd never tried and left it as in FS2004. :-(

It may be that the LNAV guidance in PMDG's 747 simply cannot tolerate the high-frequency "jitter" caused by the turbulence modelling in FSUIPC and ASX. Is there anyway to "throttle" this effect (as a test)?

Well, the 747 seems to cope reasonably well with it here except when it gets to "severe", and the effect is identical for cloud and wind turbulence (as one would expect with the same code operating it).

As for "throttling" it, the parameters for this are as originally described in the release notes and now in the Advanced Users guide (see TurbulenceRate and TurbulenceDivisor). Although these are only listed under "Winds" they apply to both Cloud and Wind turbulence, as there is no difference internally -- the only difference is where the trigger comes from, a value in the Cloud parameters or in the Wind parameters (if both together the maximum of the two wins).

Regards

Pete

[jase439]

Do you have the option in FSUIPC checked to allow FSX's weather to be changed?

Yes.

This issue would be rather clearer for me if you were to first update to the current FSUIPC increment (4.287 at present, above), as this automatically suppresses FSUIPC's interference in ASX's weather when the latter is running.

Running 4.285, but will grab 4.287 and give that a go (although, it's worth pointing out I was able to recreate the problem without ASX running).

But the turbulence modelled by FSUIPC is identical for Cloud and Wind turbulence -- in fact it IS the exact same code.

Sorry, if I was unclear. I was not comparing wind vs cloud turbulence but the effect of FSX vs FSUIPC turbulence.

The changes are not "sudden" but spread using a Gaussian ("Normal") distribution. The maximum range of change is computed first, then random targets are computed. The targets follow a Normal distribution about the mean. Then the distance to that target is divided by a value (adjustable in the INI file) to provide an increment, and the values are changed by that increment at a frequency also adjustable in the INI file until the tasrget is attained, at which point another target is computed.

I was seeing deltas of apx. +/- 2 kts in wind speed, but it was changing rapidly (in comparison to FSX's own turbulence modelling). It didn't seem to me that the magnitude of the shift was as important as the frequency of the change.

The FSX turbulence was described by the 747 test pilot as rather unrealistic.

I would agree with that assessment. But it is curious to note that the 747's guidance systems appears to tolerate it better. In fact, when the wind did shift (as a result of turbulence in FSX), the magnitude of the shift was larger than that modelled by FSUIPC but much less frequent.

As for "throttling" it, the parameters for this are as originally described in the release notes and now in the Advanced Users guide (see TurbulenceRate and TurbulenceDivisor).

Let me play with these (in addition to trying the 4.287 build) and see if I can't conjure up a better result.

Thanks for the feedback.

J

[Pete Dowson]

Running 4.285, but will grab 4.287 and give that a go

But 4.285 also had that same change, so FSUIPC wasn't interfering with FS's weather when ASX was running, and therefore its "suppress cloud turbulence" shouldn't have had any affect at all on ASX's options.

... (although, it's worth pointing out I was able to recreate the problem without ASX running).

Yes, but that wasn't relevant to the difference with FSUIPC trying to manipulate ASX's weather.

I was not comparing wind vs cloud turbulence but the effect of FSX vs FSUIPC turbulence.

So you've not actually tried or had any wind turbulence?

The point is, if you look up through the thread, that folks have been saying the wind turbulence is okay, it's only the cloud turbulence with a problem. and this is where is gets silly, because they are not differentiable anywhere in my code! They are the same, the same code, the one entry point from both causes.

I was seeing deltas of apx. +/- 2 kts in wind speed, but it was changing rapidly (in comparison to FSX's own turbulence modelling).

The changes underlying the ones you see are at about 3-4 times the frame rate and the 2 knots changes, whatever, will be arrived at through many small increments (as per the parameters). When you say "changing rapidly" how do you mean? If they are slow enough to be seen on screen, they are probably as they should be.

As for "throttling" it, the parameters for this are as originally described in the release notes and now in the Advanced Users guide (see TurbulenceRate and TurbulenceDivisor).

Let me play with these (in addition to trying the 4.287 build) and see if I can't conjure up a better result.

Okay, but it worries me that the parameters I ended up with were the result of a lot of testing and feedback from PMDG, and they were very pleased with the end results which they said were realistic. I really would hate to mess that up for the majority for the sake of a few who seem to have problems for, so far, unknown reasons. I would like to know what is affecting the few so adversely. The fact that they see a difference in cloud vs wind turbulence is the really mysterious part.

Regards

Pete

[jase439]

I repeated my earlier tests with 4.287 and as you surmised, the results were not any different than 4.285. I then decided to fool with the turbulence parameters. To help rule out some variables I:

1. Left ASX running but disabled all wind and cloud turbulence as well as wind shear functions; only random turbulence in wind and clouds were enabled in FSUIPC.

2. I suppressed wind variance via FSUIPC so the only "perceivable" effect of turbulence was jitter in wind speed.

With the default turbulence settings (TurbulenceRate/TurbulenceDivisor), the aircraft was uncontrollable. I find this curious because my earlier suspision was the shift in wind direction was throwing the 47's control system out of whack. However, wind direction was steady here, and only wind speed varied (+/- 3 knots or so) and the aircraft went bananas. Indeed, suppressing turbulence via FSUIPC immediately corrected the problem. Unsuppressing it, the aircraft began its slalom maneuvers.

Adjusting TurbulenceRate only affected the magnitude of the wind shift. Not surprisingly this is what the documentation says it does :) Dialing it down, didn't appear to have much impact - it just took a little longer for the plane to begin rocking uncontrollably. Dialing this value up just made the aircraft roll over on its head that much sooner.

Adjusting TurbulenceDivisor had no perceptable impact on this problem that I could tell. I tried adjusting the first two divisors from 20,20 to 30, 40, 60, and 100. I couldn't tell you 100 was any better or worse than 20 with respect to this issue.

When you say "changing rapidly" how do you mean?

Well, given - for example - a 5 knot shift in wind speed, that transition happens in about 1/2 second to a second. My frame rate was around 20 in this example. Your document says the update rate is about 5-10Hz, and I would say it seems to be about that. I guess it's rapid compared to my 1kt(degree)/second wind smoothing. It might be helpful - from a diagnostic standpoint - if there was a config parameter to override the frame-rate-based logic and specify an explicit update frequency.

I dunno. It's a tough call. Personally, it "seems" like a fundamental problem with the lateral navigation control logic in PMDG. It tries to overcorrect as if it's making it's corrections based on the immediate values of wind direction and speed. I would expect a system like this to employ some kind of hysteresis to absorb or filter the "noise" . I know the Level-D 767 has something like this and it is entirely unperturbed by turbulence and wind shear. Based on the observation, I don't think FSUIPC is "causing" this any more than ASX is causing this. I think it's just poor (or bugged) control logic in the aircraft - but again, I can only speculate (maybe FSUIPC is tainting some simulator variable in some way that the PMDG controller depends on)...who can say? It does strike me as odd to me that even 1kt turbulent shifts in wind speed (speed not direction, mind you) alone can cause the 747 to start rocking violently (but only turbulence seems to induce this behavior - the aircraft handles normal wind shifts - moving from one wx station to the next - just fine). Initially - in turbulent winds - the degree of banking may only be +/- 5 degrees, but it quickly escalates to +/- 15, and once you're beyond this event horizon you're about 60 seconds away from rolling the aircraft upside down and becoming bug squat on the countryside :)

I'm open to ideas, but I'm stumped as to where to go from here (beyond disabling turbulence entirely).

[Pete Dowson]

Adjusting TurbulenceRate only affected the magnitude of the wind shift. Not surprisingly this is what the documentation says it does :)

Yes, its the divisors which control the speed of the changes -- they set the number of increments approaching each computed target.

Adjusting TurbulenceDivisor had no perceptable impact on this problem that I could tell. I tried adjusting the first two divisors from 20,20 to 30, 40, 60, and 100. I couldn't tell you 100 was any better or worse than 20 with respect to this issue.

Well, with 100 it should take 5 times as long for the computed targets to be reached -- the increments will be that much smaller.

It might be helpful - from a diagnostic standpoint - if there was a config parameter to override the frame-rate-based logic and specify an explicit update frequency.

I can't really control that exactly as it has to be in-line with the FSX SIM code, not in a separate thread, and I am dependent upon SimConnect calls for synchronising. The timing is only controlled by assuming I'm being called often enough and then limiting things based on elapsed time since the last update. The evenness of the updates is then obtained by altering the increments according to how long it was since the last.

You can see this in operation in the FSUIPC4.LOG if you enable the Smoothing logging. To do this add:

Debug=Please

LogExtras=512

to the [General] section of the INI file. The log will become pretty big, but it will log the increments it is using, as they are computed.

Personally, it "seems" like a fundamental problem with the lateral navigation control logic in PMDG. It tries to overcorrect as if it's making it's corrections based on the immediate values of wind direction and speed.

Maybe, but it is odd consdidering it was they who most thoroughly tested my algorithms for me, and approved them. Maybe something got chasnged their end in some recent update? Has there been one? I've not purchased the thing myself (no use for it in a hardware cockpit) but I have one version they supplied for my testing only.

It does strike me as odd to me that even 1kt turbulent shifts in wind speed alone can cause the 747 to start rocking violently (but only turbulence induces this behavior - the aircraft handles normal wind shifts just fine!).

But it is also odd that it seems fine here. but as I say, my version of the 747 may be out of date? Is there any way I can tell?

I'm open to ideas, but I'm stumped as to where to go from here (beyond disabling turbulence entirely).

Well that would be the same as FSUIPC3 wind smoothing + FS2004.

Regards

Pete