Serious FDE bug in FS2004?

[Bob Scott 7,721]

So far two factors seem to be critical to reproducing FS2002 performance in FS2004:First is the table 404 shift that I think we all agree now will correct the AoA/pitch differences that result from using a 2002 parameterization in 2004.Second, the induced drag constant Kdi also seems to be an issue. FS2002 used a constant in table 2404, which is now ignored, and instead now appears to be calculating one...theoretically using 1/(Pi*e*AR). Leaving e as set in the config file, I tried experimenting with the aspect ratio, after noting that even for rectangular wings, wing_area didn't always equal span * chord, and is often significantly off. I adjusted the chord figures in both the FS2002 Cessna 208 and the PMDG 737-700 models to chord = wing area / wing span and restored near-identical pitch and drag performance between the two models. I think what MS calls "root" chord is used synonymously with "mean" chord.I did see trim setting changes with the PMDG model, which required a positive delta in the Cmo value to correct an excessive requirement for nose-down trim after the table 404 shift. Right before quittin' time last night I tried some preliminary testing with horizontal stab incidence angle, and it appears we may have another orphaned parameter on our hands there as well.CheersBob ScottATP IMEL Gulfstream II-III-IV-V L-300Washington, D.C.

[Guest]

Yes, and according to the contributor in:http://www.forums.simflight.com/viewtopic.php?p=53320#53320... we also have no drag from cowls.What's going on here? I understand the desire to reduce the workload on the processor, but did they forget several things in reinventing the wheel? Are these just teething problems that will be ironed out for FS2006? (If I can wait that long. :( )Andrew

[Guest]

I tested a revised model that had been rotated 6.5 degrees nose down against the original model. Airfile and config were identical.There was no discernible difference at all in the flight attitude between the two models. Could it be that FS9 just ignores the pitch change in the visual model? Appears to be the case here.

[Johan_Dees 84]

>Okay, After several more hours of testing ... I have a few>discoveries.>>Some changes do not take effect unless you "Reload Aircraft">twice ... Yes, that's right, two times.>>In adjusting CG%, the first reload (using wsieffert's routine)>did not take, ... did it again, and there she was. The same>was true with suspension adjustments. I will do more testing>on this.>>Though I have not yet found a solution to incidence and twist,>hmmmm, I found that moving wing_pos_apex_lon forward 1.5 feet>took me from -1.2% to 20.6% CG. A 2 took me to 30.6~%.>>With this simple change, my CoG was in the valid 18-38% range>for the AC520 and felt really good. Had to adjust front>suspension and she feels great (relatively speaking :-) on>takeoff and landing. Rotated on the numbers and climbed out>on the numbers.>>Still requires about +6 over cruise trim on short final, and>that may be close.>>The test panel helped tremendously to sort this out, but would>have loved the AFSD as well.>>More later.>>MiltonFinally got FS9, and I can confirm that above method works very well for setting the COG.More to follow.Johan[A HREF=http://www.phoenix-simulation.co.uk]Phoenix Simulation Software[/A]-----http://www.people.zeelandnet.nl/johdUnofficial PSS Website

[Guest]

I think that was covered and dismissed somewhere further up the thread.

[Guest]

>I think that was covered and dismissed somewhere further up>the thread.Where?

[Guest]

Can't recall, and the thread's gotten too bloody long for me to go looking. :-lolAndrew

[Guest Ron Freimuth]

>Yes, and according to the contributor in:>>http://www.forums.simflight.com/viewtopic.php?p=53320#53320>>... we also have no drag from cowls. There never was 'cowl flap drag' in MSFS. Unless added by a 3'rd party gauge.>What's going on here? I understand the desire to reduce the>workload on the processor, but did they forget several things>in reinventing the wheel? Some of us call the MS guys who work on the FM's and code "Airheads". Besides not understanding and setting rather elementary parameters and tables in AIR files correctly, they always break things that worked in a previous version of FS. The CPU load from the flight model is fairly small, these changes have virtually no effect. However, FS9 appears to have better memory management and multi-threading. My AC no longer pause when the new GPS is loading data. >Are these just teething problems that will be ironed out for>FS2006? (If I can wait that long. :( )>Andrew Forget it. The basic FS has been around since 1982 or so. I've seen these sorts of problems in each new version of MSFS since 5.0. The Airheads wreck about as much as they fix in the FM stuff in each new version. Sometimes they wreck more than they fix. They wrecked a lot of the FS2K FM and autopilot code, and only added one thing of value: "fuel_flow_scalar". A way to set SFC that probably would not have been added if I hadn't given a long list of 'bugs and improvements' to the FS2K2 beta testers. I don't think any of the other bugs/deficiencies were fixed, but the Airheads sure messed up a lot that worked in FS2K. On top of that, MS refused to fix these things in FS2K2, even though it should have been simple (for non-MS programmers) to fix the FM coding errors. It has been suggested that if MS can get away without an FS patch(s), the project is considered to have been more sucessfully developed. IOW, political. Maybe B.W. gets a bonus from MS if he can avoid having patches done. So, instead of fixing the problems, MS gives BS explainations at their 'FS Insider' site on how to make an ILS/GS landing with a bad autopilot. Etc, etc. Much cheaper than doing what should have been done. While the incidence/pitch problem can be fixed in the AIR file if one knows what he is doing, I now found that 'prop effect on rudder' appears to have disappeared in FS9. 'prop effect on horiz Stab' died between FS2K and FS2K2. While the 'GS used for rate FB' in SPD Hold that appeared in FS2K2 was not fixed in FS2K4. Ron

[Guest]

>Can't recall, and the thread's gotten too bloody long for me>to go looking. :-lol>>AndrewI looked and can't find any test results for this particular adjustment.This is significant, in that it appears there is no need to adjust the model to create a proper a/c attitude in flight.Everything can be done in Table 404.Anyone who's come up with different (or similar) results, please post (or repost, in the event I somehow missed it).

[Guest]

Curious George,Perhaps I was talking out of my hat. All I can add (as a complete novice) is that FSAviator said,"The consequences of the missing variables can be patched by FDE authors making an appropriate alpha shift of the REC404 lift curve."... in his final conclusion. And discussion seems to have centred around that. I gather his above comment was courtesy of his discussion with Microsoft. Andrew

[Guest]

OK, I see where we got off track. I was just referring to offsetting the visual model a certain number of degrees to make up for the missing angle of incidence whereas this thread is discussing offsetting Table 404.Offsetting the visual model appears to make no difference.

[Guest]

My investigation of the new bugs in the FS2004 flight model appears to be reaching a conclusion. Unfortunately I have a potentially new and serious bug to report. This new bug appears to be a direct consequence of the failure to process the REC1101-50 variable. I find it difficult to test what follows with certainty as the tools available for testing in FS2004 are fewer than in FS2002. Please perform appropriate tests and report your findings.As has been pointed out in various posts MSFS calculates induced drag from AoA rather than CL. This is not a material problem if AoA is positive or rendered positive. In REC404 of the air file it has always been possible to declare CL=0 at a negative AoA. REC1101-50 was then set equal to that negative AoA and modified the induced drag equation thus;Induced drag = AoA value from REC404 minus negative value in REC1101-50 multiplied by (everything else)In FS2004 the position appears to be as follows;The flight model has no such modifier. If the current AoA is negative and the current CL is positive it calculates the induced drag as negative since the 1101-50 correcting variable has been removed.We have been discussing whether the bugs in the FS2004 flight model may be patched by an AoA rotation in lieu of an AoI rotation. I think the answer is that it may appear so if you test only for high AoA. Consider a worked example for a bomber or transport of Whitley vintage with 6 degrees of mean incidence.According to the AoA shift proposal the FDE author would set CL=0 at AoA = minus 6 to compensate for the missing incidence in FS2004. The aircraft might cruise pitched one degree nose down with an angle of attack of plus five degrees. In FS2002 we could set 1101-50 to minus 6. With the MDL displayed at 5dAoA the FS2002 flight model would look up a negative AoA ( minus 1) in REC404 then subtract minus 6 from 1101-50 to yield plus 5. This is the correct AoA and the induced drag would be calculated correctly.The FS2004 flight model has no 1101-50 modifier and so it presumably calculates the induced drag using minus one degree yielding a negative induced drag. With current tools it is difficult to tell if it truncates the impossible negative result to zero.If you test in FS2004 at an IAS which forces AoA negative, but CL positive in REC404 what do you all see?My tests appear to confirm zero or negative induced drag when it should be positive and potentially very positive in an aircraft of large mean incidence flying at significant AoA < mean incidence. In the absence of an 1101-50 variable I believe this happens as soon as a run time look up associates a positive CL with a negative AoA in REC404. I therefore believe that those of you who are reporting that the AoA shift works are not testing across the whole AoA range. I conclude that it only works while AoA > mean incidence. As soon as AoA < mean incidence in FS2004 the induced drag equation fails. It produces negative drag (thrust) when it should produce positive drag. The maximum size of the error appears to depend on the mean incidence of the real aircraft as does the propensity to generate the error.Where FS2002 would correctly calculate substantial induced drag at 5dAOA the bug in the FS2004 flight model apparently causes it to see the AoA as minus one degree causing it to calculate instead either a small negative drag (a thrust) or maybe truncates it to zero.The decision to remove the 1101-50 variable from the induced drag equation seem to have caused disastrous consequences for the calculation of induced drag when AoA < mean incidence. I urge you all to test this finding, which is that that AoA rotation cannot substitute for AoI rotation due to induced drag errors in the FS2004 flight model which that proposed solution exacerbates due to the increased need to associate negative AoA with positive CL in the absence of the 1101-50 modifier.Note that if the result of further testing leads to the conclusion that FS2004 is correcting the REC404 look up from negative to positive then we must conclude that Microsoft have replaced the 1101-50 air file variable with an autoset variable whose value is set to match the AoA value at CL=0 automatically. I find this very unlikely and it does not match what I think I see. If other experts confirmed this new report the only realistic conclusion is that the FS2004 flight model needs to be patched. If Microsoft fail to patch it then, following this new discovery, we are back where we were before it was suggested that an AoA rotation could substitute for an AoI rotation.1) The MDL author will have to rotate the MDL nose down by the mean incidence to correct the loss of the incidence and twist variables2) The FDE author will have to code REC 404 using a lift slope in which CL=0 at AoA=0. Above that the real world lift slope can be coded. By this means the correct induced drag will be produced over (almost) all of the lift slope. This two step process appears to be the correct answer to Tom Goodrick's main question, which was;New build aircraft would need to be produced in exactly the same way. This process has the advantage that it cannot cause any unknown FDE consequences from the rotation. Curious George I confirm that MDLs may be rotated by degree and display rotated in FS2004. You may have had a load / reload problem. This solution was off the table while we investigated AoA rotation. I know believe that AoA rotation fails as soon as a negative AoA is looked up in REC404 so MDL rotation is back on the agenda now.There appears to be a requirement that a negative AoA must not be associated with a positive CL in FS2004 unless the flight model is patched by Microsoft. This precludes use of an AoA rotation to solve the displayed incidence error. Accurate flight dynamics can be created for use with the unpatched flight model, but only if the MDL author corrects the missing rotation variables, whilst the FDE author corrects the missing drag variable.MEAN INCIDENCE.Since mean incidence is a very important number in FS2004 I offer the following explanation of how its value may be estimated.Wing twist is the means by which wing incidence is 'washed out', but twist is no longer relevant in FS2004. Washout is used to ensure that the angle of attack of the outer wing is less than the angle of attack of the inner wing so that the outer wing which has the aileron attached stalls last. A washed out wing stalls progressively from root to tip giving a progressive rather than sharp stall. The real aircraft designer decides whether to design this expensive production option in, or not.'Janes' normally declares root and tip incidence if they differ. Taking the example of a BE36 Bonanza the root incidence is 4 degrees and the tip incidence is 1 degree.The average incidence is (4+1)/2 = 2.5 degrees By default you will settle for using average incidence as an approximation of mean incidence. In reality the average incidence and the mean incidence are only the same if the wing twists smoothly from the mid span (middle of fuselage) to the tip. Therefore if you seek greater accuracy you should consult a three view drawing and note whether the wing has a centre section. If it does the centre section is unlikely to be twisted. The washout normally begins where the outer wing is attached to the centre section. By careful measurement, (of % twisted area, not % twisted span), you may be able to refine the average incidence towards a superior evaluation of mean incidence.OTHER BUGS.Taking into account the information in this post I conclude that failure to process REC1101-50 accounts for all of the difference in observed dynamic performance between FS2002 and FS2004. Reviewers and users alike have reported that there are power or thrust bugs in FS2004. I conclude that they are wrong, and that there are no such bugs, the changed pitch display and loss of performance being solely due to the removal of the three key variables discussed in this thread. I have addressed all the other bugs that have been reported in the FS2004 FDE code within the thread 'Resolving CG issues in FS2004'. The only other FS2004 flight model bug which I can find after extensive testing, and which cannot be patched by using means explained in the two threads, is the 'distributed load MOI bug' , (see post 4 of the CG thread ). A number of posts in this thread report that FS2004 is also not processing this or that, or incorrectly evaluates something. I have read those reports carefully and I believe that every one relates to changes which took place with the transition from FS2000 to FS2002. I am as certain as I can be that the only new bugs in FS2004 relate to the three key variables which Microsoft have admitted removing.There are some new variables, some old bugs are fixed, but no other new FS2004 bugs have come to light.--FSAviator.

[Guest]

FSAviator,Regarding you last couple of paragraphs,If the following is correct,http://www.forums.simflight.com/viewtopic.php?p=53320#53320then we also have another new bug, no drag from cowls.Andrew

[Guest]

There has never been a drag factor introduced from the use of cowl flaps in any of the MS sims, unfortunately. So while this is not news, it continues to be an omission, since cowl flap drag is significant.

[Guest]

As I posted earlier, I've tested the effect of rotating the model forward by an angle equal to the angle of incidence. There was no difference at all in the cruise attitude at identical IAS between the model with zero rotation and one with 6.5 degrees. So this solution appears not to work.Again, if someone has found otherwise, please post the results.Also, in the case of the Whitley mentioned earlier, when Table 404 is adjusted downward by 6.5 degrees (that is, where zero lift occurs at -6.5 degrees AoA), the AoA and pitch are identical in cruise, both coming in at -2.9 degrees at a cruise speed of 145 mph IAS. The implications of this are interesting, as one would expect to see a positive AoA and a negative pitch, as FSAviator has explained above. Yet this is not the case in my testing.