The tech thread

[LHookins 3,102]

2 hours ago, domkle said:

...would it make sense to have the whole flight model in a .txt format for an easy access to all, said .txt files being compiled by the sim at run time ? Same for sceneries files. Nothing hidden in binary files.

The data would be no less complex and it would add a lot to the loading times.  "Easy access" is relative here. 🙂  

I personally prefer to have data, no matter how complex, accessible in a file that can be edited by a text editor.  Failing that, an editor program that displays and allows changes to the data in an easy to read format.

Occasionally the creator doesn't want you to have direct access to the data.

Hook

[Claviateur 789]

6 hours ago, FDEdev said:

That's what they usually told us of course as well.  I had the honor to fly a few month with a brilliant Dash8-300 test pilot and it was an eye-opener during that time to learn how dangerously wrong even a Level-D sim can be.

I remember at least 2 aircraft types where it took you a few minutes during every simulator check to get used to the noticeable more sensitive and spongy handling of the sim. If you were flying the real thing the day before, you inevitable rotated right into the stickshaker during the first takeoff if you tried to fly it like the real one.

 

Yes and I read an article a while ago, I don`t remember where I found it online, with pilots mentioning that home computer flight simulators today and according to them, offer with some "pocket money", surprisingly refined flight models and systems that can be more convincing than the Level D equivalents. They compared the time they spent flying that aircraft addon to the sessions they had in the training facility of their airline companies...

But as we know, it's usually the lack of surround peripherical visual, hardware cockpit, the low cost controllers with their low cost mechanics, electronics, the lag, fps etc that punish the well crafted PC flight sim flight models.

[Gulfstream 416]

7 hours ago, LHookins said:

The data would be no less complex and it would add a lot to the loading times.

No, it wouldn't.

And they've already stated settings files are being stored in XML, which is a "text file", and the most verbose possible one at that (as compared to something like JSON).

Loading and parsing even extremely large text files is nearly instantaneous on today's computers, and these particular configuration files are not gigabytes in size.

[b737800 470]

15 minutes ago, Gulfstream said:

Loading and parsing even extremely large text files is nearly instantaneous on today's computers

Tell that to Notepad

(EditPadPro and NotePad++ are of course what I use)

[LHookins 3,102]

45 minutes ago, Gulfstream said:

And they've already stated settings files are being stored in XML

I personally like XML format and have done a lot of work in it in the last few years.  JSON looks good, thanks for the heads up.  It's clean, like INI files, whatever that particular format is called.

Loading one file is quick, loading many thousand is less so.  It will always be quicker to load a large binary file containing numeric data than it will be to load and parse the same data as text.

Compare the time of the first load of P3D after a reboot to subsequent loads.

Hook

[tweekz 2,165]

11 hours ago, FDEdev said:

Designing a flight model in x-plane and FSX/P3D is actually easy and accessible to everyone. Designing a high quality flight model is the problem.

 

To start out - I am not a native speaker, so please excuse my bad wording on certain topics.

The big question mark regarding MSFS is how those "1000 surfaces" will actually be calculated. Lookup tables, parameters / functions, CFD like calculations?

Compared to X-Plane I'd be interested in your expertise. Afaik blade element theory (BET) splits a wing into smaller elements (usually 10) and then calculates aerodynamic data considering its geometric shape. This however is only done for lifting surfaces. Things like the fuselage is considered with parameters.

First, I'd like to know if this is roughly correct.

Secondly, I was thinking that hypothetically lookup tables could even lead to a more accurate flight model as long as the provided aerodynamic data is accurate enough. BET gives you a quite good approximation quite fast. But it will not respect every little aerodynamic detail. So if you have exact performance numbers, those should lead to a better result.

That in mind, I tend to think that the more complex the airframe gets, the greater the approximation of BET (with a quite low resolution of 10 elements per foil) deviates from the actual result. Especially with the complex aerodynamic geometry on modern airliners. But that is not where it ends. Aircraft manufacturers also use different materials to influence aerodynamics. Lot of things that need parametrical consideration.

How accurate is the calculation of each of the sections in BET in first place? Resolution wise.

So basically, before I drift too far away - what's your stance on those things?

Edited February 3, 2020 by tweekz

[zalox 54]

My concerns are related to old SSD and delete cicles when downloading lots of textures.

Lots of GB/Flight (or TB/year), those SSD will have heavy wear.

[Murmur 3,941]

Tweekz the pondering in your post was pretty much correct on everything you said.

What can be added, is that a real time CFD simulation is not possible on a home pc. In other words, increasing the number of points/surfaces can increase accuracy, but only up to a certain level, because you can't use the fully fledged aerodynamics equations on all those points and still run a real time flight model.

The solution to this, is that a predictive flight model (like in XP, AFS2, and FS2020) would need some empirical/parametrical inputs to work around that, and allow designers to tune it in order to get the best accuracy compared to the real aircraft. So in effect, predictive flight models need to have some part of "table lookup" to make them more manageable in terms of achievable accuracy.

Asobo said very little about the inner workings of their new flight model, so I guess we'll know more after FS2020 will be out and people will play with its flight model.

 

[Gulfstream 416]

8 hours ago, LHookins said:

Loading one file is quick, loading many thousand is less so.  It will always be quicker to load a large binary file containing numeric data than it will be to load and parse the same data as text.

 

Yes but obviously these complex simulators already load hundreds if not thousands of independent binary files (libraries) so the constant disk access is a wash, whether it is text files or binary files.   It's all just 0s and 1s in the end.

For example, I just wrote a quick parser in C# to parse the entire history of the NTSB accident database records.

It's a 23.2MB CSV file and it parses in 0.12 milliseconds.   And that's even putting it into C# classes and adding them to an in-memory store. 

And yes, this locks up Notepad if you try to open it, but that's because Notepad is not designed to handle 23MB text files.

It's all fast.  Very fast.

Edited February 4, 2020 by Gulfstream

[188AHC 2,138]

Why? I locked this topic so it would give time to remove posts with personal attacks. When that's done we'll probably give it another try.

It's cleaned up. Have fun. 

[G-RFRY 1,708]

Could MSFS end up on this,

https://www.tomshardware.com/uk/features/nvidia-geforce-now-launches

Priority streaming.

[vortex681 806]

On 2/3/2020 at 3:39 AM, FDEdev said:

You are overrating the quality of civil Level-D sims. They are not much more than motion equipped procedure trainers. Many are simply way off when close to, or beyond the edge of their usual operating envelope. Even worse, sometimes functions in the real aircraft aren't even simulated in the multi million dollar Level-D sims, see MCAS.

That's not my experience at all. I worked in the development office in a flight sim complex (6 sims) for a few years so got to see first hand how the testing and acceptance process works. Level D sims are certified as "zero flight time" and, as a result, crews can be type rated in them without actually flying the physical aircraft. All level D sims are supposed to accurately simulate all systems accessible from the cockpit which are critical to the operation of the aircraft so it's possible that some non-critical functions may not be simulated. Many (most?) sims don't accurately simulate the edges of the operating envelope because the training carried out in them should prevent the crews ever getting that far.

On 2/3/2020 at 7:28 AM, FDEdev said:

I had the honor to fly a few month with a brilliant Dash8-300 test pilot and it was an eye-opener during that time to learn how dangerously wrong even a Level-D sim can be.

Did he report his observations? If so, then someone wasn't doing their job correctly. Our sims were periodically checked by test pilots and any handling or system errors were immediately referred back to the manufacturer who was expected to correct them in the next sim software load (or sooner if it was a critical error). You'd be surprised at the seemingly insignificant things they were expected to fix. Anything "dangerously wrong" would have caused the sim to lose its certification.

On 2/3/2020 at 7:28 AM, FDEdev said:

I remember at least 2 aircraft types where it took you a few minutes during every simulator check to get used to the noticeable more sensitive and spongy handling of the sim. If you were flying the real thing the day before, you inevitable rotated right into the stickshaker during the first takeoff if you tried to fly it like the real one.

Were these sims level D? The FAA, for example, specifies the control forces, displacement and response time (usually in hundredths of a second) in level C and D sims based on actual aircraft data from the manufacturer. I can't see how they would be allowed to retain level D certification if they were as bad as you describe.

Edited February 4, 2020 by vortex681

[Paraffin 2,197]

4 hours ago, rjfry said:

Could MSFS end up on this,

https://www.tomshardware.com/uk/features/nvidia-geforce-now-launches

Priority streaming.

Not a chance. Microsoft has their own streaming infrastructure. 

[G-RFRY 1,708]

26 minutes ago, Paraffin said:

Not a chance. Microsoft has their own streaming infrastructure. 

More exposure more customers why put it on steam then just have it XBOX exclusive and PC.

[FDEdev 955]

6 hours ago, vortex681 said:

1. Level D sims are certified as "zero flight time" and, as a result, crews can be type rated in them without actually flying the physical aircraft.

2. Many (most?) sims don't accurately simulate the edges of the operating envelope because the training carried out in them should prevent the crews ever getting that far.

3. Anything "dangerously wrong" would have caused the sim to lose its certification.

4. Were these sims level D? 

1. This stupid money saving idea obviously wasn't invented by pilots. We were the first (and the last) bunch of pilots who transitioned to a Heavy without doing at least a few circuits with the real aircraft before starting line ops. Since we were the only group of pilots where they tried that, you can draw your own conclusion about the handling qualities of a Level D sim.

2. If the sims don't accurately simulate the edges of the operating envelope, then we are talking about expensive procedure trainers. What's the use of a simulator if I can't simulate everything correctly, especially the edge of the envelope? In this case one could do the whole type rating without a sim on the actual aircraft, which is was happened to me on my first 4 engined turboprop, because a simulator for that aircraft simply didn't exist. 

3. That's wrong because you correctly observed that 'Many (most?) sims don't accurately simulate the edges of the operating envelope'. Unfortunately you have to fly the aircraft at the edge of the envelope during some of the semi annual sim sessions.  

4. Yes of course. 

Edited February 5, 2020 by FDEdev