Instant Scenery V3

[John Young]

I’ve never been a great fan of instant scenery-type programs, firstly because of the difficulty of keeping track of object libraries and secondly the extremely bad effect on frame rates if you overdo it. That said, it’s a very good tool to get a quick result if you don’t want to spend 6-24 months designing custom scenery in Gmax. I have though, just tried the demo version of Instant Scenery 3 and there are some new features that are really useful.

IS3 will place Airport 1 and 2 grass polygons visually live in FSX. You could also use some of the other grass textures too to give a bit of variety. The real benefit though is the customisation of the Landclass scenery surrounding an airfield and the placement of lakes, rivers, roads (with or without traffic), railroads and telegraph poles. It took me just an hour to understand the program and do this test enhancement to an area of Seattle that falls within the 50 mile demo area:





The program has no flatten facility, so I used ADE for that when I added the runway in the same program.

When I’ve changed Landclass before, I’ve used EZ_Landclass by Russell Dirks. It works with an Excel spread sheet and the tiles are rectangular. IS3 just does it visually in FSX and with any shape you want, it’s so easy. That said, there has to be a need to change the Landclass in the first place. If you use photo-scenery or if you are content with the enhanced Landclass appearance that OrbX offers then the need may not arise. However, the program will manipulate Orbx Landclass just as well.

Although there is some capability to change coastlines, it is not possible to extend land into sea or even create an island. You can however extend sea into land, because sea has a higher priority. Adding roads with traffic is though, extremely easy. None of these terrain features are available in FS9 unfortunately, which is a real pity.

It’s well worth having a play with the demo version. Two tips though - when I launched the program the first time, it told me that the trial period had expired. After a Google search of the problem I discovered that uninstalling it and installing it again got it working. The second thing to note is that you may be mis-led into thinking that the polygons float. They do initially, but you need to refresh the scenery (re-launch FSX) to integrate them with the default scenery.

http://www.flight1.com/products.asp?product=iscen3

John

[gsnde]

I have bought it and agree the new features are cool. I don't use it very often, and for me it feels that it is a bit less intuitive in handling placed objects compared with version 2 (but that might be just me).

[BadPvtDan]

John, how did you place your scenery before?

[DaleRFU]

it looks a great tool, how does it compare to others such as ezscenery?

[John Young]

I use ADE to place 3-D objects, but it might be helpful if I explained why.

I normally use a photo polygon for the airfield ground, overlaid with runways and taxiways etc made with AFX. It's why I was interested in the new ground features in IS3 for FSX, because making ground tiles is really quite difficult and costs money. My 3-D objects are custom made with Gmax and are placed as models, not library objects, with ADE.

There are 3 advantages to that:

- Models are built as large Gmax scenes each scene sharing a 1024 x 1024 texture sheet. So for example, the whole of the weapons complex for an RAF station might be 1 Gmax model with just 1 texture sheet. That will limit the draw calls to just 1 for that whole area.

- For objects like whole plantations of trees, several hundred trees can be built as one model and aligned with the shape on the ground. The trees are attached to each other and, very importantly, all the common texture vertices welded, because texture vertices are more critical than polygons for performance. That means you can have a whole forest of trees with just 1 draw call and they are super efficient frame rate-wise.

- If you put an object into a library file and distribute it, anyone can use the model for whatever project they like. I don't really want to see my custom scenery models appearing inappropriately in different parts of the world, especially without asking me first. It tends to kill the whole point of making custom scenery unique.

There is a huge drawback to this approach of course and that's time. Airfields like Duxford took two of us 2 years to build with perhaps 4000 hours of effort between us. It took Manfred and Pete 3 years to build Gutersloh. It's like designing into a black hole because 99.7% of people who download the finished project will just grab it and never acknowledge it. It's really difficult to re-motivate again after experiencing that year after year. We've lost two top rate designers over at ACG in the last couple of months because of that and it's why I switched to AI aircraft. I can complete a project in 2 months and not get bogged down or worry if it's not been time well spent.

Instant Scenery is a far far quicker tool to use, especially with third party libraries and most certainly has enormous value. I've tested scenery built that way for people over the years and it can look good. It's just a pity that some designers who asked me to test didn't take frame-rate measures as they were going along and realised earlier that they were producing a slide show, trees being the biggest influence. With so many libraries available, I find them impossible to manage. I have 2806 library files on my PC taking up 352MB. I don't know what scenery they relate to and I newer really use them, but just keep them in case they might come in useful some day. I am really reluctant to add more.

Just my personal rationale. It's a hobby and enthusiasts should of course use whatever tools suits their interest, level of skills and available time.

John

[LEBTowerGuy]

I use ADE to place 3-D objects, but it might be helpful if I explained why.

I normally use a photo polygon for the airfield ground, overlaid with runways and taxiways etc made with AFX. It's why I was interested in the new ground features in IS3 for FSX, because making ground tiles is really quite difficult and costs money. My 3-D objects are custom made with Gmax and are placed as models, not library objects, with ADE.

There are 3 advantages to that:

- Models are built as large Gmax scenes each scene sharing a 1024 x 1024 texture sheet. So for example, the whole of the weapons complex for an RAF station might be 1 Gmax model with just 1 texture sheet. That will limit the draw calls to just 1 for that whole area.

- For objects like whole plantations of trees, several hundred trees can be built as one model and aligned with the shape on the ground. The trees are attached to each other and, very importantly, all the common texture vertices welded, because texture vertices are more critical than polygons for performance. That means you can have a whole forest of trees with just 1 draw call and they are super efficient frame rate-wise.

- If you put an object into a library file and distribute it, anyone can use the model for whatever project they like. I don't really want to see my custom scenery models appearing inappropriately in different parts of the world, especially without asking me first. It tends to kill the whole point of making custom scenery unique.

There is a huge drawback to this approach of course and that's time. Airfields like Duxford took two of us 2 years to build with perhaps 4000 hours of effort between us. It took Manfred and Pete 3 years to build Gutersloh. It's like designing into a black hole because 99.7% of people who download the finished project will just grab it and never acknowledge it. It's really difficult to re-motivate again after experiencing that year after year. We've lost two top rate designers over at ACG in the last couple of months because of that and it's why I switched to AI aircraft. I can complete a project in 2 months and not get bogged down or worry if it's not been time well spent.

Instant Scenery is a far far quicker tool to use, especially with third party libraries and most certainly has enormous value. I've tested scenery built that way for people over the years and it can look good. It's just a pity that some designers who asked me to test didn't take frame-rate measures as they were going along and realised earlier that they were producing a slide show, trees being the biggest influence. With so many libraries available, I find them impossible to manage. I have 2806 library files on my PC taking up 352MB. I don't know what scenery they relate to and I newer really use them, but just keep them in case they might come in useful some day. I am really reluctant to add more.

Just my personal rationale. It's a hobby and enthusiasts should of course use whatever tools suits their interest, level of skills and available time.

John

Well said John, I agree about the models vs library objects, while I started out using some common library objects to build airports (thanks to the guys here and the ABO library). I have found that I really don't want my custom objects that I have spent months researching their design, modeling, and texturing, only to have show up in some half-assed rip-off project. This appears to be becoming the norm, at least here, there are others that handle their scenery in a similar fashion. I have also found that in the end it is easier to place the models in the scenery, the ADE page looks less crowded with a few large objects rather than dozens of tiny individual ones.

[BadPvtDan]

I rarely make library objects by the,selves an combine them with other things t make that a less desirable option. Thanks for the detailed explanation!

[Sophie_Westenra]

I’ve never been a great fan of instant scenery-type programs, firstly because of the difficulty of keeping track of object libraries and secondly the extremely bad effect on frame rates if you overdo it. That said, it’s a very good tool to get a quick result if you don’t want to spend 6-24 months designing custom scenery in Gmax

Hi,
I'd just like to say that I've used all of the Instant Scenery programs and all of them allow custom made and default models to be placed onto your custom made airport or default airport and all of them allow easy tracking of your model objects.
I've been making custom models and objects for years and while I use to have to watch how many objects were being placed back a few years ago when the Pentium 4 CPU was the best PC on the market, these days it's just chalk and chesse to make a comparison.
Instant scenery will view any model or as in the case of John Young, a series of grouped models, as being a single model for placement. As long as you use a single texture sheet as John young has mentioned you will not have any issues with texture mip maps going blurry. Down side in FS9 is that 1024p texture sheet fills up pretty quick if you want fairly high res textures on your models. Remembering DXT1 is a highly compressed format and naturally causes much less resolution to occur in the simulator than what you will see in your drawing program.
You can of course opt not to use mip maps on your textures; this will not have much of an impact on your PC resources if you have a fairly new CPU like an i5 or i7 with a moderate graphic card. Again if you have the PC power, like most late model PC's straight off the self now have, you could possibly even use 32 bit color textures or part thereof.
Worrying about draw calls on scenery models these days seems a little out-dated. I've seen single payware FSX airports that are over 1GB of photo quality models and the draw is huge but some how the PC is able to handle it without to much effort.
I believe in this day and age there is far to much apprehension about draw calls, unless you're still using old tec then it's time to test your limits of your PC to see what you can and can't have but most people who have upgraded there PC in more recent years will have no effort running any FS9 scenery, no matter how many draw calls it has.
Sophie

[John Young]

Hi Sophie,

Yes I agree, IS programs can place large merged scenes as well as singleton models. That's even more of a reason why I prefer not to use them for that purpose. Large scenes might comprise a control tower/fire station/vehicle complex, a third of an RAF technical site, a 5-C-type hangar "Waterfront", a HAS Squadron complex, or a weapons compound for example. They are all coherent models that take a long time to make and while I'm happy to help if people ask me, I prefer not to put the models in a library and had them to the world on a plate.

I find that Mipmaps are essential with 1024 x 1024 texture sheets, not because of frame rate but because of stuttering as the sheets load. Try flying a tight circuit around a HAS site with 12 AI aircraft without mips and see what I mean. Draw call management with large scenes of attached objects, welded common texture vertices and LODs is still very important for good performance even with an I7 processor, if you happen to have one. This is a very useful reference from Microsoft as to why that's the case:

http://blogs.technet.com/b/torgo3000/ar ... atter.aspx

Here's an example of a small scenery that I was asked to test a little while ago:



The frame rate at the time of the screen shot is an unusable average of 10.7 fps and has not exceeded 12.5 fps at any time over the minute I ran the test. This was with an I7 2600K, 3.4Ghz processor and Nvidia GTX 580SC/1536MB card.

The scenery was built with a "pepper pot" approach from generic libraries. Not all of the models are authored well, there is no opportunity to attach into composite scenes so that performance can be optimised and there are no LODs that I can detect.

It's really the ease at which users can get carried away with instant scenery programs and third party libraries without periodically checking the frame rate that I think is the pitfall. That happens too of course with custom building and placement with any tool. It's why Gutersloh took 3 years to build. All credit to Manfred who was a die hard IS/library designer until he realised 2 years into the project that performance was suffering, particularly after allowing another 10 fps for AI and testing with my I7 processor. A large chunk of Year 3 was spent re-engineering in the way described, to bring the performance up.

IS programs most certainly have their place, but an I7 processor is not a magic bullet for poorly authored, singleton models, scattered in too great a quantity with the ease that these programs allow, particularly for the 60% of users who run FSX.

Just my view.

John

[Sophie_Westenra]

That is fine John; I'm not in disagreement with your personal view of the product.
I’m just one person whom has enough some self control to handle the products huge capabilities without being in fear of it destroying my simulator experience.
I would not advocate that mip maps not be included on AI aircraft texture sheets at any time, that will almost certainly cause shutter issues on lesser CPU's.
In my experience, I have found 99.9% of Airports operate at a high end frame rate and have zero shutter effect up until the time that you pack the airport with poorly made AI aircraft that have poorly designed lods, if they have actual lods at all.
Most Airports, even those with dozens of models and texture sheets are only equal to a hand full of AI aircraft in both poly count and draw call.
Like I said, most of the issues caused in poor perforance are a result of AI Aircraft, not the Airport it self.
That's my experience but may be I have a bit more CPU power than most users ?
Sophie

[Firebird]

Speaking as a user of scenery rather than a developer I can appreciate what both of you are saying. After spending a lot of time discussing optimal scenery configuring with John's colleague, Ian, I saw first hand the benefits that these optimisations can bring and the difference they make to the users experience.

What was equally apparent was the amount of time and effort that goes into making these magnificently detailed scenery. A round of applause for all those that have the skill and patience for these masterpieces. A consequence though is that we would only get a few of these sceneries and with the ever depending user, I include myself here, for more accurate/realistic scenery there is a need/market for simpler/quicker scenery enhancements which the modular library system is ideal for. I commend people like Stew and Brian for producing these enhancements for my system.

I can understand that these are not optimised in anyway. As I understand it this is for two reasons, one the mdls in the library could be made more system friendly and second that scenery made using them use individual mdl placements rather than multiples melded together, when applicable.
The current perfect example of this is when near Alconbury with Wittering in shot (both with full traffic) and my FS9 almost stops dead, 3-5 fps, on occasion and my system is fairly high end.

I am guessing that the people that built the libs are probably not going to go back and redo the libs to make use of the latest fps friendly methods. I can't really blame them as every modeller I have aided from time to time has always wanted to move onto new subjects rather than revisit ones they have finished.
Taking this into account if newer versions of these libs were produced, or new ones produced, what techniques should be applied to the mdls, or even the library compilation, to make them system healthier?
Is there a better way or using these libs to make healthier enhancements or is it a case that nothing can be done because of the library system?
Maybe a more interesting question is there a better way of producing system friendlier scenery enhancements than using the library system, taking into account the reasonably rapid creation time and the nature of our method where a lot of developers would need access to the mdls?

Or is it simply the case that the method that we use is simply the best we can do and we have to live with the limitations if we want quick enhancements?

[John Young]

I'm not sure there is much you can do when using third party generic library objects. Many will work fine, others won't. By their very nature, they can't be attached together to share a common texture sheet, common texture vertices welded and LODs added. If on the other hand, the models are being custom made, all of these measures are possible if you choose to use them. Whether you then place them with an instant-scenery type program or something else is a case of personal preference. If you are doing the groundwork in ADE anyway, you might like to place with that. I like the Gmax and ADE combination. Others will like FSDS and EZ or Instant scenery. There's no right or wrong way, just different characteristics for each.

I wouldn't assume that every user of our output has an I7 processor and high end graphics card. What I would do, is measure the frame rate of the design regularly and if it's heading below 30 fps, to pause and re-think the content. As Sophie says, it's a case of self control. I say 30fps because, 20fps is a comfortable minimum for the scenery, leaving 10 for an AI package. My target for detailed custom scenery is at least 50 fps because I do have a large safety margin with an I7 processor. I've just measured Duxford at 300ft just before the 06 threshold (the critical point for frame-rate, particularly in a cross-wind) and returned 95 fps with no AI and 110 fps with 20 different AI aircraft types.

If anyone is interested and doesn't know how, the minimum, maximum and average frame rate measure (shift ZZ) in FS9 can be switched on by adding the following lines to your FS9.cfg file (first entry in the [MAIN] section:

AVE_FRAME_RATE_DISPLAY=1

The tweak for FSX is done in the same way, but is a bit more complicated. PM me if you would like it.

You need to measure in Windowed mode in both sims.

John

[Sophie_Westenra]

I agree, the best way to make scenery is to continually test the output as you progress. You will notice any large hits as your producing it. If you’re worried about frame rates with AI on the ground then simply add some parking to an afcad early in the development stages and assign some AI to sit on the ground and work around them…
Don't be scared to have a go, you can always remove troublesome models from a scenery and Instant Scenery 3 is fabulous at doing that.

[The current perfect example of this is when near Alconbury with Wittering in shot (both with full traffic) and my FS9 almost stops dead, 3-5 fps, on occasion and my system is fairly high end.

Steve, do you get poor frame rates when the AI traffic is removed from one of those Airports???

Sophie

[Firebird]

Hi Sophie, I admit that I hadn't tried it w/o AI but when I did I found that the two addons by Brian were able to co-exist happily with my locked fps set at my default 25. So it would appear that the AI traffic at the two bases are the cause, or at least the mixture of two sceneries and two lots of AI traffic.
I must admit that I was surprised at that. I had never experienced that before neither at Heathrow nor at the Woodbridge/Bentwaters twin bases.

[John Young]

Sophie has a very valid point about the quality of some AI and the effect on frame rate performance. I've seen recently an "AI" offering included in a pay-ware scenery package that had just 3 LODs and 23,000 polygons in the first one for each aircraft. That first LOD count is about the same as a good quality flyable aircraft. MAIW have a reputation for good quality AI and I'm as surprised as Steve, if it's MAIW aircraft causing the performance loss.

Steve's mention of Woodbridge and Bentwaters is really relevant to this discussion. It was built by Ian and I for the first time using the principles in the Microsoft Texture Vertex/Draw call post. It was a complete breakthrough in performance for what is not one, but two very detailed sceneries that are just 2 nautical miles apart at the closest point. Both bases together return a frame rate figure of 84 fps without AI and 74 fps with it, with my I7 processor:



The 3 performance techniques were critical - large Gmax scenes sharing the same texture sheet to reduce draw calls and to enable Tvert welding and LODs to take out the detail within a short range, so that one airfield doesn't have an undue influence on the other.

What's also quite remarkable is that there were 84 AI aircraft in the scenery when I took the "with AI" measure, comprising A-10s, F-16s and C-130s. It works because like the scenery, there is a crucial final "paper aircraft" LOD that kicks in to reduce the aircraft to just 4 polygons (mesh triangles) before they can do any damage. Here's the final LOD from the A-10.



The Bentwaters/Woodbridge illustration is a good example why instant scenery-type programs didn't feature in the design. ACG scenery designers are interested in airfield construction, particularly historical ones. The "Twin" bases are an accurate representation of two USAF bases in the UK that closed in 1993. They were made with several visits to the airfields, thanks to the museum staff at Bentwaters, to get the photographs necessary for the photo-textures. It is just not possible to recreate airfields accurately like this with 3rd party object libraries, never mind achieve the level of performance we did. We also didn't want to place the custom models in libraries to be placed with one of these programs because we didn't want to make the models available for anyone to re-use any where in the world. It would kill the unique nature of the scenery and isn't particularly fair given the amount of time spent creating them and when very few people are likely to thank you for them.

As I said before, that's not to say that instant scenery-type programs don't have a place. They do of course. One thing about building the "Twins" is that it wasn't "instant".They required a lot of time, skill and determination to get to the end. Not everyone can do that and IS3 and the like offer a far faster alternative, particularly if time and skill is lacking. It's a really good way to get into scenery design in the first place. It's why I posted the link to IS3 in the first place.

John

[Sophie_Westenra]

Steve,

It would suggest going by John's observations that you have at least one poorly made AI at one or both of those Airports. As it would be very highly unlikely for MAIW to have packaged a poorly made AI, you are most likely looking for an aircraft/s that you have downloaded else-where. It sounds like this aircraft does not visit your other busy Airports like Heathrow so that may nail it down for you.
I'm glad this discussion has helped you in some way.

John,
I'm not sure if your aware or not but most developers know about a little program that I won’t name here, which is essentially a GUI decompiler for FS9 and FSX bgl scenery files for which can easily provide a path to convert them to library files or alternatively into the old fashion xml code. From there they can be authored to appear anywhere in the simulator using lat/lon and height.
You can not stop people from using your custom models if they are very determine to do so.
It’s really the unauthorized use of ones own photography turned textures that tends to upset me for it’s the textures that take the bulk of your time, energy and sometimes out of pocket costs collecting them. All you can do is make it harder for people to find your textures by hiding them deep in the world scenery folders with file names liken to MS default files, but then there are ways to find them as well without to much effort if you know a few tricks.

Sophie

[Firebird]

John, I can confirm that I am only using what we would call true AI aircraft. F-5E, A-10 and TR-1's for WZ and someone famous did the Harriers at XT.

[John Young]

Yes indeed Sophie, I'm aware that a .bgl file can be interrogated like that. Not much I can do about that, but I can at least not make it easy for users to grab the models from visible library files.

John

[sr01]

John
Since you slid the question of AI into this thread .Would a separate small Texture file for canopies have an undue effect ?
Can that be incorporated onto one Texture file ?

[John Young]

Using one texture sheet for several objects in scenery design reduces draw calls, but probably more importantly, it enables common texture vertices to be welded. So for example the hundreds of texture vertices on say control tower handrails, steps and railings (if you build them in 3-D) could be welded together with those on an adjacent buildings with rails so long as they are attached together and use the same mapping on the sheet. The Tverts can be reduced from the initial several hundred on the rails to just 4. It's still a large number of polygons, but as Microsoft say, polygons don't matter, Tverts do. The opportunity to weld Tverts on aircraft is nowhere near as great because not a lot of them are common and it's hardly worth the effort and the risk of screwing up the mapping.

I've been building scenery for 17 years but I'm relatively new to AI. I take a lot of my learning from the masters at MAIW and note that the norm is to use just one 1024 x 1024 texture sheet for all parts and I try and stick to that. However, I built the B-29 with two sheets because of it's size. I couldn't detect a real penalty in doing that. If the canopy is simple, I usually include the frame colour and the Perspex on the main texture sheet as two small squares (mental note for myself - must remember to label them for the re-painters). However if the canopy frame has text or symbols on it - like the pilot's name or rescue arrows, I sometimes put them on a separate 128 x 128 texture with it's own alpha channel if things are tight on the main sheet. I did that for the Harriers and they seem fine.

Hope that makes sense.

John