AF99 vs. SCASM : The Ultimate Confrontation... well... not really...

[hubbabubba]

This evening, Ivan sent me an e-mail about a topic that, I think, may be of interest to some of you. With his permission, I'm reproducing it here;

Hello Guy, (note: yeah, that's my first name; Guy, pronounce it like the "gui" of guitar)

Earlier, I described the necessity of minute adjustments and being able to SEE the results of these adjustments changing contour lines. I found that I had to put an additional bulkhead at the front edge of the windscreen. I found that I had put the top edge of the fuselage about 0.03 foot too low at that point because I had done a single unbroken line from windshield's vertical frame to the next bulkhead forward.

....Concave.jpg shows the situation as I added the new bulkhead at the front edge of the windshield. As you can see, the outline then became CONCAVE which is not right. It should be convex or straight at all points. (I named this file wrong: The part is actually Y1.)


Attachment 13385

....Convex.jpg shows the result after adjusting the top points of the next two or three bulkhead upward to re-create the convex shape. (I named this file wrong: The part is actually Y1.)


Attachment 13386

....FrontalCheck.jpg shows a contour and streamline check of stringer Y2. (Y1 is on the centerline and doesn't show in this screenshot.) Note that absolutely NONE of the parts in this screenshot will appear in the model. These are the equivalent of my building templates and jigs.


Attachment 13387

This check and adjustment was done in about a half hour. If this were a purely SCASM project, I don't believe I could have done this check and adjustment at all. Do you know of a way this can be done in SCASM? This "misalignment isn't usually visible in the simulator except via odd shadows in unusual places. I fix them because I KNOW they are there.

Take Care,

[Ivan]

Hi Hubbabubba,

As you were pointing out, this isn't really a "Confrontation". It is more about the limitations of each method of building. I believe that a combination of both methods produces the best results.

AF99 has a pretty stupid assembly process. SCASM does allow for visual checks.

That third screenshot with all the front bulkheads displayed gives a pretty good impression of how I built the nose section of the P-40C. You might be wondering why some of the bulkheads are not symmetrical:

Sometimes this is because I am making a change that I am unsure of. The right side retains the pre-modification dimensions.

Sometimes the left side has a series of points that need to be as co-linear as possible. The missing points on the right side forces that segment to be a straight line which I can use as a reference to scoot a point fore or aft slightly for best alignment.

- Ivan.

[hubbabubba]

Well, Ivan, I thought that "Visual vertices adjustments under AF99 compared to SCASM" was a bit dry for a title...

I had pretty much figured-out what you meant by "seeing" your model but, just to be sure, I assume that the line going left from the nose-spinner junction to the last bulkhead is the one we're talking about?

If so then, yes, AF99 is doing, for such job, a far better job than SCASM would, no contest.

AF99 is doing many things; it is a project-maker, a 3D CAD, a texture placer, a basic animator, and an assembler, all rolled into one program.

I simply think that it is an excellent 3D CAD.

The "Project Wizard" is a necessary evil (how many projects did I erased with that contraption in the beginnings!).

The texturing tool was created to make us cry and give up. Which I almost did until Signore Andrea Cini, creator of many AF99 add-ons, came to my help in pointing to me that American numeral notation is different than French method. Grazie mille Andrea! Once I had set my "Paramètres régionnaux" to take that into account, I was still crying... but, now going somewhere, I persevered!

The "one time" animator is doing his job, as long as you don't bother about gears popping in and out and silly prop discs. After all, Aircraft Animator wasn't created for nothing.

I could go on with the help files (not helping much), the keyboard functions that you have to learn by literally tripping on them (oh! so it can do that!), or the tutorial which show you how you will never do anything except a balsa model, but I won't.

As for the assembler, until our exchanges with Bretoal, I was living in a state of partial bliss born out of ignorance. I was, like you Ivan, slightly annoyed by AF99 parts and components quotas. I was also mystified by models that would not show properly, even when still under the parts/components limits.

But, after Bretoal's teachings, and the now famous "Ivan's conga", I started exploring models and was horrified, it is not too strong a word, by the incongruous mayhem coding procedure. I understand that in a "one size fits all" assembler, some unnecessary code is present. But it does not explain all of the "quirks and quacks" I found.

Enough ranting. So AF99 is good. That's why I'm using it pretty much in the same way you do. Here is proof of that;

Attachment 13398

Of course, like you, it is mostly as jigs or templates that these parts were used. Look, for instance, at the two wood blocks on the hood. Their actual shape on the released models is quite different. But I try to solve any possible visual issues before stepping into SCASM. In the worst case scenario, I can always return to the "blueprints" and modify, or redo, a part/component (I never use structures).

As for concavities, the jeep is full of them! In your own example, I'm not even sure that your original concave parts would show in game. You would know it is there, but you would probably be the only person on this planet aware of this ugly blemish! LOL! In "limit cases" like this one, AF99 will probably do as if nothing happened, and gamers would be totally oblivious of that "imperfection".

Next time, I will tell you why I "abandonned" AF99.

[Ivan]

Hi Hubbabubba,

I didn't know that your initial build for your Jeep was with AF99. I figured that your issue with relocating the floor of the Jeep was an issue because it was all SCASMed. From AF99, that relocation would be easy!

I DO notice something that needs a slight clarification though, or perhaps I already know: Your Jeep in AF99 has a rear wheel well. In your actual model, it appears not to have a wheel well because the cutout is a transparent texture. Is this because when the concave part is textured, the concavity (wheel opening) disappears?

Actually I have an interesting solution for my next large project which I anticipate will be a B-26 Marauder in honour of my neighbour Col. Wallace Boblitt: The aircraft will NOT be a single "Assembly", and will not be entirely "Buildable" using AF99. Its assemblies will require manual final assembly inside SCASM. That will allow the separate AF99 assemblies to be built and flipped around but each won't be the entire aircraft. They will have 2D templates to represent the parts that are missing so I can get an idea of the actual shape of the final product though.

Perhaps I should apply this technique to my incomplete Heinkel He 177? The B-25C won't get this treatment because it was a proof of concept that a decent twin engine bomber can get built within AF99 limitations.

BTW, you asked earlier (facetiously I believe) what my motto was. I told you, "As Real as I can Make It". THAT is why I keep adjusting the contours on my P-40C when I know there is a problem even though I am fairly certain the results are not really visible in the simulator. That is also the reason why my projects take forever.

- Ivan.

[hubbabubba]

I DO notice something that needs a slight clarification though, or perhaps I already know: Your Jeep in AF99 has a rear wheel well. In your actual model, it appears not to have a wheel well because the cutout is a transparent texture. Is this because when the concave part is textured, the concavity (wheel opening) disappears?

I'm still a bit puzzled by that one. "Your Jeep in AF99 has a rear wheel well."; no, it doesn't have a wheel well. The inside panel is the internal polygon of the back-sides. The external panel is a single trapezoid polygon which is textured with a transparent bitmap to create the "well". The wheels (right and left) are "glued" to their internal side steps (the slated front part and the horizontal part where the "toolboxes" are) and the "step-wheel" sub-assembly is "sandwiched" between outside and inside panels mentioned above. This is the trick here. Nice or what?!

The idea of sub-dividing a project in sub-assemblies was at the base of my exploration of SCASMing. Your conga was the other part of the equation. I simply pushed those principles to their logical conclusion.

In AF99, the architecture works a bit that way;

Is the POV over or under the the CoG?

If the answer is "under", then is the POV right or left of the CoG?

If the answer is "right", then is the POV front or aft of the CoG?

If the answer is "front" then show tail parts first, left wing parts second, cockpit parts third... etc...

Of course, this is only one branch of a very dense tree.

With my method, the architecture goes like that;

Glue part A to part B (a "part" can be a polygon, a component, or already glued "parts")

Glue part AB to C

Glue part ABC to D

Glue part Z to Y

Glue part ZY to X

Glue part ZYX to W

Glue part ABCD to ZYXW

Show part ABCDZYXW

This is, too, only a short sequence.

Try to imagine a jeep made with AF99 architecture. I've tried, but I can't. My imagination doesn't go that far .

I'm preparing a post where I will talk scale and precision of the two techniques. But feel free to ask other "impromptu" questions. Likewise for anyone interested.

[Ivan]

I'm still a bit puzzled by that one. "Your Jeep in AF99 has a rear wheel well."; no, it doesn't have a wheel well. The inside panel is the internal polygon of the back-sides. The external panel is a single trapezoid polygon which is textured with a transparent bitmap to create the "well".

Please see the attached image. I highlighted the outline of the wheel well that I can see in your AF99 model. Perhaps you didn't actually put this in the actual model?

In AF99, the architecture works a bit that way;

Is the POV over or under the the CoG?

If the answer is "under", then is the POV right or left of the CoG?

You just described one of the seriously "Goofy" things that AF99 does. The first viewing decision is Top-Bottom and should not be. Fore-Aft and Left-Right decisions should be made first. Also there is no possibility of a Fore-MID-Aft sequence when all the viewing decisions are made before any parts are displayed.

I will continue in another message.

- Ivan.

[Ivan]

I was having trouble with dropped connections in my last message. Not sure why.

In showing your gluing sequence for ABCDZYXW, I believe you left out a very useful SCASM technique that I am sure you know:

With AF99, all pieces must be separated by "Glue" (viewing plane) that determines when to show part AB and when to show BA. Sometimes this determination doesn't need to be made, but there is no way to avoid this with AF99. This is one of its serious weaknesses.

Let's use the example of the common Radial engine. I attached a screenshot of a A6M2 Zero to illustrate:
The engine is made up of the following parts:
A. The face of the frong cylinder bank
B. The Interior of the cowl in front of the cylinders
C. The Reduction Gear Housing (Cone in the center of the engine)
D. The Exterior of the Cowl.

The viewing sequence is ALWAYS ABCD. There is NEVER a case in which a decision needs to be made. A never displays in front of B which never displays in front of C which never displays in front of D. SCASM can handle this. AF99 can't.

Now there ARE some screwed up ways to do this in AF99 such as putting the engine face in the Tail assembly, but you can't do this when there are a bunch of parts that need this treatment. I have seen folks put a Left side cockpit wall in the right wing assembly, Right cockpit wall in the left wing, etc. with each part as an insignia facing the opposite direction so that it doesn't show when you are on the side of the assembly, but this is seriously non-intuitive and a pain to track down.

- Ivan.

[hubbabubba]

Please see the attached image. I highlighted the outline of the wheel well that I can see in your AF99 model. Perhaps you didn't actually put this in the actual model?

They say "an image is worth a thousand words"! How true! Yep, you're right, this part was only made to localize the "best" angle to cut between the rear and central "parts" (as in ABCD part). That obviously concave polygon never made it in the final model.

The viewing sequence is ALWAYS ABCD. There is NEVER a case in which a decision needs to be made. A never displays in front of B which never displays in front of C which never displays in front of D. SCASM can handle this. AF99 can't.

This is probably a case where my background in scenery macros was of help. We (scenery designers) often make entire buildings without using a single glue part. For example, the sequence of a simple house with a slanted roof would look something like that;

:simplehouse

call( :fronthouse )
call( :backhouse )
call( :leftsideHse )
call( :rightsideHse)
call( :leftroof )
call( :rightroof )
Return

Since the internal parts are never to be seen, there is no possible interference from the external parts. Better yet; suppose you want to be able to get in that house;

:simpleHseinout

call( :fronthouseins )
call( :backhouseins )
call( :leftsideHseins )
call( :rightsideHseins )
call( :leftroofins )
call( :rightroofins )
call( :floorins )

call( :fronthouse )
call( :backhouse )
call( :leftsideHse )
call( :rightsideHse)
call( :leftroof )
call( :rightroof )
Return

By placing the internal parts first in the showing order, the external parts cover them from the exterior. Once inside, you see only the interior parts. With transparent textures you could even have a look inside the house and, if a window is on the opposite side internally, a view of what is behind the house. Not a single glue part in that.

Which brings me to my next question; if your "one-way" ABCD sequence for A6M2 Zero can't be done in AF99, how did you do it? I know that AF99 can apply parts as "insignias", which are not glued. Was is your trick?

[Ivan]

Hello Hubbabubba,

The "Trick" with Radial engines is something I worked out way back when I built my first Corsair. In this case, it is better illustrated with screenshots from the P-40C which uses the same technique with its radiator opening:

The Keel shot shows an entirely medium green nose up to the opening. One frame back from the opening is where everything comes together. The front face of the radiator is there. It is glued to the cowl not at the edge of the component but inside it. The cowl interior wall component is also glued at the same location sequenced AFTER the radiator. The radiator baffles are glued after the cowl interior, so up to this point, if we are forward of this frame, we see the baffles in the foreground followed by the cowl interior followed by the radiator face. The last component is the dark red cowl exterior that is also glued to the same frame so that when we are forward of that frame, we see it over everything else. If we are behind that frame, we see the green nose over everything else.

That pretty much illustrates the technique. There is one other issue: Because the baffles are glued forward, the last one glued will bleed through the opposite side. That is why there are insignia versions of each side (facing out) that are glued after the ones facing in.

The old P-40E was built with a much more complicated sequence and has a couple very slight bleeds when viewed from below. Knowing what I do now, it could have been built without those bleeds.

- Ivan.

[hubbabubba]

PRECISION

AF99 works to the 1/100th of a unit. With the usual settings, the unit is one foot. Therefore, with AF99, you can, in theory, adjust to 1/100th of a foot.

But this is only in theory.

You see, when the assembler within AF99, a commercial flavor of SCASM btw, is doing its job, it has to adjust to SCASM scale: the meter. It also has to adapt to the internal scaling. Without going into details, the scaling command is "SetScalex", and it is set at "7". The mathematical relationship of this is that each units represent 1/512 of a meter, or 0.001953125 meter.

How much does a foot represent in meter? According to my converter program: 0.3048 meter.

How many SCASM units does AF99 take to represent a foot? 156

156 x 0.001953125 = 0.3046875

Pretty close... but...

To get an inch, you must round up: 1 foot divided by 12 = 0.0833333333333... , or 0.08 within AF99 limits.

How many SCASM units does AF99 take to represent an inch? 12

But wait a minute! 156 divided by 12 = 13! More than that; 12 x 12 = 144, not 156. In fact, 156 - 144 = 12! So AF99 foot is 13 inches long!

Now, AF99 can render 1/100 of a foot, or 0.01 foot.

How many SCASM units does AF99 take to represent 1/100 of a foot? 2

SCASM units are integers. You can't have 1.56 (156/100) for a unit, so it is rounded to 2.

Of course, with SCASM, you can go to one unit. The following animated gif shows an AF99 square foot (white), an AF99 square inch (blue), an AF99 smallest resolution of 2x2 and, finally, a "SCASM only" 1x1 square.

Attachment 13470


The difference in precision may look trivial, but it is not. In lengths, it double the precision, in surfaces, it quadruple it and, in volumes, it multiply it by eight. If you remember, Ivan, the octagonal cross section of the MG canon holder was "lenticular" but, in AF99 blueprint, I can vouch that it was circular. At very small scale, AF99 starts to "lie" to you. Only SCASM precision could correct the visual imperfection at that level.

BTW, when I made the above GIF in AF99, all squares, although properly sequenced, were all glued together along impossible angles, creating a flickering effect. I just removed the gluing sequence and called all squares starting with the 1 foot square and finishing with the 1x1 square. What you see from the left side is exactly reproduced on the right side. AF99 can't do that.

My daughter was on the PC all evening and I had to wait at this hour to finish this post. Time for some rest now.

[Ivan]

Hello Hubbabubba,
I believe your description of the scale differences between AF99 and SCASM are somewhat misleading: There is no doubt that the unit in SCASM is more precise than that of AF99 (1/512 meter versus 1/100 foot).

For best results, one should be working to the native scale which is that of SCASM. I don't believe that working in AF99 units actually loses very much though. The resolution of AF99 is 0.01 foot or 0.12 inch (roughly 1/8 inch). The native scale in SCASM is 0.077 inch (roughly 1/13 inch).

Given an AF99 dimension, the conversion to SCASM internal units is never more than 2/512 meter or 1/6 inch off. As a result, sometimes a good curve might become less so, but the change won't be great.

Regarding Left-Right symmetry, I believe the calculations used by AF99 are generally stable. If you doubt this, pick any reasonable AF99 aircraft and view it in DPED. DPED works in meters but I haven't seen it show a difference between left and right sides of my models when I have been working on DP files. (....except for the BV 141....)

See the attached conversion chart. First column is AF99 units (0.01 foot). Second column is raw conversion to SCASM 1/512 meter. Third column is integer rounded. As you can see there are numbers that are skipped, but the error isn't cumulative.

A couple years ago, You and I were working on calculating the cockpit viewpoint for adjustment using SCASM. I refused to accept that a "guess" was the correct method to make this determination. In the final test, to prove that the SCASM unit was 1/512 Meter, I created a cube component about 1 inch on a side and put it centered on the estimated viewpoint. The result was that the viewpoint ended up within the cube because the cube was invisible. AF99 components (smooth) have all visible polygons facing outward which is why the cube was invisible. With a 1 inch cube, if the viewpoint was as much as a half inch off in any direction, the cube would be visible. That test showed that the cumulative dimensional error in AF99 is within that tolerance at least for my test.

I am not really sure what you mean about the barrel sleeve (? I believe a canon is a barrel?) being lenticular. If I remember right, you were doing your own conversions for dimensions for which I never knew the details. If you were using a AF99 structure to create the basic shape, I can see how that would be a problem because the cross sections are often not as good regular polygons as they should be.

Regarding "Inches": They are not an AF99 unit! I only use them to quote numbers that folks can intuitively understand. If I say 0.01 foot, folks will have to think a while. If I say 1/8 inch, most folks can see that. If they can't a 1/8 inch marking on a ruler is easier to find that a 1/100 foot marking. When working with drawings for a project, I convert all inch values to hundredths of a foot.

BTW, for folks who might want to test out the "Units" we are discussing, Use FDE to adjust a flight model's cockpit viewpoint in inches. (I believe this is record 301.) You will find that the values "snap" to discrete values. You can't just plug in any arbitrary value. AirED also snaps to values, but the problem is that their datatype is defined incorrectly to be an exact match to SCASM internal units. If your FDE doesn't show the right numbers, email me at Ivan1GFP@yahoo.com and I will send you my updated FDECTRL file which has the correct conversions.

- Ivan.

[hubbabubba]

Hello Hubbabubba,
(...)
Given an AF99 dimension, the conversion to SCASM internal units is never more than 2/512 meter or 1/6 inch off. As a result, sometimes a good curve might become less so, but the change won't be great.
(...)
I am not really sure what you mean about the barrel sleeve (? I believe a canon is a barrel?) being lenticular. If I remember right, you were doing your own conversions for dimensions for which I never knew the details. If you were using a AF99 structure to create the basic shape, I can see how that would be a problem because the cross sections are often not as good regular polygons as they should be.
(...)
- Ivan.

Hi Ivan

I had to search into my "leftovers" of the MG jeep to find this;

Attachment 13531

This is the part I'm talking about (pale blue), whatever its real name. To make a round octogon, I would make a temporary part and have it rotate by 45° and "save as". Then I would create the polygon using the "P" key. Up-down and right-left came OK, but the other vertices were shortened by AF99. The blueprint, as you can see, was not showing that; the octogon appears to be "round". This is when you told me that the model in game wan't round and, by gosh!, you were absolutely right!

The yellow octogon is my own representation of what was really happening. This is why I'm telling you that AF99 lies to us, especially close to the inferior limit of its precision.

The solution was to redo all the vertices under SCASM using a grid on a bitmap. This way, I could "elongate" the 45° angles that AF99 had "ignored" in rounding up.

The jeep is, compared to an aircraft, a very small object. The gun on top of such a jeep is even smaller. On an aircraft, a precision of ±2 SCASM units will probably do. With the gun, it wasn't enough.

I'm not trying to be misleading regarding the scale used by AF99, I'm simply thinking of the average designer that have to live with inches. When I was making the Taifun, I would translate all measurements, originally in metric system, to the imperial system! I'm not sure I would take the same path now btw...

If someone is up to transferring all measurements in hundreth of a foot, the better.

Precision and scale was not the main reason I "defected" to SCASMing anyway. More on that later...

[Ivan]

Hello Hubbabubba,

I have a pretty fair idea of what a Browing .50 Caliber machine gun looks like, so I decided to try to build a component similar to what you are using for your barrel extension / barrel sleeve. The sizes may not be exactly the same, but should be within range. I didn't do anything particularly strange here. I just imported, rotated and resized the 8 point circle from the storehouse and used two different sizes as references to build a component. If you read the dimensions in the point editor, they are within reason for the precision of AF99.

It took a while to respond because I needed to actually try this out for myself. The end result is that I am getting more or less reasonable results in the simulator as I see it. I am not sure why your results are so strange. This thing is kind of small, so I had some trouble getting the alignment perfect in the simulator.

Easiest way to see if it looks right is to tilt your head slightly and see if the sides look "Square". They don't look exact but they don't look bad either.

BTW, I still think you should have used a Hexagon instead of an Octagon because of the three equally spaced rows of cooling holes.

- Ivan.

[hubbabubba]

Hi Ivan,

I just verified your two octagons and they are rather "round" indeed.

To make the barrel sleeve, I didn't used the "magnify" tool. For this octagon, I made a vertical line 0.24 AF99 units long and "saved as". Using the "rotation" and "bank" tool, I would turn the line by 45° increments and "save as". Three rotations later, I would have a four lines asterisk-like star. Using the "draw" and "P" tools, I would simply draw my octagon. Unfortunately, it is only after your acute eyes had discovered the deformation that I realized that the rotation tool was shortening the line during the rotation. For some reason, it cuts one AF99 unit at each end of the line, whatever the length of the vertical line. So my 0.24 units diameter in height (Up-Down) was only 0.22 units diameter in width (Right-Left). The 45° angle lines were also shortened.

But, using the "magnified" tool, I was not as lucky as you were with your 0.25 and 0.07 AF99 diameter units octagons. By chance or by design, these two octagons are almost looking the same once "saved as". Not so for my 0.24 units diameter octagon. Once "saved as", height and width are equal, but 45° angles are clearly too large. Here is a composite image of the three 0.24 units octagons;

Attachment 13743

The narrowest octagon is the one I made with the "rotation" tool. The top vortex is at +0.12 and the bottom vortex at -0.12, but the rightmost vortex is at +0.11 and the leftmost vortex at -0.11. The 45° angled vertices are also too short when compared with the unsaved octagon made with the "magnify" tool (highlighted blue lines).
But, once "saved as", the octagon made with the "magnify" tool is also deformed. The 45° angled vertices are too long when, again, compared with the "ideal" octagon shown by AF99 before being saved.

I did the same exercise with a 0.25 and 0.07 octagons like yours and got almost exactly the same shape. Was it by chance or because you choose these sizes on purpose, only you can tell.

But the smaller it gets, the greater the chances AF99 will deform the "ideal" octagon. This animated GIF shows the muzzle bore hole as it would look before and after being saved; a tilted square or a lozenge, depending on your inclination (pun intended)! If you have a close look at my MG jeep, you will notice that it is fairly round. It is so simply because, under SCASM, I was able to put vertices in places AF99 couldn't.

Attachment 13744

[Ivan]

Hello Hubbabubba,

I actually have no idea what the diameters are that you are using for your barrel sleeve. I just picked a couple arbitrary diameters that were in the range that I estimate the diameter is on the real gun. The smaller one is to make sure that I tried something smaller than you were likely to be using.

I now know the major diameter is .24 foot. What is the minor diameter so I can try the same test again?

- Ivan.