Vertical Scaling, a new approach?

[JP Schriefer 5,591]

We all know BATters usually scale their models by 133%, but to be honest I never knew how exactly this number was reached and I'd like to know.

Today I was reading about how types of projections work and I noticed SC4 isn't really isometric, but trimetric just by 1 degree. I think it's acceptable to consider it as dimetric since it makes calculations way easier and it's an insignificant error. Actually I believe it's more likely that I got it wrong due to the prospector error or any other thing and it's indeed dimetric.

 

So this is a cube made with 3ds without vertical scaling. As you can see the height must also be equal to 433px. It means we need to scale it by (433/317)*100 = 136,6%.

It makes sense to me, but does it to you as well? It's always good to ask, sometimes other people notice something wrong we are not seeing

[rsc204 21,867]

Without getting into the accuracy of the results, there is one thing that does come to mind.

One of the biggest issues custom content has is playing nicely together. Often what's more important to the aesthetics is not getting a real-world scale, so much as making something that fits with all the existing content. This is already varied enough to cause problems and altering the scale once more may add to the problem. Of course and extra 3.6% isn't exactly a huge change so this isn't necessarily something that will cause problems. But equally I think the rough 133% has been adopted pretty widely and even if it's technically not perfect, I've never seen the final models and had issues with the vertical scaling when done this way either.

[Handyman 791]

You may be Interested in this article by rivit. It delves quite deeply into this topic.

http://users.tpg.com.au/rivit/rvt-SimCity-articles2.htm

[JP Schriefer 5,591]

11 hours ago, rsc204 said:

Without getting into the accuracy of the results, there is one thing that does come to mind.

One of the biggest issues custom content has is playing nicely together. Often what's more important to the aesthetics is not getting a real-world scale, so much as making something that fits with all the existing content. This is already varied enough to cause problems and altering the scale once more may add to the problem. Of course and extra 3.6% isn't exactly a huge change so this isn't necessarily something that will cause problems. But equally I think the rough 133% has been adopted pretty widely and even if it's technically not perfect, I've never seen the final models and had issues with the vertical scaling when done this way either.

Yeah, I know it's not a big difference and the buildings looks great the way they are done, I just find it an interesting subject since it's math and I like it. I intend to do tests with my buildings to see if I continue to 133% or if I go to 136,6%  

7 hours ago, Handyman said:

You may be Interested in this article by rivit. It delves quite deeply into this topic.

http://users.tpg.com.au/rivit/rvt-SimCity-articles2.htm

That's really interesting! Thanks for sharing

[mattb325 28,892]

I like it!

It actually vindicates the scale I've been using for quite some time which has always gone against 'standard' wisdom.

If I take the arbitrary floor to ceiling height often cited by US players in these forums as being 10' (3.048m) and apply the 133% I get 4.05m....which always appeared too small for me (Maxis single floor residentials use this scale). 

If I take that same number (3.048m) and multiply it by 136.6% I get 4.16m which is much closer to what I actually use (4.5m for floors 2-10 and less per floor thereafter if the building is to be higher than that)  Of course, 0.1m = 1 pixel in game, so we aren't talking huge differences.

In terms of the standard custom content, W2W for European buildings = 5m per floor (think the likes of xannepan/porkissimo etc) and was also adopted by Simgoober and JBSimio for US W2W buildings and Jmeyers used this for his residential buildings as well. I understand that other US batters do not necessarily use this scale but it is only a matter of pixels, after all.

 

And to answer rivit's question (in his article) about the scale of a sim - it is almost 3.8m...which means that realistically a door has to be at least 4m high which is why I (and I think most batters are similar) - model the ground floors of buildings at least 5-6m in height and apply the more traditional scaling on subsequent floors.

[MushyMushy 3,618]

Rivit's article was really informative. I can say that I already thought the way he did about scaling, though I had a lot less factual/mathematical reasoning to back my opinion up.

The reason I don't really like vertical scaling is because as an artist who does a lot of work with perspective (of buildings no less)... I see foreshortening as normal. If there is a circle on the side of a building in the game, at the viewing angle we have the circle should logically appear somewhat flattened. I noticed the problem more when I started BATting. I tried scaling my cars up by 133%, and it didn't just make the tires look like circles... it stretched them into ovals that were taller than they were wide. After going to a lot of trouble (in most cases) to make sure my models are accurate in size, it just looks off for them to be stretched upwards.

The only altered scaling that I really think looks normal is messing with the ground floor height, as matt said. For some reason Maxis made their sims horrific monsters so we have to at least make them able to get in the first floor of the building. Beyond that they can just crouch I guess.

[rsc204 21,867]

1 hour ago, MushyMushy said:

For some reason Maxis made their sims horrific monsters

Basically so they'd be a thing we could actually see. If they were to scale, you'd have to play tiny-pixel where's Waldo to find them.

When I made my UK Phonebox, scaling was a nightmare, because at any reasonable scale you simply don't see any details, just a small silver oblong. So I ditched the rule book and just made it so you could make it out for what it was and so a sim standing/walking nearby would appear to be able to fit inside. Despite it's huge over-scalling, I think I achieved balance in terms of the limitations of the game world.

[Jasoncw 11,954]

On 6/17/2017 at 1:18 AM, JP Schriefer said:

We all know BATters usually scale their models by 133%, but to be honest I never knew how exactly this number was reached and I'd like to know.

Today I was reading about how types of projections work and I noticed SC4 isn't really isometric, but trimetric just by 1 degree. I think it's acceptable to consider it as dimetric since it makes calculations way easier and it's an insignificant error. Actually I believe it's more likely that I got it wrong due to the prospector error or any other thing and it's indeed dimetric.

...

So this is a cube made with 3ds without vertical scaling. As you can see the height must also be equal to 433px. It means we need to scale it by (433/317)*100 = 136,6%.

...

...
It makes sense to me, but does it to you as well? It's always good to ask, sometimes other people notice something wrong we are not seeing

I'm guessing 133% was a more "mathematical" looking number than 136.6% . I've also done the same diagram as you except for 133%, but yours is more accurate. 

I'm personally going to keep on using 133% because it's a convenient number and it's what's already been used. I also have to admit that for an unknown amount of time I was using 130% because I misremembered the number.  

Except for the times when I completely forgot to scale altogether.  

It's very silly that we've been using the wrong number for so long though. Good work.  

8 hours ago, mattb325 said:

I like it!

It actually vindicates the scale I've been using for quite some time which has always gone against 'standard' wisdom.

If I take the arbitrary floor to ceiling height often cited by US players in these forums as being 10' (3.048m) and apply the 133% I get 4.05m....which always appeared too small for me (Maxis single floor residentials use this scale). 

If I take that same number (3.048m) and multiply it by 136.6% I get 4.16m which is much closer to what I actually use (4.5m for floors 2-10 and less per floor thereafter if the building is to be higher than that)  Of course, 0.1m = 1 pixel in game, so we aren't talking huge differences.

In terms of the standard custom content, W2W for European buildings = 5m per floor (think the likes of xannepan/porkissimo etc) and was also adopted by Simgoober and JBSimio for US W2W buildings and Jmeyers used this for his residential buildings as well. I understand that other US batters do not necessarily use this scale but it is only a matter of pixels, after all.

I think the problem with that is, like you said, those floor heights are arbitrary. Scaling up an arbitrary number will result in an equally arbitrary number. The scaling issue relates to things that have realistic dimensions to start with. 

Postwar American residential/hotel floor to floor heights are going to be 9 or 10' (~3m) most of the time. Prewar will tend to be higher and more recently might be a little higher too. Office buildings will be more like 13-15' (4-4.5m). European office towers tend to have shorter floor heights. So a shiny new American office building scaled up might have 6m high floors (6.15m at 136.6%).

7 hours ago, MushyMushy said:

Rivit's article was really informative. I can say that I already thought the way he did about scaling, though I had a lot less factual/mathematical reasoning to back my opinion up.

The reason I don't really like vertical scaling is because as an artist who does a lot of work with perspective (of buildings no less)... I see foreshortening as normal. If there is a circle on the side of a building in the game, at the viewing angle we have the circle should logically appear somewhat flattened. I noticed the problem more when I started BATting. I tried scaling my cars up by 133%, and it didn't just make the tires look like circles... it stretched them into ovals that were taller than they were wide. After going to a lot of trouble (in most cases) to make sure my models are accurate in size, it just looks off for them to be stretched upwards.

The only altered scaling that I really think looks normal is messing with the ground floor height, as matt said. For some reason Maxis made their sims horrific monsters so we have to at least make them able to get in the first floor of the building. Beyond that they can just crouch I guess.

It's absolutely true that in a perspective environment you'd expect foreshortening, but SC4 isn't a perspective environment. Buildings just look stumpy without getting scaled up. This was a super common issue in the early days of BATing, people would put a lot of work into modeling something and they'd do a preview render and it would just look *wrong* and they didn't know why and would get frustrated and then someone would come along and tell them to scale up. Scaling them up by 133/136.6% turns the front facades into proportional elevation drawings which is what I think the eye expects. It does cause other problems, like you said especially with circles and other curves, which cars have many. But buildings are almost entirely orthogonal. 

Also, even though foreshortening might be expected in the situation, isometric projection doesn't have correct foreshortening anyway (which is why it looks weird in the first place). It's a weird kind of abstraction or approximation. If it followed correct foreshortening rules it would be true perspective. I mean the isometric problems are why M C Escher was able to make all of those drawings. 

It's a "pick your poison" kind of situation. 

Which is why 3d games are so nice.  

[mattb325 28,892]

Quote

1 hour ago, Jasoncw said:

Postwar American residential/hotel floor to floor heights are going to be 9 or 10' (~3m) most of the time. Prewar will tend to be higher and more recently might be a little higher too. Office buildings will be more like 13-15' (4-4.5m). European office towers tend to have shorter floor heights. So a shiny new American office building scaled up might have 6m high floors (6.15m at 136.6%).

Without wishing to sound contrary, I would seriously check your figures on those commercial height assertions: I do an awful lot of travelling for work (US, Asia & Europe) - all new commercial offices irrespective of the continent on which they sit seem to be around 2.7m floor to (false) ceiling height; then allowing a .3m - .5m void space for services + ~.25m for slab which gives at the very most a 3.25m - 3.55m floor to floor height. This, when scaled is 4.32 (133%)/4.44m (136.6%) - 4.72m (133%)/4.85m (136.6%).

My assertions of 4.5m is exactly within this scale range.

[T Wrecks 10,210]

Meh, it's really a difficult topic because there's little certainty, and we're trying to squeeze stuff into an environment that lacks a uniform and coherent scale anyway, even among buildings of similar type/size. That's the one point where rivit's article - a very good read, and a thoroughly presented and backed up argument! - kinda fizzles, because IMO Maxis scaling is by far not as thought through and carefully controlled as he seems to think. The way the buildings look (coupled with common sense) make me think that it was a team of artists and not just one person, and looking at the result makes me doubt that this team had very strict specifications to adhere to that were thoroughly checked by QA.

There will have been some general guidelines and agreement, for example considering the overall downscale to squeeze more buildings into a city, or some fundamental scaling principles such as the Sims being overscale so we can see them, or the standardised building footprints for building families. However, I don't think a nondescript 3-storey commercial building, for example, will have been carefully scrutinised following strict mathematical guidelines, and rejected if it turned out that floor height was .4 metres off spec. It appears more likely to me that they eyeballed everything by plopping the building in game, checking whether the footprint was in line with the requirements for the building family the model would belong to, and tweaking only if the building looked so visibly and obviously off that it really stood out at the first glance.

RL can be tricky, too, I agree. The house I'm living in (European pre-war W2W) has floor heights that decrease the further you go up. (This used to be pretty common, because historically these houses would often be home to a shop, its owners, and their family, with the floor heights echoing status. So the ground floor would be the shop, the first floor would house the owners, the second floor their children, the third floor maybe the grandparents, and the top floor any servants/assistants.) The first floor has a ceiling height of 3.4 metres, second floor is 3.2 metres, third floor is lower again (around 2.8 to 3m, I guess), and the fourth floor is maybe 2.4 metres tops.

So, debating over 3.6% may be academically interesting, but IMO it's overthinking.

(You can tell that I'm not a MINT guy, can't you? )

[Jasoncw 11,954]

12 hours ago, mattb325 said:

Without wishing to sound contrary, I would seriously check your figures on those commercial height assertions: I do an awful lot of travelling for work (US, Asia & Europe) - all new commercial offices irrespective of the continent on which they sit seem to be around 2.7m floor to (false) ceiling height; then allowing a .3m - .5m void space for services + ~.25m for slab which gives at the very most a 3.25m - 3.55m floor to floor height. This, when scaled is 4.32 (133%)/4.44m (136.6%) - 4.72m (133%)/4.85m (136.6%).

My assertions of 4.5m is exactly within this scale range.

Yeah it could definitely be lower than 4m, I was thinking about steel skyscrapers which are higher. 

This building is 4.4m http://www.archdaily.com/774528/dominion-office-building-zaha-hadid-architects. This one (Roppongi Hills) is 4.1m but 4.5 on some floors https://www.mori.co.jp/en/office/japan/spec/images/ceil_roppongihillsmt/image_01.gif. My Inland Steel Building BAT turned out to be 4.2 unscaled with a few of the office floors being taller. Leadenhall Building is 4m http://www.skyscrapercity.com/showpost.php?s=f6b2db1325d665ad6a21f3428c723211&p=86483915&postcount=726. My US Bank BAT is 4. Swiss RE London HQ is 4.15 http://we-aggregate.org/media/files/86862bbe67ae8d2a1182093ae993dad7.jpg. 7 World Trade Center is 4.1 http://www.som.com/FILE/14896/7wtc_1400x800_plan_02jpg.jpg. Meanwhile the GSW building in Berlin is given credit for integrating the services into the floor slab to make the most out of a low 3.3m height. My Telus Tower BAT was 3.5m.

[SimCoug 5,976]

Since this thread is mostly about scale, I'll just drop this here...
What size do you folks typically model doors?  Specifically the ground floor doors for commercial buildings.  
Just form my own tests, It looks like anything between 3.5m to 4m high works pretty well with Maxis scale, but then again, some of their buildings don't match up.  
What about stairs? If I try to model them in RL terms, you can barely make out the steps.  
Thanks for any input... Trying to get a handle on the scale of things when modeling seems to be more art than science.  

[Jasoncw 11,954]

I personally just model even the ground floor to real life dimensions. It might look weird next to some of the street props but those are already so inconsistently scaled that no matter what is done things at that level will be out of scale. At least my building will be consistent, and I think that changing the ground floor can affect the composition and impression of the building as a whole. It's also hard to know exactly how to change it while still having it "go" with the rest of the building. 

For stairs if it's a real building I stick to real life, and if it's not then I do 0.2 by 0.3 because it's a convenient number that's close enough to what stairs normally are. For 3ds max this shows up pretty well but in gmax you're right that the steps fuzz together. 

For things like railings I try to do real dimensions but I make them bigger if they're not legible. For things like fire escapes, that's greatly exaggerated because otherwise the fire escape will look like a translucent material instead of grate. And then for actual grates, like ornamental ones covering up vents, I try to do it realistically and then if it doesn't look good I try to fudge it a little to make it more legible but often times it doesn't work and you're just stuck with it being fuzzy. Sadly sometimes things just can't be seen at SC4 scale. For vents, like on roof junk or on mechanical floors, I also exaggerate, just enough that the vent pattern is clear (otherwise it all blends together and looks like a solid surface). 

Something which I think is important is to make sure that all of the exaggerations are localized. So for example even if the stairs themselves are made bigger the overall staircase should be the same size. Or if window frames are exaggerated the overall size of the window should stay the same, even if the bigger frames make the window look jumbled. 

The nice thing about max is that it renders fine details a million times better, so you don't have to worry much about adjusting the building from real life. Small details tend to be more legible and if they're not they tend to blend together more attractively than in gmax. You can even texture bricks at actual size and the brick texture can still be legible. 

[mattb325 28,892]

Hi Simcoug, I've sent you a PM with links to the Maxis scales (as listed not just by a Maxis employee, but one of the lead artists) and the Maxis model that was shipped with the Bat, as well as some other handy heights, extrusion dimensions and the % scale that they used which makes batting quick and effortless for me.

I don't wish to appear contrary or shrill in these forums, but it becomes a little tiring having to defend everything publicly, hence the PM.....which you can use or dismiss as you like