Southern Railways "Schools" V class [WIP]

[Gumboots]

Have done a bit of playing around with sizes for ships and consists, using WP&P's groovy custom cargo cars (which is what I always run).

If I'm going to use electric track for the shipping lanes, that rules out any ships before 1904 if you want to lay your own shipping lanes during the game. I've tried making electric locos available earlier by event, and the locos do become available, but you still can't lay electric track for them before 1904.

What will probably work for 19th century shipping (haven't tried it because just thought of it) is if the scenario author sets the date in the editor to post 1904, pre-lays shipping lanes with electric track, then sets the date back to 18-whatever before saving the scenario. If that works, then no worries for 19th century shipping.

(Edit: Ha. Doesn't work. Just gave it a quick test. The electric track will be there in 1830, and the electric loco will be available, but the game still insists the track that was laid as electric can't be electric before 1904. That screws that idea.)

Anyway, I've set up a basic Blender model with imported body .3dp's for all the express coach eras, along with a single AutoB freight car for a quick comparison. The express coaches seem to be the longest single cars in any era, and the AutoB is the longest freight car.

For this Southern Railway scenario I'd be stuck with the C era stock. I could finangle things again to change dates and sizes, but the C era carriages are about right for scale for SR in the 1930's so I figure use them. The result is that for a full 8 car consist, the length required is greater than an actual SR ferry of the period. These were in the 300-350 foot range (about the same as a naval destroyer of the period) and an example is shown at the bottom of the picture, alongside a standard RT3 dock. It fits the dock nicely but is too short for a full express consist.

To get a full express consist in safely, I'd need the larger ship shown at the top of the pic. This is more like a trans-ocean liner in size, rather than a cross-Channel ferry. It's a bit long for the dock but would be fine once underway. Since this ship is getting pretty big it's probably close to the limit of what should be used, so if I do multiple freight cars for this scenario I'd sorta be working backwards from a reasonable maximum ship size, then seeing what fits in that. This will mean some freight cars will stay as singles, while some may be ok as doubles or triples.

Obviously it would be possible to have both faster express ships with great pax appeal, with cheaper, slower and uglier ships for freight. It would also be possible to do smaller and cheaper ships for shorter consists, if there turned out to be any use for them in the game. These would have to be coded with low free weight, so their speed dropped off dramatically if they had a consist sticking out the back, to stop people using the cheap ones to haul long trains and looking stupid.

Basic_ships_shot.png

Edit: Just had another thought for added fun. I'd already figured that you could do also do paddle wheel ships just by setting the paddle wheels as RT3 "drivewheels" in the coding. That's easy. There's no way of coding propellers, but they should be underwater anyway so you wouldn't bother. However, ship rudders are often partially above water and they could be coded as trucks. By setting the truck attachment point correctly relative to the ship's rear track attachment point, it may be possible to get the rudder behaving properly around corners. Not that anyone would give a rat's when the thing is going 35 knots in a straight line out in the ocean, but having thought of it I'll have to try it sometime anyway.

[Gumboots]

Ok, got together all the cargo cars for the relevant period and sized them up. Without going to crazy big ships, which I don't want to do right now, the story is that for this size of ship the automobiles, flat cars, goods wagons, stock cars, tankers and all express (troops, mail, passengers and dining car) are all too long for doubles and will have to stay as singles. I figured express would only ever be singles anyway and I'm not too fussed about the others.

The good news is that boxcars, fruit cars, all hoppers (covered and open) and reefers will all fit in the ships as double units if we want them (and we does want them) providing the bauxite hoppers are changed to use the same body as the other open hoppers. I figure this is ok since they can still be skinned as bauxite hoppers, they'll just be a bit shorter than they are now.

So the upshot is that about half of all freight can easily be done as double consists and still fit in reasonable sized ships.

All_cargos.png

[Gumboots]

Ok, this is back on. No good roughing out five million locos and not finishing one of them. Since I'm now over the initial buzz of actually being able to model things conveniently, I'll get back to working on the Schools class. The plan is to get it finished by the end of this month.

I have managed to get my hands on some works drawings for a range of components, including a GA of the whole loco. I'll do some corrections on these for distortions tonight, then use them to do a final check on the model.

[Gumboots]

Ok, I now have the thing set up with about 30 different accurate drawings for various views. This is cool.

This may seem over the top, but the good thing about it is that there's no need to guess. All I have to do is build over the drawings and everything will look right. This saves a lot of trial and error, as well as a lot of head scratching and swearing. It also gives me more of an insight into how the whole thing was put together. I find this interesting in its own right, but an advantage of understanding the construction better is that when the drawings aren't clear on some particular point it becomes easier for me to figure it out. Everything just starts to make sense.

It turns out that I had a lot of bits right, but other bits wrong. The wrong bits won't be hard to fix. This should give a more convincing model for people who are familiar with the class. There should be fewer of those niggles where you know something aint right, but you can't quite tell what it is.

The drawings I've been working from are scans of old works drawings, and these often have various distortions in them. Often a GA will be split into two scans, and each one may have a slightly different orientation and scale. Various bits of each scan might be stretched or bent or skewed slightly too. This all sounds like trouble, but I've found it's easy to fix if you just give it a couple of hours in Photoshop. GIMP would be just as good, if you prefer to use that.

The first thing I do is to start by taking lines which I know should be straight and horizontal. The ground level in a GA profile is an obvious case. Getting it straight and horizontal is usually just a matter of arc warping the whole layer to some fraction of a percent, and/or giving it a slight rotation. In bad cases it may be necessary to chop one section into a new layer and treat that separately, before merging it back into the first layer. This doesn't take long to do in practice. The next bit is to take the lines you know should be straight and vertical, and fix them so they are. It's basically the same process as fixing the horizontals, except that you use a skew transform instead of rotation.

The next thing I do is to look at things I know should be round, and check them for height and width. This will often require a slight scaling one way or the other to get heights and widths matching. If the drawing is in two halves, I'll then cross check both to make sure they are scaled to the same size before I join them. I'll also use a calculator to do proportional checks against known dimensions, such as comparing wheel diameters and wheelbases to overall length and height, etc to make sure the whole thing works together.

Once I'm happy with the drawing I'll load it into Blender as a background image, then position it and scale it to suit the units I'm using. Blender allows very fine control over position and scale. With a bit of messing around, it's not hard to set things up so that what started out as a comparatively rough drawing ends up being accurate to a fraction of an inch, in any direction, over the entire locomotive. At this point, building the model mesh becomes as easy as playing with Lego blocks.

[EPH]

What a cool modding/skinning concept.

'British Miracle' as you say does run simple track over blue-painted land to simulate ocean traffic - it is hard to talk about the British economy without overseas trade.

But I didn't invent the idea. I seem to remember I borrowed it from 'Men and Seals'. Just to give correct attribution.

[Gumboots]

Ok, having had a break and tackled some easy stuff, finishing this mongrel of a thing is back on the agenda.

I've fired it up again and done some more testing on the rough beta model. Stats are now where I am pretty happy with them. IRL they were a moderate sized choofer rather than a heavy duty one, and generally ran with a 5 car express consist. I've used that as a basis for my latest round of testing.

Pulling power and free weight have been wound back slightly. On my Royal Tour map the Schools will now get to 70-ish between Sydney and Newcastle, and 80-ish between Newcastle and Port Macquarie, which is about right for what they should do with 5 express cars. Going down a 3% or 4% grade (like Dunedoo down to Muswellbrook) it'll still touch a top speed of 94. That was only ever done briefly on a downhill stretch by the real thing, so that fits too.

Performance up grades is still useful, meaning it can actually be run just about anywhere. I've also increased the maintenance cost from $8k to $11k. This will still allow a fairly long economic life (15 years or so) but is a bit more in balance with stats for some other locos. The Schools is now as good as it ought to be, but not over the top.

If anyone wants a zipped copy of the latest .lco file, it's attached at the bottom of this post.

---

Ok, so next bit is the gnarly stuff: deciding which details can be kept, and which have to be ditched or simplified to keep the verts and tris count under my chosen limit. I went through it yesterday and have knocked it back to 150 under the limit. It'll still look just as good when finished, but there's now scope for adding a few more bits that were missing.

I've also decided I'll have to re-scale the UV mapping to give a bit more space for things, so will be knocking that back from 24px/RT3 unit to 22px/RT3 unit. This part isn't actually that big a deal. Working out how to UV map everything and pack it all in is most of the work. Rescaling bits is easy.

Should have a new beta version fairly soon, with tender and all sorts of stuff.

Edit: Old zip removed. Use the one in this post if you want to try it out.

[Gumboots]

I was going over this again and had a brainwave. Said brainwave resulted in a lot of verts and tris disappearing without having to compromise the looks at all. The total count is now about 240 less than it was on Thursday.

The brainwave was to do with the modelling of some details for the tender. I had modelled the leaf springs and axle boxes on the tender, since with this style of old English tender those are quite prominent details. They're not modelled to the nth degree, just blocked out enough so that graphics can take over fooling the eye.

Anyway, this resulted in quite a lot of verts and tris since there are three leaf springs and three axle boxes each side, plus six large bolts each side to hold it all together. What I just realised is that with the tender having its own skin image, which has quite a lot of available space on it, there was no need for me to model each side separately for these details. I could simply join them up underneath the tender, and thereby not only halve the numbers of verts and tris, but also halve the number of separate meshes to be rendered. This number of separate meshes is, AFAICT from reading up on this stuff, pretty much as much of a performance hit as the numbers of verts and tris.

So anyway they're joined up now. They look solid at the moment because they aren't skinned yet, and just have an overall material applied in Blender. Once they're skinned the alpha channel will be doing its job and they'll look like the real deal.

Cheltenham_tender_detail.jpg

Tender_detail.jpg

[Gumboots]

Ok, so 2 and a 1/2 years later... ... I might actually finish this thing. I've been over it again and have managed to knock the poly count back to something reasonable, while still keeping it looking good enough to keep most sane people happy. Poly count is now down around 1940 (may change slightly) which is about 60 less than the infamous double Mikado.

That should be ok for performance in most games, but the only way to find out is to try it. The poly count may less than the double Mikado, but this is a faster locomotive, so it may get choppy when it's winding out past 80 mph. I'll run some tests with it before I get into serious skinning, just to check the frame rate is reasonable.

[Gumboots]

Got all the bits hooked up, and it's working now. This is after I made every beginner's mistake in the book, including exporting the tender files when they were set back behind the locomotive for a loco+tender display in Blender, instead of having the tender body separately placed at the geometric origin (0,0,0 on X, Y and Z axes) which is what you need for exporting.

Having decided to give up for the night, go to sleep, and then try again in the morning the mistake became obvious. Frame rate seems fine with basic testing just running a few trains. I think it's the overall amount of stuff in a scenario that really loads the game engine, and not so much the poly count of a few individual locos. Unfortunately I cleverly deleted all my old large scenario saves a while back (no idea why, and can't remember) so I'll have to build up another one or two for more intense testing.

Anyway it works now.

Ok_it_works_now.jpg

[Gumboots]

Just did a bit more testing to fine tune a couple of things. Skinning is still very basic, and the UV mapping is just a quick "get it all running" job, but it's usable if anyone wants to play around with it while they're waiting for better skinning.

Stats are set to match my custom express car pack in the pre-1950 era, so it will be a little bit underpowered with default PopTop express cars. It's set up the way the real thing was set up: as a moderate-sized choofer for five car consists on secondary express services. I'll balance the running costs so the economics works for that, once I get some more testing done.