Editing Vessel Tutorial 1

This article intends to guide you through the construction of a new Vessel DLL, by using the sample code from the OrbiterSDK. 

This documents my adventure transforming the stock ShuttlePB into a reasonably accurate model of the Surveyor lunar lander. There was a model on OrbitHangar, but it had wildly inaccurate mechanics, due partly to the fact that it was based on Vinka's Spacecraft.dll.

This tutorial assumes you have a working compiler which you know creates valid working DLLs from the OrbiterSDK sample code. I happen to use the free Microsoft command-line compiler, together with an editor called TextPad. Setting up the [[Free Compiler Setup|free compiler]] and the editor is beyond the scope of this document. Check the rest of this Wiki, or the Orbiter forums.

= The Goal Spacecraft: Surveyor =

Surveyor was a series of unmanned lunar landers launched by NASA in 1965-1967, before the first Apollo. It is a particularly simple, yet interesting and highly successful design.

Functionally, Surveyor consists of a large solid-fueled main retro-rocket, three medium-sized throttleable liquid fueled vernier engines, and six cold-gas Reaction Control System (RCS) attitude control thrusters.

The Surveyor flight plan in its final minutes is one exciting ride, and obviously not good for humans. There is no possiblity of abort (Two of the seven Surveyors, 2 and 4, crashed). The maximum acceleration is 9G's. And, it is very difficult to fly manually. It's a perfect job for a machine.

Surveyor is launched on top of an Atlas-Centaur rocket, on a direct impact trajectory to the moon. Impact will occur after about 66 hours. About 16 hours after launch, it performs a single mid-course correction maneuver, which aims it to precisely impact at the desired landing point. 

Thirty minutes before impact, Surveyor maneuvers into a retrograde attitude. From here, its Altitude Marking Radar (AMR) will eventually be able to see the surface and measure the distance. The AMR locks on to the surface at about 200km, one minute before impact. Once the altitude reaches a pre-determined value called the altitude mark, it starts a short timer. 

At the altitude mark, the spacecraft is hurtling towards the ground at over 2.5km/s and will impact in less than one minute. For Surveyor 1, the altitude mark was about 110km, and the timer was about 7 seconds. Once the timer expires, the vernier engines light up, then one second later the main retro ignites for its 40 second burn. The main retro gets rid of 2.3km/s of speed, and leaves the spacecraft between 5km and 20km above the surface and travelling downwards at between zero and 230m/s. During this stage, the vernier engines are used to stabilize the spacecraft and compensate for any thrust misalignment in the main retro. The AMR is automatically jettisoned at main retro ignition, as it is no longer needed. In fact, it is stuffed into the nozzle of the main retro, and is blasted out by the retro ignition.

The spacecraft then drops the spent retro motor and continues down on the vernier engines. Using a seperate four-beam Doppler radar and a simple analog computer, it throttles the engines such that it will reach a downward velocity of about 1.3m/s about 4m above the surface. It then cuts its engines and falls the rest of the way.

= Why a custom DLL =

Many spacecraft can and in fact have been simulated with Vinka's spacecraft.dll. This method uses a text file to describe a spacecraft, and a general purpose DLL to interpret it.

The reason that this spacecraft must be written from scratch, and not use something like spacecraft.dll, is that when the vernier engines are on, they are used to control attitude as well as provide thrust. Two of the three engines use differential thrust to control pitch and yaw, and the third is mounted on a pivot to control roll. This effect cannot be done with spacecraft.dll, and therefore demands a custom DLL. Besides, I felt like writing a DLL, just to learn how to do it.

I hope that this article is useful to anyone trying to write a custom vessel DLL. I find it much easier to understand a concept when 1) I need to use it and 2) There is a well-documented example of using the concept. You have to provide the need. I provide the example.

= The Rules =

# '''''ALWAYS''''' test each change, no matter how small
# For historical or real vessels, get good reference material. 
# '''''NEVER''''' proceed to the next change if the last one doesn't work yet
# Start with something you know works
# '''''ALWAYS''''' '''''ALWAYS''''' test each change, no matter how small
# There is NO rule 6
# '''''NEVER''''' '''''NEVER''''' proceed to the next change if the last one doesn't work yet
# Only make one change at a time
# '''''ALWAYS''''' '''''ALWAYS''''' '''''ALWAYS''''' test each change, no matter how small
# '''''NEVER''''' '''''NEVER''''' '''''NEVER''''' proceed to the next change if the last one doesn't work yet

= Historical Spacecraft Documentation =

The best place to get references for NASA spacecraft is the [http://ntrs.nasa.gov/ NASA Technical Reports Server]. For this vessel, applicable reports are things such as:
*[http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19670029911_1967029911.pdf Surveyor Spaceraft A-21A Model Description] for spacecraft systems and such. 
*[http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19650014299_1965014299.pdf Final Report: Surveyor Vernier Thrust Chamber Assembly Development, Phase III] All the other documents refer to the fact that the Surveyor Vernier engines specific impulse varies with thrust, but never says in what manner. Page 86 shows an Isp vs Thrust curve for the vernier engines
*[http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19660026658_1966026658.pdf Surveyor 1 mission report. Part 1 - Mission description and performance] Tells about the actual performance of the Surveyor 1 lander in excruciating detail.
*[http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19660082318_1966082318.pdf Surveyor Terminal Guidance] Outlines a rather simple and surprisingly robust system for landing a Surveyor spacecraft with a stupid analog computer and a 4-beam landing radar
*[http://www.google.com/help/features.html#calculator Google Calculator] Just type in the search box, some calculation. The calculator especially excels at unit conversions. You don't have to go to that link, you can type a calculation in the Google search box on its [http://www.google.com/ main page].

= The Starting Point =

So, how do you start with something that works, if you are creating a brand new vessel? Good question. The answer is to look through the samples in Orbiter\OrbiterSDK\Samples for the ship which is closest to the one you have in mind.

Sometimes nothing is particularly close. It was in my case. In this case, just start with the simplest possible ship, the ShuttlePB. Make a copy of this folder, then rename it to your new vessel name. If you are using Visual C++, you probably need to do some weird stuff to rename the project and all the files in it. I use the free compiler, so the project files are no good to me. If you use it also, feel free to just delete all the files except for ShuttlePB.cpp, and rename it to a new name, like Surveyor.cpp.

Also, copy the source spacecraft's config file. The config file for Surveyor is shown here, and includes an attachment point, for use for attachment to a future launch vehicle.

 Classname=Surveyor
 Module=Surveyor
 
 ; === Attachment specs ===
 BEGIN_ATTACHMENT
 P 0 0 0  0 -1 0  0 0 1  XS
 END_ATTACHMENT

Now, edit Surveyor.cpp. Do a global search-and-replace to change all references from ShuttlePB to Surveyor. Go ahead and modify the copyright message at the top, to say that this is a work derived from the original ShuttlePB. By the time we are done, almost all the old ShuttlePB code will be replaced.

In keeping with Rules 1, 3, 5, 7, 9, and 10, compile this file, copy the resulting DLL to the Orbiter\Modules, and use the following scenario. It should run perfectly, and look and fly exactly like the old ShuttlePB. 

 BEGIN_ENVIRONMENT
   System Sol
   Date MJD 51982.6685767396
 END_ENVIRONMENT
 
 BEGIN_FOCUS
   Ship Surveyor1
 END_FOCUS
 
 BEGIN_CAMERA
   TARGET Surveyor1
   FOV 50.00
 END_CAMERA
 
 BEGIN_SHIPS
 Surveyor1:Surveyor
   STATUS Landed Moon
   BASE Brighton Beach:1
   HEADING 90.00
   PRPLEVEL 0:1.000
 END
 END_SHIPS

If at any point in this tutorial, if it doesn't work, track down and try to identify what could have gone wrong. Since the most common symptom of failure is a Crash to Desktop (CTD) you won't have much help debugging, so keeping the changes small from version to version will help out a lot.

= Meshes =

Creating meshes is beyond the scope of this document. Since I am in a hurry and not primarily interested in artwork, I borrowed the mesh from ''Surveyor1-1.zip'' by Jim Williams, posted on OrbitHangar. 

If you are following along, get that add-on, and get the meshes out of it. Rename ''Surveyor-deployed.msh'' as ''Surveyor-Lander.msh'' . Copy this mesh to Orbiter\Meshes , and copy all the .dds files to Orbiter\Textures .

Now, to use this mesh. In Surveyor.cpp, look for the method Surveyor::clbkSetClassCaps . Near the end, change the line (Note that it doesn't say ShuttlePB, remember that you searched-and-replaced that out.)

  // visual specs
  AddMesh ("Surveyor");

to

  // visual specs
  AddMesh("Surveyor-Lander");

Compile and test it. Notice that the spacecraft now looks like Surveyor, but it flies like ShuttlePB. Also note that the legs are not on the surface. Go ahead and take Surveyor for a spin, and put it in orbit around the moon. It will be much easier to test the verner and RCS engines in space. Do a QuickSave, rename the QuickSave to something sensible like ''Surveyor in Orbit.scn'' and use the new scenario from now on.

= Vehicle coordinate system =

As described in the Orbiter API guide, vessels use a left-handed coordinate system. Even though Surveyor looks nothing like and is flown nothing like an airplane, an airplane is still a useful model to describe the coordinate system. Imagine yourself sitting in the pilot's seat of an airplane. The X axis is the "left-to-right" axis. +X points out to the right wing, and -X points out to the left wing. The Y axis is the "up-to-down" axis. +Y points up, in the direction of the vertical tail fin, and -Y points down in the direction of the wheels. The Z axis is the "front-to-back" axis. +Z points to the nose, and -Z points back to the aft end of the plane. In an conventional airplane or rocket, the engines thrust vector points towards +Z. The exhaust streams out the back towards -Z. We are going to use this coordinate system, also, as it almost matches the historical Surveyor. On it, +Z and -Z are reversed and the whole thing is a right-handed system.

So, imagine yourself again sitting in the pilot seat of Surveyor. The engines and legs are behind you, on the back of your seat, and the antenna and solar panel stick out ahead of you. When you are on approach to landing, you will have your back to the surface of the Moon and your face straight up towards the stars.

= Vernier Engines =

These will be the "main" engines controllable with the NumPad+ key. Even though these vernier engines are the very reason we are using a custom DLL rather than spacecraft.dll, we are not going to do anything fancy to them yet.

The three vernier engines are arranged symetrically in a triangle around the Z axis. Engine 1 is near leg 1, on the +Y axis. Engines 2 and 3 are out on legs 2 and 3. They are each mounted 0.86 meters from the Z axis, 0.5m behind the origin.

First, search-and-replace variable name th_main to th_vernier, and hpr to ph_vernier. This re-names the variables for the main engines and main fuel tank.

Find this code in Surveyor::clbkSetClassCaps . 

 void Surveyor::clbkSetClassCaps (FILEHANDLE cfg)
 {
   THRUSTER_HANDLE th_vernier, th_hover, th_rcs[14], th_group[4];

Change it to 

 void Surveyor::clbkSetClassCaps (FILEHANDLE cfg)
 {
   THRUSTER_HANDLE th_vernier[3], th_hover, th_rcs[14], th_group[4];

since there are 3 vernier engines, and we need to keep track of them all.

It is best to put as many vessel parameters into constants as possible, to keep them all in one place and easy to adjust. The parameters for the verniers are Maximum thrust, Isp (Fuel efficiency), and engine location. So, add the following constants near the top of the file.

 const double VERNIER_PROP_MASS = 70.98;
 const double VERNIER_ISP = 3200;
 const double VERNIER_THRUST = 463;
 const double VERNIER_RAD = 0.86;
 const double VERNIER_STA = -0.5;

Find the line in Surveyor::clbkSetClassCaps which reads as follows:

   th_vernier = CreateThruster (_V(0,0,-4.35), _V(0,0,1), PB_MAXMAINTH, ph_vernier, PB_ISP);
   CreateThrusterGroup (&th_vernier, 1, THGROUP_MAIN);
   AddExhaust (th_vernier, 8, 1, _V(0,0.3,-4.35), _V(0,0,-1));

Change them to:

   th_vernier[0] = CreateThruster(_V(         0.0*VERNIER_RAD, 1.0*VERNIER_RAD,VERNIER_STA), _V(0,0,1), VERNIER_THRUST, ph_vernier, VERNIER_ISP);
   th_vernier[1] = CreateThruster(_V( sqrt(3.0)/2*VERNIER_RAD,-0.5*VERNIER_RAD,VERNIER_STA), _V(0,0,1), VERNIER_THRUST, ph_vernier, VERNIER_ISP);
   th_vernier[2] = CreateThruster(_V(-sqrt(3.0)/2*VERNIER_RAD,-0.5*VERNIER_RAD,VERNIER_STA), _V(0,0,1), VERNIER_THRUST, ph_vernier, VERNIER_ISP);
   CreateThrusterGroup(th_vernier, 3, THGROUP_MAIN);
   for(int i=0;i<3;i++) {
     AddExhaust(th_vernier[i], 1, 0.1);
   }

This creates the three vernier engines, arranges them in an equilateral triangle, groups them into the "main" thruster group, and adds an exhaust display. The vernier engine geometry is adjustable with the constants VERNIER_RAD and VERNIER_STA. VERNIER_RAD determines how far away from the Z axis the thrusters are, and VERNIER_STA determines how far in front of the Center of Mass they are. In this case, VERNIER_STA is negative, so the vernier engines are behind the center of mass. 

CreateThruster takes five parameters:

# The center of thrust in body coordinates
# A unit vector in the direction of thrust, in body coordinates (Note that the exhaust travels in the opposite direction)
# The maximum thrust of the engine, in Newtons
# The propellant resource (Fuel tank) this engine draws from
# The Isp (Fuel efficiency) of the thruster, in N*s/kg. In other words, how many Newtons does burning 1kg/s of propellant generate?

It returns a handle to the engine. This handle is a meaningless number to the vessel, but identifies the engine to Orbiter.

AddExhaust has a simple form which takes three parameters:

# The handle of the engine to attach to
# The length of the exhaust flame
# The radius of the flame

It automatically points in the direction the engine points (Actually since this is exhaust, it points the opposite direction) and is kept in sync with the engine direction if this changes, as it will below when we add differential thrust.

Compile this and test it. Burn the main engines and look outside. You should see three engines pointing backwards (in the direction of the legs) firing. Notice that they are extremely weak and inefficient compared to the old main engine. You probably cannot launch this from the surface anymore, so it's a good thing we put it into lunar orbit first. Don't worry yet that the engines are not aligned with the mesh.

= Differential Thrust =

Now for the moment you've all been waiting for, and the justification for making this a custom DLL.

During the mid-course correction and vernier descent, the three engines work together to control pitch, roll, and yaw. 

Notice that there are two engines below the X axis, and one twice as far above it. Engine 1 balances the torque around X generated by both engines 2 and 3, when they are all at the same thrust. So, in order to pitch, we just burn engines 2 or 3 a little bit more or less than engine 1. Burning them more generates a nose-down pitch, and vice versa.

Yaw control involves differential thrusting on both engines 2 and 3. To yaw one way, Engine 2 thrust is increased and Engine 3 thrust is decreased the same amount. Because of where Engine 2 is mounted, it generates both pitch and yaw torque all by itself. However, with this differential thrusting, the other engine generates an equal yaw torque, but opposite pitch torque. Therefore, this differential thrusting generates a pure yaw torque.

For roll control, Engine 1 is mounted on a pivot. It pivots around the Y axis, up to 5 degrees in either direction. 

So, in order to do all of this, we have to reorganize the code quite a bit. We will break the change into three independently-testable parts. First, move the following line from Surveyor::clbkSetClassCaps to the Surveyor class definition:

   THRUSTER_HANDLE th_vernier[3], th_hover, th_rcs[14], th_group[4];

This changes these variables from local variables to class fields, so they can be accessed by other class member functions. Compile and test this. It should still work identically.

Now we are going to add a brand new method to the class, clbkPreStep. Orbiter call this function each time step, after it has figured out the time step's length, but before it does the physics. This is an ideal place to put the differential thrust logic. clbkPreStep is passed the current sim time (The number in Sim 17s in the upper right corner, but with milisecond precision), the length of the next time step, and the current MJD. We don't need any of these for Surveyor, but they are available for a more complicated vessel. 

Add the following line to the class definition:

   void clbkPreStep(double SimT, double SimDT, double MJD);

Add the following code somewhere near the bottom to give this method a stub.

 void Surveyor::clbkPreStep(double SimT, double SimDT, double MJD) {
 }

Now compile and test this. It still won't do anything different, but it is significantly different under the hood now. clbkPreStep is being called for each frame of the simulation.

Now the actual differential thrust code. Add this to Surveyor::clbkPreStep :

   double P,Y,R;
   P=GetThrusterGroupLevel(THGROUP_ATT_PITCHUP)-GetThrusterGroupLevel(THGROUP_ATT_PITCHDOWN);
   Y=GetThrusterGroupLevel(THGROUP_ATT_YAWRIGHT)-GetThrusterGroupLevel(THGROUP_ATT_YAWLEFT);
   R=GetThrusterGroupLevel(THGROUP_ATT_BANKRIGHT)-GetThrusterGroupLevel(THGROUP_ATT_BANKLEFT);
   sprintf(oapiDebugString(),"Pitch %f Yaw %f Roll %f",P,Y,R);

   SetThrusterDir(th_vernier[0],_V(0.087*R,0,1));
   SetThrusterDir(th_vernier[1],_V(0,0,1.0+0.05*(P-Y)));
   SetThrusterDir(th_vernier[2],_V(0,0,1.0+0.05*(P+Y)));

Here we are using the GetThrusterGroupLevel functions to get the level setting of each of the six RCS groups. The subtraction makes us treat a turn one way as just the negative of turning the other way. 

oapiDebugString is a very useful debugging tool, in fact one of the only ones we have. This provides a pointer to a string deep inside Orbiter which it prints in black and white at the bottom left corner every time step. In this case it will print a number between -1 and 1 for each of the three axes. 

SetThrusterDir is kind of a cheat here. This sets the thrust direction for an engine, and should be passed a unit-length vector. However, nothing inside of Orbiter normalizes the vector, so if you set it for example to a length 2 vector, then fly the spacecraft and use that engine at full thrust, it will generate twice its rated thrust. I haven't looked closely enough to know if it uses twice as much fuel as well.

So, we are going to manipulate the thrust vector. For engine 1 (th_vernier[0]) we are just adding an X component. We scale it by 0.087=tan(5deg) to simulate the 5deg rotation range of the engine. For engines 2 and 3, we add a factor to the Z component. Since the P term is positive for both engines, it will run them in parallel for pitch commands. Since the Y term is positive in one and negative in the other, it will run them in opposition for yaw commands.

Compile and test this. When the main engines are on, you should see them react to rotation commands. Pitch, and both the bottom engines will get hotter or cooler. Yaw, and one bottom engine will get hotter and the other colder. Roll, and the top engine will twitch from side to side.

We now have a spacecraft which is maneuverable just on the thrust of its main engines. This ends part one.

Continue to [[Vessel Tutorial 2]]