TerraStereo User Manual
TerraStereo is a software tool used for visualization of large point clouds.
You need a PC with 64-bit Windows OS and a point cloud data set to use TerraStereo. Data set can either be a TerraScan project saved as LAS1.2/1.4 or FastBinary format or it can be a set of LAS1.2/1.4 files that have the same scale and the same origo. To view points in stereoscopic mode you also need to have compatible stereo display hardware or then use anaglyph stereo.
To increase the speed of rendering, you need a compatible GPU (see hardware requirements).
Handle up to 100 000 000 000 points
High quality rendering
Digitizing shapes manually
Interactive distance measurement with continuous snapping
Stereo support
Advanced shading options
Automatic TerraScan -trajectory based fly-through mode
Stereo image recording option for short animations and still stereo image pairs
Support for GPU for increased performance
View reference vector models using shape files or ASCII files
CAD plugin option for Bentley Microstation
TerraStereoe uses the same setup and licensing system than the other TerraSolid suite of software. The main difference is that the installation of Microstation is not required. This also means that the program cannot use the environmental variables defined inside the Microstation and thus the setup is slightly different.
Three cases of installation are explained here, select the one that is appropriate for your case. In all cases the setup-program checks and installs, if required, some necessary Windows components.
Regardless of which of the two cases our installation belongs to the installation is executed running setup.bat from software distribution folder.
Standalone installation
Enter any directory you want for the installation. The directory should have read/write access for everyone enabled.
Do not enter anything for Microstation CE -directory. Setup will pop up some errors due to this but you can ignore them.
Copy the license under this folder under tstereo -directory
Other Terrasolid suite software and MicroStation CE already installed in the same workstation
Enter the Terrasolid base directory (eg. c:\terra64) for installation directory during the setup
Enter MicroStation CE directory
The software will be installed into tstereo -folder under this base terra-directory
Copy the license into the license folder of the Terrasolid base directory. The program will automatically search for the directory of ..\license\tstereo.lic for a license and thus found it in the same location than other terrasolid licenses.
You can also use license manager to get pooled licenses into the workstation using this setup
Other Terrasolid suite software installed in the same workstation with Microstaiton V8/V8i
Proceed as in the above option but ignore the Microstation CE -directory and the resulting setup error dialog.
Other software setup is related to GPU drivers and their settings. To get stereo viewing working it usually requires some time to configure the different systems. A walk through is explained in the appendix. If your are using Schneider PluraView with a preconfigured workstation everything should be already set up for you.
Load the MDL-application by typing <mdl load tstereo> from the keyin or use mdl loader dialog.
Click help icon on the opened toolset.
Alternatively you can open a file TStereoUG.pdf from docs folder in the Terra -installation directory using file manager (eg. c:\terra64\docs\TStereoUG.pdf).
To start a program in stand-alone mode double click TerraStereo.exe icon in the installation folder (or create short-cut for it somewhere else and use that).
Note – where there is a reference to LAS1.2 format it is there for the biggest compatibility reasons. LAS1.4 files will in general also work with TerraStereo, but the extra features (eg. extended classification flags etc.) will not be used.
Create a visualization:
Open a previously made TerraScan project with LAS1.2 or FastBinary storage blocks and around 25 million points per block (other block sizes will also work; 25 million points is just the recommended for best performance) or create a new project from LAS12 files
Check the required visualization channels (typically rgb, classification and density)
Set classification mask according to the specific visualization needs from file/mask
Create required additional channel(s)
Adjust the visual appearance of the selected channel using channel tab from the main window
Makes a proejct file from user selected LAS1.2 files. The files need to be of suitable size (eg. preferably less than 2 GB each) and stored with same offset and scaled either to cm or mm. If you create the project with TerraScan, you will get this kind of project automatically. This is most useful for already processed and stored LAS blocks (eg. map blocks in the city), when only a subset of the total dataset is needed.
Loads a previously converted TerraScan project files or converts the project if that has not been done. All the converted files are under <octree> -folder inside the TerraScan project folder. Large projects may take a while to convert.
The current version supports LAS1.2/1.4 or FastBinary based files under TerraScan project.
Use 'Create 3D Shadow Project' option to create 3D blocks from the original TerraScan 2D blocks. This is useful for projects that have multiple floors or otherwise wide local elevation range or just large block sizes. The resulting visualization blocks will usually be much smaller than original blocks. With this option there will also be a project file named project_shadow.prj on the same folder than the original project file (the original project file will be left intact). This shadow project itself cannot be opened with TerraScan or TerraStereo; you should always use the original project file when opening the project.
If you have a very large project (10-200 billion points), start with the lowest detail setting and then investigate the results in black&white coloring before starting to generate any channels. For optimum performance the block size in TerraScan should be around 25 million points per block (a typical value that also works well on TerraScan).
To clear the visualization project and all channels just empty the <octree> folder manually or use File/Update/Reconvert project -menu.
This menu item is enabled if the classification channel has been generated previously for the current project. Setting the classification visibility will not affect current coloring but it will mask off all points in the unwanted classes from display.
The classification channel is stored as a 8-bit channel, but the octree can currently handle only 32 first classification codes. This means you are able to adjust the visibility of all 256 classification codes but the effect of this is only seen in a point blocks that are close to the viewer. For faraway points the octree visibility function has to be used and there you only can adjust 32 first codes.
If a TerraScan ptc -file is available and stored as part of the terrascan project (a reference needs to be found from the .prj file) the classification cisibility window will use that instead of the default one.
Set the visualization channel to a previously created channel. Note that Black/White and Elevation channels are implicitly defined and thus need not to be generated. Note also that all channels are not available to create on GPU mode.
Channels |
|
---|---|
Black/White |
No coloring – maximum amount of points can be loaded and the channel is implicitly available right after project conversion. |
RGB |
Normal RGB coloring of points. |
Intensity |
Raw intensity coloring. The range of adjustment is 0..1024 (eg. 12 bits) |
Surface |
Surface like shading of the points. Does not work on low density areas. If timestamp channel is available also aligns the surface direction correctly. Prerequisite: Local density channel |
Local Density |
Shows a local density map of the point cloud calculated for each point. |
Classification |
Colors points based on their classification. |
Elevation |
Show the absolute elevation of the point in multicolor spectrum of monochrome. In the current version there is no user defined spectrum option. Note! In current version the far-away view is drawn using octree elevation which is not the same as absolute elevation. This channel is generated on the fly so it is also available without channel generation. |
Surface with Elevation |
Instead of basic white color of the surface rendering uses elevation color for shading. Useful for hires image creation of aerial data. Prerequisite: Surface channel |
Surface with RGB |
Instead of basic white color of the surface rendering uses RGB color for shading. Very usable with ortho3d point clouds. Prerequisite: RGB and surface channels |
Classification with Intensity |
In addition to basic classification color uses average intensity value for shading those. Prerequisite: Average intensity and classification channels |
Average Intensity |
An intensity channel that is created by filtering the raw intensity channel. Useful if the intensity is noisy and cannot be improved further by recalibrating. If the original intensity signal is superb then this channel should in general not be created. Prerequisite: Local Density channel |
Average RGB |
A filtered RGB channel analogous to the average intensity channel. Useful to clean and smoothen up the final point cloud for presentations; especially with mobile data. In general not useful for aerial data sets colored with high quality orthoimages. Prerequisite: Local Density channel |
Intensity RGB |
A rendering mode that takes the luminosity from the intensity and color from the rgb. Use slider to set the brightness of the intensity to the suitable value and then the rgbi slider to set the amount. Prerequisite: RGB and Intensity channels |
Creates a visualization channel from the points to be used later on. Creating a channel can take a long time, especially the arithmetic channel. To stop the processing mouse click on the progress bar.
Reconverts TerraScan project. Use this to update changes into the visualization if you have made changes to the project.
Here you can set different user settings that define the operation of the program. If a setting is marked with an asterisk (*) changing it will require a restart of the program.
This directory is used to search the license and to digitize points into TerraSurvey.
The recording tool makes individual bitmap frames that take a lot of space; on the other hand it provides a very flexible way of inputting those frames into most movie makers or animation tools also in stereoscopic mode, since the left and right frames are exported separately.
A same frame buffer is used both for windowed real time visualization and hires image creation. An exceptionally large frame buffer will decrease system performance with GPU! Both real-time rendering speed and amount of blocks that can be loaded are reduced.
If the values given here are smaller than the screen size the window cannot be extended beyond these limits.
If the values given here are smaller than required for on-screen hires images the corresponding short-cuts are disabled from the camera menu.
For hires mode to work these sizes must be twice the window size in both directions.
A CPU memory defines the maximum amount of CPU memory the software will use to cache blocks of data before starting to drop the unused blocks. This should in general be at least 2 Gigabytes lower than the available system memory. If your system has a lot of ram; for instance 64 Gigabytes you can increase this since it allows more smooth visualization of bigger dataset.
By default the mouse wheel forward / backward movement is slow and rotation with right mouse button is performed around the snapped point. If one wants to move faster and rotate the camera a shift key must be pressed.
If this option is checked the reverse of the above comes into effect. Mouse wheel forward / backward movement is fast and rotation with right button targets the camera. Pressing the shift key slows the movement down and rotates the model around a snapped point.
Note that currently in parallel projection mode the model is always rotated when the right button is pressed on the mouse regardless of this setting.
The program automatically detects the GPU and if it does not find a suitable one it starts the operation in CPU mode. This can be forced by setting checking this option. You can either set it here or from tstereo.ini -file. Only use this if you are having problems with the graphics board driver resulting a hangup. Most features are not available in CPU mode.
The setting of a stereo hardware is sometimes quite tricky. To make installation ans setup troubleshooting easier, TerraStereo by default starts with anaglyph stereo option. Only enable OpenGL stereo if you have stereo capable hardware correctly installed and configured, because even enabling it affects performance even if you toggle the stereo mode off from the camera menu during normal operation.
If you only intend to use TerraStereo in non-stereo environment then you should set stereo mode to none. This makes sure that the smallest amount of dedicated memory is reserved for frame buffer and this can make operation faster and smoother with smaller GPU.
Digitizes the shape and sends it to TerraSurvey for further editing. De-activates the shape. Exporting can also be done by pressing <e> on the keyboard. The export actually generates a text file (singlepoint.txt under tsurvey -directory) with the digitized points by appending so the digitizing can also be implemented with other software.
Clears either the active shape or all shapes depending on the chosen sub menu selection.
All digitized shapes are written into a selected or created ASCII text file.
The shape data is written so that the chained shapes have code 1 at the start of their point data and single points have 0 code.
All shapes are grouped by the next (autoincremented) number in the line and then there are just the coordinates.
Import an existhing ASCII file that is in the same for as exported shapes.
An inactive shape can be activated later by selecting it with this menu command. After selecting is done the shape becomes active and can be edited normally.
A single shape will either be a group of markers or a continuous line string. The mode can be changed during editing. Open and filled polygon shape types are also available when ending the digitizing with a specific keyboard short-cut.
Please note that by default all the vectors are shown in front of the point cloud. To view them in the right depth (especialy filled polygons) choose a cursor mode (from operation tabs) that has active depth buffer, for instance solid or plane.
Here you can define the operation used to define interpolated point between two points. This digitized point is different from the others in a way that it can float in the space and not be locked into points.
Under this menu item you will find all available levels of higher resolution rendering options. Use it to render high resolution image to the window (instead of a file). Works best with TerraZ enabled and big point size. The mouse can be moved on top of the still hires image, but it will be reverted back to a normal resolution once any movement occurs.
When this item is checked only the octree is drawn. This is handy when one wishes to view only the overall dataset (like a surface rendering from the ground points at the national level). If the check is removed the individual blocks start to load normally based on the camera location.
This menu contains commands and settings related to camera positioning, direction and trajectory. It also contains recording tools.
Starts or stops recording individual frames into the storage path specified in the settings. All viewer movements and mouse movement that happens within the rendering area generates a new frame. Note that the recording is not at constant frame rate and when one is still and does not move the mouse there is also no frames generated.
Resets the frame numbering of the recording.
Records a single (current) frame only. Useful for making a still images that can then later be converted into a suitable form depending on the presentation stereo display hardware.
This creates a big raster image (.tiff format) using maximum point size settings. This feature is currently experimental and only creates mono images.
Minimum recommended hires image size is 5k x 5k, and you need to experiment to find a correct distance to view the points so that they are scaled in a suitable way – in general it is not useful to have too much void area between points.
The maximum total size of the image is 10k x 10k; but you can have other dimension larger if the other one is then smaller, like 15k x 3 k for visualizing a road surface.
To make things easier adjust the window to about the same aspect ratio than your hires image settings before creating the image.
TerraStereo automatically tries to load all trajectories of TerraScan project if it finds them. If this menu option is disabled it means no trajectories were found.
If the trajectories are available selecting this menu will fit the view and show the trajectories in green color. The program enters trajectory selection mode in which the user can still move around the dataset as usual, but snapping to points is disabled and instead the snapping targets the different trajectories. Double clicking on the point on the selected trajectory will choose that trajectory and set the camera into that point facing the direction of the trajectory OR top-down with 15 m above the trajectory depending if the 'Use top-down view on Attach' is checked or not.
Now the backward/forward movement is directly attached to the trajectory line so instead of moving directly in space the camera is moved along the trajectory. The viewing angle is also automatically adjusted according to the pitch, roll and heading values stored in the trajectory.
User is not restricted to being on the trajectory – the relative viewing angle can be adjusted using basic mouse control and with <ctrl> key pressed the selected trajectory and user's anchor along it are shown while simultaneously enabling the user to offset the distance and position from the trajectory. A good example is to set the view to follow trajectory from suitable distance looking top-down on a highway data. First select trajectory and location along it, then press shift+T to orient top view and then hold down ctrl -key while adjusting the distance from the road with mouse (normal backward movement control for mouse). When the cntrl -key is released one can move along the road with arrow keys or with mouse.
If you want to view the data that is left or right relative to trajectory it is usually best to view it from some distance. In typical urban mobile dataset this results a view that is blocked by the walls on the opposite side of the road. Use front clipping adjustment (with point snapping, mouse wheel movement and ctrl -key) to clear the view from unwanted point data.
Detach command detaches the camera from the selected trajectory and goes back to normal movement mode. The last anchor point on the trajectory is stored and using Re-Attach command one can snap back to the previous location on the trajectory. The Re-Attach command can also be used to reset the offset and angles when already attached to the trajectory so that the camera location is on the anchor point and the view orientation is reset.
Controls the initial location and direction of the camera after attaching or re-attaching to the trajectory. If checked the camera is viewing top-down and its' location is 15 meters above the trajectory.
Removes any clip setting and shows all points.
You can use this command to save position and camera viewing angle into a file <position.txt> that is then written into recording -directory (given in the settings). One can then quickly restore the camera back to this position afterwards.
To view a cross section you must first have a two point digitized in the active shape. Those will define the line along which the cross section will be viewed initially. On the secondary view the top view of the cross section will be shown and in this view the thickness of the cross section can be adjusted using keys 5 and 6.
Use this command to view the point cloud in full screen mode. Use <esc> key to return to normal two view mode.
Note that you cannot adjust any settings on full screen mode – only the menu commands are available.
This locks your camera 1.7 meters above the elevation of the lowest dense layer of points found under camera's XY position. If you only have ground points visible, for instance, your position is locked to elevation. You can only pan sideways when in this mode.
Makes a horizontal cross section view setup. The secondary view shows the camera and it's viewing frustrum and a horizontal cross section of 5 meters BELOW the camera. The primary view will show the view of the camera and the views are locked so one can steer the camera from the secondary view (use ctrl-wheel to adjust elevation).
This menu contains settings related to input controls.
Set the operation mode of mouse wheel either to elevation highlight or zoom/move (default).
Display |
|
---|---|
View |
View orientation – relative to trajectory forward direction if attached to one |
Rendering |
Quality of the rendering. Hires requires processing power and bigger frame buffer. |
Point Size |
Size of a single point adjusted – this is adjusted by density and distance |
Contrast |
This is the type of TerraZ required; either with big borders and small color points (strong), or thin borders and big color points (medium) or full color (none). |
View Lock |
When this is switched on the dual views are organized in a way that the camera in the second view is moved to 10 meters above the camera in the first view facing top -down. The camera positions are now connected in a fixed vector but their directions can be changed. Whenever the camera direction is changed from the 2nd view the camera direction in the 1st view is restored relative to the original angle between views when the lock was activated. When switched off the views move independently. |
Camera |
Toggles projection between perspective and parallel. You might need to fit the view after changing this. |
Stereo |
Toggles configured stereo mode on/off. Stereo depth slider is used to adjust the screen distance. |
Cursor – Standard |
Basic non-transparent and always on top cross hair |
Cursor – Solid (active depth buffer) |
A variation that has a white transparent center that highlights the depth map of points |
Cursor - Cross Plane (active depth buffer) |
A filled circle with a cross hair that orients itself according to a estimated cross section of the possible planes derived from the selected area's points. The area is selected with ctrl+shift+wheel. The same logic applies for digitizing and plane lock as for plane cursor (see below). |
Cursor – Plane (active depth buffer) |
A planar filled circle with a cross hair that orients itself according to a plane derived from the selected area's points. The area is selected with ctrl+shift+wheel. To keep the plane and still move the cursor around (for digitizing) hold ctrl down. You may freely move and rotate around and the digitizing plane will remain the same until ctrl is released. |
Channel - Intensity / RGB |
|
---|---|
Brightness |
The intensity base level. Not used in RGB. |
Contrast |
The contrast in intensity channel and gamma of RGB channel |
RGB vs RGBI |
The ratio of intensity mixture from around 50% to 100% |
Channel – Surface |
|
Contrast |
The amount of light (from the camera) that is not reflected back when the angle of surface is kept constant. Bigger contrast reveals small details but one has to observe surface facing directly towards it, otherwise the scene will be quite dark. |
Level |
The amount of surface shading applied |
Channel - Elevation |
|
Start Level |
The starting level of the sliding color spectrum |
End Level |
The ending level of the sliding color sprctrum |
Trajectory |
|
---|---|
Movement Speed |
Adjust the speed of the movement related to key pressing and mouse/automatic movement |
Movement Control with Mouse |
|
---|---|
Mouse wheel |
Move forward and backward either along the camera view or trajectory direction. |
Left button pressed + motion (drag) Middle button pressed + motion (drag) |
Pan left/right and up/down (leaves the cursor visible during drag operation) |
Left button pressed + motion |
Move forward and backward along the trajectory (only available when attached to one) |
Right button pressed + motion |
Adjust the angle of the viewing (not available in parallel projection mode) |
Right button pressed + motion + shift pressed OR Both buttons pressed + motion |
Rotate the model around. The rotation is done around the currently snapped point or (in case of no snap) around the octree coordinate. In void areas the rotation center point will be estimated. If the camera is top down then the rotation is done around z axis and center of view. |
Mouse movement actions + ctrl pressed |
Move/pan relative to the trajectory |
Middle button pressed + motion |
Rotate around viewers z-axis. The rotation will be reset (aligned with the horizon) after one turns the camera or rotates the model. Panning and zooming will not reset z-axis rotation. |
Other Mouse Actions |
|
---|---|
Mouse movement alone |
Search and snap points or search and snap trajectories |
Mouse movement alone + ctrl |
Stops snapping points and keeps the current depth plane active. The resulting plane depends on the cursor mode, for instance standard cursor just keeps the orthogonal distance to camera as constant. |
Right click when mouse is still |
Hide the mouse cursor |
Left mouse click with shift |
Appends a point into a digitized shape and also sets the distance measurement related to this point. If click is on a line between two existing shape points inserts the point between those points instead of appending it to the end of shape. |
Middle mouse click with shift |
Tries to snap to a vertex in existing shape (either active or reference). |
Right mouse button double click |
Exports the currently digitized shape and makes it inactive |
Mouse wheel movement (configurable) |
Moves fw/bw (perspective mode) or Zooms in/out (parallel projection mode). If mode is set to elevation highlight then sets the elevation highlight based on the snapped point. |
Mouse wheel + ctrl + shift |
Adjust the cursor selection radius (requires plane or cross plane -mode for the cursor) |
Mouse wheel + ctrl (view lock only) |
When views are locked one can adjust the elevation of the camera from the secondary view with this. |
Keyboard Commands |
|
0 (zero) |
Toggle stereo on/off |
Up/Down |
Move along the trajectory |
Page up / Page down |
Zoom in / out (parallel projection) Move forward / backward based on the selected cursor (perspective mode) |
1,2,3 and 4 |
Generates on-screen hires-image on 1.5, 2.5, 3.5 and 5.0 ratios. Disabled if not enough memory is reserved. |
5 and 6 |
Decrease and increase the cross section thickness. Activate secondary view when viewing the cross section to adjust this. |
C |
Clip points from the selected view that are closer than the snapped point |
E+shift |
Finalizes active shape as an open line string or group of single points (depending on the digitizing mode). |
Esc |
Removes clip from selected view and also removes any view synchronization. On a full screen mode restores the default two window screen setup. |
P+shift |
Closes active shape (if there are at least three datapoints in it) and sets it to a closed filled polygon. |
R |
Toggles recording frames on/off |
S + shift |
Closes active shape (if there are at least three datapoints in it) and sets it to a closed open polygon. |
S |
Take a snapshot image (record a single frame). |
T+shift |
View from the top and orientation reset (north in the top of the window) |
T |
View from the top without other orientation change |
U |
Clip points from the selected view that are farther than the snapped point |
Delete |
Deletes a highlighted digitized point from the shape |
W |
Toggle the mouse wheel operation |
SpaceMouse Movement : Hover mode (forced top-down orientation) |
|
Pan |
Push/pull converts to zoom in/out. Sideways panning converts to x pan. Fw/bw panning makes y pan. |
Rotation Z |
Rotates point cloud around z-axis |
SpaceMouse Movement : Basic moving mode (man in the street) |
|
Pan |
Push/pull converts to up/down. Sideways panning converts to left/right movement. Fw/bw panning moves forward or backward. |
Rotation Z |
Turns the camera by changing the heading of the viewer |
Tilt fw/bw |
Controls the pitch angle of the viewer |
SpaceMouse Keys |
|
Fit |
Fits the view |
T (top) |
Sets orientation to top-down and changes mouse operation to hover mode |
F (forward) |
Sets orienation to horizontal and changes mouse operation to basic moving mode |
Shift + Enter |
Digitize point |
Shift pressed after point digitizing |
Locks the rotation to the last digitized point instead of camera location |
The example of the ASCII file that is used to import and export digitized TerraStereo shapes is shown below:
1 0 25497314.768253 6672783.234002 2.917022 0 192 0
1 0 25497315.443959 6672783.261783 2.906883 0 192 0
1 0 25497316.429680 6672783.305057 2.896435 0 192 0
1 0 25497317.395459 6672783.355037 2.880271 0 192 0
1 0 25497318.054491 6672783.379760 2.860086 0 192 0
2 1 25497317.502050 6672783.772485 2.896018 0 192 255
2 1 25497318.151247 6672783.780747 2.881296 0 192 255
2 1 25497318.132483 6672784.453892 2.895346 0 192 255
2 1 25497317.466816 6672784.410401 2.906331 0 192 255
2 1 25497317.502050 6672783.772485 2.896018 0 192 255
The columns are:
Type The type of the shape. 0=random points, 1 = linestring, 2 = filled polygon, 4 = open polygon
StringID The identification key for the particular chain (unique for each shape in the file)
Easting Coordinates
Northing
Elevation
Red Color components in 8 bits [0..255]
Green
Blue
Note that by default TerraStereo sets all digitized shapes to green color, but when importing shapes as a reference one can create different colors for the particular feature.
A generic mapping software like TerraSurvey can be used to produce the file based on the dgn fila geometry, for instance.
In TerraSurvey this is done using a user specific format where the fields are surface/string/e/n/z/code and then using list elements command to fill the fields according to the wanted feature (using code as a color definition like '0 192 255')
Direct point digitizing feature with TerraSurvey requires MicroStation V8 and TerraSurvey v010.004 or higher – all other features work standalone.
Remember to upgrade to the latest available drivers from the GPU manufacturer if using GPU mode.
Operating Systems |
---|
64-bit Windows 10 |
64-bit Windows 8 |
64-bit Windows 7 |
Recommended Processors (CPU of the PC) |
|
---|---|
Operation mode used |
Min. recommended number of cores in the processor |
CPU (default) |
6 – the more the better -both rendering and arithmetic calulations speed |
GPU (optional) |
4 – more cores will not make rendering any faster, but they will speed up other operation |
Memory |
|
8 Gbytes of RAM is absolute minimum – at least 16 Gbytes of RAM is recommended. |
The program starts in a CPU and mono mode after a fresh installation. For these to work no special hardware setup is required and if you do not need the additional benefits of using GPU for rendering (faster speed) or non-anaglyph stereoscopic display you may skip the next section. However, even for CPU mode it is known that for some setups (typically laptops) the actual main GPU needs to be directly used and the additional integrated display-GPU (typically integrated Intel chipset gpu) bypassed, if possible. This option can usually be found from the driver or settings control of the actual main GPU (eg. AMD or nVidia).
A fully functional hardware setup consist of a single Quadro GPU either integrated into a laptop or installed into the workstation.
If non Quadro nVidia GPU's are used the software cannot run in stereo mode.
The typical settings that need to be changed from the driver before the stereo mode works are as follows (assuming nVidia's 3dvision is used, but similar steps need to be made also for other stereo display systems):
Change the monitor refresh rate to meet the requirement for stereoscopic viewing (typically 120 Hz)
Enable OpenGL stereo by setting on 3d-settings/manage 3d settings/stereo-enable (from the driver)
Set the OpenGL stereo mode by setting 3d-settings/manage 3d settings/stereo-display mode to the generic active stereo with nVidia vision or the other mode that is appropriate for your system
Note that these settings are found on a different part of the driver than the gamer's full screen 3d settings (enabled only on nVidia 3dVision systems). The gamer's full screen stereo mode is a good testing application that the hardware actually works – if one gets it working then the stereo windows will most likely work after the changing of OpenGL stereo parameters defined above.
The workstation should be configured for you to get it running right out from the box, but this reference is here just in case you accidentally change some of the settings.
Monitor displayport cables should be connected so that the one labeled A is the closest one to the motherboard in the AMD board (the upper and slimmer GPU) and the one labeled B next to it.
The driver settings from AMD FirePro advanced settings should be set to 'Enable quad buffered stereo' = checked and mode = 'Passive with Horizontal Invert (Dual Head)'
If there are problems with the stereo display even after these settings are as explainded here, try rebooting the workstation in order for changes to take effect.
GPU Hardware Setup
Since Quadro boards that have decent amount of memory tend to be quite expensive, there is an alternative approach for enabling the stereo operation. It is based on a new driver feature where one can select the GPU's for which the cuda is enabled.
Note! Not all configurations and quadro/geforce combinations work! Making a configuration as explained below requires knowledge of system configuration and may crash the system if the boards are incompatible.
The proposed hw configuration (best performance for the money) is the following:
One stereo-able Quadro board, but this can be a low end (but reasonably new) Quadro like Quadro2000. What is required, though, is that the driver for the board makes it possible to switch off the CUDA for this board. For example Quadro2000 will work.
One GeForce boards that must be TerraStereo compatible (new enough to support CUDA version requirement) and should have as big memory and as many cores as possible. These should in general be the very high end GeForce boards but their price is usually only a fraction of the similarly equipped Quadro board.
Install both boards into workstastion so that the low end Quadro board will be installed in the primary PCI-E bus and the GeForce board other after that.
Disable CUDA for the Quadro board and enable it for the GeForce board.
Connect the display(s) through the Quadro board and leave GeForce board unconnected.
GPU Memory (Settings)
The amount of GPU memory allocation defines the maximum memory the program will try to reserve for points in megabytes. Using small value will reduce the visual quality but improve speed. It also leaves more resources for other system applications like CAD programs etc. You have to restart program in order to start using the new values.
In addition to the memory reserved for points TerraStereo also allocates memory for the frame buffer from the same total memory amount (shown near 'Detected' label).
A default value of 0 will probe the GPU for the amount of memory left and reserve about 80% of the available space.
Supported Graphic Boards
GPU's are developing fast so it is not recommended to buy a board that has been on the market for more than two years even with considerable discount.
In order to increase performance over the CPU mode the rule of thumb is that the GPU needs to have at least 1000 cores (compared to the 6 core CPU).
Any OpenGL compatible stereoscopic viewing hardware (excluding the GPU) should work with TerraStereo. One can also use older quadro boards with CPU mode just to get the stereo mode working without any other performance acceleration.
The biggest performance increase when using GPU is for 4k displays and it is also nice for higher frame rates for lower resolution (like full HD stereo) displays. For high resolution displays (eg 4k stereo) it is important (both for CPU and GPU mode) that the memory bandwidth of GPU is high.
Graphics Boards |
|
---|---|
GPU |
Open GL Stereo |
nVidia Quadro [M/K/P] 6000 / 5200 / 5000 / 4200 - or newer |
Yes |
nVidia workstation Geforce GTX series Recommended: Geforce Titan or newer |
May be possible if a Open GL stereo cabable Quadro board is also present in the system – otherwise no. |