If you encounter a bug in the program, please send details of the problem along with the Crash Report to
Set of gerber files (a subset even better) allowing me to repeat the problem is substantially helpful. The files will be kept secret.
Alberto's reported bugs in flattening - rewrite cross detection to n*log(n) scan line algorithm, as it is in effects/triangulation.
There is unnatural transparency gradient, close to the board edges, when observed from some angles. Best fix would be "per pixel fog". I am slowly allowing myself to think about this more GPU power demanding method.
Sometimes board outline is routed via the drill file. This could be detected and possibly used instead of the board outline drawing Gerber file.
I have build the menu (GUI) system on top of DXUT (3D graphics). I did it to enable working in a fullscreen mode. Fullscreen mode saves resources, especially for the older GPU - I still have fear that some users may suffer on GPU performance. It could be useful to have both systems: native windows GDI-GUI and current DXUT.
Single layer board auto file to stackup assignment have top/bottom problems.
hot fix 4 of 4
One of "recently" corrected artifacts debug, caused another problem, visible as removal of a vertex in a very specific cases, see picture.
This problem is very sensitive to actual value of View / Render Effects Parameters / Fade Gradient Length. As a workaround, you can change it by a small value and retriangulate the board.
Board Outline, declared in stackup, but rendering to no "loop", i.e. no surface, can cause a crash. A simple bug fix it is not
enough, as the resulting PCB will be simply empty - that is confusing
Grouping tracks/pads/etc... into groups of overlapping objects. Creating the copper netlist and separating objects for geometry calculation.
In case of 2 outlines (polygons) touch test, one of the segments of one of the polygons was spiked, in most of the cases. It was a kind of
"how was it working at all?" error. One victim, as far as I know.
Inability to detect the footprint features will no longer result in an exception (crash), but just an undetected component type.
Crash when saving .zofzproj file containing a sparse BOM table.
Fixed problem witch the IPC356 parser option: "Long ID/Pin name correction". The option allows to read an non compliant file,
where component id and pin name text are longer then assigned space in the ipc356 specification. The option is enabled by default.
Lines containing a "M" marker are not read, also from the compliant files, if this option was enabled.
The shape recognition algorithm was unable to find pin "1" in case there were only non-plated holes in a component.
Such condition could be caused by above bug. Uninitialized value was used as index, causing a crash somewhere later in
the execution, because of a memory corruption.
In some rare cases, rendered PCB, saved as a ".zofzproj" file, was corrupted.
Component geometry detector crash on some complicated BGAs.
A lot of new component classes are added to the internal library.
Menu system for selecting component types and adjusting component parameters.
Mathematical operations for aperture macro definition: unary "-" and "+" and parentheses "(", ")" are added, as defined in GerberX2.
Some file names are longer then dialog box - implemented horizontal scroll bar.
Option to sort files by date/time.
Lighting - simple ambient and difuse light dialog.
CAM_Load_Rules - added sections OrCAD (Alexiej) and Cadence Allegro (Benjamin).
CAM_Load_Rules - must be more visible - added edit button in the autoload dialog.
Aperture Macro rotation center adjusted to the new clear Gerber definition.
Aperture Macro rotation for circle added as option - defined by GerberX2.
Save/Restore of the program window position and size may permanently prevent the program from starting, if minimized or otherwise inactive window position is saved. The Direct3D can initialize only in visible window - program was exiting with Direct3D device error.
There was a hot-fix for old PCAD inserting space characters in Gerber files. The hot-fix was also removing spaces in G04 (comment). That was making problems when the comment string was starting with digits. eg. G0433...)
Excellon routing offset matching to Gerber was not working. Slots could disapear or be miss-placed.
The macro definition was not flattened before it's use. Negative drawing was preserved and was destroying parts of drawing outside the macro.
Very short arcs were replaced by a full circle. This condition detection is now set to trigger on equal points only.
Irregular pinout (via IPC356) could hang the program.
Adding step and repeat have been changing ".zofzproj" format. This should have been accompanied with the .zofzproj version being incremented.
I missed this. Unfortunately ".zofzproj" files which are saved by version 0.5.0015, are unreadable by any other version.
Reading of the BOM file. Component info is displayed on the right plane.
Tabelarized Component Library, autodetection. In case a component is not identified, the former "extruded" form is displayed. The component tables are stil under development.
Currently, implemented are: L-lead (QFP, SO, SOT, DPACK), J-lead, DIP, chip capacitors (resistors).
GUI bug - folder selection crash. I could not see it for a long time, but it seems I have fixed it with your help.
Cross-Section of the PCB and components. Use "DEL" key, also measurement maker(s).
Hole diameter display (was forgotten while in the implementation).
Wheel zoom movement time constant setting in "Navigate" menu. Plus shift-key speed modifier. Ctrl and shift works also for keyboard controlled movements.
Possibility to switch off auto-offset matching for drills and IPC356. Priority for zero offset if matching, in this case, is detected.
PCB Outline processing, finding the correct outline, in case of multiple "T" connections, can explode in time or memory usage - possibly leading to crash. This often hapends if some tracks are misinterpreted as outline (in wrongly set meta drawing) or when mutlipe lines are not finised to exactly the same point.
IPC356 autodetection of rotation angle units, when radians are used. Why to use radians anyway if there are only 3 digits of precision assigned for rotation?! 90deg=>157=>89.954deg
Netlist related operations could crash if there were no components on a bottom side, in some cases.
Removal of board outline of rendered board caused crasch.
But report tool can handle "portable version" stored log file.
More layer types can be used in the CAM_Load_Rules.txt - autoload script. Layers can be defined in each section.
When particular section is used as a rule for filenames, associated layers are added to the program.
Alternatively, layer types previously defined in the Stackup / More Layer Types can be used. See the description inside CAM_Load_Rules.txt
If last object in an object list flattens to nothing, some other objects may be removed.
Board outline containing sequentially equal points may cause crash.
Board outline path selection could create empty loops by going back and forth the seme lines, in some cases.
IPC356: if the 317-Through-Hole line entry was used to describe a SMD feature, it was read incorrectly, causing the Component-Mesh-Modeler to crash.
IPC356: unnamed nodes are assumed to be not-connected. The node name is automatically generated in form: " N/C_nnnnn". The same should be done to nodes named "N/C".
IPC356 does not contain information about THT component placement side (top/bottom). It is now possible to swap those components to the bottom side. Use right mouse button click on selected component.
It is possible to hide unwanted components individually. Use right mouse button click on selected component.
State of component hide and swap is saved together with the PCB stackup and rendered board files.
Possibility to add more layer types. Acces via Menu: Stackup / More Layer Types.
The extra types are saved together with stackup (.camset) and with rendered project (.zofzproj).
Loading stackup file, adds to current set of layer types. Loading rendered project overwrites the types set.
Layer types alone, are saved and loaded together with the Color Presets.
All colors preset is saved and restored by stackup and rendered board files.
Beside above, rendered board save/restore contains: state of layer enable checkboxes, the Component checkbox, the Hollow checkbox and saved camera positions (0..9)
Semitransparent Solder Mask was drawn twice in the Hollow mode, possibly reducing the frame rate.
Semitransparent Solder Mask transparency (alpha) value was reversed in the Hollow mode.
Portable version cannot access its init file. Parameters are not saved for the next time the program is open. The path to the file is changed, by error, to the c:\windows directory and therefore not accessible without administrative privileges.
Sometimes, when routing and drills are on separate files, auto-fit drills to Gerbers fails. One of the drill files is shifted or not visible.
Overlapped, but not continuous routing on the drill "plate" may lose some of its components.
CPN, C, for crosscheck versus measurement selection
Ctrl */ rotate to the nearest rectangular rotation
Home rotates to the nearest rectangular position
Ctrl-Home, rotates to zero angle - as Home before
Screenshot directory and use-project-directory check box
Dialogs are not 'transparent' for the mouse click or drag
Component Display Feature.
At the moment, the only information about component is taken from IPC356 file.
Components are represented by set of basic shapes: cube, pin/cylinder, rect-chip-filling.
The actual component shape selection is guessed, based on e.g. component designators. (see IEEE 315 or Wikipedia).
Also the THT component layer is always assumed to be TOP.
Therefore, urgently, more features must be added:
AutoLoad suggests the path and name for saving the .camset file.
Camera FOV (Field Of View) setting and orthogonal projection setting.
Objects clearly outside of the Board Outline are removed.
Introducing a new features in the netlist-browsing, I have managed to block the old one.
ESC-key is not stopping the autopilot, but it is expected to.
New version check function contact the website via wrong port (443 instead of 80). Till now that was working. Due to some kind of update, my provider has removed this redirection. I will make a notice about the new version release via the website and the social media.
Above problem also disables the bug report tool.
Auto-Load fail when simplified algorithm - could not decide gerber or excelon, after 20 first lines.
If two outlines follow directly each other in the Gerber file and first point of the second is equal to the last point of the first then those two outlines will be connected. If the outlines overlap, XOR will be then applied instead of ADD.
"Netlist/Nodes (PCB)" and "Netlist/Nodes (IPC356)" board browsing. The difference between those two methods plays role in case of discrepancies between PCB-detected netlist and IPC356 netlist.
Reading the board outline from copper layers in addition to former method, that is using a separate layer. Meta image is distinguished from the layer image by using "impossible" thin tracks. The "meta-way" is now officially accepted by the Gerber Specification.
Auto-Load works from a command line parameter if you specify just the folder path. If you specify a file of ".zofzproj" or ".camset" type, it is also loaded, as it was before.
Automatic adding of core/prepreg in between copper layers, when [Set Equal Spacing] in BoardOutline / Thickness is pressed.
Drill routing of more complicated shape, a multi-segment path, does not store mesh reference. Trying to highlight some nets (jump net, error, pin) can cause program crash.
Drill routing of more complicated shape, a multi-segment path likely causes false positive detection of short circuit.
False warning about unsaved work on program exit.
Aperture Macro Comment not implemented. When encountered, causes syntax error: "Macro Primitive Type ?"
Gerber Image Polarity "Force Negative" and "Force Positive" setting in addition to existing "Default Negative" and "Default Positive" settings.
Auto-file-detection checks file content to be Gerber for layers and Gerber or Exelon for drills. This modification solves ambiguous sections, e.g. Protel99se v.s. Kicad "drl" and "txt" - Protel creates "drl" files in a rare binary format,
whose could be taken as Kicad excellon drill files. This functionality is a stub for GerberX2 "%TF.FileFunction" - based autoload.
Some Gerber export modules can generate longer numbers (more digits) than format declared in the same Gerber file header. Workaround is possible in case of trailing zeros removal mode.
Beside signed installation package, there is a companion version, running directly from a folder, without installation and associated privileges. It keeps all settings and examples in the same directory, whenever you place it.
Two more color presets from Christian Weidner, including OSHPark purple.
Ultra High resolution screenshots. x1 (current resolution) to.. x10 19200x10500. download example
Gerber parser allows for Gerber X2 codes: %TF %TA %TD codes, but ignores its content. You can read Gerber X2, but ZofzPCB makes no use of the new additional information at the moment.
Still not correctly interpreting board outline drawing when using single line as a board cut-out. this can happen in panelisation. Example: two rectangles, one inside the other, form a frame shape. This may be not intended.
Additional syntax in CAM_Load_Rules.txt autoload script:
Plated/Non-Plated option (Drill_PTH/Drill_NPTH).
drills from-to assignment.
Added section "Mentor" and section "Proteus" to the CAM_Load_Rules.txt autoload script.
Added Mentor specific Excellon drill size definition.
ver 0.4.0000 ()
I need some pleasure too. Feature-freeze dissolves. I have decided to add correct thickness for tracks and other layers. It is possible to switch it off in the view menu. This could be required if those additional triangles overheats your GPU, causing too low frame rate.
This is now an official release.
I have arranged for code certificate. That is securing the install package transfer.
I have 3 reports about error in the triangulation, it seems to be the same bug - transparency fade-out band extraction (used for preparing the hollow effect).
It is related to creating band (or ribbon) inside big surfaces, in case of bumpy track drawing e.g. horizontal tracks only. I am still searching the solution faster then n^2.
Some old software does not generate Gerber image polarity parameter %IPPOS*% or %IPNEG*%. Therefore I have added "Def. Negative" checkbox.
The checkbox defines image polarity in case the Gerber file does not declare it (default polarity). Now I found that there are some Gerber files simply declaring wrong image polarity.
This is driving me crazy. What to do?
Extremely long "flattering" time. Detailed description:
for the "flattering" I am still using bounding rectangles to limit the edges crossing algorithm.
It works for local and short tracks. But when it is not efficiently working, I have n^2 operations, and this can grow huge.
The second stage (triangulation) I have already n*log(n) – scan line sweep algorithm successfully implemented.
And I am planning to apply this to the first phase too. (edges are circles – that is the difficult point).
So, n*log(n) is much faster then n*n, I have already seen this. (n is a big number.)
Additionally, since I am using double precision and(especially for circles) I am using 3 value logic: (yes – no – do not know), there is a possibility that I need to solve binary equation system to meaningfully decide what edges are staying and what are removed. Sometimes it takes a very big matrices to solve (n^3 for the Gaussian back-substitution algorithm)!
Solution for this is a double-double arithmetic package – a very cool approach, using pair of double numbers to have almost quadruple precision. This should minimize the ‘do-not-know’ state, as it did in triangulation.
Spaces (character 32) in the Gerber files are forbidden by format definition. But somehow, there are Gerber files containing spaces, as I have seen, in the aperture definitions.
Gerber file read, statements like: $1=$2+$3 - there is a chance of exception.
Lack of pads or IPC356 test-points on one side and simultaneous lack of THT pads-test-poins causes auto-scale and auto-offset calculations for IPC356 fail, placing components and test-points almost at infinity.
IPC356 Load errors were not displayed.
Multiple IPC356 netlist test-points definitions on a single place could cause problems by error browsing.
Only first type of IPC356 errors were displayed correctly. Next sections listings interpretations were shifted.
ver 0.3.0002 ()
This is a feature-freeze version. There are new features related to IPC-356 and netlist. Alberto have reported some crashes in flatening.
ver 0.3.0001 ()
The most robust triangulation I ever had. - I have finally implemented simply a higher resolution calculations in previously unsolvable cases.
I use library from High-Precision Software Directory - really clever quad-precision constructed on FPU double precision base.
The flattening part however is still in FPU double, where the special cases are solved by a three-valued logic. All 3 Alberto's reported crashes are fixed. I am still working on some new features.
Effect bug - fix it by setting View / Render Effects / Minimal Hollow Area to 0
0x80004005 error message: Locking of the vertex buffer as update, when at least 200MB of video memory is allocated. Workaround: change the window size (buffers will be recreated)
Netlist detection for negative rectangle - short circuit falsely detected
Board outline drown (routed) in the drill file, when it is a closed loop, is acting as one big hole and making all components shorten. It should actually create 2 netlist entries instead and leave the cut-out on place. In addition there is a bug in the current version, related to simplified handling of routed loops
Alberto's reported crash. It is very likely related to the above drill-loop handling There are 2 independent bugs
User rights problem by installation - please run the installation as administrator. This can be done using right-click on setup.exe.
The setup.exe is now included in the download package. Another cause may be a default selection of installing for ALL users.
ver 0.2.0011 ()
Solder Mask semi-transparency looks better, but requires more power form pixel shader
Solder Mask is very close to the copper layer and Silkscreen, we have Z-fighting artifact when board is far from camera
ver 0.2.0010 ()
excellon G85 and excellon empty fast fix
Installation package contains DLLs from WinXP. Internet related help functions are not working on Win7. To fix this issue delete from program directory (C:\Program Files (x86)\ZofzPCB\ZofzPCB) the following DLLs:
Solder mask should be negated and semitransparent, regardless increase GPU power
Right panel check boxes "orphaned" and overlapped while manual layers build-up process
Right panel check boxes may not be visible after auto-load
Drill file in a Gerber Format: Circle=Drill, Track=cnc-Routing
Automatic drill file format recognition
ver 0.2 ()
Translucent and opaque PCB volume rendering - slider adjustable (real-time)
Individual layer visibility checkboxes
Pre-release Version 0.2.0007 () - bug fixes:
Side bar: copper layer numbers are not assigned in some cases
Download button disconnected form its handler
Hollow option badly influences measurement text display
Slot drills are seen as somehow bigger to netlist detection algorithm
Error in pads search by mouse "E" edge projection
Layer disable have no effect on mouse projection detection
ver 0.1.0010 ()
Save .zofzproj file for zero sized components - in this case file is broken.
"Polygon flash" pads (e.g. octagonal) are not seen by drill/pads matching. In case there are only "polygon" pads, drill fitting is not working.
Bug in Save .zofzproj file, when pcb contains zero-sized elements.
Camera motion breaking after releasing cursors keys is independent for each direction.
Polygon pads are now also taken for drill/gerber offset and scale matching.
Board volume done by 2 "bottoms" of the pcb (only one visible unless cam. is inside PCB) and color manipulation.
Board volume dimension adjusting accordingly to peeling layers - now "single layer" option seams unnecessary.
Layers have hollow and opaque option, pre-rendered, allowing immediate switch
Big surfaces are divided to smaller, to avoid very long triangles. Important for vertex based color error, more visible after adding above feature.
10.2012 - 05.2013: no progress
New, faster algorithm for polygon intersections, processing surfaces for triangulation stage for straight segments only.
Very small "hollows" are skipped.
Non-plated drill files, if listed at first position, could cause read error.
New Side Panel - individual layer visible checkbox, volume transparency slider, hollow/opaque options.
Mouse wheel used for zoom by default.
Color scheme concept is a bit different now - color picker menu is updated.
Color scheme save / restore - integrated into menu. Repository: "%APPDATA%\ZofzPCB\*.color".
Changes in geometry algorithms requires adjustments in .zofzproj file format
Changes in geometry algorithms requires adjustments in the layer editor
RS274X commands development:
any two character command without parameters e.g. "%AD*%"
Step and Repeat meaning "no repeat"
real Step and Repeat (other as above)
New Rendering Effects Parameters in View menu
features: layer spreading, support for 3D mouse from 3Dconnexion
New version download broke - not connecting to browser. Fixed 04-07-2012, upgrade not required, version number not changed.
Seams more testing finally taking place, several problems are detected in the core of the program:
"Complex Apertures Tracks not implemented" Gerber tracks drown by complex apertures, till now only circular apertures can be used
for drawing. Now expanding to rectangular.
Regular Polygon apertures error.
Excellon explicit decimal point format request - please send me a file example.
Special cases by rendering/triangulation.
ver 0.1 - the major update
Resizing (or other Direct3D Reset) and active net marker - program crash.
Triangulation, weakly simple polygon touch point, some special edges order and directions - program crash.
Sensitivity to aperture sizes close to assumed calculation error (0.1um) - program crash.
There is a tendency to use such tracks for "meta" object, like operator text or measurements.
Object below 1um will not be triangulated.
Zero sized tracks were crashing the program. Now objects rendering
to "nothing" are disposed.
Aperture Macro (AM) was instantiated to aperture definition using
millimetres, even when inch was specified.
Debug log file was written in the program directory. This is not
allowed above winXP.
The place is changed to CSIDL_COMMON_APPDATA
winXP: C:\Documents and Settings\All Users\Application
Gerber File Reading - multiplication sign in aperture macro
definition is 'x'. (according to RS-274X).
ZofzPcb have been treating capital 'X' as coma.
Now both, small and capital 'x' means multiplication.
Wide range of graphics hardware - exclusive use of 16-bit vertex
Support for G85 - slot drills
Two layer-dissection options - use mouse wheel click
Guessing Excellon drill format:
- decimal point - search for fit in a loop.
- first guess is a copy of gerber settings
RS274X commands development:
Step and Repeat
There are some popular graphic cards, having a limitation of 16-bit
Example is Intel's GMA 3150.
At the moment, the gerber viewer uses one vertex/index buffer per layer.
16 bits is usually not enough for this approach. However, whenever
possible, the gerber viewer will use 16bits, allowing small designs to
be displayed. Unfortunately some components of the demo file
Flexible approach will be implemented in one of the next versions.
Guessing Excellon drill format - it is often needed to put file
format options manually.
Option to use a copy of associated gerber read settings will be
Following RS274X commands cause 'debug' exception:
Step and Repeat
Load stops when reading CAM files.
Reading empty gerber files (containing no drawing primitives) creates
empty layer - this is understood as already signaled file read error.
workaround: remove empty layers from layer stack list
status: fixed in ver 0.0.0004
Problem in flattening a rectangular track fill path, where
track grid exactly equals track width.
workaround: increase filling-tracks width or decrease grid in
status: fixed in ver 0.0.0004