Inkscape pcb tracing

[Nominal Animal]

But you mentioned putting dots at every line by right clicking.

Here is the exact steps.  First, I do like I explained above, select the target layer I want the traces to be in, and lock the other layers.  Then I switch the Layers and Objects pane to the Fill and Stroke pane by clicking the icon at the top of the pane area.  Then, I select the Pen tool (the one below the spiral on the left side).  Here is what you see at this point (if you start with a blank document, for clarity):


Next, I left-click on the polyline starting point.  Do not drag; that creates a curve instead of a polyline.  (If you do an accidental click, you can always press Ctrl+Z to undo just the last path point, without canceling the entire path.  Press Esc to cancel the entire path.)  You see the starting point stick, and you can freely place the next path point:


I add each path point to the polyline by left-clicking the end of each new line segment:


Note that you can use your mouse scroll wheel and Ctrl+scroll to zoom in and out without breaking the polyline/curve draw mode.  You can even change layer visibilities.  Let's say this is the final point in this particular trace segment:


At this point, we right-click anywhere in the document area, to tell Inkscape we have added all the curve/polyline points, and this path is complete for now.  This will not add a new line segment, and will just "finish" all previously set line segments and curve segments in the path:


At this point, the newly created path is selected, and you can click on any Marker in the Stroke style tab of the Fill and Stroke pane, to pop up the marker selection tool window:


When you click on a marker shape, you can also change its properties.  (This is particularly useful when using arrowheads.)  Here, I just used the dot marker:


Clicking any other Inkscape window or tool, outside the marker selection tool window, will close the marker selection tool window; I like to click on the clear area below the Fill and Stroke pane for this.  You can then click on the other end marker, and select it the same way, and close it too the same way.  You'll end up with something like this:


If you notice your path points need a bit of moving, use the second tool from the top, Node tool.  This lets you move, add, delete, and convert (from vertex (AKA sharp corner) to curve tangent and vice versa) path points in the currently selected path.  If you have added the markers, those will come along automatically.  If you change the path color (in the Stroke paint tab of the Fill and Stroke pane), the marker colors will change as well.

[Nominal Animal]

In another post, I mentioned that it might be easier to first mark all vias and pins using the Ellipse/Arc tool (the circle one).  When you draw a circle, you can use the Rx: and Ry: to set the semiaxis lengths (half of horizontal diameter, and half of vertical diameter, respectively), and the fill color in the Fill tab of the Fill and Stroke pane.  When you have a suitable dot designed, just Edit > Cut (or press Ctrl+X), and you can paste (Ctrl+V, as usual) them to the currently active layer.  With the Select tool (topmost), you can move them.

In the upper right corner of the Inkscape window, there is a magnet icon for Snapping.  To the right of it is the small triangle for its details:


Click on the Advanced mode to see all the details, and enable only Cusp nodes and Object midpoints.  (If you make sure Bounding boxes is not checked, it does not matter if Edges, Corners, Edge midpoints, or Centers is checked; I just "cleaned up" the others to focus on the two important ones.)  Also make sure the topmost, Enable snapping is checked; you can also press % key to toggle this.


After you have placed all the dots for the pins and vias –– perhaps using slightly different colors to make it easier to see what is what? ––, drawing the trace polylines is much easier, because you don't need to hit the dot exactly; Inkscape will Snap the vertex/cusp/end point to exactly the center of the dot.


I do not know which approach of the two (polylines with end markers; or dots for pins, pads and vias) is more efficient.  One thing to note is that if you do use dots, you can right-click on them and select Object properties..., and rename their identifier (ID:) to something more descriptive, like say "Switch1_pin6".  (Again, it should start with an uppercase or lowercase letter A-Z, followed by any number of letters, digits 0-9, or underscores.  Other characters, including space, can cause issues later on.)

Furthermore, the Title: field in that same pane is also the tooltip text, when you open the SVG file in a browser.  When you hover the cursor on top of that object, the browser will show the tooltip text for you.  Consider it a comment field, as it can contain any text you like.

The same also applies to paths and even images: Inkscape will name paths as pathNNNN by default, and images as imageNNNN, with NNNN an unique number, but you're absolutely free to use better names.  The ID is also shown in the Layers and Objects pane, when you expand the contents of a layer, so functional names (like say ISO122_p5 for pin 5 of ISO122 IC on the PCB) can help a lot.


One important/useful trick: whenever you open an SVG file in Inkscape created by somebody else, or perhaps a program like Graphviz, first unlock all layers, and select Edit > Select All on All Layers (or Ctrl+Alt+A), and then Object > Ungroup (or Shift+Ctrl+G), and repeat these, until the status bar at the center bottom of the Inkscape window says No groups to ungroup.

SVG files do not actually have a concept of "layer"; they only have a "group" concept that Inkscape uses to implement layers.  This means that when you open an SVG file not created in Inkscape, it usually has lots of groups that makes editing it difficult.  Simply ungrouping everything is like separating everything so you can modify them individually.

I habitually use Graphviz to generate graphs from simple text descriptions, and gnuplot to plot curves and 3D surfaces from numerical data; and both can create SVG files you can finesse in Inkscape.  They both do group the elements, so unless you know about the aforementioned "select all on all layers, then ungroup; repeat until no groups left" trick, they can feel horribly difficult to edit.  With the trick, everything in them is editable.

SVG itself is an XML-derived format, and editable as a plain text file in any text editor.  (The current specification is here, but the vast majority of SVG files use only the path elements, with an occasional circle or rect.)
In loop123.inkscape.svg, the bulk of the contents is the two embedded PNG files, encoded in Base64.  I also used Inkscape 0.92 to create it originally, and one of its "bugs" is that it leaves a lot of unneeded metadata and definitions around unless you deliberately "clean" them out.

Instead of using image or plotting libraries to generate graphs or visualization images, I just have my scripts and programs emit SVG instead.
If you look at my home page, the image of Tux is also SVG, and regardless of your device precision (or zoom level, try Ctrl++ and Ctrl+-), it will always be sharp, never blurry like JPEG images; and it is also tiny, just a 21749 bytes total.


I am not saying this to try and make you fall in love with Inkscape and SVG images.  To me, they are only tools that help me overcome the limitations of my rather badly limited brain; that's all.  The learning curve is steep, but because of all the additional things you can do with Inkscape and SVG files, I claim it is worth the time and effort to learn for everyone who creates or modifies images conveying information.

I personally used Aldus/Macromedia Freehand for years (before the turn of the century), then Adobe Illustrator for a few years, before ever touching Inkscape.  They are all vector graphics editors, with plenty of alternatives; with Illustrator being the "professional" commercial software package everything else aspires to beat.

To repeat, I do not create a schematic from PCB images directly, because I know I will miss and mix/confuse the details.  It is just the nature of my focus; I definitely did not see the gorilla in the selective attention video.  Separating the "discovery" phase from the "understanding" phase helps with this, just like explaining a software/hardware development problem you are having to an inanimate rubber duck often helps you reorganize your thoughts and understanding (so that one usually realizes the solution midway through the explanation).

After tracing all the tracks –– and in your PCBs, they do seem to go all over, so I do not believe you can get by with doing just part of the PCB; I believe you will have to trace all the tracks on the full PCBs, even if you are interested in the circuits involving a couple of the ICs only ––, I would draw a small schematic involving those ICs, and how they are interconnected.  For the isolation amplifier, I would do the sensor signal path separately from the calibration signal generation, if at all possible.  My intent there would NOT be to draw the full schematic for the entire PCB, but to use the Inkscape traces to discover the connections, and verify them using a multimeter in ohms or diode connectivity mode.  (The reason for that is again the same: if I use a single source, I am likely to miss/mix/confuse some details.  But, when I use two separate sources, I can cross-check myself and my own work, with an eye open for any discrepancy; and if noticed, I immediately get into the paranoid "okay, I screwed something here, but what and where and why?" -mode.  I WILL make errors, but I use tools to discover them before I rely on the erroneous results, so I have at least a chance to learn from my mistakes.  While the extra steps may seem to be unnecessary work, I do them because without, I will make an unacceptable amount of errors.)

Maybe others do always perfect work in minimal time, but I definitely do not.

[loop123]

I'm puzzled. Even if you don't put the dots when drawing lines/traces and you do it globally at once by selecting all objects in all layers and using the marker only once at the end. It will only take a second to put dots in all lines automatically. You said it takes so long. What computer are you using? Celeron? Just to have idea because it will take much longer by accessing marker and putting dots at every line!

Thanks for the tips about Control-Z. It is very useful. I think i'll trace the entire circuit. But with hundreds of crisscrossing lines. Won't it get confusing. Also just because all is traced doesn't mean one can already understand the circuits.

Hope you can really write books because you are so good in explanations. Inkscape Demystified would be a good book for you to publish. Didn't you write any books before?

 

But you mentioned putting dots at every line by right clicking.

Here is the exact steps.  First, I do like I explained above, select the target layer I want the traces to be in, and lock the other layers.  Then I switch the Layers and Objects pane to the Fill and Stroke pane by clicking the icon at the top of the pane area.  Then, I select the Pen tool (the one below the spiral on the left side).  Here is what you see at this point (if you start with a blank document, for clarity):
(Attachment Link)

Next, I left-click on the polyline starting point.  Do not drag; that creates a curve instead of a polyline.  (If you do an accidental click, you can always press Ctrl+Z to undo just the last path point, without canceling the entire path.  Press Esc to cancel the entire path.)  You see the starting point stick, and you can freely place the next path point:
(Attachment Link)

I add each path point to the polyline by left-clicking the end of each new line segment:
(Attachment Link)

Note that you can use your mouse scroll wheel and Ctrl+scroll to zoom in and out without breaking the polyline/curve draw mode.  You can even change layer visibilities.  Let's say this is the final point in this particular trace segment:
(Attachment Link)

At this point, we right-click anywhere in the document area, to tell Inkscape we have added all the curve/polyline points, and this path is complete for now.  This will not add a new line segment, and will just "finish" all previously set line segments and curve segments in the path:
(Attachment Link)

At this point, the newly created path is selected, and you can click on any Marker in the Stroke style tab of the Fill and Stroke pane, to pop up the marker selection tool window:
(Attachment Link)

When you click on a marker shape, you can also change its properties.  (This is particularly useful when using arrowheads.)  Here, I just used the dot marker:
(Attachment Link)

Clicking any other Inkscape window or tool, outside the marker selection tool window, will close the marker selection tool window; I like to click on the clear area below the Fill and Stroke pane for this.  You can then click on the other end marker, and select it the same way, and close it too the same way.  You'll end up with something like this:
(Attachment Link)

If you notice your path points need a bit of moving, use the second tool from the top, Node tool.  This lets you move, add, delete, and convert (from vertex (AKA sharp corner) to curve tangent and vice versa) path points in the currently selected path.  If you have added the markers, those will come along automatically.  If you change the path color (in the Stroke paint tab of the Fill and Stroke pane), the marker colors will change as well.

[Doctorandus_P]

I suppose you just don't like answers that do not conform to your preconceptions, eh?

Nope, that's not it. I disagree here, but don't consider it a preconception. Sure, it is a bit of extra work to re-create the schematic too, but I have done it myself and it's not nearly as bad as you seem to suggest. Parts of the schematic can often beguessed and quickly drawn.  The Images of the PCB will remain the reference during the reverse engineering, and if there is a discrepancy between the schematic and the PCB, it's usually easy to spot differences between the pictures (provided they are of good enough quality) and the drawn PCB tracks.

And in the end, the schematic itself is an extra sanity check. You can analyze the schematic (you can even simulate it) to see if it makes sense. And on top of that you have DRC (and ERC to a lesser extend) as extra tools to check whether everything fits together in a proper way, (or if it does not: to indicate areas that need more attention).

And it's also not just me.
Have a closer look at the last ten minutes by the video linked to by Andy Chee. I agree the video is pretty low on information density. It's mostly a few guys fooling around and having fun. But they do have a workflow close to how I would do it myself (Note I saw maybe 5 minutes of that video).

For KiCad the workflow is mostly:

1. Creating of decent images.
2. Post processing of images. Contrast, removing distortions, etc.
3. Create KiCad project.
4. Put symbols in the schematic. Add meta info to the schematic (RefDes, resistor values, footprint info, etc).
5. Put the footprints on the PCB.
6. Use PCB Editor / Place / Add Reference Image. to add the images
7. Draw the PCB outline (preferably from making physical measurements for the sizes).
8. Pan / scale the image.
9. If it does not fit very well, maybe go back to the first steps. Fitting images are important.
10. Put a few footprints on their proper locations. (Either use the picture or physical measurements as reference).
11. When physical measurements are used, they can be used to spot image distortion.
12. Now you can spot a connection in the image.
13. Draw that connection in the schematic.
14. >[F8] to update the PCB, and then draw that connection as a PCB track.
15. Repeat until finished.

There is a bit more to it. It's just a rough outline of the workflow. Along the way you will want to shift and re-organize the schematic. If pictures are not clear enough you may have to go back to the physical PCB as a reference, maybe even measure parts, Set up net classes, maybe create schematic symbols or PCB footprints, etc.

[Doctorandus_P]

And also, there are several other youtube video's for reverse engineering a PCB with KiCad:

https://www.youtube.com/results?search_query=kicad+reverse+engineering+pcb

I quickly skimped though the video below, and it seems to be much better than the earlier posted video link.

[sparkydog]

I had to take about 80 pcb shots just to align the front and back pcb image. The key is trying to get pictures where all the sides are straight.

Go back and re-read Nominal Animal's comment on how Big Clive does it. To wit, get yourself a rig (I don't see why a pinhole is necessary) that lets you a) hold the board level on a steady surface, and b) hold a camera level relative to the board at a fixed distance. This could be as simple as a repurposed cardboard box. Use this to take many pictures of the board with small offsets. Stitch the results together in e.g. hugin. This will get you an image that isn't distorted (it may be rotated, but that's a trivial fix by comparison) and has reduced perspective distortion.

Alternatively, find a photogrammetry program that actually works for you, take a dozen or so photos at various angles, and use that to reconstruct a 3D model. Pop said model into e.g. Blender, rotate it until it's axis-aligned, and use that to generate an orthographic-projection image.

Sure, you might get a "lucky" single shot that's good enough, but proper technique and a little extra work will get much better results more reliably. (That said, just having a simple rig to ensure that the board and camera are level with respect to each other will make your life much easier.)

[loop123]

This may be the last pcb ill ever analyze. How do write text on the pcb bec I want to put label on the diodes path to make clearer. But I tried tracing again and again. The diodes really have this schematics. What could be its purpose? What happens when you have Vin+ of 10 microVolt with Vs+ of 7.2 just separated by a diode? what could be its purpose. Anyone pls give some clue.

[loop123]

After some reflections. Could it be a reverse battery  protector for the circuits? On its default mode. The diodes wont conduct. With reversed battery. It will short v+ and v-. The result being the amplifiers wont get reversed voltage. But if you continue using reverse battery with high current and no fuse. The wire or pcb trace will just burn up?

If the theory is sound. Why is the differential inputs connected to the diodes. Could it be when the input voltage is very high. It will short the inputs resulting in no output?

[loop123]

This may be the last pcb ill ever analyze. How do write text on the pcb bec I want to put label on the diodes path to make clearer. But I tried tracing again and again. The diodes really have this schematics. What could be its purpose? What happens when you have Vin+ of 10 microVolt with Vs+ of 7.2 just separated by a diode? what could be its purpose. Anyone pls give some clue.

(Attachment Link)

The Inkscape tracing is beginning to bear its fruits. The following circuit has the same 4 diodes at input. The purpose appears to be to clamp excessive input voltages to V+ and V-.

[Nominal Animal]

I'm puzzled. Even if you don't put the dots when drawing lines/traces and you do it globally at once by selecting all objects in all layers and using the marker only once at the end. It will only take a second to put dots in all lines automatically. You said it takes so long.

It is slow in Inkscape 0.92, instant in 1.3.2.

I think i'll trace the entire circuit. But with hundreds of crisscrossing lines. Won't it get confusing. Also just because all is traced doesn't mean one can already understand the circuits.

The next step is to select traces that are connected together (by say a trace on the other side), and Path > Combine (or press Ctrl+K), and adjust their color, so each set of connected traces is the same color/shade.

Didn't you write any books before?

Not books, but some teaching materials and such; and I've tutored a few friends in math and physics.

I had to take about 80 pcb shots just to align the front and back pcb image. The key is trying to get pictures where all the sides are straight.

Yep.  A rig to hold the camera is useful.  By "pinhole" –– me fail English ––, I meant that any box or tub with a hole large enough for the camera lens will work; the hole doesn't need to be very large, maybe finger thickness, but a larger hole will work too.  The idea is to put your phone on top of the box or tub, camera down through the hole, so you can move the entire rig to take the pictures.  Makes stitching them much easier.

I just want to figure out the purpose of the diodes (see photos for the third hidden one).

ESD and overvoltage protection!  Such diodes are robust, and you need to really abuse them to get them to short.  Let's say the forward voltage drop of these diodes is about 0.3V.

Any voltage on Vin+ below Vs- -0.3V is clamped to Vs- -0.3V by the topmost diode in your hand-drawn schematic.
Any voltage on Vin+ above Vs+ +0.3V is clamped to Vs+ +0.3V by the second diode from top.
Any voltage on Vin- above Vs+ +0.3V is clamped to Vs+ +0.3V by the third diode from top.
Any voltage on Vin- below Vs- -0.3V is clamped to Vs- -0.3V by the bottommost diode in your hand-drawn schematic.

In an ESD event Vs+ and/or Vs- will vary a bit for a very short time, but the voltage on the Vin+ and Vin- pins will always stay within Vs- -0.3V .. Vs+ +0.3V because of the diodes.

Anyway, is there a command inside Kicad to scan the images with Inkscape traces and just create full KiCad schematics of the components and circuits so you can run the circuit like Pspice?

I don't think so.

To repeat: I personally need the extra step between PCB tracing and schematic capture, to avoid making too many errors.
If I use KiCAD or EasyEDA or whatever for tracing the PCB from the board images, or directly drawing the schematic while examining the traces, the number of errors and details I'll overlook makes me crazy.  I don't like doing buggy work.

From the point you are now, I'd join the connected traces into nets, add the informative text labels, and adjust the trace/net/text colors according to the logical purpose of the net.  Supply rails and ground would each have their own color, so would different input stages.

Only then would I start schematic capture, placing the components first, then using the Inkscape document to check which component pin or pad is connected to which net, and a multimeter to verify non-obvious connections in ohms/resistance or diode/continuity/beep mode.  It is this double-checking that lets me trust my own work.  I know it sounds like extra work, but I do claim that it does help with the selective attention issue when doing direct-to-schematic/pcb-capture, i.e. missing connections or traces because you're too focused on the logical track at hand.  When you are focused on tracking where this track goes, it is too easy to miss the gorilla, say a track continuing through a pin/pad elsewhere.  That is the nature of human focus, and this dividing into steps with double-checking is my way to work around my own limitations.

If others can do equally good job faster or without the extra step, good for them, but I do suspect their bug rate will be higher.

I wonder how you trace the ground. It would be easier to just write Gnd in the pins connected to ground instead of drawing all traces between ground pins making such a huge crisscrossing, isn't it?

I'd use a specific color for ground pins.

I'd also add text labels nearby with the same color, describing each net.

Also how to write text because I'd like to write where Vin+, Vin- is in the inputs.

See the A four down from the spiral?  It is the Text tool, also activated by pressing T.
When that tool is active, you can click on anywhere to place a text element.  A small cursor will appear, and you can just type the text.  The text properties –– font, style, size, row spacing, superscript, subscript, and horizontal and vertical alignment –– are at the top of the Inkscape window, just below the menu bar.
The text properties affect the currently selected portion of the text, so I normally just type the text "blind", select all of it (by pressing Ctrl+A), and then adjust the text properties until I'm happy with it.  By default, the next text element I create will have those new settings.

The same tool can be used to edit existing text elements.  The mouse cursor will turn into a vertical line mark when you hover on top of a text element, and if you click there, you're back in text edit mode, at that point in the text.

Text elements are just like all other objects: they will be placed in the currently selected layer.  So, you'll typically want to lock the other layers, and select the target layer, so that you don't accidentally change the target layer by selecting an existing object on another layer.

If you use a nonstandard font, you can convert any text object into a combination of filled paths, by selecting one or more text objects (with the same text color), then Path > Object to Path (or press Shift+Ctrl+C).  If they all had the same text color, you can combine them into a single path using Path > Combine (or by pressing Ctrl+K, equivalently).  Such path/paths will have no stroke, only a fill, with the fill color being the text color.  However, it will no longer be editable as text; it will be just some elaborate filled combined curves.

Useful trick: when saving SVG files, I always save the master file in Inkscape SVG format, then a copy as an Optimized SVG after converting all text objects to paths.  The Inkscape SVG version is the one I edit, and the Optimized SVG version is the one I publish or import into other programs when needed.  To convert all text objects to paths, simply select any one text object, then Edit > Select Same > Object type (or press Shift+Alt+A), and finally Path > Object to Path (or press Shift+Ctrl+C).  When you save a file, the format is set in the lower right corner, default being Inkscape SVG.  When you use the Optimized SVG format, after you select the file name and click Save, Optimized SVG Options tab will pop up.  I tend to select Convert CSS attributes to XML attributes, Collapse groups, and Work around renderer bugs; in SVG Output tab, Remove comments, optionally Embed raster images (if you want to include any PNG/JPEG/etc. images you had linked to, in the final optimized SVG file), and Enable viewboxing if used for browser viewing – but NOT selected if used for printed media with physical measurements; in IDs tab, Remove unused IDs, and if you defined any IDs by hand, put the prefixes into the Preserve IDs starting with field, separated by spaces.
This is how you can use funky fonts in SVG images and have them show up correctly, just like I have on my home page.

[loop123]

I'm puzzled. Even if you don't put the dots when drawing lines/traces and you do it globally at once by selecting all objects in all layers and using the marker only once at the end. It will only take a second to put dots in all lines automatically. You said it takes so long.

It is slow in Inkscape 0.92, instant in 1.3.2.

I think i'll trace the entire circuit. But with hundreds of crisscrossing lines. Won't it get confusing. Also just because all is traced doesn't mean one can already understand the circuits.

The next step is to select traces that are connected together (by say a trace on the other side), and Path > Combine (or press Ctrl+K), and adjust their color, so each set of connected traces is the same color/shade.

Didn't you write any books before?

Not books, but some teaching materials and such; and I've tutored a few friends in math and physics.

I had to take about 80 pcb shots just to align the front and back pcb image. The key is trying to get pictures where all the sides are straight.

Yep.  A rig to hold the camera is useful.  By "pinhole" –– me fail English ––, I meant that any box or tub with a hole large enough for the camera lens will work; the hole doesn't need to be very large, maybe finger thickness, but a larger hole will work too.  The idea is to put your phone on top of the box or tub, camera down through the hole, so you can move the entire rig to take the pictures.  Makes stitching them much easier.

I just want to figure out the purpose of the diodes (see photos for the third hidden one).

ESD and overvoltage protection!  Such diodes are robust, and you need to really abuse them to get them to short.  Let's say the forward voltage drop of these diodes is about 0.3V.

Any voltage on Vin+ below Vs- -0.3V is clamped to Vs- -0.3V by the topmost diode in your hand-drawn schematic.
Any voltage on Vin+ above Vs+ +0.3V is clamped to Vs+ +0.3V by the second diode from top.
Any voltage on Vin- above Vs+ +0.3V is clamped to Vs+ +0.3V by the third diode from top.
Any voltage on Vin- below Vs- -0.3V is clamped to Vs- -0.3V by the bottommost diode in your hand-drawn schematic.

In an ESD event Vs+ and/or Vs- will vary a bit for a very short time, but the voltage on the Vin+ and Vin- pins will always stay within Vs- -0.3V .. Vs+ +0.3V because of the diodes.

Why did you ignore the Vs+ of 7.2V and Vs- of 7.2V? Remember the right of the diodes are connected to the power supply. The clamping voltage is about 7.5V +-  (if the diode forward drop is 0.3V) or 8V +- (if the diode forward drop is 0.7V), not 0.3V.

Lets use example of 50V input voltage.



IF the Vs supply is steady enough and able to maintain 7.2V then 50V at input will cause current through R1 and D2 to Vs.
Since forward voltage of the diode is about 0.7V and Vs is 7.2V voltage drop across R1 is

Vr1 = 50V - (0.7V + 7.2V) = 42.1V, right?

Your website doesn't work. No contents inside except your logo and email address like this.



Is your degree Physics? What part of physics do you tutor? I'm interested in physics beyond the Standard Model. And may need your help to formulate a theory.

Thanks for the rest of the Inkscape tips. But I won't continue tracing all circuits anymore. It's enough to know there is no RFI filters at the input. So I want to add RFI filters for now.

[sparkydog]

I had to take about 80 pcb shots just to align the front and back pcb image. The key is trying to get pictures where all the sides are straight.

Yep.  A rig to hold the camera is useful.  By "pinhole" –– me fail English ––, I meant that any box or tub with a hole large enough for the camera lens will work; the hole doesn't need to be very large, maybe finger thickness, but a larger hole will work too.  The idea is to put your phone on top of the box or tub, camera down through the hole, so you can move the entire rig to take the pictures.  Makes stitching them much easier.

Ah, yes, that's how I expected such a rig would work. Note that "pinhole" has a specific meaning in photography, which perhaps led to some confusion. Here, the goal is just for the rig to have enough structure to hold the phone/camera more-or-less securely. (As a gross oversimplification, the less material you remove for the lens to look through, the more sturdy that structure will be.)