Dimensions from Drawings- Software tools?

[Scutarius]

Hi,
It's very common for datasheets to include mechanical drawings with missing dimensions, where you usually have to subtract and add to get the desired value.

Do you use any special tools for this? I'm thinking of simple software with a predefined grid where you can load an image, adjust the size, and measure. I think I had one before... or it was for reverse engineer PCBs, 

[ataradov]

Often images in the datasheets are not to scale. They also often omit and compress repetitive features.

[w42i]

I often use Inkscape to load the page of the PDF I am interested in, scale the drawing to 1:1 and then use the measurement utilities of Inkscape to get the dimensions I need.

Usually I use one dimension of the drawing to set the scale and the verify it with another one or two. If multiple dimensions don't contradict another, it is likely that the drawing is to scale. But it isn't guaranteed, either. There could be just one feature/dimension that is off.

Take the results with a grain of salt, because the line widths can influence the result (this may also depend on the export settings when the PDF was generated, etc.). It is not as reliable as taking the measurement in a CAD model where the CAD knows the exact dimensions.

Anyways, often this is all you've got and you have to work with it, sometimes just a blurry scanned pixelated image.

[Scutarius]

Thanks!
What I have been doing is importing the drawing into Fusion 360, adjusting the scale using 2 or 3 dimensions from the drawing. It's tricky, but most of the time it works reliably.

I have come across many low-cost suppliers —like those for USB-C connectors—that are really a pain to deal with.

[MarkF]

I have brought photos into FreeCAD to use as a reference.

The scaling is a little cumbersome if the aspect ratio of the photo is not correct (but doable).

FreeCAD can only measure from lines, points, and planes within the drawing. 
You would need to draw an outline of the object in the photo to measure from. 
Therefore, very precise measurements would be a 'not_happening'.
I have had some good results but not to the precision I suspect you're wanting.

Then as @ataradov said, how much do trust the accuracy of the drawing?

[thm_w]

Do you have a bad example of this? I usually just break out the calculator, I would not trust the drawing.

Also a lot of these USBC connectors are copies of major brands, so you can sometimes find an equivalent 3D model from another supplier.

[PlainName]

It's very common for datasheets to include mechanical drawings with missing dimensions, where you usually have to subtract and add to get the desired value.

That is the way mechanical drawings are meant to be.

The reason is datums and tolerances. Generally, measurements will be from a datum and if you have a two complete but different measurements to the same object then they won't match if you allow for tolerance. An example might be a series of holes across a PCB: you get given the width of the PCB and the distance to the first hole, first hole to second, etc., but not the last since it wouldn't come out the same as the overall width if you allow for tolerances.

Another example is where something is centred: the datum may be one edge and there will be a measurement to the centreline, but not from that edge to either edge of the object. (At least, if you do get a distance to the object then it won't be centred). Again, it's a matter of the measurements not agreeing if you take into account tolerances. And there will always be a tolerance - you cannot have a tolerance-free measurement.

[gamalot]

This is a standard form of mechanical dimensioning. My solution is to calculate the desired dimensions mentally.

I believe this might help prevent Alzheimer's disease. 

[Scutarius]

Unfortunately, I don’t have a good example at the moment.
I understand it’s not always possible to include every dimension in a drawing.

That said, I spent about 30 minutes with an AI and put together a simple tool that does exactly what I needed.

[thm_w]

If you set your zero reference to the center of that red crosshair, makes it easier to enter in any modern EDA tool.

eg. left hole x = -5.78/2, y = 0
A7 x = 0.5, y = 0.45 + 1.15/2

I agree though it would be nice if a tool could just pull all that data in, you'd click on the pad and then it would give you the x, y. Or if the EDA tool could overlay the data to scale.

[abeyer]

That said, I spent about 30 minutes with an AI and put together a simple tool that does exactly what I needed.

I'd be curious to see what you ended up with there, if you're willing to share.

[Doctorandus_P]

That is the way mechanical drawings are meant to be.

The reason is datums and tolerances.

Indeed. All mechanical stuff has a certain tolerance and when you attempt to place all "measurements" in the drawing, then you get completely lost of where the tolerances are. So for 30+ years I add or subtract some measurements to extract the numbers I'm interested in. For the last 10 or so years I copy and paste numbers into a python console (I use python not for programming, only as a calculator).

For the rest, I use both FreeCAD and KiCad, and both can import a (pixel based) image and scale it. Then doing the calculations to get the "intermediate numbers" you need to design your (for example) footprint is easy. When half the footprint is drawn, it's often beneficial to "re scale" the image, to make a best fit with the partial drawing you have made. I still get all the "important" measurements from real calculations, but scaling the footprint and comparing the "best fit" with the drawing adds confidence that no big mistakes have been made (or it points out dubious areas).

I once had a gear with unknown parameters, I did not know whether it was metric, or based on bananas. So I put it in a flatbed scanner, imported the image in FreeCAD, drew some gear tooth in FreeCAD and this made it easy to average out the "fuzzy edges" of the scanned image, and get a good estimate of the real dimensions of the gear.

Reverse engineering from photographs is a separate topic. Especially reducing the distortions made by projecting a 3D object via a lens on a flat surface is an important step.

[PGPG]

It's very common for datasheets to include mechanical drawings with missing dimensions

I made (rather tried to made) my first dimensional drawing when I was 6 years old. After an illness, I couldn't go to preschool, so my mother took me to her class. They were drawing the crankshaft of a ship's engine (my mother taught technical drawing at a shipbuilding school). Ever since then, I've known that technical drawings shouldn't include dimensions that can be calculated from the others. This has been the rule 'ever since'.

Do you use any special tools for this?

Addition, subtraction and sometimes dividing by 2 do not require the use of tools.

[oldrev]

FreeCAD + Inkscape.

First, export your PCB from your EDA software as a 3D model, then import it into FreeCAD.

Do the main drawing in the TechDraw workbench of FreeCAD, and finally export it as an SVG to Inkscape for fine-tuning.

You can check the PDF yourself, it is fully scalable.
https://docs.borneoiot.com/hardwares/buce/#datasheet

[oldrev]

By the way, if you're interested in the styles of this PDF datasheet, here's its open-source template: https://github.com/oldrev/tids

[free_electron]

It's very common for datasheets to include mechanical drawings with missing dimensions, where you usually have to subtract and add to get the desired value.

That is the way mechanical drawings are meant to be.

but that is NOT how the PCB world works.
In the PCB CAD world there are no tolerances. (the real pcb has tolerances. the CAD does not !)
To design a footprint we need absolute coordinates. center to center of objects like pads or traces.
working off a common origin requires manual calculation , which can accumulate errors. (user typo's)

To design such footprints i let the tool do the calculation. In my CAD tool the coordinate fields allow for equations (Altium)  and i can simply move the origin during the design work. i can even do bulk edits like "add x or y to all pads"

[PlainName]

It's very common for datasheets to include mechanical drawings with missing dimensions, where you usually have to subtract and add to get the desired value.

That is the way mechanical drawings are meant to be.

but that is NOT how the PCB world works.
In the PCB CAD world there are no tolerances. (the real pcb has tolerances. the CAD does not !)
To design a footprint we need absolute coordinates. center to center of objects like pads or traces.
working off a common origin requires manual calculation , which can accumulate errors. (user typo's)

To design such footprints i let the tool do the calculation. In my CAD tool the coordinate fields allow for equations (Altium)  and i can simply move the origin during the design work. i can even do bulk edits like "add x or y to all pads"

I think you'll find that is how the PCB world works. Take, for instance, a simple SOIC footprint:

https://en.wikipedia.org/wiki/Small_outline_integrated_circuit?useskin=vector

See those tolerances? When you lay out your PCB you should be taking those into account. Not just taking the middle number and assuming that's it, but working out the worst case and making sure it is still fine. Placing hardware on a PCB is actual mechanical assembly, not software nor a virtual CAD model, and you need to allow for real life.

[Smokey]

I feel like mechanical tolerancing is different from Minimum/Typical/Maximum PCB Footprints. 

[free_electron]

I think you'll find that is how the PCB world works. Take, for instance, a simple SOIC footprint:

hold it. you are confusing part tolerances and pad geometry.
Yes, you need to take into account the pin width and position tolerances to come up with pad shape and position.
But the copper output (gerber / odb++ / 2581) does not have tolerance in the format. What comes out is absolute numbers.

I get this question often from our designers. Should we include tolerances between holes in the mechanical drawing ? No you don't. The coordinates are absolute. There is a registration tolerance of the drill strike on the pad and that is defined per board class (Class i - ii or iii ). the drill size is compensated for the pin tolerance ( this is where we need the datasheet)

The copper geometry does not have tolerances. There are mask registration and legend registration tolerances, but that is as a whole. It is not between objects on a layer, but as a whole layer against a whole layer. That is different from the mechanical world.
in the pcb world the entire layer (whether this is copper , soldermask or legend) can shift in respect to another. individual items cannot shift due to the way the board is constructed.
for drilling the tolerance is more "wander" as the drillbit strikes the copper and has to pierce the glass weave underneath. But now with laser drilling even that is extremely accurate.

[Feynman]

I use my PDF viewer (PDF Exchange Editor) for this. It has a simple integrated measurement tool that you need to calibrate once on one dimension that is actually in the drawing (ideally the biggest dimension to introduce less error).

I use it all the time for footprint creation, because manufactures appear to have a challenge going on on how to put the least useful dimensions possible into a drawing 

[jpanhalt]

Hi,
It's very common for datasheets to include mechanical drawings with missing dimensions, where you usually have to subtract and add to get the desired value.

To add all those dimensions overly constrains the drawing.  They are not needed.  I will often use a 2D CAD program, export in DXF and import into what I am using, which is usually Eagle 7.x.  I would never trust the accuracy of something measured on a printout unless I was trying to copy a design for which no dimensions exist.

[nctnico]

Late to the game but in my experience using the 3D model from a part is a good source because that comes out of the 3D design package the manufacturer uses in case the dimensions of the datasheet aren't clear. Using a Mayo (3D file viewer) or Freecad, I measure the dimensions. Either way, I always check whether the 3D model of a part fits with the outline and footprint of a part.

I actually encountered the situation where the datasheet had the wrong dimensions for a Hammond enclosure. Fortunately I tried to fit the 3D PCB outline with the 3D model of the casing and after measuring the actual casing, it turned out the datasheet was wrong.

[free_electron]

Late to the game but in my experience using the 3D model from a part is a good source because that comes out of the 3D design package the manufacturer uses in case the dimensions of the datasheet aren't clear. Using a Mayo (3D file viewer) or Freecad, I measure the dimensions. Either way, I always check whether the 3D model of a part fits with the outline and footprint of a part.

I actually encountered the situation where the datasheet had the wrong dimensions for a Hammond enclosure. Fortunately I tried to fit the 3D PCB outline with the 3D model of the casing and after measuring the actual casing, it turned out the datasheet was wrong.

the reverse also happens, especially with 3d models made by "3rd parties" like the stuff found on samacsys , snapeda and 3dcontentcentral. If it doesn't come from the manufacturer i'd take it with a bag of salt.
And, like you said : even then. I also have an issue with a hammond model right now. Their model is simplified and does not show a slight slope on the inner wall of the cabinet. The frontpanel is shown as a rectangle while in reality it is a slight trapezoid. It's only 2 millimeter  but the board doesn't go in ...