• Under 0.5mm pitch (20mil)

    100dollarhacker • 04/29/2024 at 14:34 • 0 comments

    We already saw how to create 1mm pitch (distance between IC's terminals) once we want to get down to finer details it gets much harder. If the bit too high it will just scratch the surface. If bit to low it will make deep trace. Breaking fine details or even break the bit itself. 

    I already have CNC so something should be done with it. It does really good jobs until it comes to the final details. 

    Then I got an idea what if we lift the bit a bit. And use chemical etching instead of milling. The result was really good relatively to just milling. But it still was not enough. 

    So I've just turned off CNC's motor and let the bit scratch painted surface.

    The result was much better than I have ever seen. 

    Once I started to make full board the scratch method was not that good some places left not scratched as result wires has not been created.

    So, I went to the previous method but now  instead of removing all the coper, I made just a gentle scratch of the surface. Then put it to Ferric Chloride for 10 minutes and Walla!! PCB for 0.5mm pitch ICs.


    The method to create those board is simple:

    * Make a height map of the copper board with distance of 5mm

    * Paint the board with black non erasable marker

    * Find a bit that makes tiniest scratch (of 0.1mm)

    * Bring bit up and then start lowering the bit until surface scratched. This is done by lowering the bit by steps of 0.01mm and check if it started scratching.

    * Reset current height as zero.

    * Check 'use height' button (forgot it more than once) 

    * Press send button 

    After finish with CNC machine 

    * Lower the board to Ferric Chloride deluted solution of 1:4 FC to water.

    *Leave it for 10 minutes 

    * Wash with water and clean with acetone to remove black coating 

  • Traces under the microscope

    100dollarhacker • 04/19/2024 at 20:07 • 0 comments

    After I broke some PCBs I started to get the idea what should be done to increase probability for successful PCB print

    So one thing we can notice is the trace width. If wide trace is used fine details cannot be printed.  

    Let's see three different traces: over etched, under etched and one just the right width.

    Here is example of 0.5mm chip layout, it's totally erased. Let's take a look at it under the microscope 

    The width is 0.35mm.

    Now let's look on one which is has very fine details 

    Now let's take a look on under etched one

    It's width is 0.2mm.

    Here is another one with just the right width 

    This time it about 0.25mm 

    So what is the difference? The only thing was changed are the bits and the copper plate. 

    It's possible that plates have different thickness of the copper. 

    But what about the bits are they same? 

    Let's take a look on the bits under the microscope. 

    Compare the tip statically and once rotated 

    The good: (upper part is rotating and lower is static)

    The bad: (upper part is rotating and lower is static)

    The main problem with milling is setup. Even if you run height map of 5mm steps, you will have still the problem that each bit has it's own depth configuration. Since sharpness and movement of the tip is greatly changes between different tips of the same company of the same batch. 

  • CAD design, CNC print

    100dollarhacker • 04/19/2024 at 15:11 • 0 comments

    So next step was to make CAD design and CNC print. The result was amazing compared to previous methods. (Standard 3018 with CNC shield nothing fancy)

    This how I reached about one mm resolution. It gave me ability to easy connect complex components. Boards could be manufactured in matter of minutes with many ICs on them.

    For example here I've connected MCU and gyro board.

    Soon I hit a limitation, SOIC - Small Outline Integrated Circuit. Some projects used this tiny enclosure so tiny the width between terminals is only 0.5mm.

    Smaller detail usually could not survive. After a lot of try and error I found that it's possible 

    But the problem remained why some print were successful and others failed.

    To understand that let's look on the problem from microscopic point of view.