Cheap recognition & pick & place & reflow oven solder machine for tiny PCBs

[eneuro]

I had a fun to play with ABC: A System for Sequential Synthesis and Verification available for download here and while it was designed based on experience on SIS/MVSIS (has smilar commands) after changing genlib cell library to support only NAND2 and INV1 gates (I've removed XOR while trying implement this logic only on NANDs) I've got nicely visualised network of NANDs, eg. I could use those PDF: 74HC00D Quad 2-input NAND gates 



I do not know why this ABC synthesis software crashes when tried map using only NAND2 gates instead of INV1 (inverter with one input output), but it doesn't matter, since those INV1 gates can be easy replaced by one NAND2 with its inputs tied, so it looks like five small 74HC00D SO14 (Quad 2-input NAND) should do the trick and "calculate" movement direction in pure digital CMOS NAND implementation of this gray2move function 

I need add D flip-flop or shift register  to store previous 2-bits of light sensor readings, but it could be fun if it worked  with pure NANDs implementation-I didin't check if this graph of NANDs gives correct output direction bit values, but probably ABC software did it right 
Of course using XOR gates could simplify this function as shown above, but I tried generate pure NAND version based on 74HC00D's 

[eneuro]

After some brain storm on this thread Low Cost Pick and Place Machine , I've finally found better solution for rotating XY plane around Z axis and steel ring for base like this below will be used:



Side view below: about 15mm high and 22mm wide, with 0.329 meter radius and weight... 5.5 kg, not for cheap China post mail, but... quality must weight-not too much plastic crap  in this cuctom prototype PnP machine with additional feader machine



It has ~68 cm diameter, but when we assume that only 0.5 m (( 50 cm) will be usable, than we get working piece area comparable to.... 80 PCBs at size 5cm x 5cm, 320 PCBs at size 25mm x 25mm each while total working area is >0.2 m^2 

Probably at this point you might have no idea how it will work, no problem  I will post when prototype rotating horizontal axis in XY plane and vertical in Z axis will be installed with bearings, while I'd like to use it as soon as possible for manual SMD PnP.

I design this with those linear motion motors in mind

Yaskawa Linear Motors in Motion


and probably not only one but two PnP heads on one transission line to speed up two times PnP assembly, thanks to linear motors advances in linear transmission lines, but due to the investigated possible patent pending two head version will not be shown on YT for the moment-only one head version 

[eneuro]

Today, I want make teardown of... optical mouse, since as we can see it has quite advanced signal processing builtin and maybe this thing could be usefull for precision positioning, while gray code encoder used for fast movements to find position.

Understanding Optical Mice

I need make custom encoder for my PnP base shown above and considering different options, while for the moment I'd like to make working prototype not low cost but... cheap, really cheap 
One of the simpliest one is print gray code of encoder on paper and make thermal transfer or even glue and protect from water, etc.
Probably I will try something like this, but I will also create pseudorandom noise with limited lets say 64 levels of gray on paper with black gray encoder pixels to help optical mouse sensor recognize precision movements betweem 1mm spaced gray encoder black coloured data



Those sensors even a few years ago has quite high 500 fps in their builtin cameras, but maybe someone hacked those cameras to get not CPI as output but its pixels itself? 
How many pixels it is? 
I never tried hack this and it now looks interesting, however of course microscope camera with CV could also be an option-I mean I have gray code encoder, so microscope camera reads this code between each 1mm spaced places, but than makes measurememts in its own pixels of this 1mm space between gray  code data and determines fraction of milimeters, so with help of CV we could make fast linear movements and still be able to make precision positioning based on calculated offset from the centre of CV camera and processed milimeters boundary black lines.

So, I will play with this mouse for fun only, since it looks like it can't help too much detect gray encoder boundaries without direct access to its pixels, so probably I will have to look for more professional optical encoder solutions  for this ring XY rotation around Z axis, like those used in ink jet printers, etc 

However, now realized, that if I had CV available in assembly head, than PCB could be marked and CV system could make those tiny position adjustments based on image processed position of those markers.
It will be a challenge to get decent precision positions while rotating my PnP assembly head along this ring around Z axis, but for the moment I've smallest SMD parts 0805 and SO8 packages on test PCBs to automate its assembly, so not huge precision needed.

Yep, folks hacked those tiny optical mouse cameras 
http://www.bidouille.org/hack/mousecam

Optical mouse cam
http://spritesmods.com/?art=mouseeye

Your Mouse Is A Terrible Webcam
http://hackaday.com/2014/01/14/your-mouse-is-a-terrible-webcam/
Who cares, if we were able make some processing of this data without sending to PC image, but prcessed output -position 

I looked up the datasheet of the optical sensor (which is an ADNS2610) and it told me the sensor has a tiny 18x18 CCD, which can be read out using the serial port (the one which normally interfaces to the PS2/USB-chip).

Decent  quality image from optical mouse 



Yeah, teardown time   

[eneuro]

This is a monster in many of optical mouses which we are looking for 

ADNS-2610  Optical Mouse Sensor

Nice  400cpi @ 12*0.0254m= 0.3048 [m/s]  (30 cm/s motion) 

Resolution is 400 counts per inch (cpi) with rates of motion
up to 12 inches per second (ips).

Sample typical  circuit block diagram, so now we have idea what we have to salvage from our optical mouse to keep it working in prototype

[eneuro]

I've made quick teardown of old A4 Tech OPTO 815 mouse



N1165 marked SoC (shown in datasheets as C1165-011) in the middle:



Unfortunatelly there is no nice ADNS-2610 with acess to raw sensor image data, but bigger C1165-011 USB Optical Mouse SoC , which looks like doesn't provide on its pins access to raw sensor image under mouse, but outputs directly to USB its position and buttons state, so it is useless to compute absolute position.



Anyway, this mouse had wire at the front, so I've drilled hole at the back and fixed problems with connection, so not this optical mouse better fits my needs, while still hate those bloody RF wireless mouses, when its battery deplates and is useless just in the middle of the job 

This ADNS-2610 is difficult to source , so we'll try visual camera module instead.

[eneuro]

After thinking a little bit, I realized that maybe in this quick cheap version I could try use this SPCP168A Description USB Optical Mouse SoC



which I've found during teardown of another failed (more modern ) cheap optical mouse, which works (left mouse button failed) and which is nice probably IR LEDs diode is used, since when tried connect to PC at Linux XWindows, than cursor still of course works perfectly and no redish light comes from the bottom of this thing 



I was  that there is only a few elements on this PCB and it looks like easy to reverse engineering circuit, redraw new PCB when needed and desolder, while it is strange, but difficult to find this IC SPCP168A  in any databases... how is it possible that chip from 2010 datasheet is not available for sourcing?

It has very simple example circuit, so it could be usefull in a few more my projects too, but  where to source this IC? Maybe someone knows what is going on with this IC? 



It looks like PCB layout is similar and very simple too with a few caps and resistor:



Not too much there,  so  I will try desolder Z axis (scroll buttton) encoder and.. replace with my own which will read position on this half meter ring shown above, so it might be a little bit tricky, but with the help of desoldered oryginal element and this SoC optical mouse datasheet seams doable 



It has nice lens holder with light (IR?) guides, so I will cut everything around this PCB and salvage with bottom plastic cover.



Additionally, failed left mouse button, can be replaced with another working middle button, so there will be more place to put there my custom encoder with photodiodes to detect printed black/white lines (1mm spacing for first tests) pattern below and synchronize with read XY position 

If it works, maybe even gray encoder will not be needed, since probably this Z axis tracking hacked should count quite reliable black lines spaced by white the same width, but it need to be tested and compared with this Z axis limits and sampling rates, but 1m/s rotation not needed around Z axis on ring-assembly head holder will have its own stepper motor for fast rotation of taken PnP part 

Anyway this SoC can be used to send back to PC controll software position on ring using this integrated USB driver, so that is fine and I will try write some Linux code to be able adjust its default 1000 CPI XY resolution, which is not so bad.
Datasheet says it has by default 1000 CPI, but I have to ensure and test if Linux driver is capable maybe to switch to higher resolution available, but 1000 CPi looks fine-we'll have CV microscope cameras in assembly head, so this encoder will be used to make small adjustments to place  parts in exact position calculated from computer vision system analysis of exposed copper/silver/gold pads on PCB, so maybe this encoder build from failed optical mouse might work for us 

[eneuro]

I've made some art work and now my PnP (CNC) machine encoder in Z axis rotation along steel ring looks better 



I've reinforced with hot glue USB connector wires which were not protected in any way and could easy be cracked-it were floating inside this optical mouse 

Fixed problem with left mouse button, by replacing with good wheel button and removed this Z axis wheel encoder, which send on two pins some pulses to SoC Z axis pins as ex[ected from example general circuit shown above..



Now, we need examine this Z axis part with wheel installed inside and try rotate and see what kind of pulses might be created in each direction, to be able design similar detecting printed black/white pattern.
I could do it while powered in real USB socket before disassembling this, but it must be simple On/OFF, I guess  and shouldn't be too many supprices-it is simple circuit powered from USB +5V

It is time to design machine horizontal axis with this encoder installed, since its vertical Z movements will be controlled for the moment by... DVD ROM laser diode holder with its bipolar stepper motor to get quicly some result and help in manual SMD PnP to see something doing borin tasks like moving parts from one place to another with vaccum pump 

[eneuro]

Trying upgrade this PCB, by adding additional status diodes, since it looks like this schematics fits very close to layout



Tested in XWindows (Linux) that indeed this SoC has ability to add DPI switch, which switches between 1000CPI and 1600DPI-it works by pressing this black puch button, but it is possible also to add LED indicator for this, so I will try do it since, instead of LED i can use optocoupler internal dioded if needed and try to switch to 1600DPI using small MPU, which will "press" this DPi switch button and check on status LED if it is changed.
However, according to schematics (from SoC datasheet) linked above, we need to ad 2.2k resistor and when checked tracks, I will have to cut part of one track to insert this additional resistor, since in current design they didn't prepared PCB for this CPI LED status feature (or  I missed something), but additionally something like power diode can be easy added, so I will add green for power status and yellow to test this DPI switch modes.

BTW: When we calculate what 1600 DPI means it is 25.4mm/1600 ~ 0.016 mm 
Not so bad, for precise adjustments 

Fighting with this Z-axis encoder, since when tried probe desoldered, unfortunatelly I can't see any shorting to ground on two of its pins, so I do not know, not sure what kind of encoder signal is send-2 bit gray code or something else-it is not described in this SoC preliminary datasheet.

That is interesting, that there are some pattents around this Z-axis encoding 

Mouse-associated Z-axis encoder
http://www.freepatentsonline.com/6285355.pdf

We'are  too lazy today to investigate this after 60km bike trip between hills, so I someone have some idea what we could expect from those two pins when wheel mouse (Z-axis) is uses let us know, to complete quite succesful teardown and rapair of failed optical mouse button, with additional enhancements of this tiny PCB by adding resistors and LEDs to unpopulated pads on this PCB 

Update: Probably it can be 2-bit gray code in this optical mouse Z-axis.
Interesting usage of three mouse buttons and wheel below 

CNC control with a mouse (pseudo pendant)

The wheel(encoder) will either move positive or negative 1mm for each change on the encoder
Left click changes to X axis---Right click changes to Y axis---Wheel click changes to Z axis.

There is no wheel no more, but MPU can be used and additional step up/down buttons connceted to its pins, than MPU will output correct 2-bit gray code,
so we can choose axis and eg. by holding up/down buttons make 1mm auto increments every 1 second, etc 
Yep, we'll try 2-bit gray code as inputs to those  those ZA/ZB SoC pins, without loosing time on oryginal optical mouse wheel Z-axis encoder 

[eneuro]

PCB was upgraded to support CPI LED (yellow) and PWR led (green):



After power up CPI led (1000) is off and mouse movements slower than when DPI is pressed and optical mouse switched to 1600 CPI-software reads more counts per the same distance (inch), so screen cursor movement looks faster, since probably its driver do not know that DPI button was pressed.

So, when we'll need this higher 1600 CPI resolution this experiment showed that we need add tricky circuit, which will read this CPI LED status and "press" automatically DPI button, until CPI LED is ON 

Printing tomorow custom encoder background on A0 paper to see if this optical mouse Z axis SoC pins Za/Zb feeded with this data will track machine rotation around Z axis 

Update: There is ~2.3V on those ZA/ZB between ground, and when tried short with 1k resistors cursor seams to scroll text in terminal 

I wrote small MPU software to generate 2-bit gray code on ATTiny85, so I think I will add support for incremental steps on two other, so pressing them should move this whel cursor up/down, so it could be quite interesting interface to select movement axis, then make 1mm steps in both directions. Using additional pin we can toggle steps multiply by lets say 10x so in this mode 10mm (1cm) steps like in many CNC manual control devices 

[eneuro]

I wrote small MPU software to generate 2-bit gray code on ATTiny85...

Yep, idea of using MPu between any rotational encoder and enhanced SPCP168A optical mouse PCB, was good, since when connected my development ATTiny85 MPU powered from mouse PCB USB (4.3Vz to provide a few mA of currents-now realized that I've BOD at this range enabled on this MPU.. Ups  ) it looks like eg. in PDF viewer software scrolling works fine in both directions, but seams that scrolling is update when ~ two gray code steps are made-not step by step 



I've used 2-bit gray code like this below and had a few cycles set (worked fine with 16 steps -4 cycles).
 

Code: [Select]

./mpu_gray_test: Number of steps: 4
./mpu_gray_test: step: 000  binary: 00  gray: 00
./mpu_gray_test: step: 001  binary: 01  gray: 01
./mpu_gray_test: step: 002  binary: 10  gray: 11
./mpu_gray_test: step: 003  binary: 11  gray: 10

./mpu_gray_test: step: 000  binary: 00  gray: 00
./mpu_gray_test: step: 001  binary: 01  gray: 01
./mpu_gray_test: step: 002  binary: 10  gray: 11
./mpu_gray_test: step: 003  binary: 11  gray: 10

./mpu_gray_test: step: 000  binary: 00  gray: 00
./mpu_gray_test: step: 001  binary: 01  gray: 01
./mpu_gray_test: step: 002  binary: 10  gray: 11
./mpu_gray_test: step: 003  binary: 11  gray: 10
./mpu_gray_test: step: 000  binary: 00  gray: 00

Direction is randomly choosen to simulate machine rotation changes during its operation, but it was strange, when there were no full 4 step cycles (multiply), than maybe PDF viewer buffers or makes  averaging while scrolling, since I was not able make scroll each step by step 
It is time  software which will read those mouse movements and output raw data from mouse not postprocessed, but AVR MPU itself powered with a few mA looks like can do this job, but I will add 3.3V voltage regulator, while I wouldn't like to mess with existign one on this PCB, while I might want do other more power demanding tasks, so I will take only USB 5V power to have the same ground on MPU with this oryginal SPCP168A PCB, so MPU closes Z axis wheel two pins to ground via additional resistor and now I will and my custom encoder light sensors and make PCB to read real machine encoder data, than try write a few lines of code to create in OpenGL and Blender virtual 3D model, so we should be able see on HD screen machine model 3D view in parallel with real working machine to test how amchine axis encoder data is translated by software 

So, close have something usable.
Probably I will use LinuxCNC realtime OS, since I'd like to have G-code manual mode support, etc...

[eneuro]

It is time  software which will read those mouse movements and output raw data from mouse not postprocessed,...

Using under Linux command below we were able find input device file associated with another optical mouse with wheel:

Code: [Select]

$ cat /proc/bus/usb/devices

Code: [Select]

I: Bus=0003 Vendor=093a Product=2500 Version=0110
N: Name="PIXART USB OPTICAL MOUSE"
P: Phys=usb-0000:00:1d.0-2/input0
S: Sysfs=/devices/pci0000:00/0000:00:1d.0/usb2/2-2/2-2:1.0/input/input8
U: Uniq=
H: Handlers=mouse1 event8
B: PROP=0
B: EV=17
B: KEY=1f0000 0 0 0 0
B: REL=103
B: MSC=10
This allowed us grant read permisions to input device /dev/input/event8 which with a few lines of C code detects mouse movement events and outputs time when it happens and relative position x/y changes and wheel Z axis changes, which as expected for small slow scrolls are in +/-1 steps, so some software (like PDF viewer) might buffer those changes and calculate accumulated position, but we have nice wheel Z axis increments/decrements in the case of oryginal Tracer optical mouse, so now we can test quality of MPU gray code generated using Z axis wheel simulator shown above 



Now, we can design mechanical axis, cut prototype in steel, add bearings, main rotation encoder and fun part begins, when real machine movements will be translated to Linux XWindows custom control software, thanks to aaccess to USB optical mouse data 

[eneuro]

There is  no mine PnP on YT so far, but we are looking what kind of DIY PnP folks does.
This one has decent speed, but it looks like... all those CNC machines

Youtube - DIY Pick and Place Machine Project


We'll see how my concept will work in practise, but its axis are dead simple and similar to this fast PnP 3 axis
Ultra Fast Pick & Place Robot - FANUC's New Three-Axis Delta Robot Packs Small Batteries


I was thinking about using three such arms instead of four, but it looked like four arms are needed to position picked part in space, however as we can see it works fine with three arms too, so definitelly I need creat 3D model in my CAE and try simulate something like this, since... yep it is really ultrafast and could work with my feeder machine concept

BTW: Concepts changes from day to day, but well I want make it really cheap and outperform at least myself in SMD pick & place... which will be rather easy since... handling those bloody tiny SMD parts and manual positioning is horrible 
Still rotation around Z axis and simplier two axis instead of shown three axis might work as well for me, since I do not need such ultra fast PnP like shown on video above, but this machine uses the same three arms, which means design one, let manufacture three  and have PnP ready 

As simple as it is and... fast 

[eneuro]

We decided make something similar to this and mechanical parts seams to be easy to manufacture, but control electronics and motors might be a little bit tricky, but we like challenges



Our machine will be ~60cm height and ~70cm width when arms positioned like in image above.
Ordered steel balls in local bearings supplier expected delivery time in one-two days, so can't wait make prototype axis using custom made spot welder for such "fireworks" 
I'm working on mathematical modeling of those three axis movements with arms ~50cm (half meter) long.
This numeric simulation will give us idea of needed steppers resolution to achieve decent assembly head accuracy in the range of 1mm-0.1mm in first prototype with CV help to make more precise adjustments based on at least 10fps image processing to fine tune assembly head positioning, which seams doable while I was able use OpenCV with 30fps live augmented video camera objects detection, etc, not mentioning huge speedups when GPU Nvidia support accelerated CV will be implemented for critical parts of PCB pads/pins/vias  image processing .

Since we estimated our machine arms size, CAE simulation will also allow estimate expected workpiece area size-it can be also done using basic geometry analysis   

[eneuro]

Steel balls are in my hands right now, so it is time try change them into PnP machine assembly head holder, like in concept photo above 


We need 12 such balls:



Connected in pairs, those steel balls  will create 0.500m long arms holders 


While, we know its dimensions, prototype arms and assembly head holder can be designed and welded.

Steel, copper, nice in touch wood-no crappy cheapy plastics