Open Source Pick And Place Has A $450 BOM Cost | Hackaday

Crowded hands dingy, get rid of the rest of the filler panel with surface mount components. This is the job of the pick and place machine. Over the years, these printed circuit board assembly (PCBA) trading tools have become closer and closer to the reality of the home store. The diving price of some models fell below $10,000. However, if you do it unprofessional, it is still unscrupulous.

On the other hand, this cost can be explained as the project itself. You are not buying a $450 shopping tool, you are buying

. There are two main parts here, an X/Y/Z machine tool can also rotate the vacuum-based part picker, while the feeder can pull out the parts to be placed. All of this is working, but there is still a long way to go before it becomes a machine that is set up and forgotten.

Rubber hits the road surface in two ways by pick and place machines: feeder and optical placement. Where Stephen Hawes did a lot of work,

Started in January. The stack of PCB and 3D printed parts is hung on the front guide rail of the gantry assembly, the width of the tape can be adjusted, and the paper feeding can be fine-tuned using the interesting PCB encoder wheel and worm gear. [Stephen]’s main controller board, RAMPS shield and Arduino Mega running a customized version of Marlin can be used with up to 32 feeders.

So far, although he doesn’t seem to solve the visual system problem, although

It does include a "down camera", confirming that this is a planned feature. Vision is very important in commercial products. At least one downward camera is required to accurately locate the circuit board, and an upward camera is usually used to ensure the position and orientation of the components (if not more than one for each purpose). Cameras). Without these, the machine will not be able to calculate, and may cause deviations in the overall circuit board size, the duration of the placement process, and axial misalignment. However, [Stephen] chooses to use to increase the field of view should not be a fixed job

Drive the machine and the project

. Compared with the complexity of the feeder, it will be much easier to add.

[Stephen] admits that there is still a lot of work to be done, and he is happy to help improve the performance of the feeder design and enrich the functions on the road. Although we suspect that as in the early days of guiding 3D printers, such projects can never be truly completed. At least

Much easier to manufacture.

[Thanks Nils! ]

Smoothieboard is the official controller of the OpenPNP stack. You can make RAMPS work, but this is not the most supported/developed one.

Over the years, we have added/modified many things based on OpenPNP project/user feedback.

For those who own RAMPS, OpenPNP RAMPS support is a more way to use it for certain purposes.

I considered making my own SMT machine, but I bought one from Madell. It cost me $14,000, but the machine includes parts rotation, top view, bottom view and their feeders do a good job. Sometimes it depends on the time and cost of the decision, and I made the right decision. In the past 6 years, I have owned this machine and the time saved exceeds its cost.

It also involves backflow. The oven is a great reflow oven. I used it for several years before upgrading to a $300 desktop temperature-controlled reflow oven. Then I found out that the electric oven is the same! ! !

As far as I can remember, OpenPNP supports all of these.

Ok, if you have 14K

Breeders have always been the pain point of these things. Blush first this looks good. The best choice and location will be with odrive and server. I am waiting for a machine with a price of $4K, which has a good paper feeder and 0402 size parts placement with servo.

You may spend a lot of time on these things. In the end, it’s better to assemble by hand or use manual pick and place or manually perform the first few times in the loop pick and place, and then pay to the assembly factory to rest. I am now an evaluation . I may update later.

I want to make a feeder that only requires you to dump a bunch of parts into a flat container. Use ocv to identify the direction and pick it up. If the part is upside down, it will automatically shake without placing the part on the head (using the automatic sliding cover), or drag and drop until it is correct (it may take less operation time than setting the tray). This can be quickly set up and provide more "feeders" (although based on the gantry area...maybe the conveyor belt can be used to reduce the footprint).

For me, the value of PNP is limited-I rarely do more than 20 boards on any board, and most of the time I get 3 boards from OshPark.

Therefore, what I really want is not PNP, but PND-Pick and Place. This is a machine driven by voice commands. I only need some components to magically display at the designated position on the workbench.

Alexa, give me 6 100nm 0603 cover

Hey google give me a 10k 0603 resistor

Even if it is limited to the chicken part, it can save me a lot of time.

I like this idea and hope it can place the components in the right way. I swear, when you try to flip its right side with tweezers, the 0402 resistor has 12 white sides and one black side.

Yes, for whatever reason, they always tend to turn peanut butter upside down. Statistically speaking it doesn't make sense

I have manually loaded a very large board (each board has more than 500 components, maybe hundreds of boards), I am pretty sure this is a confirmation bias, because although I agree with this approach, I also spend time calculating When I shook the container they entered, how many parts were black side up and how many white sides were up side, only about 50%. I think it looks worse than it actually is, because in most cases, since a bunch of parts on the right have been pulled out, more upside-down parts will be seen.

Yes, but if there are no markings on the part (such as 0402 and 0603), then it doesn't matter which way you solder it?

The resistance of most SMD resistors is on the top of the chip. Mounting it upside down will reduce the cooling air flow, thereby reducing the rated power. During PCB vibration, it may also damage the mechanical strength of parts. That being said, I am happy to solder them upside-down for prototyping in the lab, but not when others want to check my work.

Yes, but we are talking about manual welding 0402, so its output and reliability are not high.

In fact, I don’t feel it, but it may be that the heat dissipation of the resistive layer is closer to the thermal quality of the pad and trace, which is better than air...How to test?

This is because as things get smaller, all these folded sizes will come into play.

I also like this idea, but you can’t provide a feeder for every component. Therefore, you need to buy some kind of mint dispenser or something, and a robot to extract and distribute.

"This is because as things get smaller, all these folded sizes will come into play."

(laugh)

If you need many things of the same kind, just throw something on the table. Pick up and place objects that land correctly until they are all upside down. Then press down with your fingertips to lift it up, and then wipe it off. Place the fallen ones immediately. repeat.

If you only need one or two, I find that using rubber antistatic pads can flip them well. Use tweezers to gently push the edge down to turn it over.

Another option is to put them in a small container with an open top and place a clickable dome of some kind of metal under it so that you can push down and make them all jump.

Put them on a piece of paper, tap the paper with your finger, and then they will jump

They jump on the carpet!

FTFY

B ^)

Either way will do.

Years ago, when I was partially responsible for the Fuji CP6 PnP machine I used to work on, I made some kits for DIY manual SMT projects. I created a program to place parts on a "board", which is an array of 3×5 cards, printed with a screen printing of PCBA pictures, and then covered with PCBA size double-sided tape.

The program places 14 different parts on each card neatly on the PCBA, and places 4 "spare" parts on one edge. After finishing, I got a card with the parts and "instructions" needed by the manufacturer for my hands to place. The smallest part is 0805, so it is very easy. The hardest part is to use all the double-sided tape to make the "wooden board". Finally, I will cover each "board" with a transparent material (used old paper protector) to hold the part in place and cover the adhesive. Tuck them all into a 3×5 ESD bag for transportation. My customers liked it, but my boss closed the site I was working on, so while I was looking for a job, other orders were put on hold.

I have always wanted to replicate this idea with a container like a container for plastic parts. Put each part into the cavity of each BOM, then close the top (obviously seal the compartment with a compartment). Unfortunately, most of my products have not fully supported DIY kit models recently (not everyone cares about QFN)

good idea. For the cavity method, card bottom layer, laser cut corrugated cardboard, plastic top layer may work well

Or make a vacuum-formed plastic sheet into a tray with partial grooves?

If needed, do some Alexa skills for this.

You can pay for some assembly services at any time...

and also:

Mike has a second-hand but professional pick and place machine, but still uses manual assembly.

He said that for small batch production, using manual assembly is faster than installing SMT machines.

He also played a 15-minute video with tips on how to perform manual assembly:

I just want to use more highly integrated parts in SOIC. What is particularly lacking is a device equivalent to an FPGA for power processing. It only needs to make a buck/boost charge, provide one or two buck/boost outputs, and perform I2C control of the "or" logic of the nanogrid device, and a manual solderable device.

But sadly all of these are BGA.

If we had more mass-produced configurable analog gears, so many designs would have required one-third of the parts. Where is the ESD protection ADC/amplifier/comparator chip with 60v function?

I am engaged in analog power supply design. Large companies are willing to pay prices beyond the minimum, and it is extremely difficult to provide a convincing business case for a broad market segment. According to the pricing structure, we must sell 500,000 parts to reach the breakeven of the development, so if you cannot provide a compelling reason for this, it will not be approved. (When I say the minimum, I mean that we have put the parts into production, and possible customers say "I don't want to pay for all these extra things", and then we pass by actually not doing anything. "Rotate" the mold put a different mark on it and made a data sheet, which did not mention the features they didn't want, and reduced some prices, and then they bought it. If you buy enough, it is worth it for us Price reduction.)

We have an i2c chip that contains multiple buck/boost controllers. Marketing is extremely difficult, because when you start talking about how to set up registers to control the phase relationship between different channels, the people who need the hardware the most will be completely stuck. On previous similar chips, we spent about 70% of the engineering time to support the software required to demonstrate the unit and provide support to customer designers who tried to use it, while we spent 30% of our time developing it. It is expensive and expensive and even harder to recover because it is not a one-time expenditure. Just like every new product purchase will increase the ongoing cost.

It is also difficult to do a lot of digital work in the process of higher voltage (not impossible at all, but very inconvenient). Due to process requirements, it takes up a lot of mold space.

Oh, I don’t know anything about chip design!

My workload is small and I am usually the only engineer, and mostly work at the consumer level, so my first task is to replace every component I can use with software, no matter how much complexity it adds. More hardware means that once it is abused, more things will fail; and due to the timetable, one attempt and completion of the circuit board manufacturing will make more things mess up. Plus the hardware has been repaired, and someone wants to complete half of the design before the design is complete and know what they want.

I don’t know that hardware engineers hate handling registers! I think it makes sense, I even met software experts who really don't like software. Many engineers seem to prefer the things they feel they know the lowest level.

Of course, it seems that this problem can be solved through sufficient economies of scale. If we have a fully integrated "default choice" that everyone agrees and can be made through billions of dollars, then in a period of time, 555 timer and 7805 will appear everywhere, I believe someone will figure out the cost issue .

Not that there are people who seem to want to standardize everything now! Everyone seems to hate standards and prefer their temporary "perfect for this application" solution, at least in terms of software.

How to use the reverse image of the placement,

Place the component upside down on the sticky "plastic" board above the image on the back,

Place the thin board with all components in the correct direction on the circuit board with pre-flux paste,

Send it to backflow,

Badda-boom, Badda-bing!

PCB with components and conformal coating​​!

(I'll go to the door to see now)

In terms of the number of plates produced, when can PNP machines usually save labor? I need to do 20 to 30 designs occasionally, but I often do 1-5. For amateur projects, professional assembly is usually too expensive, up to 50-100 boards. JLCPCB’s assembly service is cool, but given its limitations (for example, they cannot be required to order more out-of-stock parts), the design based on it seems risky.

This actually depends on the number of components, types and the number of different parts. It may take 1-4 hours to set up my older Juki (for example, feeder, programming, etc.), so for a few boards, this does not make any time sense. However, if you use frequently used components (capacitors, resistors, etc.) for PnP settings, you can greatly reduce the setup time. In many cases, I use PnP to fill standard parts that have been set on the computer, and then manually place the remaining parts. In this way, compared to manually filling the entire circuit board, I will save about 60-70% of the time. If I have to assemble more than 20-25 boards, I usually send them to my contract manufacturer (CM) because they can finish it faster with new equipment. However, the delivery time of my CM is about 3-4 weeks, so if the time is tight, then I will run PnP and make larger runs internally.

I commend those who choose to design their own PnP machines, but I have to ask, is it better to buy an old professional PnP and then refit it with OpenPnP? This can provide a good starting point for many mechanical details and the supply of various feeders. Feeders are essential to the use of PnP. If you look at commercial feeders, you will find the complexity of these devices. Therefore, using off-the-shelf products will save a lot of time and you can buy as many things as you need. ...In some cases, the cost is much lower than self-made.

This is definitely what people do. In the past year or so, I have helped more than a dozen people (in the "smooth board" aspect of those projects). Interestingly, compared with the accompanying DOS software, these machines are usually more capable of converting to OpenPNP at one time.

Indeed, I have seen some ways of doing this. However, many people decide to build everything from scratch... It's a lot of work with mixed results. The complete custom plug and play will have a certain level of feeder restrictions (quantity, type, etc.) without overly mentioning the placement speed. Using a professional machine as a basis can indeed improve the results... and will undoubtedly be faster to create.

Yes, but it is not open source. This benefits him more than himself, and we know that when people plant the shade that they know they will never sit in, society will grow.

Actually, it will be open source... As my comment is about using OpenPnP and the controller board, but using the original equipment infrastructure... In other words, it will be hacked for bidding. Like I said, making your plug-and-play function great, but in the end its usefulness is not all the same, because there are many things to consider. I never considered all these details before getting my own PnP, so I strongly recommend starting with an existing old device and starting with all its infrastructure in place. Doing so will make the device more versatile and useful. Before anyone starts making their own PnP, I suggest you let someone else guide you through all the steps to set up and operate PnP... Besides simply picking parts and placing them, there are many other functions.

If your goal is to have affordable and functional machines in a short time, then buying second-hand products may be a better choice than doing it yourself.

If your budget is very tight, you can use it as a hobby and assemble simple parts into a machine. This is not only a fun challenge, but also a learning experience to learn how it works, then buy some parts and build it yourself .

In addition, the Chinese will manufacture desktop computers of the size shown in the video above, and you can buy a new one at a price of around 2000 Euros or 3000 Euro

The English menu with built-in LCD is optional. If you want to run it on openPNP, maybe you can reach an agreement and buy only the technology.

I started to make rotatable vacuum nozzles. I have more concepts, but they are not listed on my channel.

Just want to know if those awkward and cheap "USB keyboard cleaners" are strong enough to pull small components, or are they complete?

When the manual pick and place system was up and running, I spent a while looking for vacuum pump options.

I finally modified an aquarium air pump.

Someone documented the process here:

The availability of parts depends on:

*The weight of your parts (almost none)

*The "effective" diameter of the nozzle.

*Vacuum pressure difference.

With a perfect vacuum, you can suck about 10 meters of water through the tube.

Then go back and assume:

*The specific gravity of ceramic components is 3 times that of water.

*Low pressure 10kPa

*The size ratio of nozzle to component is 1:10

Then, you can still lift the 3 cm high component stack.

So almost any vacuum pump will do.

However, if the pressure difference is large, then the size of the nozzle is less important, for example, you can use the same nozzle as the 1206 resistor to pick up a 12*12mm QFP. Such a high ratio will reduce reliability and accuracy. Components may move during transportation due to vibration.

Thank you everyone. Although I am not looking for the best vacuum pump, it is cheaper, lighter, low-power, and a small semi-automatic robot that can use it when you are hovering on the bench to grab things for you. Or automatically organize random piles of components into types or something.

Just consider that it is a well-thought-out design, so that your concept can be quickly discovered by testing and good at development, and then let you discover its advantages, and then nail your feet to the floor with an oscilloscope. Coupled with the trivial things that work with more than a dozen small problems, it is far more fun than what you can do in a few months, and there is only a single problem that can be solved by the trivial, weird band-aid.

But I am only doing brain simulations.

>So far, it seems that he has not solved the vision system problem. [...] Without these, the machine will not be able to calculate, which may cause the entire circuit board size, placement duration and axial offset misalignment.

Wait what

There are a large number of systems that can repeatedly position 0.01 mm on a larger area than amateur PCBs or even professional PCBs. They usually use completely open-loop "dead reckoning" steppers, and basically never use computer vision. CNC machine tools. Even 3D printers. Machine vision seems to be a large-scale Rube Goldberg method. Cut your plank accurately and clamp it to a corner.

What am I missing here? I'm sure there must be something, but I have no *clue*.

You are missing random variations of parts in the tape. For large parts, this is not a problem, but the smaller you are, the more difficult it is to pick up a resistor close enough to the center to work.

Not long ago, I fixed a bug that caused the Arduino mega support to not work properly. (It seems that many have these, and many IOs are very cheap). Do you know if it still applies to Arduino mega?

(This reply is in response to the first comment, namely openPNP Ramps vs Smoothie board.)

Please be kind and respectful to help make the comment section great. (

)

The site uses Akismet to reduce spam.

.

By using our website and services, you expressly agree to our placement of performance, functionality and advertising cookies.