Why is COB cheaper than soldering?

[analogo]

In cheap electronics like calculators various chips are attached to the board using COB techniques instead of soldering. It is often said that this is done because COB is "cheap". And how much cheaper it is, in practice?

Soldering seems like a very cheap thing to do, and many other parts on these boards are soldered anyway (capacitors, resistors, mechanical parts), so you cannot really avoid a soldering pass. Why is it then COB used?

[Rerouter]

The encapsulation for an IC is not free. Nor is the wire bonding. So they do it through the assembler at a cheaper cost

[firewalker]

It is a bare die under the epoxy. I guess the gain is from "packet free" ics.

Alexander.

[EEVblog]

The chip itself is cheaper because it's bare die. But you have to factor in that cost saving with the cost of COB placement at your assembly house.
I don't have actual figures, but it must be cheaper otherwise they wouldn't do extra steps at the assembly house that could otherwise be done essentially "for free" with the existing PnP machine.

[analogo]

The packaging must be quite expensive then, because COB looks like a labor-intensive technique compared to soldering with a pick and place machine, at least according to these videos from SparkFun: https://learn.sparkfun.com/tutorials/how-chip-on-boards-are-made

[The_Explorer]

This wire bonding is done in any case. That's how the die is internally connected to the leads of the package. So this step isn't additionally.
It is just moved from the packaging house directly to pcb manufacturing.

You probably safe the ceramic (or plastic) package of the chip itself and the intermediate testing of the packaged chips.
This could also mean you have one player less in the game who wants to earn some money (if packaging is not supplied by the semi fab).

 

[Ian.M]

See last year's thread on the subject:
https://www.eevblog.com/forum/chat/chip-on-board/

[filssavi]

COB us used to shave every cent from the final price, I know for a fact that for small IC’s on older nodes (say 180nm up) the packaging can cost more than the whole Die production, in that case you use COB and save quite some money, end to end the process is cheaper since you are merging 2 different steps

That reasoning is only valid for mass production with extremely high volumes since only at that scale the high upfront cost for tooling engineering etc is a non issue (it is so diluted due to the huge volumes to be a non factor)

[krho]

We tried to do COB as we've gotten the boards schematics,gerbers... we've gotten the offers from several manufacturers here in the Europe.. At the end it was cheaper to redesign the board to use TQFP instead of COB. That's for approximately 1500-2000 boards per year

[Alex Eisenhut]

We tried to do COB as we've gotten the boards schematics,gerbers... we've gotten the offers from several manufacturers here in the Europe.. At the end it was cheaper to redesign the board to use TQFP instead of COB. That's for approximately 1500-2000 boards per year

You're not the target market.

[langwadt]

We tried to do COB as we've gotten the boards schematics,gerbers... we've gotten the offers from several manufacturers here in the Europe.. At the end it was cheaper to redesign the board to use TQFP instead of COB. That's for approximately 1500-2000 boards per year

you probably need a volume with a few more zeros at the end for it to make sense

[coppice]

For many small parts the packaging cost exceeds the die cost. It varies a lot, though. Some fairly similar packages can have very different costs, so package choice is a key step during the chip design process. There can be considerable savings by not packaging the die. If you are approaching die bonding for the first time the startup issues might look daunting, but once you are set up and running with the appropriate equipment and staff skills the cost per chip is very low.

COB is only appropriate for high volumes, or special uses. If you try to buy die in small quantities the costs are dominated the costs of customer service, rather than the cost of the die. It usually only makes economic sense when you are dealing in numbers that mean you are directly served by the silicon vendor (i.e. not through distribution).

[krho]

We tried to do COB as we've gotten the boards schematics,gerbers... we've gotten the offers from several manufacturers here in the Europe.. At the end it was cheaper to redesign the board to use TQFP instead of COB. That's for approximately 1500-2000 boards per year

You're not the target market.

The firm that we bought had approximately the same volume per year as us, so we said let's get some offers and we were badly surprised.

[james_s]

3,000 boards a year is small volume, the sort of products where you're likely to see COB is going to be several hundred thousand to millions per year. I see it used on very low cost stuff that is made in huge numbers.

[KL27x]

This wire bonding is done in any case. That's how the die is internally connected to the leads of the package. So this step isn't additionally.
It is just moved from the packaging house directly to pcb manufacturing.

I would think it is much more expensive to do it in the pcb manufacturing process. Doing it on a standard size, tiny, smooth IC package would be much more convenient. Less space, less setup, less hassle/hangups/handling. You can easily feed IC's through tubes or reels or whatnot.

The cost of the IC package must be pretty high to make up the difference and come out ahead. I bet it only pans out for very high pin counts.

[coppice]

This wire bonding is done in any case. That's how the die is internally connected to the leads of the package. So this step isn't additionally.
It is just moved from the packaging house directly to pcb manufacturing.

I would think it is much more expensive to do it in the pcb manufacturing process. Doing it on a standard size, tiny, smooth IC package would be much more convenient. Less space, less setup, less hassle/hangups/handling. You can easily feed IC's through tubes or reels or whatnot.

The cost of the IC package must be pretty high to make up the difference and come out ahead. I bet it only pans out for very high pin counts.

As long as the board is not too big, the same die bonding robots do the work, whether its on a board or in a lead frame. Why would it be more expensive to do it on a board? For many devices the package cost is comparable to the die cost. It's not insignificant.

[james_s]

Consider if you are making a run of say 2 million units. Now work out the cost of the IC lead frame, epoxy potting, then the time/effort of inserting the IC into holes in a PCB, the cost of drilling those holes, the cost of the solder used to hold the IC to the board, the shipping cost of the added weight of the product. On a small scale that is all insignificant, but now multiply each of those numbers by 2 million and see how it looks.

I don't know offhand how much solder it takes to install a typical IC but over the production run I would not be surprised if it is several tons.

[noras]

What is COB? Its full name is chip-on-board, which is a chip-on-board package. It is a new type of package that is different from SMD surface-mount package technology. Specifically, the LED bare chip is adhered to the PCB with conductive or non-conductive glue, and then Wire bonding enables electrical connection and encapsulation of the chip and bond wires with glue.

This kind of packaging is not a package, but it integrates upstream and downstream enterprises. The production from LED chip package to LED display unit module or display screen is completed in one factory, which integrates and simplifies the production of packaging companies and display manufacturers. Process, production process is easier to organize and control, product point spacing can be smaller, reliability is multiplied, and cost is closer to civilians.

[noras]

The COB packaging technology is categorized as a package-free or provincial-saving package. However, this package does not eliminate the packaging process, but saves the packaging process. Compared with the SMT process, the COB packaging process is saved. Going to a few steps saves time and process to a certain extent, and also saves costs to a certain extent. SMD's production process needs to undergo solid crystal, wire bonding, dispensing, baking, stamping, splitting, color separation, tape, and patch. The COB process is simplified on this basis. First, the IC is attached to the circuit board. Then solid crystal, wire bonding, testing, dispensing, baking, and become a finished product.

In terms of the production process alone, several steps have been omitted, and industry insiders say that this can save a large part of the cost. It is worth noting that the COB package does not require reflow soldering, which is one of the advantages of COB.

[rs20]

This point on the other thread seems relevant, and hasn't been mentioned here yet:

COBs may also allow the rest of the board to be assembled using a cheaper process than a conventional package really permits.  In cheap electronics you'll occasionally see COBs mounted on little two-layer carrier PCBs that slot into the main PCB.  Now there's no need for careful alignment of a many-legged IC package, the main board can be much coarser pitch and the whole thing can be wave soldered instead of stenciled and reflowed.

[KL27x]

As long as the board is not too big, the same die bonding robots do the work, whether its on a board or in a lead frame. Why would it be more expensive to do it on a board? For many devices the package cost is comparable to the die cost. It's not insignificant.

Leaving out the setup for the specific board.

Simply size/logistic. When you mass produce something, the smaller it is, the easier it is. You don't build a wire-bonding robot for each job you want to do. You use this robot as much as possible, for many jobs.

If you can stack a job in a smaller space, load the robot infeed and collect in the outfeed many x more pieces, you do less work and have less human labor and overall easier time. PCB is way bigger and doesn't feed in a tube or reel. They probably do large panels at a time. Sure, there may be robots that can take a giant stack of PCB and feed them and stack the output. But then it's a more expensive robot and that goes into the cost, too. You still have to move that giant stack in and out to clear the way for the next job. Maybe then you have human operating a forklift. 

Ask anyone who operated a PnP if they will buy your resistors and caps that are 1/5th the price, but they only come 500 pieces on a reel. When robots are so sophisticated they don't need tuning and tweaking, then maybe. Once you have this robot tuned in and running, you don't want it to stop.

[flasonsmts]

I know this COB well because I know the COB is widely used in USB flash drive mainbaod production. The USB flash drive solder in a large amount per year, 10 billions pieces , so it is worth to produce in COB format to reduce the cost.  each time the PCB board factory produce millions pieces of USB flash drive COB chip and sell it sperately to USB flash drive assembly factory. This make sense and profit.