PCB black blob of glue ICs during the 70's-90's

[RoGeorge]

I'm curious if those chip were of a particular technology, or if they were whatever size/technology was available back then, just that it was glued on a PCB instead of heaving their on case with pins and datasheets and everything else a standalone chip use to have.

Might be a personal bias, but I think those COB (Chip On PCB) were very low power compared even with today tech (batteries in such devices used to last for 5-10+ years).

What technology/size was that?  Was it CMOS or something else?

[Ranayna]

I think those CoB could have been just about anything. They were used because in mass production the price of an IC package was relevant, and CoB was cheaper, and repair/rework was absolutely irrelevant.
I do not remember where i read it, but supposedly labor costs in china were so cheap at some point, that the bonding and blobbing of such a chip was done by hand. And it was still cheaper than using a packaged IC.

[tom66]

CoB is still common in cheap consumer electronics manufactured in ridiculous volumes.  For instance my 1yr old Casio FX-83GTX calculator, which retails for about £15,  has a CoB chip.  The process is probably no different to a standard microcontroller (100 - 200nm range)

[tooki]

I do not remember where i read it, but supposedly labor costs in china were so cheap at some point, that the bonding and blobbing of such a chip was done by hand. And it was still cheaper than using a packaged IC.

There are videos on YouTube showing how it’s done. A human places the die on the PCB and loads it into the wire bonding machine, which uses machine vision to identify the bond pads and then bonds them.

[RoGeorge]

I suspect those chips were less advanced, something like um rather than nm tech.  I wonder how far fetched would be to do that in a DIY manner, without ordering the dies from a chip fab.

It's been about 50 years since the blob chips started to be a thing, should be achievable with today DIY tech, isn't it?

Those chips were very low power, and for small automation there is no need for high integration scale.  um would be enough.  I'm still waiting for DIY chips similar with the nowadays 3D printing tech but for ICs, don't you?  How far is that?

[magic]

Before the '90s everything was µm tech and for another decade or two everything ultra cheap still was

Any small chip with not too many connections can be mounted that way. Many of them may be MCUs with mask ROM.

They are still present in every low cost DMM. Some are the traditional ICL7216 equivalent IC, others some mystery meat delta-sigma with autoranging and TRMS. I have also seen blob chips in cheap digital thermometers.

Sometimes there is a TQFP footprint surrounding the blob. Probably they used a packaged chip for R&D and couldn't be assed to modify the production board design.

[magic]

I'm still waiting for DIY chips similar with the nowadays 3D printing tech but for ICs, don't you?  How far is that?

Far, I guess. It's not exactly like 3D printing, you start with a clean silicon wafer and need to keep it clean and diffuse controlled impurities into controlled areas. Lots of fun chemistry and fun temperatures
Prolly wouldn't even be legal in the EUSSR these days.

There are news about DIY semiconductor fabrication every now and then. I think the state of the art at this time is a few transistors on noncomplementary MOS process.

[wraper]

Those chips were very low power, and for small automation there is no need for high integration scale.  um would be enough.  I'm still waiting for DIY chips similar with the nowadays 3D printing tech but for ICs, don't you?  How far is that?

Many of them were not low power at all. It could be just about anything. Like SoC with a built in 6502 or 6800 CPU, mask ROM, calculator SoC.

[RoGeorge]

The ones I have in mind were the blob chips for clocks, remote controls, pocket calculators, thermometers, wrist watches and such, with LCDs.  Those were very, very low power.  Still have a few such devices.  Last time I've changed the batteries was 3-5 years ago, don't remember, and they are still working fine.  Maybe those were not CMOS, like a 6502, that's why I'm asking.

[edavid]

The ones I have in mind were the blob chips for clocks, remote controls, pocket calculators, thermometers, wrist watches and such, with LCDs.  Those were very, very low power.  Still have a few such devices.  Last time I've changed the batteries was 3-5 years ago, don't remember, and they are still working fine.  Maybe those were not CMOS, like a 6502, that's why I'm asking.

Of course they were CMOS.

(And older, larger feature size CMOS processes had lower leakage than more modern processes.)

[eti]

I'm curious if those chip were of a particular technology, or if they were whatever size/technology was available back then, just that it was glued on a PCB instead of heaving their on case with pins and datasheets and everything else a standalone chip use to have.

Might be a personal bias, but I think those COB (Chip On PCB) were very low power compared even with today tech (batteries in such devices used to last for 5-10+ years).

What technology/size was that?  Was it CMOS or something else?

I believer the proper term is "glob tops"

[eti]

I do not remember where i read it, but supposedly labor costs in china were so cheap at some point, that the bonding and blobbing of such a chip was done by hand. And it was still cheaper than using a packaged IC.

There are videos on YouTube showing how it’s done. A human places the die on the PCB and loads it into the wire bonding machine, which uses machine vision to identify the bond pads and then bonds them.

Sure are:

[tooki]

I suspect those chips were less advanced, something like um rather than nm tech.  I wonder how far fetched would be to do that in a DIY manner, without ordering the dies from a chip fab.

It's been about 50 years since the blob chips started to be a thing, should be achievable with today DIY tech, isn't it?

LOL no. Even manufacturing the first generation of discrete transistors is fundamentally out of reach for DIY.
Here are some videos on the history of the transistor that explain the process. The first link is a playlist of a 2.5h documentary that I highly recommend to anyone despite the length:
https://youtube.com/playlist?list=PL2JLMUvYJIRI_OBwm9BW3Wh2M4GXbtDhB
This one focuses on Fairchild:
https://youtu.be/z47Gv2cdFtA

Yes, there are one or two folks who have made transistors and extremely basic ICs on their own, but it’s extremely difficult, and difficult to do consistently. But semiconductor factories aren't called “foundries” without reason. You need ovens. Special equipment. And lots of toxic chemicals.

Semiconductors are a modern miracle. It’s miraculous that integrated circuits work at all*, but the truly miraculous part is that they can be manufactured at scale and affordably, never mind at the low cost we have now.

*The principles are simple. The difficulty is in actually making the damned things. The levels of cleanliness and materials purity required are absolutely mind-boggling. The silicon used in the most advanced semiconductors today (the processors in our phones and computers) has “eleven nines” of purity: 99.999999999%. That’s one atom of contaminant per 100 billion. To the best of my knowledge, there’s no other consumer product in existence that requires raw materials even remotely that pure. In one of the videos above, they told how Fairchild figured out that one reason for wildly fluctuating yield in batches of their transistors (first generation!) was determined to be whether workers had washed their hands after urinating — even though nothing was handled with bare hands.

[RoGeorge]

Seen the videos from the playlist and added them already to this collection https://www.eevblog.com/forum/chat/fun-for-nerds/msg3297026/#msg3297026, but didn't know the Fairchild one, thanks!

In the playlist they were showing about the beginnings of manufacturing germanium transistors in Japan during the 50's, and how they were controlling the purifying of the crystalline germanium rod with a pierced bucket of water and a piece of floating wood in it. 
https://youtu.be/ihkRwArnc1k?list=PL2JLMUvYJIRI_OBwm9BW3Wh2M4GXbtDhB&t=1165 (at minute19:25)

Closer to our times, DIY ICs in a garage:

"Z2" - Upgraded Homemade Silicon Chips
Sam Zeloof

[SiliconWizard]

This is still used as others have mentioned. It's just basically a naked die with wire bonding directly to the PCB. For large volumes, it can be worthwhile. May also have some benefits in terms of reliability in harsh environments, at least compared to conventional SMD packages.

These days, flip-chip is more common than wire bonding on PCB. It's basically a naked die with balls on the die pads and mounted like a BGA. Underfilling is also a common practice in this case to improve reliability.

But back in the 70's, SMD was virtually non-existent. Hence those CoBs. Why CoBs became "popular" in small consumer electronics (such as watches and calculators) before SMD, such is history.

[tooki]

Amusingly, I posted my earlier reply on the way to work, and then today at work, I finally got to see wire bonding done in person. One of the PhD students I’ve assembled tons of boards for is working on silicon radiation detectors. We assemble the readout boards for him, onto which he then attaches a die and performs the wire bonding. I’d asked him to let me watch the next time he did some, and that ended up being today!

[james_s]

There are videos on YouTube showing how it’s done. A human places the die on the PCB and loads it into the wire bonding machine, which uses machine vision to identify the bond pads and then bonds them.

Even today? Surely a machine could place the die on the PCB at least as effectively as a human.

[tooki]

I’m sure it’s often done by machine, especially in higher-wage countries than China. To me, it’s one of those manufacturing steps that I would never have guessed was ever done by hand, but China does by hand anyway. I suppose it also depends on the volume.

[Miyuki]

I’m sure it’s often done by machine, especially in higher-wage countries than China. To me, it’s one of those manufacturing steps that I would never have guessed was ever done by hand, but China does by hand anyway. I suppose it also depends on the volume.

I saw in an old documentary from Czechoslovakia how they switched from manual bonding to semi-automatic (the operator chooses just orientation and first point) in the 70s and early 80s
But the main reason for them was to speed up production as there were not enough people to do it.
So I can see how it can work in China with an army of poorly paid workers. They have no problem finding more people.

[tom66]

Probably also depends on the number of bonds.

A small "toy chip" that just generates sound when the "toy trigger" is pulled might have 5 or 6 pads, "easy" to bond.

A Casio calculator IC with hundreds for the matrix display and keypad?  Forget about doing that manually on any economical scale.

[Alex Eisenhut]

I saw one in the '80s on a VIC-20 game cartridge, so basically a 8Kx8 ROM or maybe EPROM. FWIW.