Some of the common IC types and what's in them.

[BShan]

Found this little gem of an app note on different package types from TI

www.ti.com/lit/an/snoa280/snoa280.pdf?ts=1692120964222&ref_url=https%253A%252F%252Fwww.google.com%252F

Ah yes, "Pressed Ceramic Package" description fits exactly what I experienced. The "filler" acted almost as hard as glass, and every lead from the leadframe was embedded in this thick filler layer.

Aluminum has a much higher electrical resistance than gold, it's MUCH more chemically reactive and not resistant to oxidation, and it has a rather low melting point.

For chips which will likely get hot, it's no good for bond wires. Copper has similar properties to gold, therefore they're much more interchangeable.

I didn't think about the thermal aspect. I reckon a bond, already potentially weaker due to aluminum oxide maybe interferring with a physical bonding, would also be less resiliant to the effects of thermal expansion mismatches.

 

[macfixer01]

At the risk of being pedantic, EPROM is erasable programmable read-only memory. The original ROM had to be programmed during manufacture (hard coded). PROM could be programmed once after it was manufactured. Thereafter, it could only be read. EPROM allowed the package to be erased by exposure to UV light and reprogrammed. EEPROM eliminated the need for the window and the UV light by allowing the memory to be erased electrically.

That's the problem with being old. Your brain gets filled with useless trivia that's just waiting to escape.

Dave

Yup, PROM’s were also called OTP or one time programmable. Also are we supposed to believe there is a mythical 40 pin CERDIP that yields 20% more than a Pentium Pro? Even if it was some prototype, I think he missed a decimal place there? Maybe .05 gram, but certainly not 0.5

 

[Alondro]

Yup, PROM’s were also called OTP or one time programmable. Also are we supposed to believe there is a mythical 40 pin CERDIP that yields 20% more than a Pentium Pro? Even if it was some prototype, I think he missed a decimal place there? Maybe .05 gram, but certainly not 0.5

This mythical CERDIP was supposedly found in King Tut's tomb and was left by AAAAAAALIENS after they built the pyramids!

 

[lazersteve]

I'm thinking that 20% higher statements is likely referring to grams per pound yields which pentium pros are not nearly close to the top of the grams per pound scale as far as ceramic style cpus/DIPs are concerned.

Steve

 

[lazersteve]

Here's a picture of the lower yielding "large" (40 pin) CERDIPs. The highest yields per pound come from the smaller packages (8-16 pin style).

I ran this batch last year. It's a mix of 28 thru 40 pin cpus.

Steve

 

[lazersteve]

The above cpus are not PROMs for the most part, there are a few EEPROMs in there, but it's mostly Z80/8086 style ceramic cpus.

The glass windowed PROMs (with the frit between 2 ceramic halves) are much lower yielding and labor intensive to process than those in the box.

Steve

 

[Brett Novak]

Correct Lazer - a good find none the less. That's all I basically run since I have them in abundance. (And, what I based my original % yield off of so it's a bit of cheating )

Those are old MCU's - 8 bit devices that, some of them could be new since the goldies are still used for test chips and validation (amazingly) Or they could be old Intel or original MCHP, NXP, TI or ST goldies for 8051's, 8085's or similar.

And they are heavy - much much much better yields than silver plastic DIP packages. And way way way better yielding than the newer TQFP (flat pack with little legs on all 4 sides) by 2:1 at a minimum.

 

[macfixer01]

I'm thinking that 20% higher statements is likely referring to grams per pound yields which pentium pros are not nearly close to the top of the grams per pound scale as far as ceramic style cpus/DIPs are concerned.

Steve

Lazersteve,
You seem to have conflated some things here. The 20% comment was mine. The only reason I mentioned PROM’s also being known as OTP’s was in response to what Bshan had said, not that there were any shown in the photos.

No offense but I suggest you read Brett’s post again. Talking about the 40 pin CERDIP’s he said the average weight is 5.0 Grams and they have a yield of 0.5 Grams recoverable gold Per Device!

I compared that to gold cap Pentium Pros since they’re one of the higher yielding chips Per Device. Based on multiple reports here and elsewhere, they only yield at best about 0.4 grams? That’s why I suggested his claimed number is 20% higher than a PPro. It seems incorrect for what’s only a 40 pin chip, with a much smaller gold cap, and that weighs only 1/18 as much.

 

[Brett Novak]

Mac - I'm falling on my own sword here.

I did a test this weekend running 105g of goldies, 105g of goldies SHREDDED, 105g of plastic package DIPS, and 105g of plastic TQFP packages.

Yields were (+/-10%)

105g goldies straigt - 4.5g
105g goldies ground - 4G (may be a rounding error or little extra water when I weighed it)
105g old PDIP (plastic silver leads, 2 rows of pins) - 1.5g
105g TQFP - couldn't measure it, so I'll try a run of 1kg.

So with goldies we are cheating and getting ~0.03%

Silver PDIPS - ~0.014%

and TQFP's - not measureable

goldies fall in line (IIRC) with the old pentium yields.

Now - on the weight, yes - a 4 pin CERDIP is about 5g but I didn't sort them LOL.

I will seperate 100g of 40, versus 28 versus 16, versus 8 pins. I think I have all of those, but yeah - the bigger packages will yield less by WEIGHT since the amount of ceramic goes up exponentially.

 

[Brett Novak]

OK broke out a few sample cases. Each on is 100g +/- 2%

Big ones - which are the wide body 40 pin monsters.

Narrow long ones - I think they are 28 lead's

Small ones - I didn't sperate them out but they are 6, 8 and 16 leads.

bed time - will play with acid tomorrow.

 

[Brett Novak]

and...guys - keep the math easy don't mix grams per pounds - keep it in grams,KG and (t)oz You'll save a lot of head scratching. We don't all read the same weights...so everything I do is in grams. but my numbers are still close.

 

[macfixer01]

Mac - I'm falling on my own sword here.

I did a test this weekend running 105g of goldies, 105g of goldies SHREDDED, 105g of plastic package DIPS, and 105g of plastic TQFP packages.

Yields were (+/-10%)

105g goldies straigt - 4.5g
105g goldies ground - 4G (may be a rounding error or little extra water when I weighed it)
105g old PDIP (plastic silver leads, 2 rows of pins) - 1.5g
105g TQFP - couldn't measure it, so I'll try a run of 1kg.

So with goldies we are cheating and getting ~0.03%

Silver PDIPS - ~0.014%

and TQFP's - not measureable

goldies fall in line (IIRC) with the old pentium yields.

Now - on the weight, yes - a 4 pin CERDIP is about 5g but I didn't sort them LOL.

I will seperate 100g of 40, versus 28 versus 16, versus 8 pins. I think I have all of those, but yeah - the bigger packages will yield less by WEIGHT since the amount of ceramic goes up exponentially.

So if I follow that correctly you actually processed 21 of the CERDIPs (21 x 5 =105G) and got 4.5 Grams of gold total or about .214 Grams Per Device? Also can you actually specify if that was your prototype chips? Or was it ordinary average gold cap, gold pins, 40 pin CERDIP’s?

While we’re an it, are you actually saying you got less gold from the same type/number of chips that you ground up?

 

[lazersteve]

I calculate all cpu/dip yields in grams per pound not per chip.

The pentium pro is such a common cpu that it's yields are very well documented at 0.33g/cpu (90g cpu weight). That's roughly 1.66g/pound which is not that high up in the yields per pound as ceramic cpus/dips go. The tungsten heat spreader is 30g average which is why they yield so low per pound along with their low housing weight to pin count ratio.

The 40,32, 28 pin cpus shown in the box average about 3.6 to 4 g/pound depending on the mix ratio, plated legs, black lids, and gold lids.

If you are getting 0.5g per cpu of any size that's astonishing to me and I've conservatively run 10,000 pounds of assorted purple ceramic cpus/dips in sorted and unsorted lots in the last 15 years.

Brett do you have photos of the chips that you got your reported 19.45g (454 ÷ 105g = 5 x 4.5g = 19.45) per pound yields from? That's better that even the smallest xtal oscillators I've run and far better that the average 9g+/ pound that mixed Intel 186 postage stamps produce.

Not doubting you, but I'd like to see these chips for reference. The only comparable yields to that I've seen has been from special aerospace "test cpus" which produced 28.5g from 225g that equates to around 57.5g per pound, but they were factory test pieces as far as I was told and therefore would not be found "in the wild".

Steve

 

[Brett Novak]

Mac - for all intents & purpose, these are 1980's style gold chips with plated legs, plated chip carrier, plated cover, and gold bond wires. $20 each 25 years ago, so, add inflation.

The 4.5g was on the heavy side, but it had some stock left in the pot from an AP experiment, which is why I said ground versus whole yielded about the same 4g per 105g stock, and this is repeatable,

For plastic packages, that number drops off fast to 1.5g per 105g of material. For TQFP it's even lower, but this is a point of the evolution of IC's. (More) thinner bond wires, and incorporating different bond wire material such as CuPd

I will try and run a BIG batch of new IC's but I'll betcha I won't get much.

 

[Brett Novak]

I was just replying and taking pictures LOL. Here's ~100g of 40 lead wide body CERDIPS, 28 lead thin body, and 8 to 16 pin thin body.

 

[lazersteve]

I've run untold pounds of each of those types. Typically I run then in 4 to 6 pound and as high as 20 pound batch sizes.

Steve

 

[Brett Novak]

and these may never be found in the wild ;-)

I'll run a batch of OTS devices.

 

[Brett Novak]

PM'd you with an offer Laser! Happy to oblige to an independent study by someone smarter than me!

Cheers!
BN

 

[Brett Novak]

I've run untold pounds of each of those types. Typically I run then in 4 to 6 pound and as high as 20 pound batch sizes.

Steve

I'd love to pick your brain - I have tried....

200g with straight AR and it was a nightmare.
1kg with AP, waited a month, got a ton of flakes, and re-ran the rest through AP... a nightmare.

100g or so at a time in small batches seems to be manageable but I've yet to get high volume down.

 

[lazersteve]

Awesome posts guys...this is what makes the forum great!

Steve