UK Vintage Radio Repair and Restoration Discussion Forum

UK Vintage Radio Repair and Restoration Discussion Forum (https://www.vintage-radio.net/forum/index.php)
-   Websites (https://www.vintage-radio.net/forum/forumdisplay.php?f=38)
-   -   Die pictures (https://www.vintage-radio.net/forum/showthread.php?t=165425)

Noopy2014 5th Apr 2020 8:12 am

Die pictures
 
Hi all!


Are you interested in die pictures of some older Chips?

Yesterday I posted a Burr Brown DAC709 ("U.K." ;)):

https://richis-lab.de/DAC01.htm

Sorry, text is german but Google translator does a good job and there are a lot of self-explaining pictures.
Of course you can ask me whatever you want here in english.


I have also taken pictures of voltage references, mouse sensors, harddrives, MEMS,...

https://www.richis-lab.de

Not everything is vintage...


Have fun! :)

Nuvistor 5th Apr 2020 12:14 pm

Re: Die pictures
 
Excellent photos of the chip internals. ��

Noopy2014 5th Apr 2020 12:16 pm

Re: Die pictures
 
Thanks! 8-)

Guest 5th Apr 2020 1:01 pm

Re: Die pictures
 
And people say engineering is boring.

mole42uk 5th Apr 2020 1:51 pm

Re: Die pictures
 
Quote:

Originally Posted by merlinmaxwell (Post 1231719)
And people say engineering is boring.

No-one on this forum though!

I used to work in the semiconductor manufacturing area with machinery to inspect wafers for problems. AMD in Dresden had some fascinating processes, IR in Newport, Plessey in Swindon, Intel in Texas too but I never was allowed to keep any images :-( I do have a 4” wafer somewhere that has some interesting chips etched upon it, I might see if I can take a photo.

Radio Wrangler 5th Apr 2020 2:52 pm

Re: Die pictures
 
Fascinating!

I've decapped a number of devices just out of curiosity, as well as replacing individual transistors on microwave thin-film hybrids.

Thank you very much for posting.

David

Keith956 5th Apr 2020 4:07 pm

Re: Die pictures
 
Many years ago when I worked for Fairchild Semi, one of the chips I worked on was the F558 which was a 8 bit parallel multiplier. Well to cut a long story short another company (name escapes me) had an LS part out, so we bought one, took the lid off and put it under the microscope. Took about 16 overlapping photos, then stitched the photos together and traced the logic. It was a slightly modified Booth's algorithm job, but our implementation had to be a bit different as the Fast process didn't allow multiemitter transistors which were common in LS gates.

In those days such forms of copying were quite widespread - I can recall another problem chip was the 74LS109, it was difficult to meet the performance of the TI part and we realised that the parasitics were critical, so once again the microscope was put into service...

Radio Wrangler 5th Apr 2020 5:20 pm

Re: Die pictures
 
Having a look to see how they did it is fine. Copying it is sinful. Working out how to do it even better is the best thing.

Copying is playing catch-up. If you want your business to thrive, you gotta play leap-frog :-)

An oriental competitor brought out a box competing with one of ours. Shall I just say that investigation of a unit showed an uncanny resemblance in the key areas, even down to a couple of bits of suboptimal design in ours.

David

TonyDuell 5th Apr 2020 5:39 pm

Re: Die pictures
 
Quote:

Originally Posted by merlinmaxwell (Post 1231719)
And people say engineering is boring.

But boring (whether it's sinking a well or making a cylinder on a lathe) is certainly part of engineering (Sorry, could not resist).

Getting on-topic, have you found any little cartoons or text on the dice? I am told that the MicroVAX CPU chip (at least one version) has a comment in Russian on it, for example.

TonyDuell 5th Apr 2020 5:41 pm

Re: Die pictures
 
Quote:

Originally Posted by Keith956 (Post 1231771)
Many years ago when I worked for Fairchild Semi, one of the chips I worked on was the F558 which was a 8 bit parallel multiplier. Well to cut a long story short another company (name escapes me) had an LS part out, so we bought one, took the lid off and put it under the microscope. Took about 16 overlapping photos, then stitched the photos together and traced the logic. It was a slightly modified Booth's algorithm job, but our implementation had to be a bit different as the Fast process didn't allow multiemitter transistors which were common in LS gates.

As an aside I was recently working on a 4800 bps leased-line modem unit that does digital signal processing using an '558 parallel multiplier and a couple of 2901 ALU slices. Controlled by a set of microcode PROMs sequenced by 3 2911s. Fascinating bit of design.

Noopy2014 5th Apr 2020 5:45 pm

Re: Die pictures
 
I took pictures of a LM399:
https://richis-lab.de/REF02.htm
Tesla copied the design as you can easily see here:
https://richis-lab.de/REF02a.htm
But they changed some small parts of the circuit.
Probably the soviet union had some intelligent scientists and a lot of knowledge but they used the western designs to accelerate the development.

emeritus 5th Apr 2020 5:48 pm

Re: Die pictures
 
Thanks for posting this. I was involved in custom IC design in the early 1970's, a time before advances in photolithography allowed the manufacture of the high density chips shown in the link, and before automated computerised layout. Our chips were laid out by hand, using large sheets of mm graph paper, 6H pencils, eraser shields and decent plastic rubbers. Oh, and scalpels and steel rulers to cut out and insert photocopies (actualky dyeline contact prints - photcocopiers were not accurate enough) of standard elements like logic gates and D-type flip-flops. Actual cut and paste! The masks for the considerably less complex bipolar chips were cut and stripped in our lab, by hand, by a draftsman, in Rubylith, but the more complex P-MOS chips' masks were made by CGI in Glenrothies. This involved defining x-y axes on our drawings and, for each layer, listing the x-y coordinates of the corners of every shape on that layer on reams of A4 sheets. CGI would enter the data on their CALMA system, return to us plots for checking, and after correcting the inevitable errors, prepare the Rubyliths and masks used for production.

When an urgent redesign became necessary ( the only remaining suitable processing plant having been destroyed by fire!), we still had to draw the new designs by hand, but were able to hire Emihus's equipment that allowed the coordinates to be read directly onto 1/2" computer tape. As they were very busy, we could only use it outside their normal hours, whch meant starting at around 6.00pm and leaving in the early hours ( by climbing out of the window into the shrubbery and sliding it closed behind us!). The computerised system had no keyboard, commands and data setting the parameters were entered via a row of eight centre-biassed minature toggle switches, flip up for"1", down for"0" and press a button to enter each 8-bit word. These days I think you can simply specify your design parameters and a computer will do all the donkey work of laying-out.

The colours you can see in the pictures ( possibly apart from the more vivid fluorescent hues) are what you can actually see through a microscope that is equipped with a suitable illumination source. The diffusion layers are thin enough to be transparent, and the colours are produced by optical interference, the same mechanism that produces the rainbow colours in a thin film of oil on water and Newton's Rings in two touching glass plates. You can judge the thickness of a diffusion by its colour.

Noopy2014 5th Apr 2020 5:49 pm

Re: Die pictures
 
Quote:

Originally Posted by TonyDuell (Post 1231794)
Quote:

Originally Posted by merlinmaxwell (Post 1231719)
And people say engineering is boring.

But boring (whether it's sinking a well or making a cylinder on a lathe) is certainly part of engineering (Sorry, could not resist).

Getting on-topic, have you found any little cartoons or text on the dice? I am told that the MicroVAX CPU chip (at least one version) has a comment in Russian on it, for example.

I found some artwork... Have to search again on my website... ;D
The STM32 has some strange figures:
https://richis-lab.de/STM32_02.htm
https://richis-lab.de/images/STM32/02_05.jpg

Noopy2014 5th Apr 2020 5:57 pm

Re: Die pictures
 
Quote:

Originally Posted by emeritus (Post 1231801)
Thanks for posting this. I was involved in custom IC design in the early 1970's, a time before advances in photolithography allowed the manufacture of the high density chips shown in the link, and before automated computerised layout. Our chips were laid out by hand, using large sheets of mm graph paper, 6H pencils, eraser shields and decent plastic rubbers. Oh, and scalpels and steel rulers to cut out and insert photocopies (actualky dyeline contact prints - photcocopiers were not accurate enough) of standard elements like logic gates and D-type flip-flops. Actual cut and paste! The masks for the considerably less complex bipolar chips were cut and stripped in our lab, by hand, by a draftsman, in Rubylith, but the more complex P-MOS chips' masks were made by CGI in Glenrothies. This involved defining x-y axes on our drawings and, for each layer, listing the x-y coordinates of the corners of every shape on that layer on A4 sheets. CGI wokd enter the data on their CALMA system, return to us plots for checking, and after correcting the inevitable errors, prepare the Rubyliths and masks used for production. When an urgent redesign became necessary ( the only remaining suitable processing plant having been destroyed by fire!), we still had to draw the new designs by hand, but were able to hire Emihus's equipment that allowed the coordinates to be read directly onto 1/2" computer tape. As theycwere very busy, we could only use it outside their normal hours, whch meant starting at around 6.00pm and leaving in the early hours ( by climbing out of the window into the shrubbery and sliding it closed behind us!). The computerised system had no keyboard, data setting the parameters was entered via a row of eight centre-biassed minature toggle switches, flip up for"1", down for"0" and press a button to enter each 8-bit word. These days I think you can simply specify your design parameters and a computer will do all the donkey work of laying-out.

The colours you can see in the pictures ( possibly apart from the more vivid fluorescent hues) are what you can actually see through a microscope that is equipped with a suitable illumination source. The diffusion layers are thin enough to be transparent, and the colours are produced by optical interference, the same mechanism that produces the rainbow colours in a thin film of oil on water and Newton's Rings in two touching glass plates.

A very interesting story! :clap:

I take the pictures with a "normal" camera:
https://richis-lab.de/Howto_Optik.htm
You can get these nice colours by flashing "into the lens".
https://richis-lab.de/Howto_Licht.htm
The light is reflected partly on the lens and you get images as with a reflected-light-microscope. A very cheap way to take such pictures but you have to practise a lot! ;D

emeritus 5th Apr 2020 6:57 pm

Re: Die pictures
 
Re#8, while most countries incorporated "Mask Protection" legislation into their respective copyright laws a couple of decades ago in order to benefit from the reciprocal protection provided by US law, it has proved to be pretty ineffective. AFAIR, copying was established in the first case in the US, but this was because when the layouts of the original and copy were superimposed, everything lined up, even the bonding pads. From the court's reasoning, it appeared that, had the copier used the same topology, but with the physical locations of the features moved around, then copying would not have been held to have occurred. Actually, the presence of sub-optimal features can be considered to demonstrate that copying has taken place, but it all depends on the facts. When I was with GEC I occasionally gave presentations to engineers about the law relating to copying, and, if it could be done without affecting performance, advised them to include circuitry or features that served no purpose, such as a gate that would never change state, or pairs of invertors in a signal path that were not there to provide a signal delay, and to document the fact.

I recall a case successfully brought by GI against Plessey, who had poached an engineer from GI and got him to design a telecommunications chip that was essentially identical to the GI design. It was a bit of a Phyrric victory, because their only customer was BT, BT always insisted on a second source for anything they bought, and because Plessey was the only other company capable of making it, GI had to allow them to manufacture under licence.

The law is always developing, and since retiring I'm afraid I have not kept up to date.

Noopy2014 5th Apr 2020 7:24 pm

Re: Die pictures
 
It´s amazing how many of you developed semiconducters!

Let me ask a off-topic-question: Does anybody know the Plessey MS1007?
Perhaps you have already seen that I´m collecting and analysing Gould 4070 oscilloscopes. They use a CCD-chip named MS1007. I have found some information about smaller and slower CCDs (MS1002, MS1003,...) but nothing about the MS1007. Perhaps it was a special chip for Gould or for the military?

emeritus 5th Apr 2020 8:29 pm

Re: Die pictures
 
It always seemed to be difficult to get info from Plessey Semiconductors, even if you worked for other Plessey divisions! Plessey Semiconductors used to do a considerable business as a foundry for third party designs. For some customers they simply supplied packaged, untested, ICs from masks provided by the customer, and so did not necessarily know what functions a chip was performing. The families of ICs used in the projects I was involved in were for various defence projects with high degrees of security classification, so no info would be publicly available about them.

Noopy2014 5th Apr 2020 9:00 pm

Re: Die pictures
 
Interesting, thanks for your input!
As far as I'm informed CCDs were very important for radar stations. That would explain why I wasn't able to find any datasheet or document.

Keith956 5th Apr 2020 9:34 pm

Re: Die pictures
 
Quote:

Originally Posted by TonyDuell (Post 1231795)
As an aside I was recently working on a 4800 bps leased-line modem unit that does digital signal processing using an '558 parallel multiplier and a couple of 2901 ALU slices. Controlled by a set of microcode PROMs sequenced by 3 2911s. Fascinating bit of design.

At the time I was living in the US and had just bought an Exidy Sorcerer with a S100 bus extender, and took a sample 558 and built a hardware multiplier card for it. I always had plans of building a proper floating point unit using 94xx Macrologic but never quite got round to it.

Oldtestgear 8th Apr 2020 10:35 am

Re: Die pictures
 
1 Attachment(s)
I hope that this wafer picture may be of interest. it is an early 8"(prototype) wafer diffused by Cypress Semiconductor, circa 1999. I was given this at a meeting on Silicon valley when these were still very rare as freebies. It survived travelling back in my luggage despite being in a plastic sleeve. I scrounged the wafer box later from work.

I also visited Zelenograd, home of the Russian semiconductor industry, in 1996 when we were looking for suitable new suppliers. The Russians were very capable engineers but they had absolutely no scruples about copying any device that they wanted or possibly obtaining the design by more dubious means. As far as they were concerned, they were effectively at war with the West & any way to get the parts required was fine.My visit was after the collapse of Communism & Russia seemed pretty lawless at the time. As they had a command economy parts were diffused with little thought about yields & some wafer fabs were really quaint compared to Western foundries.

I was lucky enough to have worked for several USA semiconductor companies as well as a small niche distributor here in the UK that repackaged die for all sorts of programmes.

Phil


All times are GMT +1. The time now is 6:37 am.

Powered by vBulletin®
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Copyright ©2002 - 2023, Paul Stenning.