California Devices Inc and the HCS10017

In this forum you can write about anything that does not fit in other forums.
This includes generic Oric talkings and things that are totaly unrelated but want to share with people here :)
User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

California Devices Inc and the HCS10017

Post by dreamseal » Sat Jul 14, 2018 5:22 pm

I thought I would start a new topic to collect information on Calfornia Devices Inc, the company that manufactured the Oric ULA. I have spent the last few hours searching the Internet for information on this company and it appears that there is quite a lot of information out there about the company and there different series of CMOS gate array products. I have a feeling that Mike Brown was discovering back in 2014 some of the same documents that I discovered this afternoon. I found a Google groups discussion between Mike and others about the CDI HCS Series of CMOS gate arrays, and the information Mike posted suggested he had read some of what I found today: https://groups.google.com/forum/#!msg/s ... vIUfAv00UJ

Before looking at the information on CDI though, I should start by linking to a couple magazine articles from back in 1983 that confirm California Devices Inc. as the manufacturer of the HCS10017 used in the Oric:

https://archive.org/stream/popular-comp ... 9/mode/2up

https://archive.org/stream/popular-comp ... 3/mode/2up

When I started searching the net for information on California Devices Inc, I started to discover a number of very interesting things. Firstly, from the Wikipedia entry for "Gate array":

https://en.wikipedia.org/wiki/Gate_array

In here we read that a man named Robert Lipp founded California Devices Inc back in 1978. We also read that he was the person that developed CMOS gate arrays, and on the following write up on his achievements, it claims that he was the inventor of the CMOS gate array:

https://www.clusteredsystems.com/management/

and that California Devices Inc was the first world class ASIC company dedicated to CMOS gate arrays. The following document has a page showing their product ranges:

http://archive.computerhistory.org/reso ... 01-acc.pdf
cdi_products.png
and it also shows when their various products were introduced:
cdi_evolution.png
The following document also has some information about California Devices Inc:
http://archive.computerhistory.org/reso ... 01-acc.pdf
The CDI series of metal gate CMOS arrays is a low-cost family of seven gate arrays that range in complexity from 60 to 880 equivalent gates and from 22 to 72 pins. The HC series of silicon gate CMOS arrays is a high-performance family of seven devices that range from 300 to 1782 equivalent gates and from 40 to 92 pins. The HCS series is compatible with the HC series but offers higher performance and density. Seven HCS series devices are available, ranging from 300 to 1782 gates in complexity and from 40 to 92 pins.
and this table from that document is very interesting:
cdi_products_2.png
And this advert describes it a bit more:
https://books.google.co.uk/books?id=tjl ... HCS+series
If it's VLSI you want, we offer the fastest single layer metal high-density CMOS gate array available - a 4 micron silicon gate digital array - our HCS Series. Our HCS Series has the best performance record of any single layer semi-custom product available and it leads the competition in VLSI technology.

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Sat Jul 14, 2018 11:25 pm

After a fair bit of searching, I've managed to find specific details about the seven devices that made up CDI's HCS Series, from two different sources:
cdi_hcs_series.png
That info is from a publication called IC Master 1984, Volume 2, and is from pages 4461 and 4463. There is copy on archive.org.

And I found the following snippet from :
cdi_hcs_series_2.png
cdi_hcs_series_2.png (8.81 KiB) Viewed 1790 times
Unfortunately that was all that could be seen in books.google.com for this second image (above):
https://books.google.co.uk/books?id=0kw ... wQ6AEIKTAA

You'll notice that there is a discrepancy with the name of the 1260 gate device. I suspect that the second image is the correct number, i.e. HCS12600, because that makes sense and is consistent with the others.

The HCS10017 might be the realisation of an HCS10000 with Oric's logic gate design applied to it? Perhaps that is how they managed the part numbering, e.g. by having some kind of incremental numbering. Just a guess. Something that would appear to support that though is the numbering of the Telestrat HCS chips, which are HCS3119 and HCS3120. By this theory, those two would be HCS3100 devices with the custom logic applied.

If the HCS10017 is an HCS10000, then it has up to 1000 2-input gates. From my earlier post, we know that it is CMOS, that it has a single metal layer, and uses silicon (probably polysilicon) for the gates, and that they're 4 micron silicon gates.

The final image I found tonight was this full page advert from CDI:
1983_IC_Master_Volume_2_1318.jpg
Really interesting. I notice at the bottom of the advert that there is an address in Newbury, Berkshire, UK, for a company called Guiltspur Micro Systems that is a supplier within the UK. Oric also had an address in Berkshire, UK, in Ascot. It's probably a 45 minute drive between the two. I wonder if they did their business through the UK company, or whether they worked directly with CDI in San Jose, California.

User avatar
NekoNoNiaow
Pilot Officer
Posts: 119
Joined: Sun Jan 15, 2006 10:08 pm
Location: Montreal, Canadia

Re: California Devices Inc and the HCS10017

Post by NekoNoNiaow » Wed Jul 18, 2018 5:21 pm

Nice and very interesting bit of research! Thanks for putting in the effort to dig that up! :wink:

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Wed Jul 18, 2018 10:38 pm

NekoNoNiaow wrote:
Wed Jul 18, 2018 5:21 pm
Nice and very interesting bit of research! Thanks for putting in the effort to dig that up! :wink:
I have kept digging, and have managed to discover a bit about the design of the HCS10000. The eagle eyed may have noticed something quite interesting in the image of the advert that I posted in my last message. What I'm referring to is that insert with the close up of the chip surface with the text underneath saying "The new CDI HCS CMOS Silicon Gate Array". It is amazing how much such a small image can reveal. It turns out that this small section appears a great many times across the surface of the chip (without the logic interconnects), due to the way in which a Gate Array repeats a "basic cell" in both horizontal and vertical directions. Here is that small section of the chip surface rotated:
hcs_series_closeup_5.png
I think what they've done here is to invert the image. In my opinion is easier to view it like this, where I've inverted the image with an image tool:
hcs_series_closeup_4.png
By continuing to search and search some more online, I managed to dig up a deck of old lecture slides from a 2006 university course that contained this image (note the reference below to CDI):
cdi_gate_array_structure_2.png
Already this is starting to look very similar to the earlier image that I extracted from the California Devices advert.

In my next post (because I can't seem to attach more than 3 images to this one), I'm going to show just how close these images are to each other.

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Wed Jul 18, 2018 11:00 pm

As promised, here is the original image from the advert placed next to and aligned with the same section from the university lecture slide image:
hcs_series_closeup_side_by_side.png
There are a few things missing in each case, but they are certainly the same structure. In the case of the left hand image, the boundaries of the p-type and n-type "Active"/Diffusion areas are not shown. In the case of the right hand image, there are some metal lines that are not shown. The reason for the difference in metal lines is because the right hand image is of uncommitted cells, whereas the left hand image has custom logic programmed into it (i.e. by adding the necessary metal lines to form the logic gates required).

Those same university lecture slides happened to have a few more CDI related images, including this one:
cdi_unprogrammed_array.png
What I've done below is to extend the side-by-side image to include the same section from this third image, and then to colour the original two images to highlight the same parts:
hcs_series_closeup_side_by_side_colour_2.png
So a reminder again, the left most image comes from the California Devices advert, the middle image from the first black and white image of the CDI gate array structure from the lecture slides, and the right most image comes from the "CDI Unprogrammed Array" image.

Given that the image from the California Devices advert is described as the "HCS Series", then there is no doubt in my mind that the images from the lecture slides are a good representation of what we would expect to see on the surface of an HCS Series CMOS chip, such as the HCS10000.

I'm sure you're all keen to see these lecture slides. This is the URL of where I found them:

http://www.csit-sun.pub.ro/courses/vlsi ... pt0607.pdf

Not a lot mentioned about CDI, but there are a number of images relating to CDI gate arrays. Look at the last two pages of the PDF, also in Chapter 13.

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Wed Jul 18, 2018 11:29 pm

From having studied the "CDI single metal layer gate array structure" image above for a bit, the following is what I think the "tile" is that would repeat across the chip in both horizontal and vertical directions:
cdi_gate_array_tile.png
This image contains four cells. Two of them are 2 input cells, and the other two are 3 input cells. So what this means is that the cells probably toggle like that across the chip, i.e. 2 input cell, 3 input cell, 2 input cell, 3 input cell, etc. The images from the lecture slides appear to imply this.

From reading a bit on how CMOS gate array sizes are measured, they are usually given in what is known as "equivalent gates". For example, the HCS10000 has 1000 equivalent gates. An equivalent gate is defined as two pairs of transistors with which a two input NAND or NOR logic gate can be made. What this means is that 1000 equivalent gates is 4000 transistors. To get an idea of whether that is enough for the HCS10017, the 6502 CPU apparently has something like 4528 transistors (if you include the depletion mode pull up transistors, which really you should for the comparison to be fair). So the HCS10017 seems like it should have enough equivalent gates to achieve the functionality we're expecting.

One last note about the above image. It has 10 transistors. A transistor is formed where the polysilicon (red lines in the earlier images) pass over the Active/Diffusion areas. If we're expecting 4000 transistors, and there are 10 in each of these "tiles", then it might give us an idea of how many times this tile is repeated in each direction. Presumably the shape after repeating in each direction is roughly square. I'll have to play around a bit to see what some likely numbers are, and then use GIMP to generate an image of that size with this small section tiled the necessary number of times in each direction to achieve a roughly square image with 1000 equivalent gates (4000 transistors). The end result might be getting close to what we're expecting the uncommitted CMOS gate array part of the surface of the HCS10000 to look like. Then its only the custom metal lines that make up the "Paul Johnson designed" HCS10017 specific logic that we'll need to wait for the silicon surface photographs (i.e. die shots) to reveal the placement of. I'm really hoping to arrange for such photos to be made over the coming weeks.

User avatar
Dbug
Site Admin
Posts: 2701
Joined: Fri Jan 06, 2006 10:00 pm
Location: Oslo, Norway
Contact:

Re: California Devices Inc and the HCS10017

Post by Dbug » Thu Jul 19, 2018 6:51 pm

dreamseal wrote:
Wed Jul 18, 2018 11:29 pm
Then its only the custom metal lines that make up the "Paul Johnson designed" HCS10017 specific logic that we'll need to wait for the silicon surface photographs (i.e. die shots) to reveal the placement of. I'm really hoping to arrange for such photos to be made over the coming weeks.
A decap of the ULA would be awesome, specially if we can reverse the logic accurately.
Who know, maybe we would discover some traces of some features they were planning to add but did not have time to :)

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Fri Jul 20, 2018 7:52 pm

Dbug wrote:
Thu Jul 19, 2018 6:51 pm
A decap of the ULA would be awesome, specially if we can reverse the logic accurately.
The nice thing about it being a CMOS gate array is that the diffusion areas and polysilicon lines are in pre-defined areas, as are the VDD and VSS lines. I'm used to looking at fully custom NMOS die shots, and for those you need to identify everything. But for this CMOS gate array, all we'll need to do is identify where the smaller (i.e. smaller than the VDD and VSS) metals lines are. Still a big task but seems more straight forward than reversing the 6502 or 6522 or AY-3-8912. So hopefully what we'll end up with will be accurate. And its all digital as well, which makes it easier.

User avatar
NekoNoNiaow
Pilot Officer
Posts: 119
Joined: Sun Jan 15, 2006 10:08 pm
Location: Montreal, Canadia

Re: California Devices Inc and the HCS10017

Post by NekoNoNiaow » Sun Jul 22, 2018 5:58 pm

dreamseal wrote:
Wed Jul 18, 2018 11:00 pm
What I've done below is to extend the side-by-side image to include the same section from this third image, and then to colour the original two images to highlight the same parts:
Huhu, reading the first paragraphs I was going to suggest coloring, only to be rewarded on the next paragraph with exactly that. ;)
Again, thanks for your work!
dreamseal wrote:
Wed Jul 18, 2018 11:00 pm
Given that the image from the California Devices advert is described as the "HCS Series", then there is no doubt in my mind that the images from the lecture slides are a good representation of what we would expect to see on the surface of an HCS Series CMOS chip, such as the HCS10000.
Given the evidence in the colored version it is clear that the differences are simply the mask used to program the gate array, the bottom layers are almost identical to the manufacturer's unprogrammed ones.
dreamseal wrote:
Wed Jul 18, 2018 11:00 pm
I'm sure you're all keen to see these lecture slides. This is the URL of where I found them:

http://www.csit-sun.pub.ro/courses/vlsi ... pt0607.pdf

Not a lot mentioned about CDI, but there are a number of images relating to CDI gate arrays. Look at the last two pages of the PDF, also in Chapter 13.
To be precise, the CDI gate array section of Chapter 13 is at Page 73 and even though it is short it is complete enough to understand the process.
And holy kitty, this PDF is impressively exhaustive. I pity the poor souls who have to follow this class in under one year. ;)

NightBird
Officer Cadet
Posts: 33
Joined: Sun Nov 24, 2013 7:00 pm

Re: California Devices Inc and the HCS10017

Post by NightBird » Mon Jul 23, 2018 9:10 pm

Dbug wrote:
Thu Jul 19, 2018 6:51 pm
A decap of the ULA would be awesome, specially if we can reverse the logic accurately.
Who know, maybe we would discover some traces of some features they were planning to add but did not have time to :)
Do you know Deus Ex Silicium?
www.dexsilicium.com

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Tue Jul 24, 2018 12:32 am

Dbug wrote:
Thu Jul 19, 2018 6:51 pm
A decap of the ULA would be awesome, specially if we can reverse the logic accurately.
Some very exciting news today :D

Last week (or it was probably the end of the week before), I started making a few enquiries about this. I spoke to Mike Connors (who is the same man who decapped the ZX Spectrum ULA for Chris Smith as part of his excellent investigation into that chip). My email was asking simply for a quote as to how much it might cost to get the chip decapped and photographed. Mike said he'd need to take a look at the chip to get an idea. So I couriered him a few chips that I'd already bought for this purpose. Mike confirmed that he'd received the chips Tuesday last week.

Mike Connors works at Datel, and he has a web page here describing their reverse engineering services:
https://www.rawscience.co.uk/

This is a commercial lab, so I was preparing myself for a quote that I probably wouldn't be able to justify paying by myself. I wanted to get an idea though.

I hadn't heard anything since Tuesday last week, so I sent Mike an email late last night. About mid-morning today, I received a reply from Mike to say that his engineer had already taken some photos. This was more progress than I had been expecting. I was still expecting a quote of some sort. I tried not to get my hopes up and assumed that it was perhaps just a test to give me an idea of what they could produce. Then as I was on my way home from work this evening, I received another email from Mike with a link to a file asking if I could access that link. I tried and I could.

At this point I had my mind on getting home as quickly as I could to see what was in the file. I didn't fancy trying to download it on my phone. It was rather big.

To cut a long story a bit short, the file I downloaded was a package containing various files, including the full HCS10017 die shot in perfectly clear quality.

:D

I certainly wasn't expecting this. So I wrote back to Mike Conners asking how much I owed him for this. He replied saying that he isn't able to charge for the work (perhaps commercial labs can't charge for this type of reverse engineering), and he asked as payment only that I mention his name and the company, i.e. he did it as a promotional favour. So I would like to say a very big thank you to Mike at Datel for doing this work for us, at no cost. It really is far more than I was expecting, and the quality is amazing. And if any business is considering having some professional reverse engineering done, then check out Mike's Raw Science webpage mentioned above.

Without further ado, here is a link to download it from (a warning that it is 100 Meg in size):
https://drive.google.com/open?id=112Xjl ... KFMJqFL9PA

@Dbug, the above link will probably die at some point. So I was wondering if you have space on your Oric file server to give it a more permanent location? I will probably also make various dieshot websites aware of this so that they can store copies. We need to preserve this for all time now.

NightBird
Officer Cadet
Posts: 33
Joined: Sun Nov 24, 2013 7:00 pm

Re: California Devices Inc and the HCS10017

Post by NightBird » Tue Jul 24, 2018 5:47 am

Wow! Bravo!

User avatar
Dbug
Site Admin
Posts: 2701
Joined: Fri Jan 06, 2006 10:00 pm
Location: Oslo, Norway
Contact:

Re: California Devices Inc and the HCS10017

Post by Dbug » Tue Jul 24, 2018 8:16 am

dreamseal wrote:
Tue Jul 24, 2018 12:32 am
Mike Connors works at Datel, and he has a web page here describing their reverse engineering services:
https://www.rawscience.co.uk/
(...)
I certainly wasn't expecting this. So I wrote back to Mike Conners asking how much I owed him for this. He replied saying that he isn't able to charge for the work (perhaps commercial labs can't charge for this type of reverse engineering), and he asked as payment only that I mention his name and the company, i.e. he did it as a promotional favour. So I would like to say a very big thank you to Mike at Datel for doing this work for us, at no cost. It really is far more than I was expecting, and the quality is amazing. And if any business is considering having some professional reverse engineering done, then check out Mike's Raw Science webpage mentioned above.
So I guess it means we can advertise this die shot on facebook, twitter, etc... copy it, modify it, etc... all we need is to say that Datel allowed that to happen?

Sounds good to me :)
dreamseal wrote:
Tue Jul 24, 2018 12:32 am
@Dbug, the above link will probably die at some point. So I was wondering if you have space on your Oric file server to give it a more permanent location? I will probably also make various dieshot websites aware of this so that they can store copies. We need to preserve this for all time now.
There you go: http://defence-force.org/download/oric/ULA/

- hcs10017_dieshot.jpg (100MB), that's the original file, which believe me is not easy to open on half of my programs :)

- Details_WindowsPhotoGallery.png, an example of tool that can open the file but mangles it beyond recognition...
- Details_WindowsPhotoViewer.png, an other program that opens it, but that time in full detailed glory

- Dieshot_thumbnail.png, a zoomed out view of the die in reasonable resolution (1415 x 1245 pixels)
- Dieshot_thumbnail_zoomed_details.png, the same, but with some zoomed in details to show easily how it looks on the large quality one.

Dieshot_thumbnail_zoomed_details.png
ULA die shot with detail highlights

And yes, great work!

Now just need to figure out how this monster works :D

User avatar
Chema
Game master
Posts: 2282
Joined: Tue Jan 17, 2006 10:55 am
Location: Gijón, SPAIN
Contact:

Re: California Devices Inc and the HCS10017

Post by Chema » Tue Jul 24, 2018 10:51 am

I am following this shivering with anticipation....
Last edited by Chema on Tue Jul 24, 2018 10:17 pm, edited 1 time in total.

User avatar
dreamseal
Officer Cadet
Posts: 61
Joined: Sat Mar 17, 2018 6:14 pm

Re: California Devices Inc and the HCS10017

Post by dreamseal » Tue Jul 24, 2018 7:36 pm

Below are download links for two additional images that Mike Connors sent me with the original die shot image.

The first is an image of the die shot that has been annotated, showing the pin numbers, and with many of the lines coloured:

https://drive.google.com/open?id=1aezVU ... 4LcPdEK4lw

Unfortunately it is a bit too low res, so difficult to make things out.

I was also sent a close up section from what appears to be the same annotated image:

https://drive.google.com/open?id=1BsQZP ... c7gSt9wCb1

This one is much clearer and gives an example of what the hi res annotated image would look like. I have asked Mike Connors if he could send me the larger copy of the annotated image, since it seems like they must have created one. Hoping to hear back soon.

Also I've been in touch with Mike Brown, who has in the past done a lot of thinking about what the internals of the ULA might look like. I knew he would be very interested in the die shot, as would any Oric enthusiast. Mike Brown has put another copy of the die shot images on his web site below:

http://oric.signal11.org.uk/html/ula-dieshot.htm

and he has started producing his own annotated images, such as this one linked to below:

http://oric.signal11.org.uk/files/pub/u ... otated.jpg

Post Reply