PDA

View Full Version : homebrew 8008 project



~llama
December 12th, 2003, 05:18 PM
Well, my homebrew project (the Neptune) will probably be a very close near-clone of the Mark 8 with one major difference: the processor will be homebrew too; I want to try to build an 8008 out of TTL chips :D i know, i'm crazy...

Anywho... I have the 8008 user's manual and block diagrams, i know how it works, but i have pretty much zero experience with designing TTL circuits (this was intended to be how i learn that)... so i'm calling on you guys with experience with TTL stuff (i hope)...

Are is there anything out there comparable to a schematic for the 8008? I know that Datapoint implemented it in TTL in the 2200, so there have to be schematics out there -somewhere- but... does anyone know where anything like this might exist?

Super-Slasher
December 16th, 2003, 05:39 PM
Making your own processor out of TTL chips? Can you fill me in on what this is, like what TTL chips are and how the heck they can be made into a processor unit? You've perked my ears, hehe...

~llama
December 16th, 2003, 05:44 PM
TTL chips are basic logic IC's, like registers, flip-flops, etc... so if you assemble enough of them you get a working CPU. the 8008 was originally created for some datapoint terminal, implemented in these logic chips, and intel condensed all the logic onto one chip, and the 8008 was born.

Basically you design the processor and implement the parts of it with chips of less logic integration than a microprocessor.

Erik
December 16th, 2003, 06:35 PM
I believe that the Scelbi 8H was an all TTL machine, for example.

Erik

CP/M User
January 5th, 2004, 03:21 AM
"~llama" wrote:

> TTL chips are basic logic IC's, like registers, flip-flops,
> etc... so if you assemble enough of them you get a
> working CPU. the 8008 was originally created for
> some datapoint terminal, implemented in these logic
> chips, and intel condensed all the logic onto one chip,
> and the 8008 was born.

That seems like quite a trickly thing to do building a
microprocessor, or is it actually easier to go through
the process than to actually describing it?

> Basically you design the processor and implement
> the parts of it with chips of less logic integration than
> a microprocessor.

The only thing I understand, is that it's all microscopic
& supposed to be quite a delicate operation.

Cheers,
CP/M User.

hwhall
January 7th, 2004, 07:53 AM
I believe that the Scelbi 8H was an all TTL machine, for example.

Actually, the Scelbi 8H used the "8008 CPOU on a chip". Some of the original documentation is at http://users.techline.com/cmlove/scelbi/scelbi.htm
--Wayne

Erik
January 7th, 2004, 08:02 AM
Actually, the Scelbi 8H used the "8008 CPOU on a chip". Some of the original documentation is at http://users.techline.com/cmlove/scelbi/scelbi.htm

Hmm. . . you're right, of course. Maybe it was the Kenbak-1 I was thinking of. . .

I know there is an early 1970s, pre-Mark-8 machine that is considered by some, at least, to be a PC that was all TTL.

Erik

CP/M User
January 7th, 2004, 11:51 AM
"Erik" wrote:

>> Actually, the Scelbi 8H used the "8008 CPOU on a
>> chip". Some of the original documentation is at
>> http://users.techline.com/cmlove/scelbi/scelbi.htm

> Hmm. . . you're right, of course. Maybe it was the
> Kenbak-1 I was thinking of. . .

> I know there is an early 1970s, pre-Mark-8 machine
> that is considered by some, at least, to be a PC that
> was all TTL.

How about this one:

DATAPOINT 2200 Computer Terminal Corporation (CTC)

according to old-computers.com/museum this machine
came out in 1971 & uses TTL logic which is said to be
equivalent to the 8008. According to this site, the Mark 8
came out in 1974.

Cheers,
CP/M User.

Erik
January 7th, 2004, 12:08 PM
The Datapoint was a minicomputer, though. The Kenbak, the Scelbi and the Mark-8 were very early microcomputers.

Erik

CP/M User
January 7th, 2004, 01:21 PM
"Erik" wrote:

> The Datapoint was a minicomputer, though. The Kenbak,
> the Scelbi and the Mark-8 were very early microcomputers.

So this is only a part of the machine then?

http://www.old-computers.com/museum/photos/ctc_datapoint-2200_1.jpg

Cheers,
CP/M User.

Erik
January 7th, 2004, 01:26 PM
I'm not sure if it is or it isn't the whole machine. I am speaking more to the intent of the design.

There's always been a fine line between mini and micro but the basic premise of a micro is "one user" while the basic premise of a mini is "one workgroup."

From what I understand of the Datapoint products, they were workgroup type machines.

I could be wrong. It's happened before at least once in this thread alone! :)

Erik

CP/M User
January 7th, 2004, 05:18 PM
"Erik" Wrote:

> I'm not sure if it is or it isn't the whole machine. I am speaking
> more to the intent of the design.

> There's always been a fine line between mini and micro but the
> basic premise of a micro is "one user" while the basic premise of a
> mini is "one workgroup."

> From what I understand of the Datapoint products, they were workgroup
> type machines.

> I could be wrong. It's happened before at least once in this thread
> alone! :-)

Oh okay, I can only go by what old-computers.com has on their website &
unfortunately it doesn't speak in terms of what the user base is. It
does say in their website that the company came up with a LAN, but I'm
unsure if they made it for this machine or for others. Also even
though the Datapoint 2200 used TTL logic for the CPU, it was
responcible for the success of the 8008. But apart from that, I don't
know much more, or if it is a type of Minicomputer (since
old-computers.com is looking at machines which pre-date the
Microcomputer).

Cheers,
CP/M User.

Erik
January 7th, 2004, 07:19 PM
Speaking of oldcomputers.com, I really do like that site.

When I first started with www.vintage-computer.com I considered implementing a database driven site with user-contributed machine information. In the end I decided that I didn't really have the time to implement that with all of the other things I had going in life.

It's still in the back of my mind, though, and someday I hope to get enough free time to take a shot at it.

Erik

CP/M User
January 7th, 2004, 08:58 PM
"Erik" wrote:

> Speaking of oldcomputers.com, I really do like that site.

It's fine when they get their information right.

But when they fail to mention about the TI 99/4 & that it
was a seperate system from the TI 99/4A they look
stupid, which doesn't help from the fact that they're been
told & it still remains uncorrected!

Instead they insist that the TI 99/4A came out in 1979
which is a mistake & that the TI 99/4A is a 16bit computer
& the TI 99/4 was an 8bit!

> When I first started with www.vintage-computer.com I
> considered implementing a database driven site with
> user-contributed machine information. In the end I
> decided that I didn't really have the time to implement
> that with all of the other things I had going in life.

That's okay, I see someone has tried to correct the TI99/4A
page by adding comment to the TI 99/4, but since the info
is inacurate in the Technical Info section, they should be
made to correct that, since people like myself & perhaps
others have made mistakes based on their mistakes!

> It's still in the back of my mind, though, and someday
> I hope to get enough free time to take a shot at it.

Perhaps just make sure that's correct before posting it
on your website. Due to the mistakes done to that
computer the TI 99/4x series, they should really have
2 seperate pages which talk about these 2 systems. As
it is, they have to do a bit of work to get it right & since
I've told them & nothing has been done, then I feel they
are too lazy to get off their behinds & fix it! :-(

But apart from that, their okay.

Cheers,
CP/M User.

Terry Yager
January 8th, 2004, 04:01 AM
"Erik" wrote:

> Speaking of oldcomputers.com, I really do like that site.


Instead they insist that the TI 99/4A came out in 1979
which is a mistake & that the TI 99/4A is a 16bit computer
& the TI 99/4 was an 8bit!


Cheers,
CP/M User.

Really? I didn't know that the 99/4 was an 8-bitter. Do you know what processor it used? Was it still the same 9900whatever (partially disabled) that they used in the 99/4A?

--T

Erik
January 8th, 2004, 06:14 AM
Really? I didn't know that the 99/4 was an 8-bitter. Do you know what processor it used? Was it still the same 9900whatever (partially disabled) that they used in the 99/4A?

I always thought that the 99/4 and 99/4a differed mainly in the keyboard.

I've never used a 99/4, though. Just the "a" model.

Erik

CP/M User
January 8th, 2004, 12:33 PM
"Terry Yager" wrote:

>>> Speaking of oldcomputers.com, I really do like that site.


>> Instead they insist that the TI 99/4A came out in 1979
>> which is a mistake & that the TI 99/4A is a 16bit computer
>> & the TI 99/4 was an 8bit!

> Really? I didn't know that the 99/4 was an 8-bitter. Do
> you know what processor it used? Was it still the same
> 9900whatever (partially disabled) that they used in the
> 99/4A?

Yes it was the same processor as a matter of fact. Someone
told me that the 99/4 was 8bit, most of the other technical
specs are also the same, except for the year it came out.

Since TI created their own CPU, I'd imagine that the
modified this for the 99/4A.

Cheers,
CP/M User.

barryp
January 8th, 2004, 05:58 PM
Really? I didn't know that the 99/4 was an 8-bitter. Do you know what processor it used? Was it still the same 9900whatever (partially disabled) that they used in the 99/4A?

I always thought that the 99/4 and 99/4a differed mainly in the keyboard.

I've never used a 99/4, though. Just the "a" model.

Erik

Both the TI-99/4 and the TI-99/4A use the TMS-9900 microprocessor, (not partially disabled at all, the same chip used in their business systems) which is a 16-bit CPU in the same way that an 8088 is 16-bit.

The keyboard is the main physical difference, the /4 had a calculator-type keyboard which was disliked immensely. The reason for the A in TI-99/4A is that the Video Display Processor was changed from the TMS-9918 to the TMS-9918A, which is a much more significant difference.

With a few exceptions, the two computers run the same modules.

CP/M User
January 8th, 2004, 09:39 PM
"barryp" wrote:

>>> Really? I didn't know that the 99/4 was an 8-bitter.
>>> Do you know what processor it used? Was it still the
>>> same 9900whatever (partially disabled) that they
>>> used in the 99/4A?

>> I always thought that the 99/4 and 99/4a differed
>> mainly in the keyboard.

>> I've never used a 99/4, though. Just the "a" model.


> Both the TI-99/4 and the TI-99/4A use the TMS-9900
> microprocessor, (not partially disabled at all, the
> same chip used in their business systems) which is a
> 16-bit CPU in the same way that an 8088 is 16-bit.

> The keyboard is the main physical difference, the /4
> had a calculator-type keyboard which was disliked
> immensely. The reason for the A in TI-99/4A is that
> the Video Display Processor was changed from the
> TMS-9918 to the TMS-9918A, which is a much more
> significant difference.

> With a few exceptions, the two computers run the
> same modules.

The 99/4A would have came out later than the 99/4
though wouldn't it?

Okay, so I mixed up the CPU as being the difference,
though I did know it was some hardware difference
about those machines. Unfortunately, the books I
have don't go into this very well, well since they were
mean't for the real hardware 'buffs.

Cheers,
CP/M User.

barryp
January 9th, 2004, 05:57 PM
The 99/4A would have came out later than the 99/4
though wouldn't it?

Yes, of course.


Okay, so I mixed up the CPU as being the difference,
though I did know it was some hardware difference
about those machines. Unfortunately, the books I
have don't go into this very well, well since they were
mean't for the real hardware 'buffs.

There are other minor differences too, not worth mentioning.

CP/M User
January 9th, 2004, 08:01 PM
"barryp" wrote:

>> The 99/4A would have came out later than the 99/4
>> though wouldn't it?

> Yes, of course.

>> Okay, so I mixed up the CPU as being the difference,
>> though I did know it was some hardware difference
>> about those machines. Unfortunately, the books I
>> have don't go into this very well, well since they were
>> mean't for the real hardware 'buffs.

> There are other minor differences too, not worth
> mentioning.

I'm guessing that those are cosmetic! :-)

Cheers,
CP/M User.

netbios
April 4th, 2004, 07:28 PM
"Erik" wrote:

> The Datapoint was a minicomputer, though. The Kenbak,
> the Scelbi and the Mark-8 were very early microcomputers.

So this is only a part of the machine then?

http://www.old-computers.com/museum/photos/ctc_datapoint-2200_1.jpg

Cheers,
CP/M User.

The picture is indeed a complete Datapoint 2200 with a disk pack sitting on top. On the right top were two data tape cassette drives. The OS was loaded on one, and you used the other for your program store.

I programmed one of these in college, sometime around 1974~1976.
I wrote a program in assembly, yes it was 8008 compatible, that downloaded (via a serial port) form definitions and did data entry and then uploaded the results. This was tied into a hospital database system we were developing on a Cascade 80 minicomputer.

The Datapoint OS came with an editor and assembler and other program development tools and utilities. Since we wrote our own data entry application, I don't remember what else they had.

BTW: while you can make comments about the circuit technology deteriming whether something is a "mini" or a "micro" that can be misleading, particularly during cusps in the technology. A better deteriminate is the system I/O configuration and target application set.

Both this Datapoint 2220 and the Cascade 80 were built out of MSI TTL logic. The Cascade however was in a desk sized unit with a printing console, a line printer, and two disk drives (one removable).

Dave.

Erik
April 4th, 2004, 08:20 PM
One of the Datapoint 2200s just sold on eBay for about $170 US (http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=4121047707)

Not too bad a deal if you can find the boot tape or a copy. . .

Erik

CP/M User
April 4th, 2004, 11:41 PM
"netbios" wrote:

> The picture is indeed a complete Datapoint
> 2200 with a disk pack sitting on top. On
> the right top were two data tape cassette
> drives. The OS was loaded on one, and
> you used the other for your program store.

> I programmed one of these in college,
> sometime around 1974~1976.
> I wrote a program in assembly, yes it was
> 8008 compatible, that downloaded (via a
> serial port) form definitions and did data
> entry and then uploaded the results. This
> was tied into a hospital database system
> we were developing on a Cascade 80
> minicomputer.

> The Datapoint OS came with an editor and
> assembler and other program
> development tools and utilities. Since we
> wrote our own data entry application, I
> don't remember what else they had.

> BTW: while you can make comments
> about the circuit technology deteriming
> whether something is a "mini" or a
> "micro" that can be misleading,
> particularly during cusps in the
> technology. A better deteriminate is the
> system I/O configuration and target
> application set.

> Both this Datapoint 2220 and the
> Cascade 80 were built out of MSI TTL logic.
> The Cascade however was in a desk sized
> unit with a printing console, a line printer,
> and two disk drives (one removable).

Thanks for your interesting comments Dave,
glad to see that someone has used one of
these machines & know what it's like. The
applications you've written sound failrly
impressive (even for this time)

Cheers,
CP/M User.

CP/M User
April 4th, 2004, 11:44 PM
"Erik" wrote:

> One of the Datapoint 2200s just sold on
> eBay for about $170 US.

> Not too bad a deal if you can find the
> boot tape or a copy. . .

WOW!! Do they normally sell for more, Erik?

$170 US sounds fairly expensive, but I guess
it would be rarer to come by! :-)

Cheers,
CP/M User.

Allison
April 2nd, 2005, 11:13 AM
[quote="CP/M User"]"Erik" wrote:

> Speaking of oldcomputers.com, I really do like that site.


Instead they insist that the TI 99/4A came out in 1979
which is a mistake & that the TI 99/4A is a 16bit computer
& the TI 99/4 was an 8bit!


Cheers,
CP/M User.

The TI9900 (chip) is a 16 bitter, one of the first. The TI99/4 (and 4a) are based on the TI9900 chip and therefore 16bit. Now here's the twist as a cost reduction the 99/4 computer series had logic to trnaslate the 16bit data bus into two sequential 8bit reads/writes to 8 bit wide memory and IO. However the CPU is 16bit. The closest analogue to that is the intel 8086 and the 8088 as they are the same chip programatically but the bus interface differes. There are others that could do either 8 or 16bit such as the DEC T-11 wich was base PDP-11 on a single 40pin dip. The last odditiy for is the Z280 wich is really an 8bit (z80) cpu but has the ability to be either an 8bit bus or 16 bit bus for performance.

So it's easy for the non technical to not understand bus interface size and cpu data width.

Allison

Allison
April 2nd, 2005, 11:19 AM
The schematic for the internals for an 8008 is far logical only for Pmos logic and would not translate to TTL directly. It's about 5000 transistor equivelent to implement. For example most of the registers in the 8008 were not Flipflops but dynamic ram cells (it would crash if the closk was too slow).

To create a TTL design you will have to develope a functional specification and design from that. For the 8008 that should not be too bad as it was a pretty primitive device. What is more of a consideration is if you wnat the resulting TTL to result in a 18pin plug to the 8008 socket or if it can be a less of an emulation where its' interface to ram, IO and other logic is on an as needed basis. THe former is the more comples task as the 8008 had a multiplexed data bus where all the major control, address and data traveled in and out on those 8pins with te rest supplying clock, power and some state information.

Allison

Allison
April 2nd, 2005, 11:37 AM
Making your own processor out of TTL chips? Can you fill me in on what this is, like what TTL chips are and how the heck they can be made into a processor unit? You've perked my ears, hehe...

In the days before Microporcessor "Chips" the elements needed to create a computer were in more elmental forms. TTL is a logic family that is defines by levels and interchip interface rules. The TTL family contain basic logic elements and a few more complex ones but not enough to be a "computer" in itself. So it would take from 25-1000s of chips to create a "computer" with the number of chips required dependent on how complex a computer your trying to make.

Before TTL (around 1970ish) there was DTL and RTL logic families and then cards with raw transistors interconnected to yeild logic level functions.
For example a TTL NAND gate is rougly 5 transistors and 4 resistors.
There are famalies of logic that MOS (metal oxide gate) transistors as well
but the idea of package logic still applies despite differences in internal construction.

An example of a all transistor computer was the PDP-5 (1965) with the early PDP-8 (1967) being a mix of transistors and some early logic chip.
By 1976 the main logic of a PDP-8 was reduced to a single chip that contained enough transistors and internal "wiring" to do what the 1967
machine did.

Computers can be implemented with any device that exhibits two states and can be controlled to obey basic logic rules (AND, OR, NOT). In past years Tubes, Magnetic cores even neon lamps have been tried. Transistors and ICs based on transistors however proved most power efficient and durable over time. The only difference between ICs and transistors, is that transistors became smaller and then multiples on
one peice of silicon were interconnected to do what a board of transistors did. As techniques were developed and refined the number of transistors
and their assosciated interconnects reached the point where the 4004 and later 8008 chips were born. However the path to those chips had started years before.


Allison

CP/M User
April 21st, 2005, 12:10 PM
"Allison" wrote:

>> Instead they insist that the TI 99/4A came out in 1979
>> which is a mistake & that the TI 99/4A is a 16bit computer
>> & the TI 99/4 was an 8bit!

> The TI9900 (chip) is a 16 bitter, one of the first. The TI99/4
> (and 4a) are based on the TI9900 chip and therefore 16bit.
> Now here's the twist as a cost reduction the 99/4 computer
> series had logic to trnaslate the 16bit data bus into two
> sequential 8bit reads/writes to 8 bit wide memory and IO.
> However the CPU is 16bit. The closest analogue to that is
> the intel 8086 and the 8088 as they are the same chip
> programatically but the bus interface differes. There are
> others that could do either 8 or 16bit such as the DEC T-11
> wich was base PDP-11 on a single 40pin dip. The last odditiy
> for is the Z280 wich is really an 8bit (z80) cpu but has the
> ability to be either an 8bit bus or 16 bit bus for performance.

> So it's easy for the non technical to not understand bus
> interface size and cpu data width.

Yes, I since come to realise the original TI 99/4 uses a 16bit processor while dealing with the 8bit reading & writing. Unfortunately, oldcomputers.com doesn't seem to acknowledge the two different machines - which is what I'm more fussed about. If you're into your Star Trek it's a bit like comparing Kirk's Enterprise & Enterprise A, they may look the same, but were two different ships (if you eliminate the cosmetic changes between movies!).

Cheers,
CP/M User.

carlsson
April 21st, 2005, 03:07 PM
Now here's the twist as a cost reduction the 99/4 computer series had logic to trnaslate the 16bit data bus into two sequential 8bit reads/writes to 8 bit wide memory and IO.
I suppose there were other computers - minis and alike? - that used TMS9900 series to its full power? Would it have meant any improvement in sales of the TI99 computer if Texas Instruments had designed a more expensive machine from the beginning? How about a more powerful Basic, a more direct and 3rd party friendly way to expand the system etc? Probably they were ahead of their time. Would it not have been possible to compensate a little if the 99/4A model had become 99/5 with or without a 99/4 compatibility mode?

Oh well, all computer manufacturers did one thing or another "wrong" in retrospect. I guess it is a result of opening a new branch of trade.

Terry Yager
April 21st, 2005, 03:58 PM
The TI 990-series were marketed as "mini-computers" even though they were small enough to sit comfortably on a desktop, and would prob'ly fit most people's definition of "micro-computer" by today's standards. Actually, I'm told that the TMS9900 chip is closer in archetecture to a mini-computer than to most micros.
When the 99s first came out, they were pretty expensive, till TI figgered out that the market wouldn't support a thousand-dollar home computer, and began drastically slashing the price in order to stay competitive with other home computers that were beginning to appear about that time.

--T

barryp
April 22nd, 2005, 05:34 PM
Now here's the twist as a cost reduction the 99/4 computer series had logic to trnaslate the 16bit data bus into two sequential 8bit reads/writes to 8 bit wide memory and IO.
I suppose there were other computers - minis and alike? - that used TMS9900 series to its full power? Would it have meant any improvement in sales of the TI99 computer if Texas Instruments had designed a more expensive machine from the beginning? How about a more powerful Basic, a more direct and 3rd party friendly way to expand the system etc? Probably they were ahead of their time. Would it not have been possible to compensate a little if the 99/4A model had become 99/5 with or without a 99/4 compatibility mode?

Actually there WAS a TI-99/5 (and, I think, a /6 and a /7) but the next most popular to the 99/4x was the TI-99/8, of which there are several in existance. That was to be "the next big thing".

Allison
April 24th, 2005, 09:26 AM
The TI 990-series were marketed as "mini-computers" even though they were small enough to sit comfortably on a desktop, and would prob'ly fit most people's definition of "micro-computer" by today's standards. Actually, I'm told that the TMS9900 chip is closer in archetecture to a mini-computer than to most micros.
When the 99s first came out, they were pretty expensive, till TI figgered out that the market wouldn't support a thousand-dollar home computer, and began drastically slashing the price in order to stay competitive with other home computers that were beginning to appear about that time.

--T

The concept of minicomputers was from a time (mid 1960s) when computers wer typically 3 or more 6ft racks. Anything that would fit on
or under a desk was mini. For example in 1961 the PDP-1 was 4 6foot racks and that was the base CPU. By 1965 the PDP7 a later version of
the PDP-1 was half the size and many times faster. The PDP-8 family
started [ca1967] as typically a full rack in size [6 foot] and progressed down to a box of 0.5cuft (Decmate-III with floppy and hard disk). That trend was seens with nearly every vendor that was around long enough.

The later development of chips in some cases meant minicomputer versions of hardware could be reduced to a microcomputer chip or
set of chips. Examples are TI9900[TI990], 6100/6120[PDP-8], LSI-11. F-11,T11, J11[PDP-11], MicroVAX[32 bit VAX series went down to chips], uNova and Fairchild uFlame[DG nova] and no doubt I've missed a few.
With the reduction in the number of chips, smaller power needs and cooling the boxes got smaller. When they got small enough engineers started stuffing more into them.


Allison

Allison
April 24th, 2005, 09:31 AM
The TI 990-series were marketed as "mini-computers" even though they were small enough to sit comfortably on a desktop, and would prob'ly fit most people's definition of "micro-computer" by today's standards. Actually, I'm told that the TMS9900 chip is closer in archetecture to a mini-computer than to most micros.
When the 99s first came out, they were pretty expensive, till TI figgered out that the market wouldn't support a thousand-dollar home computer, and began drastically slashing the price in order to stay competitive with other home computers that were beginning to appear about that time.

--T

Well the TI9900 is a minicomputer that was reduced to a chip. It's very minicomputer in the way it's interfaced. The Bit addressed IO was as odd
by micro standards as they come and the only other one that stands out as very mini is the 6100/6120 copying the PDP8 IO. If you look at micros designed orginally as micros rather than micros realized from mini designs
there are sometimes unique things that are very mini. Those oddities are
a result of the cost of putting IO on minis, IE: the printer or disk could cost more and have a higher chip count than the CPU.


Allison

machine
May 5th, 2005, 11:11 AM
Computers built using TTL devices (the 74 and 54 series of TTL) was common prior to useful microprocessors. They are called 'Discrete Processors'.

But it is not an easy task to build your own discrete processor. It takes a lot of skill and it will take probably 200 to 400 chips. The board would be large and just making the PCB will cost a lot as it will be so large. Doing the PCB layout by hand will simply be practically impossible, you would need software to do this like Protel.

Most 'home brewers' use PAL chips which contain huge numbers of uncommitted gates in one chip. They write the software to implement their particular computer design in the chip and programming hardware programs the PAL chip's gates in a particular fashion resulting in a home built microprocessor. There's quite a few hobbyists doing home brews. But even doing it in PAL chips is very skilled work.

You would need to know how to implement 'microcode' also which is the code embedded in all microprocessors which tells the chip how to behave internally. Microcode is not the program you apply externally to the microprocessor to get it to do useful things.

Allison
May 5th, 2005, 07:02 PM
Computers built using TTL devices (the 74 and 54 series of TTL) was common prior to useful microprocessors. They are called 'Discrete Processors'.

But it is not an easy task to build your own discrete processor. It takes a lot of skill and it will take probably 200 to 400 chips. The board would be large and just making the PCB will cost a lot as it will be so large. Doing the PCB layout by hand will simply be practically impossible, you would need software to do this like Protel.

Humm, I did it and it was less than 200 chips. A really interesting design was EGO presented in BYTE Sept 1985, a 16 bit machine. It used no PALS
and the densest devices (other than rams) were 74381 ALU and 74189
register array. Rough guess is under 100 TTL packages. It's not microcoded.

I have a very crude machine I did in less than 50 chips. Didn't have Protel nor use it. The base design is 16bit word and a vague resemblence to PDP-8 but a few things like the Data counters in ram left out. All ttl and
fast 400ns fetch/execute cycle. No gals/pals no microcode. Paper,
Logic maps, And the TTL data book.

There are several minimalist CPU designs like TOY, VSC and others. See
http://www.homebrewcpu.com/links.htm One desing listed was only 16
pieces of logic if PALs are not used. Granted it's a trivial machine but
if you can conceive and program that Colossis is just a big wirewrap
project away. ;)




You would need to know how to implement 'microcode' also which is the code embedded in all microprocessors which tells the chip how to behave internally. Microcode is not the program you apply externally to the microprocessor to get it to do useful things.

Microcode once you understand the simplest forms is trivial and can be
easier than a sequence engine. However it's not required. The upside of microcode is it can be easier to move bits than wires. The simplest uCode
machine is two parts. A clock source like a 555 timer chip, 74374 octal latch and a EPROM of atleast 256x8 in size. Code optional. Using just a pair of 374s and a 2732 I can easily make a state machine that can alter it's state based on inputs. Thats the core of a programable system.

I once believed as an engineer designing computers was black art. One day I decide to found out how they work and found out that taken one step at a time they are less forbidding than many say. What was missing for me was all those tutorials of how gates and FFs work were only a prelude
and then they skipped over how sequential systems happen to work. Computers are only very big sequential systems and in some cases
sequential systems within larger sequential systems. The latter being
very terse descritption of a microcoded computer.


Allison

machine
May 19th, 2005, 09:49 PM
"Microcode once you understand the simplest forms is trivial and can be easier than a sequence engine. However it's not required."


No processor discrete or integrated can run without microcode. Microcode is NOT trivial. Microcode controls the entire processor's functioning. Very few engineers produce microcode. It is a rare skill.

For every function there will be up to 100 switches enabled. A discrete processor will typically have 30 or more functions.

Allison
May 20th, 2005, 05:37 AM
"Microcode once you understand the simplest forms is trivial and can be easier than a sequence engine. However it's not required."


No processor discrete or integrated can run without microcode. Microcode is NOT trivial. Microcode controls the entire processor's functioning. Very few engineers produce microcode. It is a rare skill.

For every function there will be up to 100 switches enabled. A discrete processor will typically have 30 or more functions.

Well, your wrong. Sure there are a lot of thing going on in a machine
but they can be digested easily and understood once you have the basics clear.

There are two ways to implment the control system thats the core of a computer. One being a microcoded state machine and the other is a
sequential machine. They are very differnt in design, construction and even the debugging process s different.

RCA1802 is not microcoded
PDP-8 is not microcoded.
Z80 is not microcoded
PDP-11 in the LSI-11 form is microcoded
most other PDP-11s are sequential
VAX is microcoded
EGOII (sept 1985 p229, Byte magazine is a sequential machine)

TOY (a teaching design) has been rendered in both a sequential
and micro coded forms. Byte had an article showing the design of both.

Shall I go on.. There are no shortage of examples.

Generally speaking a two state RISC machine is less likely ot be microcoded than a CISC machine. That however is not absolute. There are trade offs for both styles of design.

Microcode is easier than doing a 22 state sequential machine. I've done both.

Microcoding and sequential machine programmers are not rare, they are
just not noticed anymore. With FPGAs and CPLDs being so cheap and dense every CSEE student has to do a processor implmentation of some sort these days as a lab project. There are a lot of students. The other part of it is with logic so plentyful figuring out how to save a package or
two doesn't compute so long as it all fits in the selected CPLD or FPGA.

I've studied some of the simplest examples as well as the complex. I've been designed and built a few of both over the last 30+ years. The simplest microcode engine I've done was less than 8 peices of common 1980s TTL and a 8bit wide by 256 byte PROM. It isn't a pentium but the principles remain visible.


Allison

ka7ftp
October 10th, 2006, 03:15 PM
Here is complete info to build an 8008 computer.

len

http://www.8008chron.com/

Chris2005
October 10th, 2006, 03:19 PM
FWIW there's a book by Robert Grossblatt (perhaps pardon the spelling) that details the description of an 8088 version of a similar thing. Called "the 8088 Project Book". Probably could be found on Amazon for next to nothing.

Mike Chambers
October 10th, 2006, 09:58 PM
why don't you just build an athlon 64 out of transistors diodes and wires from radio shack :D

just kidding, that's obviously near impossible. just try a pentium 3.

Mike Chambers
October 10th, 2006, 10:57 PM
wait a minute... the 8008 is pre-altair?

Chris2005
October 11th, 2006, 01:11 PM
it's a hardware thing. Some peeps enjoy building their own *rudimentary* puters.

Terry Yager
October 11th, 2006, 02:11 PM
why don't you just build an athlon 64 out of transistors diodes and wires from radio shack :D

just kidding, that's obviously near impossible. just try a pentium 3.

How big would that be? (I'm not sure my house could contain a monstrousity like that).

--T

Chris2005
October 11th, 2006, 02:25 PM
that's how they designed chips in the beginning. An 8088, and this is purely a guess, would take up say the entire floor of a garage. It would be cool to try something like that sometime ;)

carlsson
October 11th, 2006, 03:13 PM
The 8086 has 29,000 transistors. The i386DX has 275,000 transistors. The Pentium MMX has 3.1 million transistors. The Pentium 4 as of year 2000 had 42 million transistors.

Assume you need at least one square centimeter per real transistor, at least if you wire them together manually. The 8086 would take almost 3 m (31 sq. ft). The Pentium 4 would take 4200 m, or roughly one acre. If you construct the CPU with more space, it will be larger.

chuckcmagee
October 11th, 2006, 04:10 PM
In 1965, I made a single bit flip-flop on a breadboard. That was enough of TTL the hard way for me.

Of course, if you put in some feedback, make a great square wave generator.

Mike Chambers
October 12th, 2006, 03:01 PM
The 8086 has 29,000 transistors. The i386DX has 275,000 transistors. The Pentium MMX has 3.1 million transistors. The Pentium 4 as of year 2000 had 42 million transistors.

Assume you need at least one square centimeter per real transistor, at least if you wire them together manually. The 8086 would take almost 3 m (31 sq. ft). The Pentium 4 would take 4200 m, or roughly one acre. If you construct the CPU with more space, it will be larger.

wow.... 31 sq. feet for an 8086 or an 8088.... that would actually be very fun to make haha.

that would be so cool to make a gigantic room-sized 8088 by hand and then wire each of the 40 wires into the appropriate pin positions on an 8088 motherboard.

just think how cool you would be! somebody comes over for a visit.

visitor: "what is this gigantic thing? some sort of high-powered supercomputer??"
you: "no, it's an 8088. you wanna play Janitor Joe on it?"
visitor: ":eh:"


yeah... i think i'll just start making room in the basement right now!

carlsson
October 12th, 2006, 05:01 PM
Maybe 1x1 cm per transistor is optimistic if you wire it yourself. If you need 1.5x1.5 cm, the total theoretical size increases to 6.5 m, or 70 sq. ft.

The Wikipedia entry on the list of Intel microprocessors didn't mention how many transistors the 8080 consists of, but the 8008 had 3500 and the 8085, assembly language compatible with the 8080, had 6500. So, assuming 1.5 cm per transistor, a remake of 8008 would only take 0.8 m or 8.5 sq. ft.

Btw, does an integrated chip only consist of functional equivalents to transistors?

chuckcmagee
October 13th, 2006, 04:22 AM
No WAY. There are little resistors and other components in there too. When a flip-flop is "flipping", there is a pico second or nano second (real short interval) where both transistors are conducting. Basically a direct short between the +5v supply and ground. That's where "despiking capacitors" come in. They provide a very short term source of extra energy local to the "short". These keep the power supply from being loaded down by all these shorts.

P.S. If there are some electronic engineers out there, correct me if this info is wayyyy out dated. I haven't studied TTL since the mid 70s.

grant
November 2nd, 2006, 03:10 PM
It would not be very much fun to make one of these after the first 800 transistors or so. I completed a project similar to the this. It was very fun until after I soldered the first 1000LEDs. In all, I soldered 19,008 LEDs which is 38,016 solder joints.

BUT, I still want to make a transistor or at least TTL based 8080. :) I will probably make a TTL 8080, and then just make little transistor blocks that replace the TTL components. This way it will be easier to view the schematic, AND the individual blocks can be tested on a TTL logic tester. Otherwise it would be the worst thing in the world to debug!

Here is the display. Its about 80x41. It took me about 100 hours to assemble. Every single LED has been tested.

http://www.stockly.com/images2/060129-LED_Display_Front_2718.jpg

http://www.stockly.com/images2/060129-LED_Display_Back_2716.jpg

atari2600a
November 2nd, 2006, 03:13 PM
damn!!!!!!!
Um, I mean...DARN!!!!!!!

Erik
November 2nd, 2006, 04:12 PM
damn!!!!!!!
Um, I mean...DARN!!!!!!!
That is an impressive combination of completely insane and too cool for words!

Awesome!

Terry Yager
November 2nd, 2006, 04:17 PM
We love blinkenlights! When do we get to see it lit up? (I'm assuming a sign of some kind?)

--T

atari2600a
November 2nd, 2006, 04:23 PM
(I'm assuming a monochrome LED dot-matrix IBM-PC display?)

EDIT: I see I'm not the only one who buys Safeway Select(TM) soda! :p

grant
November 2nd, 2006, 10:22 PM
I want to program pong for the Altair to go on the screen. My plan is to use some dual port SRAM for video ram. Just memory map the display...

I have designed the driver PCBs, I just haven't found the time to order them. I was talking about the project and Henry from gse-reactive told me I couldn't solder it in a year. So I got it done in 15 days. :) That was last January! Now he says he meant "finish" it... There are about 8,000 solder joints required for the drivers. I found that the most efficient duty cycle over time was 1/11. The display is organized as 1728x11 for refreshing.

Anyway, here are some pictures from the first work party. I was able to get help on about 2000 of the LEDs, but they weren't much help. They soldered 300 LEDs in the time I could do a little over 1500. :)

Why did I do this? Well, back in 1998 I found 50,000 LEDs on ebay for $300. We all know how dangerous ebay is!!! ;)

http://www.stockly.com/images2/060108-1-RamWall_Rodger_John_2616.jpg

http://www.stockly.com/images2/060108-2-RamWall_Rodger_John_Back_2617.jpg

http://www.stockly.com/images2/060108-3-RamWall_John_2612.jpg

The first panel is done! :)

http://www.stockly.com/images2/060108-4-RamWall_Mine_is_done_2618.jpg

Some of you will find this funny... I saved ALL the LED legs just to see how much 38,016 led legs weigh... Well?

http://www.stockly.com/images2/060129-Weight_of_39584_Legs_2696.jpg

Awww, man...dat kicks azz! Keep on doin it...

--T

atari2600a
November 3rd, 2006, 12:21 PM
I did NOT just see a projector hooked up via composite!!!!! Come on, couldn't you at least use S-Video!?

Anyways, I hope to see a video of this online! :p You should take this to VCF 10.0!

ahm
November 3rd, 2006, 12:36 PM
Hmm, I was sorta hoping it was a bigger version of this (http://www.sparetimegizmos.com/Hardware/Life_Game.htm).

grant
November 3rd, 2006, 03:24 PM
I did NOT just see a projector hooked up via composite!!!!! Come on, couldn't you at least use S-Video!?

Anyways, I hope to see a video of this online! :p You should take this to VCF 10.0!

Its a .1" diameter cable used to connect a Chrysler 10 CD changer to the radio. About 25 feet long. :)

It kind of "softens" the video for large format viewing. ;) Everything is recorded on a first generation TIVO so I loose some anyway...

I have a 50 foot VGA cable for it, but my ATI card is acting up and won't display TV any more.

I found some .1" spaced 5x7 blocks on ebay for $.60 a piece, and some .2" spaced 8x8 blocks on ebay for $.63 a piece. In order to have a 320x240 pixel display (76,800 LEDs) I would need a little over 2194 5x7 blocks ($1320) or 1200 8x8 blocks ($756). Obviously the 8x8 is cheaper, but the 5x7 would have half the dot pitch. I've seen the original space invaders ported to the TRS-80. Aparently it was written for the 8080. An Altair running arcade space invaders on blinking leds... Its hard to want to finish this one! :)

I'm waiting to get a quote on RGB 8x8 blocks. Full color anyone? If you give a mouse a cookie...

I would never buy a flat panel LCD, yet somehow I'm able to justify this... :D

atari2600a
November 3rd, 2006, 05:38 PM
Yeah, but on composite, the chroma & luma signals overlap, causing defects! (ever notice that checkerboard effect on flat colors & that rainbow effect on detailed patterns?) You don't get this w/ S-Video, as the signals are sent over their own leads, leading to virtually perfect 480i! (...& then using VGA would result in crisp RGB 480p, but as the (I'm guessing MPEG2) compression would probably crap up the image, it wouldn't be that big of a deal...):p (Sorry, I get obsessive compulsive over things like this...)

If I recall correctly from an episode of How It's Made (Canadia's greatest show about manufacturing! :p), RGB LED jumbo screens are made up of smaller compositions (probably in the 30X30-ish range?), each with their own microcontroller?

grant
November 3rd, 2006, 06:40 PM
I'm a perfectionist too, but ever since December 2004 the projector has been on loan. I never thought it would be here this long. ;)

That's what I've been thinking about with the RGB display. Right now the 8x8 blocks are $240 in quantities of 120. I am waiting for a quote at the 1000-2000 level.

Even if I decided it was worth it, each display has to be updated close to 70Hz*256 shades = about 18kHz. This requires a microcontroller for small blocks of the display like you said. Things start to get expensive!

I would probably settle for a monochrome display. I would REALLY like to build one with the 5x7s, but they seem to cost too much. :( An 80x20 serial terminal made out of LEDs the size of a 50" TV would be cool. :)

My current display is the size of a couch, as you can see! :)

I will try to sell a kit when I am done. The display out of individual LEDs with ALL the PCBs (drivers and LED panels) as well as all the driver chips will be about $1200. Think anyone would bite? ;)

granzeier
July 27th, 2010, 06:36 PM
The 8086 has 29,000 transistors. The i386DX has 275,000 transistors. The Pentium MMX has 3.1 million transistors. The Pentium 4 as of year 2000 had 42 million transistors.

Assume you need at least one square centimeter per real transistor, at least if you wire them together manually. The 8086 would take almost 3 m (31 sq. ft). The Pentium 4 would take 4200 m, or roughly one acre. If you construct the CPU with more space, it will be larger.

Something that is not only possible, but has been done, is to recreate the 4004 in discrete transistors. Very cool project: http://www.4004.com/

Art

Marty
July 31st, 2010, 06:03 AM
Hi All;
I have built a small TTL Cpu, using state timing and not microcoded logic... As I have wantd to build an 8080 from TTL logic, and only recently found the 4004 35th Anniversary site and Schematics... I know what it would take based on what I have already done... But maybe for help in understanding, I might try building the 4004 from 7400 TTL logic, as a first step... I know that the 4004 is not compatible with the 8080, and as a second step then doing an 8008, before doing the 8080.. and I have a sim 4 -42 Intel board and prom board and Parallel board and an old Prolog M900 that usues the 4004. I have also build an 8008 microcumputer based on the Titus Mark 8, all wire wrapped, and it works... So It can be done... You have to keep at it and regularly do something on the project, before you forget what you have done.. When I first strated my simple Cpu, I started it , put it down , waited for a couple of months and had to start all over, tearing out what I had done, But when I worked on it on a regular basis, I got it done. and It also worked, after a couple of fixes... not understanding how something worked.... So a few of us are nuts, but I would like to do this, can't tell you why, but, just I want to...
THANK YOU Marty

k2x4b524[
August 10th, 2010, 07:04 PM
whatever bacame of the LED CPU?? Did it work or did it flop?

fin
August 25th, 2010, 08:43 PM
Ok, this goes back a few years in this thread, but being a thread of historical sorts, I hope that is ok.

I was talking to a colleague the other day, and told him one of those "aha!" moments for me was reading a Byte magazine article that described how to build a functional computer from a bunch of TTL stuff you could get mail order. For me, it was the point where the CPU was no longer a black box, and was just logic.

When I left for college, I dumped my stack of Bytes in the, well, dumpster. Foolish mistake, probably. But Googling around, I think the article might have been about the "EGO" (Allison, still out there?). All I remember was that it used a TTL RAM, might have had a four-bit word, or used a four-bit instruction, and had a "store negative" as the only store instruction in order to economize on the number of instructions.

Ring any bells? Flip any bits? Anyone know where I can find old Byte articles, other than the dumpster?

-Fin