OT-ish: Build-it stuff

[woodelf]

>> All my life I've been interested in clocks, and one day I intend to 
>> make a real mechanical clock. I'll have to get a lot better at 
>> machining first, but I still want to do it.
>>
>> Economically, this makes no sense at all. To make a clock from scratch 
>> probably needs \pounds 5000 to \pounds 10000 worth of tools (many of 
>> which i have, it's true, but I'd still need to get gear cutters, etc 
>> and they're not cheap). The metal to make said clock is around \pounds 
>> 100. And the result is a much worse timekeeper than a 5 quid quartz 
>> clock from the local household shop.
> 
> 
> Of course.  Nor does collecting old computers!  :>

Well they have had wooden clocks before that!
Even with a bell -- Monks had to get up early to pray you know!


>> Now lets get back to computers. The difference in performance between 
>> a homemade mechanical clock and a quartz clock is probably comparable 
>> to the difference in performance between a modern PC and a homebuilt 
>> processor. The tools needed to make a homebuilt processor are cheaper 
>> than those needed to make a clock, thouygh. And yet, I've not seen one 
>> book or magazine article on making a processor from scratch in the 
>> last 10 years. 

This is because of the PC -- everybody wants clones and anything
other than the 8088/8086 group of processers never impacted
the market well. Other than the Mac that shot itself in the foot
with only 128K memory total and a closed architecture.

> I think once the 2901 "fell from grace" (?), this became a thing
> of the past.  I've designed two processors "from scratch" (TTL
> with bipolar ROMs for the microcode store) and found it quite
> an interesting exercise.  Not just the "logic design" but
> actually thinking about what the instruction set should be
> for that particular application domain, etc.

I have never been able to find a 2901 processer instruction
set I liked that I created. The lack of prom programmer was
also a factor preventing me from building a 2901 cpu.
(Note a 12/24  cpu looks nice to do with bit slice)


> But, nowadays, I think it would be a lot less tedious if you
> could do it in a big FPGA using synthesis tools.  You could
> get better test vector coverage (and *generation* of those
> by the toolchain!) instead of having to design vectors from
> scratch and single step the microcode (hint:  fully synchronous
> designs really off big, here!)
> 
I find learning the software and the logic design languages
can be a real chore since you really may never know just what
the FPGA/CPLD is doing.

>> If some people think it's worth making a clock from scratch, why not a 
>> processor.
>>
>> I will end with a wuote from a book entitled 'Every Boy his own 
>> Mechanic', published about 90 years ago. It's the start of a chapter 
>> on a home-made telephone system
Umm Soup can's and string ... :)
PS. I had a print called "Before high tech there was LOW TECH"
and the artist of this art had fun with all the old and forgotton
techology for fictional early space craft.The pilot with no space
suit had just the WWI goggles and leather jacket but did have a
I think a scuba air tank. Wings looked like from the 15 century
art. But he did have a radio dish -- connected to a Cambell's
Soup Soup can with string.

I can't remember the artist how ever.