-
Hi guys,
I am new to electronics but I would like to learn, I found the forum through the EEV videos on youtube. As you all seem a friendly bunch I figure that I can get some guidance and tips with my project.
At the moment I am still in the research/planning phase. Before getting to far down the rabbit hole I wanted to seek some advice and guidance.
Specification: https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407449/#msg407449 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407449/#msg407449)
What about Mechanical Calculators? https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407457/#msg407457 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407457/#msg407457)
What about Analog Electrical Calculators/Computers? https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407459/#msg407459 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407459/#msg407459)
Why use ICs at all? https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407462/#msg407462 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407462/#msg407462)
Why bother learning about all these old obsolete technologies and devices? https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407469/#msg407469 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407469/#msg407469)
Target Machines/Technology's for "inspiration" not in any way a rip off...https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407493/#msg407493 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407493/#msg407493)
Component selection:
Parametron devices https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407478/#msg407478 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407478/#msg407478)
TTL familys: https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406856/#msg406856 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406856/#msg406856)
Input devices https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407501/#msg407501 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407501/#msg407501)
Non Binary memory storage https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407962/#msg407962 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407962/#msg407962)
Enclosure Design https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406246/#msg406246 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406246/#msg406246)
Display Module Analysis https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406145/#msg406145 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406145/#msg406145)
Keyboard Module Analysis https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406222/#msg406222 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg406222/#msg406222)
Memory Module Analysis https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407496/#msg407496 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407496/#msg407496)
Logic Module Analysis https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407838/#msg407838 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407838/#msg407838)
Background:
Where to get electronics/electrical kit in the Philippines? https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407486/#msg407486 (https://www.eevblog.com/forum/projects/vintage-calculator-project/msg407486/#msg407486)
A calculator project (bah you think, 1 calculator on a chip, a LCD and a matrix input and you are golden..) but wait... what fun would that be. No I envisaged a project that used "simple" logic components to create a basic 4 function device. Now at first I considered the ICs to be cheating but then i think to myself "you will learn way more revising the old school designs to implement basic logic ICs then soldering thousands of diodes/transistors/relays."
Its a bit ambitious but even the planning stages its great as now with the net there are hundreds of books/papers from this period available to study and its all there just waiting to be used.
There are a whole bunch of worthy projects for beginners to learn from but my thoughts are doing a "real" project that had to conform to a spec, be functional and be non-trivial. This presents lots of opportunity to learn a bunch of skills and get an appreciation for electronics that's not really possible with today's black box approach.
First design decision was to use the 74LS series of ICs, Widely available and they are cheap.
Know what you are thinking, how big will it be? Whats the spec?
I want to make a desk calculator machine, not using a fpga type solution it was always going to be pretty bulky. I wanted to make the system modular from the outset so I could easily correct stupid design decisions as I learned more, it would also enable easier access for testing.
Like a hopeless newb, I walked into the electronics supply store and bought a bag of 74ls chips, some passive components, and a bunch of 7 segment displays (Much as I love the vacuum tube displays, they are hard to find here - and thats surprising. Anyway I want to make this as power efficient as I can.)
This last cast off statement might need some additional explanation.
Yes - I wanted to see what difference 50 odd years of development of component technology has brought us. Would a calculator of only simple ICs and components be more efficient? and if so by what margin?
Like all pie in the sky projects, I had a bunch of ideas but not much know how. Archive.org was a good source of information on "retro" electronics books, so I have been looking through these and gradually trying to absorb the info. The various calculator museum sites on the net provided some great background information.
The main headaches I can foresee is there is a lot of "old school" magic at work with the early calculators, I start taking out the discreet components and replacing them with ICs its going to mess with control and timing of the device. So a whole bunch of issues await there. But as the programming teach used to tell us "I hope all of your programs fail. You will learn much.".
Sitting down and planning out the different steps for the project, what I wanted. I figured the first - easiest module to complete would be the display module. I had hoped to do some insane display of segments (you laugh but 37...) I settled on a 20 digit output. I could have gone for a LCD option but I wanted to try and stay true to the design ascetic as much as possible. But thats always an option for later, and with a modular design I can just swap it out. no big deal.
The next thing to think about was the internal data representation, how I was going to encode the signals. In the very early patents I had examined these devices were copies of mechanical adders, so they were using decimal encoding in there circuits. This seems problematic and really would be a pain to implement. So its internal data is stored in binary (or possibly signed.. I am still thinking about this..)
Power source, for the moment - as its a desk device mains is fine. Plus I had a bunch of adapters lying around the house that could be brought out of retirement and given some use.
At this stage it all seems like hokem Bs, What I hoped was that as I continued with the project I could get some tips and input from you guys.
-
Hi,
Yea I did consider the 74181, I was surprised to find just how much you can do with it. It would take a lot of the headaches out of the project. But on the other hand implementing the functions is a big part of the learning experience.
But point taken it would simplify the project a great deal.
-
74181 is a much easier way to go. we had to do it both ways. First with the ALU and then in just gates. Mercifully with just gates it only had to be a 2 digit system and we used a bank of 2 4 pos dip switches to input the bcd values directly. With the ALU most of us made keypads of sorts..
-
I recently read the book "But how do it know" by J. Clark Scott (ISBN 9780615303765), explaining how a computer works for "everyone" and plan to recommend it to my students.
He describes a simple 8-bit system and how everything in a modern computer goes back to NAND gates as building blocks.
Actually I am somewhat inspired to recreate his computer - at least in Logisim (http://ozark.hendrix.edu/~burch/logisim/).
And starting with a simulation is also what I would advise zimzom to do...
-
Yes - I wanted to see what difference 50 odd years of development of component technology has brought us. Would a calculator of only simple ICs and components be more efficient? and if so by what margin?
More efficient in what respect?
- Power consumption? - Not a chance!
- Speed? - Not a chance!
- Useability? - nja - No way
- Use of space? - haha
- Use of materials? - recycling the material: yes - otherwise: NO
- Understanding electronics? - YES
Definitely worth while looking into!
-
And starting with a simulation is also what I would advise zimzom to do...
Definitely, I reckon you could even do this in the free version of FPGA toolchains. I made some 8-bit calculators (add, subtract with positive and negative integers only) and a 4-bit micro as student projects, it's certainly doable but difficult to get right on the first attempt. I would definite verify all the logic before even thinking about PCBs or IC families...
-
Some things that might be of interest.
Look up the history of the HP 9100A/B calculators. This is a programmable scientific calculator built out of discrete components; it predates ICs. Here's Tom Osborne's story:
http://www.hp9825.com/html/osborne_s_story.html (http://www.hp9825.com/html/osborne_s_story.html)
RPN may be easier to implement.
You may want to operate on BCD digits rather than binary. Conversion of internal binary to displayable digits is not trivial using straight logic.
Here is an interesting calculator simulator that exposes the internal workings:
http://www.clivemaxfield.com/diycalculator/aboutdiy.shtml (http://www.clivemaxfield.com/diycalculator/aboutdiy.shtml)
Look up Finite State Machines and Algorithmic State Machines. If you're not building a general purpose computer, you're building a state machine of some sort. It will help you minimize the number of ICs.
After you've worked out your design, you'll want to prototype subsystems before committing to a pile of PCBs. You'll want to make sure your understanding of how the logic works corresponds to reality. Missing that a line is active high rather than active low can mess up a design and be difficult to track.
You'll probably want to get some type of logic analyzer to troubleshoot and verify the design.
You'll definitely learn a lot
-
RPN will almost definitely be easier to implement. enter first number into register, enter second, operate on them.
-
74181 is a much easier way to go. we had to do it both ways. First with the ALU and then in just gates. Mercifully with just gates it only had to be a 2 digit system and we used a bank of 2 4 pos dip switches to input the bcd values directly. With the ALU most of us made keypads of sorts..
Hi Calexian, Thanks for the tip. I am gravitating towards this aporoach - it would be a great way to benchmark, just swap out the relivant modules and run the test. Ahh if only there was no carry function it would be so much easyer lol.
I recently read the book "But how do it know" by J. Clark Scott (ISBN 9780615303765), explaining how a computer works for "everyone" and plan to recommend it to my students.
He describes a simple 8-bit system and how everything in a modern computer goes back to NAND gates as building blocks.
Actually I am somewhat inspired to recreate his computer - at least in Logisim (http://ozark.hendrix.edu/~burch/logisim/).
And starting with a simulation is also what I would advise zimzom to do...
<<merged...>>
<snip>
More efficient in what respect?
- Power consumption? - Not a chance!
- Speed? - Not a chance!
- Useability? - nja - No way
- Use of space? - haha
- Use of materials? - recycling the material: yes - otherwise: NO
- Understanding electronics? - YES
Definitely worth while looking into!
Hey uwezi, Cheers for the book recommendation - I will add it to the reading list. And also for the simulation software link. I will try and get a list of the books I found on archive, some of them were excellent in describing logic. About the component stability/speed. I was not being clear. What I meant to say was given the development and refinement of discreet components I wonder if there would be a performance diffrence compared to 50 years before. I know its never going to be as fast or reliable as a system on chip solution.
Definitely, I reckon you could even do this in the free version of FPGA toolchains. I made some 8-bit calculators (add, subtract with positive and negative integers only) and a 4-bit micro as student projects, it's certainly doable but difficult to get right on the first attempt. I would definite verify all the logic before even thinking about PCBs or IC families...
Hi Krivx, Once I get to the stage of testing the designs I will have to look into a computer based prototyping program.
Some things that might be of interest.
Look up the history of the HP 9100A/B calculators. This is a programmable scientific calculator built out of discrete components; it predates ICs. Here's Tom Osborne's story:
http://www.hp9825.com/html/osborne_s_story.html (http://www.hp9825.com/html/osborne_s_story.html)
RPN may be easier to implement.
You may want to operate on BCD digits rather than binary. Conversion of internal binary to displayable digits is not trivial using straight logic.
Here is an interesting calculator simulator that exposes the internal workings:
http://www.clivemaxfield.com/diycalculator/aboutdiy.shtml (http://www.clivemaxfield.com/diycalculator/aboutdiy.shtml)
Look up Finite State Machines and Algorithmic State Machines. If you're not building a general purpose computer, you're building a state machine of some sort. It will help you minimize the number of ICs.
After you've worked out your design, you'll want to prototype subsystems before committing to a pile of PCBs. You'll want to make sure your understanding of how the logic works corresponds to reality. Missing that a line is active high rather than active low can mess up a design and be difficult to track.
You'll probably want to get some type of logic analyzer to troubleshoot and verify the design.
You'll definitely learn a lot
Cheers for the links PaulAM, computer aided design is probably going to be one of the biggest differences between the old method of prototyping, with my budget its also going to be essential to be able to test these designs "free" on the computer before commiting to PCBS. A good logic probe is on my list of things to get.
RPN will almost definitely be easier to implement. enter first number into register, enter second, operate on them.
Hi Chen, I am still unsure what to do with the RPN. But sure it has many positive points that would greatly simplify the design of the calculator. And its HP's tech solution - I figure if HP used it it must have had many advantages.
Thank you everybody for your kind advice and tips. Will adapt the spec to reflect this..
immediate changes will be to go for BCD internal representation - thats true to the design methodology of the time and should make life a whole bunch easier.
I wanted to study the RPN more and look at the alternatives before plugging for this solution. But my hunch is that this will be the superior method.
If the costs are low, I will probably go for larger registers. If i plug for 64bit registers I can do a lot more later on down the road. Additionally I can cheat and split them later thus calming additional registers for use...
Additional registers would give more wiggle room for expansion of functions later on, more consideration needs to be given to this step as it might be easier just to get a big sram and go with that. This would probably be the most expansion friendly option in design terms. (a 256k/512 sram would give me all the memory I needed for registers and scratch memory.)
That said its hardly a vintage solution.
-
Let us assume that you are going to build a 4 function using RPN. I would type in the number, and press enter. This would put it into the first (X) "stack." Next, we type in our second number. This pushes the first number into the 2nd (Y) stack, and this one is now in the X stack. By only using "enter" as a separator between different entries, we reduce the complexity. Using algebraic, +/*- would all have to be seperators, and then a final "enter" would be needed to display the answer.
-
I recently read the book "But how do it know" by J. Clark Scott (ISBN 9780615303765), explaining how a computer works for "everyone" and plan to recommend it to my students.
He describes a simple 8-bit system and how everything in a modern computer goes back to NAND gates as building blocks.
I read that book too recently. Well written and entertaining.
-
http://files.righto.com/calculator/sinclair_scientific_simulator.html (http://files.righto.com/calculator/sinclair_scientific_simulator.html) might be worth a read too.
-
Update:
Research wise...
These were the videos that got me interested in the project in the first place, they were made a few years ago but I thought the documentary was good. Its in four parts. notice the word "inspiration" is used instead of "ripped off"
Birth of The Transistor: A video history of Japan's electronic industry. (Part 1) (https://www.youtube.com/watch?v=ihkRwArnc1k#)
Circuits in stone: A video history of Japan's electronic industry (Part 2) (https://www.youtube.com/watch?v=uGRNXmWng3M#)]
The Calculator Wars: A video history of Japan's electronic industry (Part 3) (https://www.youtube.com/watch?v=ansXGewduN4#)
The technological giant of the micron world: A video history of Japan's electronic industry (Part 4) (https://www.youtube.com/watch?v=G40YwOg0_B8#)
Sure this already got posted before...
The conditions that these guys did there initial research in is just mind-blowing. You can see the disk saw break and all that fella has is a pair of aviator goggles and a few sheets wrapped around himself to protect from the fallout! (part 1)
Also observe the solution to sourcing resistor parts. :-DD considering that they had no money to set up the labs its amazing just what was achieved.
Figured the best approach was to properly research as best I could the calculator technology of the 60s and 70s. Any schematics, technical manuals and patents I could find were downloaded. The various calculator museum webpages had a wealth of tear down information that has been very useful.
http://www.oldcalculatormuseum.com/articles.html (http://www.oldcalculatormuseum.com/articles.html)
Check out the Friden Competitive Analysis 1965 for a great overview of the market during that time.
The Sinclair calculator simulator link kindly provided above is great for messing with RPN. Cheers. I will have a more in-depth look later.
Outcome:
Drastically reduced the display chip count. Bit of a Homer Simpson moment. After banging on about the keyboard input and multiplexing it didn't occur to me that I could do the same thing with the display. Thanks to Friden and calculator museum! By multiplexing the display output I can use just one BCD / Binary decoder chip instead of one for each 7 segment display. Awesome.
(time for some calculator porn.)
(http://www.oldcalculatormuseum.com/c161int.jpg)
(http://www.oldcalculatormuseum.com/locick2.jpg)
This is what happens when you don't have display multiplexing.
check out the Cannon 161
http://www.oldcalculatormuseum.com/canon161.html (http://www.oldcalculatormuseum.com/canon161.html)
or the Wang LOCI-2
http://www.oldcalculatormuseum.com/wangloci.html (http://www.oldcalculatormuseum.com/wangloci.html)
now compare and contrast between the Cannon 161 and there later 130s series (still not a multiplexing system but by the S130 the displays are truly module components of the backplane).
(http://www.oldcalculatormuseum.com/can130sint.jpg)
Straight-away a drastic reduction in complexity and wiring has occurred. From a service perspective this is great. Ask yourself what calculator would you want to have to maintain on a daily basis - the 161 or the 130s?
That said reducing circuit complexity and chip count is not free - it brings additional restrictions on the display hardware. Namely for multiplexing to work properly you need it to be fast enough to fool the human eye into thinking that its displaying the figures all the time when in reality you are fooling the user by strobing the figures across your display matrix. If your hardware is to slow you get that annoying flicker.
My gut instinct is I can get the display fast enough without having to resort to direct driving and decoding (thus the design decision on the early calculators - they couldn't drive the displays fast enough so had to go for the more inefficient choice.)
How to get multiplexing to work:
General thoughts - assuming I could get the display circuit fast enough. I would imagine that a clock circuit would be needed for the timings of the display. (yes, this is an additional complexity. But consider that on a large display if you can implement multiplexing in the design you have just saved yourself 10-20 separate display circuits. So making multiplexing work has a huge component count and cost payoff.)
That said, you already have a display register on the system. all its going to do is cycle through the register and sending the output of the decoder to the seven segment display. So besides the timing its not going to present to much extra complexity. Once timing errors/bugs are squished you have dramatically reduced the potential for component failure. (the part cannot fail that is not there...)
One dark thought that occurs to me is perhaps this was the digit constraint in later models. there has to be a limit to the amount you can strobe without flicker. perhaps a 20 digit display is to ambitious... one solution would be to have a dual driver but that would make the timings more complicated.
My guess is that this plus the power consumption issues lead to the phasing out of Nixie tube based displays.
Practical stuff...
Went back to Raon to see exactly what Ics they had available, I only had time to visit one of there stores (the one shown is known for being one of the cheapest component stores for repairing stuff.) Transformers galore, switches, passive components galore. Sadly they didn't have a lot of the 'fancy' 7400 series ics.
no 74ls181 ics :palm: no sram :palm: best they had was a gaggle of 74LS173 bit registers. but that presents a problem.
to build a register with these 4 bit bad boys would require 64 total. Price is another factor. consider that these cost 40 pesos per ic (with a wholesale discount of 5% that comes to ... 38p per IC purchased over 10 peaces..).
So to get 4 registers of 64 bits I would be looking at a whopping 2432 peso/ $55.27.
That is a big chunk of change for something I *know* I can build cheaper and with more capacity if i can find the ics.
Would be silly to rush into a component buying bonanza before doing the design, but rational planning can sometimes go south when you see all the stuff there :-DD still fairly confident if I can find a 512k or 256k chip I would be all over that like white on rice.
But its always nice to have a few lying around for testing so i broke down and got 2.
While I was there I got some more 7 segment displays. got a few bit shifters for testing, Ran out of time so that was it.
With regards to the 'fancy' Ics and memory's I should be able to track them down, its just a case of putting in the time and searching for them in the market. sulit (the PI equivalent of ebay.) is another choice for getting parts, but its usually expensive.
Inquired about military/defense surplus available and there is nothing like that here. so that idea to component/case nirvana is out.
-
Before buying or building anything I would suggest creating and testing the circuit in something like SmartSim:
http://smartsim.org.uk/index.php?page=home (http://smartsim.org.uk/index.php?page=home)
-
Digi-Key stock the 74LS181 at $7.5/piece.
-
It's funny, I've wanted to do the same thing for a while. Thought, I was planning to build a 4 bit computer entirely using discreet transistors.
I see that two different logic simulators have been recomended, Smart Sim and Logisim. Which one is more user friendly, powerful, popular, etc?
-
It's funny, I've wanted to do the same thing for a while. Thought, I was planning to build a 4 bit computer entirely using discreet transistors.
Why not just use boisterous transistors? :-DD
-
It's funny, I've wanted to do the same thing for a while. Thought, I was planning to build a 4 bit computer entirely using discreet transistors.
I see that two different logic simulators have been recomended, Smart Sim and Logisim. Which one is more user friendly, powerful, popular, etc?
I've been wanting to do that with relays. Never came around to it. The continuos clacking of chewing through an operation.
-
Before buying or building anything I would suggest creating and testing the circuit in something like SmartSim:
http://smartsim.org.uk/index.php?page=home (http://smartsim.org.uk/index.php?page=home)
Cheers for that, will have a look tomorrow. Plan to use the computer designs later before I start soldering and ordering parts, but I wanted to do the analysis first then get it down on paper/blackboard first. once that is done then get it optimized and checked with the simulators. I get the feeling there are issues with depending to heavily on the machine simulation.
It's funny, I've wanted to do the same thing for a while. Thought, I was planning to build a 4 bit computer entirely using discreet transistors.
I see that two different logic simulators have been recomended, Smart Sim and Logisim. Which one is more user friendly, powerful, popular, etc?
I have yet to experiment with them, Spice gets mentioned a lot. Regarding the 4bit transistor technology computer. Sure that is possible but its going to be a lot of heat. But with careful selection of parts I don't see why it could not be done. At least its not going to run as hot as if you had used Valves, with the added bonus of you not losing a valve every few days of operation lol.
I've been wanting to do that with relays. Never came around to it. The continuos clacking of chewing through an operation.
This is a 500 relay 8 bit computer, see by applying modern principles of operation you can do interesting projects with old technologies.
R500/7T Relay Computer (https://www.youtube.com/watch?v=67jLony0mXg#)
Ear protection would negate that problem.
Shame the power consumption is not so easily solved. I pity your electric bill. :-DD
Anything possible given enough space and $$$. With the relays you will need high quality signal types. The mother-load would be to find a old telephone switching station. Then assuming you could run off with the relays you could probably do it. Zeus machines were built with this kind of technology.
Casio developed a relay only Calculator in the 50s, it had 342 relays was called the 14-A
(http://www.casio-calculator.com/Museum/Calculator/Z-001-/14-A/1.jpg)
-
I see that two different logic simulators have been recomended, Smart Sim and Logisim. Which one is more user friendly, powerful, popular, etc?
We use Logisim with our students - it's Java and runs essentially on any platform. I don't know Smart Sim, but Logisim is quite intuitive to use and really powerful with nested subcircuits. I just opened the homepage of smartsim in a different window and will have a look at it... http://smartsim.org.uk (http://smartsim.org.uk)
...downloaded...started...
smartsim? sorry, don't like it! I think the user interface of Logisim is much more intuitive!
I will stay - and recommend - Logisim: http://ozark.hendrix.edu/~burch/logisim/ (http://ozark.hendrix.edu/~burch/logisim/)
-
I don't know what state it is in now, but the last time I looked at Logisim it didn't support bidirectional pins on subcircuits. That means it's rather useless for simulating actual circuits beyond the simplest ones.
-
I don't know what state it is in now, but the last time I looked at Logisim it didn't support bidirectional pins on subcircuits. That means it's rather useless for simulating actual circuits beyond the simplest ones.
As far as I know you are right: a pin is either an input or an output for Logisim.
However, you can use separate pins and wire them together outside of the subcircuit. This is how early DRAMs (e.g. 4164) were constructed in reality.
I just finished the first part of the CPU from the book "But how do it know?" - The program counter increments itself by using the ALU to add 1 during each cycle. For those who already have the book it's on page 122. It is in the subcircuit "CPU".
This is not an efficient CPU or any attempt to create a real CPU from scratch, just the design from the book...
-
Thanks Uwezi, I will definitely look at logisim now.
I'm quite interested in how this project turns out.
-
Display Module Analysis.
:blah:
Will put the various display solutions I considered, and why I didn't use them.
Component choices:
A veritable smorgasbord of potential display technology's exist (or exited) to display data in some form to the user of the device.
Electro-mechanical
Not an option I really considered all that much - would require lots of additional tech to drive the motor and the gears. But sure if you are looking to create some sort of a steampunk project then its the solution for you.
When I get time I will have a look for some mechanical hybrids to put here.
Outcome: Rejected, due to the complexity, size, problems with sourcing parts and noise. But no doubt some hipster will reject this and implement it in there design lol.
light-bulbs/lamps:
Ok you can stop laughing. Want to give your project that Bell labs 1920s ascetic? want to heat your lab and work on your pet project? If so this is probably right up your street.
Issues: Vacuum tube displays largely replaced there use, then led/lcd components stole the show. But when I was growing up it was very common for equipment to have lamps for readouts and state outputs. Its cheaper and easier to service than your electromechanical stuff. and vacuum display solutions were (and are) expensive and delicate power hungry beasts. lamps while not as power efficient were a logical alternative for applications where you couldn't afford a fancy nixie/pixie display.
Outcome: Rejected, while a bank of flashing bulbs (IBM 701 style) could have worked power and usability prevailed.
filament displays
Bit obscure, these components were as if your light-bulb was talking to you with filament wires.
Issues: availability, power consumption.
Outcome: Rejected for the above reasons. But a solid alternative when thinking of using nixie tubes. These are a great *lower* power alternative to nixies.
vacuum tube technologies
nixie/pixie/panaplex
Three technologies, largely all have the same issue for me, power, cost and availability put these outside of the spec. While I grant you they look great for a vintage project they present a bunch of headaches in trying to source them here in the Philippines. as they were always an expensive display solution it never really found its way here in any great numbers so now a days they are very hard to come by. Added to this power consumption is a big issue here as power is expensive. I couldn't justify there use. Despite my concession that they would look awesome.
Issues: Power consumption, availability, reliability, fragility, cost, size, heat.
Outcome: for the above reasons, rejected.
Cathode Ray Tube
The crem de la crem of VDU components, while ultra high end calculators had this as there design solution I didn't see the need when power consumption was a spec issue to have this. Added to this the problems of finding a crt (minature tv????) that would not be overkill ment this got rejected.
Issues:power consumption, availability, tuning concerns.
Outcome: Rejected, but I grant you for that "calculator where price was no object" 70s vibe it would make sense. this would be a big power hungry component.
LED technology
These fall into two separate categories separate led's and segment displays.
Separate leds:
by the mid 80s I remember there was almost overnight a switch from using panel bulbs to led's I don't know when it happened but there must have been a price/reliability development in the manufacture of these components that meant that everyone and his dog used them in there projects to commercial applications. They have many advantages over competing display technologies at the time. Low power, low heat, small size and reliability. All good stuff.
Issues: none really, other then them being very small and easy to loose. Early models were expensive but then you can say that about any emergent technology.
Outcome: Will use them for status updates (on, register in use, probably will use it to read the data-bus also for debugging.)
7 segment display
Once a high priced component that said "I don't care about project budgets - this is the 1970s and I want my front panel to look like its off the set of starwars.." I can remember my father coming home with an alarm-clock radio and we were all stunned it didn't have that mechanical Rolodex mechanism that many had at the time but it was all space age and buck Rodgers cool. 7 SEGMENTS! But these things were expensive exotica for a while, seen in high end application displays, military/government projects where cost was no object ect... Now they are cheap but a bit on the power hungry side. Calculators quickly switched to LCD modules because of this. Its a shame as 7 segment tech seems to have been a red-headed stepchild technology, to expensive to compete with the nixie displays and by the time they were becoming commercially viable for mass market the LCD breakthrough came and caused an almost overnight shift to this low power technology.
Issues:
Power, compared to the low power lcd modules they are expensive to run and they take up a fair amount of space. But I figure these will be a bit more robust.
Outcome:
Selected! Sure I know I could have gone for a lcd board, but I wanted to keep this project sufficiently retro without the nixie/pixie/panaplex components. A small voice keeps whispering "should have got more segments, you can't do proper alphanumeric.." But I can live without that with a calculator project. Also more segments would create more problems for the display driver, and it would have been slightly more expensive.
LCD
Born to late, sure this would have been the ideal component solution for the calculator project, low power, easy integration, high readability, small package size. But as its a hobby project I decided this was to recent a technology to implement in my project. In every area LCD tech beats segment displays hands down. This would be the obvious goto for modern low cost devices today.
Issues: none, To new for the project I guess.
Outcome: Rejected - due to being outside of the technology range of the spec.
Display multiplexer/strobing circuit
Found a source for 8052 chips (the very same logic that takes care of your sd cards) - this would be more than enough to take care of the display and cheap at 80p (just under $2) will only use this if I cannot multiplex the display any other way.
-
Keyboard Module Analysis
:blah:
This is the primary input module for the calculator system. On a single board the plan is to have..
1. A keypad input, (matrix scanned, clock timed.) numeric 0-9, the function keys (+-x/), enter key, on/off, memory set, memory clear, decimal point, reset.
On the actual keypad:
2. a debouncing circuit (again this would be a timing function, or use a all in one IC package.)
3. over voltage detection
4. registers used indicator (led pilot lights)
5. a data bus
6. input buffer
7. BCD encoder
Talking points:
De-bounce circuit:
(http://i60.tinypic.com/xo2sgg.jpg)
From the TI applications book...
Just encase I don't obtain a keypad with the circuit built into the package.
At this stage I am not sure what to do with the keypad the usual method seems to have used a series of reed switches, but later that moved to a membrane design. I might just be lazy and opt for an off the shelf keypad. This would reduce the complexity of the build significantly.
One interesting point is you would assume that a debouncing IC would have been on everyone's Christmas list but there are surprisingly few available and the ones that are seem expensive for what they do.
Will have a look at the various discreet debouncing solutions, then the ICs.
Have found a few membrane types available here for nothing (around $4).But these appear to be pretty flimsy. Nothing like the heavy duty keys you get on the old calculators.
4x4 membrane (160p / $4)
(http://9d29598bcd2935308956-991cdbf325666ea20f9ad28fca148ca0.r48.cf6.rackcdn.com/20140127192220-05a7cdd3b8f5d794421deb5247e92a51.jpg)
4x3 membrane (120p / $3)
(http://cdn2.sulitstatic.com/server2/images/2013/1002/055308402_81f3bf4f39995a509e327b8bdb3dbf1e42088e21.jpg)
One big plus is they both have the debounce circuit built in so that would save on component count and cost. hmmmm.....
-
Enclosure design.
:blah:
(http://www.hp9825.com/assets/images/Balsa_Wood_Case_and_Key_Labels.jpg)
Got the idea for the prototype enclosure reading the article that was recommended earlier about Tom Osborne and his work. (great article by the way! Thanks PaulAM! :-+)
I have a bunch of spare wood that is just gathering dust under the stairs, enough to make something like this.... Doesn't have to be uber pretty just work till I get the rest of the system done. While sticking to the separate logic / input / PSU seems to be the agreed development method for these early systems. According to my reading this was how it was done with the Wang models also. (with the 360e it was even a feature of the production model.)
(http://www.hp9825.com/assets/images/Ready_for_a_Showing_-_1965.jpg)
See the psu, the logic and the display/input are all separate cases.
The article I was talking about is here
http://www.hp9825.com/html/osborne_s_story.html (http://www.hp9825.com/html/osborne_s_story.html)
(its also where the images in this post are from.)
-
TTL family choices:
:blah:
Looking back to the spec for the project, one of the goals was to make a calculator using basic TTL devices. So without going to a system on a chip what can be done and what choices are available now.
Military/Civilian:
Spec is civilian temp range, so without increasing the cost significantly and finding hard to source components unless a special deal came up will just stick with the civ' rated stuff. Plus I already had a small number of 74LSxx ics kicking about so it would be nice to be able to use them.
Outcome: Civilian spec.
where you come from, whats your family?
(http://i58.tinypic.com/sma9hh.jpg)
Nothing is free, what he gives with one hand he takes away with the other. I am sliding towards a compromise, while the low power devices are nice from the power consumption perspective they are 10x slower than the speed demon Schottky. but on the other hand 3ns vs 33ns who cares right? wrong.
I like the speed of the Schottky, but I wish I had some of the low power features also.. just not so slow... but wait whats that.. there are Low power Schottky... mmmmm liking that speed/power tradeoff.
Issues: Been trying to get an up to date version of the design documents from the TI website but either its my browser, gremlins or who knows. But I just get blank pages trying to get to the TI docs page |O. anyone have a link to the current design pdfs?
Outcome: If i can source the parts I would like to go with the Schottky low power ics. This will require more thought and planning before committing to an order.
General project updates:
Bit of a slow day, but on the plus side finally understand the difference between open collector ICs and totem pole. Spent time looking at ic familys and specs, the design book has tons of advice and best practice. I now see the true extent of the challenge ahead. :phew: But even if it breaks me the amount I will have had to learn will be well worth it. Important thing I think is to accept its going to take a long time, but anything good always does. Even in the BS analysis research portion none of it is a waste of time, can all be used elsewhere afterwards - even the rejected technologies/electronic archeology, perhaps its an illness but its cool to look at these crusty old tech solutions you get the feeling you might be the first person to have read this in years. Nice thing is doing it this way I have a clear set of things I need to learn.
Bibliography:
(https://ia601708.us.archive.org/27/items/SemiconductorCircuitDesignVolumeI/TexasInstruments-SemiconductorCircuitDesignVolumeI.gif)
https://archive.org/details/SemiconductorCircuitDesignVolumeI (https://archive.org/details/SemiconductorCircuitDesignVolumeI)
-
Project Specification:
:blah:
(I wanted to have some figures to work with, sure they are pretty forgiving.. but its better to have something than nothing.)
will operate in the following temp range 0c - 70c conditions.
Mains filtering + input filtering + output filtering
<100W
4 registers 32 64 bit.
a modular system, with a back-plane design so the main unit can sit under the desk with a keyboard and display on the desk (think Wang 360e).
4 functions (add, subtract, divide, multiply)
clear, on/off.
make use of digital electronics, use the 74ls family of logic ics.
While not using a calculator on a chip solution make the unit as power efficient as possible.
Use the least amount of components without impacting on usability or functionality of the system.
make a project that's easy to fix, debug, and upgrade as I learn.
It should cost less than p20,000 (around $500) parts. that's the maximum budget for all components.
it should be able to be moved by hand (less than 50kg)
it should be able to be run from the mains.
there should be a decimal input and output.
the system should function using binary BCD..
So with these broad strokes, I figure that's enough to go ahead with.
-
Why not build a mechanical calculator?
:blah:
That's a fair question.
The tooling required to be able to build one of these devices is way beyond anything I could hope to do. While there is extensive documentation available. Its just something that never interested me. Where could you apply the skills after such a project and think of the costs and specialist equipment.
When I get time I will add more to this, but there is quite a body of existing work on mechanical calculators and computers if you find it interesting. Its good for re-enforcing concepts that are found in the electronic side of life. There was a book from the 20s that had hundreds of mechanical calculators and specialist devices with writeup. Will have to find the link again.
U.S. Navy Training Film - Basic Mechanisms in Fire Control Computers (MN-6783a/b, 1953) (https://www.youtube.com/watch?v=bOwKh9XLO1c#)
Why its important to examine mechanical calculators.
For a start Babbage and his ideas cover the whole field of calculating machines, I didn't really understand just what he did in designing the difference engine. Looking at the early relay and valve machines his ideas and concepts are everywhere. To see what ideas/methods were left on the table and why is not a wasted effort. never know whats feasible or possible today compared to the early 30s 40s...
-
What about analog electrical calculators?
:blah:
Because I am lazy, it would be a fascinating project to implement. But my gut tells me it would be far far harder to do, I like digital because analog is hard. But sure in the future why not. I have seen the analog computer project on this board and it looked great. I guess it comes down to lack of documentation and I am not confident in my analog 'skilz' to get me through.
Update: To give each possibility a fair shake, I investigated the possibility of analog computation for simple maths, however...
(http://i58.tinypic.com/30c8wut.jpg)
So my hunch was right... but if it were to have scientific style functions where I didn't require beyond 1% :-+
Not to say Analog computing is better or worse compared to digital generally as a tool. They all have there place and there advantages and disadvantages should determine the selection based on that application.
The analog museum site was fantastic, That has to be the single best reference on the internet about analog computing and applications. They have every book you could ever need to understand the subject. am busy converting the pdfs to djvu so I can read them on my old machine.
Thing is though... I look at this and I think... thats one component... even high spec its going to be cheap...
multiplication
(http://i57.tinypic.com/5lrs6a.jpg)
This requires further investigation. If the parts could be found that had the precision necessary it would be an interesting implementation. First thought is in theory it could work but the precision of the components would have to be very good. I just don't know enough about this to comment if it would be feasible to produce a 4 function calculator in analog that would have acceptable an error level.
On further investigation, Ics such as 7497 do this. interesting. So my hunch is there are a bunch of devices we would not commonly think of as analog/digital hybrid computers that are..
(http://i62.tinypic.com/2q35bx4.jpg)
its a shame that the references in the TI applications book are so expensive, no doubt they would have been an interesting read. Even today they are asking $20-$50 dollars just to read a paper from the 60s.... Seems a counter intuitive way to get people to sign up for a subscription to your journal by first trying to gouge them for cash to read an article written over half a century ago....
-
Why use ICs?
If you are not going to go down the system of a chip route why not be a total purist to vintage design?
While its true its not *really* going to be a totally legit vintage design, I had to compromise somewhere and there was an issue with what I was going to get out of the project. I wanted to learn digital logic well, have a system that was going to be 'easier' to debug than a pure discreet calculator. Looking at even the basic plans for these early calculators you have thousands of components, multiple boards. Especially as a beginner that's a big ask soldering thousands of components perfect first time. Plus a big part of this was to learn about circuit design, so I wanted to see if i could take an old schematic and update it. Finally component cost was another factor, power issues and latency.
Pictures probably explain better than 1000 words (and I will add them later...)
-
Why bother learning about all these old obsolete technologies and devices?
:blah:
Its surprising what these dusty old papers/journals/books have to offer. I noticed that many times the explanations were way more straightforward like the authors were not trying to sell some sort of certification or speaking in jargon. I also found that the examples were much more "real world" and not lost in a sea of abstract maths equations. To my mind electronics is a practical subject that seems to have been hijacked by mathematicians and standards/cert guys. I often wonder how many people put the electronics books back on the shelf in disgust when the first chapter of many of the books is all about Maxwell equations. I can't think of a way to kill peoples interest in a subject than from the jump start throwing maths and physics at them. What harm would it do to have a practical real world example like making a battery/bulb setup work with a switch. Get people doing something and learning and gaining interest first. Then hit them with the theory.
These technologies though obsolete might have some interesting new take on a subject that you found difficult before, and sometimes they are a missing peace in your understanding of a later technology. Its all a progression I guess. Understanding transistors is way easier once you have a working understanding of valve technology. But in the end its good to see the way people approach a problem and what there solutions were. People don't change much and the problem solving approach is going to be similar. What its good for is reminding us that no technology is a panicia. there is always some disruptive new discovery on the horizon that will relegate much of the current day kit into the dustbin of history. I think it gives you an appreciation of the need to be flexible and open to new ideas.
And its curious how these ideas seem to go in and out of fashion, look at nano tech, that's seeing research that was first developed for magnet line tech in the 50s. So you never know when this stuff will become the "new now."
-
Parametron devices
:blah:
Think of a oscillating magnetic encoder. It takes up less space than your magnetic line. Bit obscure but here is the patent US 3553476 A. Oi Electric used them. But power consumption seems to have been a big issue despite stability being good compared to the early transistors. (its easy to forget that the average daily yield for a transistor production line at the beginning of commercial production was in the 0.3 range!) Again imagine the cost and reliability issues and its little wonder why these power inefficient solutions were found that were later discarded when transistor reliability and price stabilized.
Much like the American calculators of this period they leveraged design innovations from larger computing projects of the previous decade. Parametron based computers were in use in Japan during the 50s.
(http://www.thocp.net/hardware/pictures/parametron.jpg)
The inventor with his creation..
more info on how this technology worked...
http://www.thocp.net/hardware/parametron.htm (http://www.thocp.net/hardware/parametron.htm)
Computer museum writeup.
http://museum.ipsj.or.jp/en/computer/dawn/0016.html (http://museum.ipsj.or.jp/en/computer/dawn/0016.html)
-
Where to get Electronics/Electrical kit in the Philippines
:blah:
One big advantage here is in the Philippines there is a fair amount of junk that is just left here as old/new stock. These guys never threw anything away. And its pretty cheap.
Deco (Dee Hwa Liong Electronics) is your budget choice, they generally will stock cheap grey market components as well as lower end and recycled parts. Arduino stuff, and basic parts.
(http://3.bp.blogspot.com/-F-zqo7-8lPw/UptliI_DNjI/AAAAAAAADdo/YrZbnKKPu28/s640/DEECO.jpg)
Raon, Quapo Manila, The Akihabara of the Philippines.
Newport Electronics Trading, If you have a DVD player or TV ect you need fixed. Odds are these guys will have all of the components needed for the job. They have a wider selection of Ics - about the same price of Deco give or take a few pesos. They also have a website and a price list :-+ You would be surprised to find that not many places actually have a website, let alone a price list you could browse.
General comments on Raon:
I have found in the past that fair pricing and customer service are going to be way way down the list of priorities here. Essential its a burn once model of business. Your average place here will try and take you for as much cash as possible and not worry about the consequences. There is very little concept of repeat business example, the lack of price lists - why? I can believe a big supplier will be shifting thousands of items of stock to big buyers but the little stalls you find in Raon don't roll like that. Generally its just a way they can gouge. Especially as a foreigner walking into these places you need to haggle hard and you will still get screwed on price. My approach is to get a local person to go in after you found what you want and get the real price for stuff.
One of the biggest gotcha's for all of these places is to check that they have applied the volume discount agreed on the bill. This will not happen unless you check and insist. You should get a volume discount on the bigger stuff at around 10 peaces (you are looking at a saving of 5% in most cases). All places will offer volume discounts or ask for "wholesale" this means 10+ items here... its like getting blood out of a stone but it can be done - Just takes perseverance. Check your change also - you would be surprised how often this will be in error (incidentally always in there favor..) .
One other piece of advice, you will get angry doing any sort of financial transaction here, its important not to show your anger. Its not America/Europe. Even if you know they are trying to rip you off you cannot lose your temper . As a foreigner all that will happen if you do is they will call the cops and you will pay big to get out. Silence then asking for the thing you want again will get you much further. The honey/vinegar thing comes to mind in short "don't act like a banker".
If you are really desperate you could order from RS or Farnel but the pricing compared to what you can get in the market is around 10-100x more. If you were going to do a big order there is a point at around $500 total that its cheaper just to go to China and get your stuff and come home with it in a suitcase. (hint, you do not want to be dealing with shipping and customs here...)
Pretty much every knock off you can think of is available here, or old stock. Unfortunately like most places you have to buy what you need when its is there as stock availability is a problem even here. Stuff you would expect to find but is not here are items like fpga boards, (besides arduino style kits.) ect high end memory components and precision test equipment are also a pain to obtain. That's where its a big departure from the electronics markets of China and Japan.
(http://media.tumblr.com/tumblr_lro5tbwftm1qa2m3l.jpg)
As you can see it gets pretty busy, tip to travelers watch your belongings in the crowds.
-
Target Machines, "inspiration" sources
:blah:
To get an idea of what I am talking about, perhaps its best to look at the types of devices I am trying to emulate. Perhaps some of you have never seen the types of beasts I am talking about.
Take this for example, at the time this would have been a state of the art calculation device. Produced by Wang.
(http://www.oldcalculatormuseum.com/wang360e.jpg)
http://www.oldcalculatormuseum.com/wang360e.html (http://www.oldcalculatormuseum.com/wang360e.html)
Now In no way can I hope to replicate this beast of modular design. But probably a more realistic target machine spec would be something like the early 4 function models produced by Sharp or Casio.
(http://www.oldcalculatormuseum.com/c121a.jpg)
-
Memory Module Analysis
:blah:
Registers:
The Internal registers presented a few issues - early machines sometimes used a wire to store register data, While keeping component numbers down was an overall goal (without going down the fpga rabbit hole.) components are so cheap and reliable now the design decision though at the time of these old machines was totally rational.
(http://www.vintagecalculators.com/assets/images/Epic3000_5.JPG)
a delay line from a munroe epic 3000
Now this does not make a lot of sense. As an aside, the kind of know how black magic needed to get one of these lines to work is just something else. Its funny how everything is relative. The guys at the time thought it was a much better solution than the other "solution" namely mercury tubes.
You have high end models like the Wang machines using magnetic core technology but that would have been top of the line/state of the art and considering probably Wang was the only one with the experience to use it properly I will steer clear of that. But sure depending on the spec it would have been an option. (but then with mag storage at that time I guess you are looking at either IBM or Wang.) I seem to remember there were really weird timing and voltage solutions needed to implement from what I had read. My guess is it would be a project killer even getting that to work.
(http://www.vintagecalculators.com/assets/images/MagneticCore_1.JPG)
A magnetic core storage array from a Casio AL-1000.
(http://i62.tinypic.com/121zt3a.jpg)
a diagram of magnetic core from a ACE report
These are two videos I found to be excellent in explaining magnetic core memory and its application. Its well worth a look if like me you didn't quite get it. I know there are some other videos floating around YouTube about this subject but none are quite as clear as this.
DoD - Magnetic Cores -Part I: Properties (https://www.youtube.com/watch?v=BtyWWswVQVg#)
DoD - Magnetic Cores -Part II: Basic Circuits (https://www.youtube.com/watch?v=brRYjAAc_qU#)
This is a good summary of the technology used in these two approaches:
http://www.vintagecalculators.com/html/calculator_memory_technologies.html (http://www.vintagecalculators.com/html/calculator_memory_technologies.html)
Its not such a mental approach if you have no access to silicone chip technology, or it was out of your project spec range/price/performance (that would explain its use in the soviet equipment well past the 1950s.), where you would get the ferrite cores today with the "square" hysteresis profile I don't know. By chance I was looking at ebay and saw a soviet era memory core for sale, sadly it was a bit pricey, but the hi resolution photographs were free ^-^. There were a number of different implementations of magnetic storage tech. You have the ram type like the ones above but also there was a rom version where during manufacture the ferrites were wound in a way that they always biased one way or the other. I believe this is what was used in the Nasa system. Thus the "knitting" analogy. Its an interesting area and would be a major project in itself to re-create one of these core memory modules. but without a bucket of square ferrite I doubt this could be done - even with the patents of a saint and the motor skills of a artist lol. No way would I be attempting the rom lines.
Vacuum tubes well if power consumption was not an issue - then sure its possible, but there are reliability and power on state issues that it would introduce that would be a real pain in the arse to fix.
Update: There was a third technology - Parametron devices.
Relays - echos of Zeus... sure these were used in very early models but the power consumption of the relay solution is insane. As far as I can tell it would work but the final unit would be huge, and be economically nonviable to run (even by 1960s cheap energy prices.) so this one got left on the drawing board.
update: (clack clack) Having thought more about this more investigation of relay computers/calculators to give them a fair shake. Have the ASCC/Mk1 user-manual. Zuse's Z1/Z4 also really deserve greater study. Will have a look this week.
Having rejected the original register storage solution, I had to find an alternative.
1. A bank of flip flops ics. - possible and probably the "high end" vintage solution back in the day.
2. something like a register Ic. - while its really pushing the brief of "simple logic only" it saves a ton of work, (a 4 register calculator that could do a decent amount of calculations would need many many flip flops to store the data.)
I figured go for the register ic unit, if I change my mind I can always just do another module. From a latency perspective I figure the register ic will be faster + more reliable than a bunch of flip flops.
* Sure I know it can probably be achieved in 3 registers but I wanted to keep thinks simple at the start.
further update:
Williams Valves
A non linear access memory device, called a Williams Valve could also be used as a type of random access memory. That said I do not believe its use outside of the Soviet Union was widespread. However its an interesting technology.
Williams' Valve (https://www.youtube.com/watch?v=b4r68iJjukU#ws)
-
Input Devices
I wanted a vintage style calculator but I can't think of why I would want the replication of keys like you have with the very early machines. so a regular key matrix will be a good compromise.
(http://i60.tinypic.com/2ihxic9.jpg)
A picture is worth a thousand words... Impressive though this is.
(http://www.oldcalculatormuseum.com/x-sharpcs10a.jpg)
that's a lot of reed switches, ok there is no matrix giggery pokery to get the input to work but still... in my view that's still only emulating the keyboard layout of the mechanical adder designs.
see what I mean? this is a typical mechanical calculator... notice the similarity. Like many devices the initial development was just a mirror of the mechanical device.
(http://www.oldcalculatormuseum.com/fridenstw.jpg)
much better just to use your (today) standard layout..
(http://www.oldcalculatormuseum.com/c121akb.jpg)
Furthermore, from a bean-counter cost/benefit standpoint. The more parts we have, especially mechanical ones that are going to be interacting with the user of a regular basis are probably going to require much more servicing down the road. Consider the worse case scenario of a 300lb gorilla of a employee or worlds strongest man having a bad day and using the input device. The part might not fail in the first year or two but thinking worst case its far more likely that the user is going to damage the keypad/keyboard. Now if that's got hundreds of components, mechanical switching and locking that's a expensive bundle of parts. You do not want to be replacing these things as part of a service agreement with a customer. Far better then to go with the cheap charlie option of a matrix keypad, if it breaks, you also have the advantage of not having to test a whole bunch of components/bits on the keyboard. so its a double whammy of cost saving on the parts and also cost saving on employee time spent fixing the unit. Now this is just my BS hobby project, but consider if this was a production device that got sold to even a trivial number of customers - say 200 units total. Imagine having to go out and fix these things, for the life of the service agreement. That's a huge cost that can potential destroy my margins. My gut instinct is that this is also what kept the early calculator designs high priced was they had to factor in the expensive input devices and there support.
Consider also the space and weight saving made going to the matrix keypad. In the spec I wanted to keep the unit weight down to something a reasonably fit person could move around if motivated. (I know the regulation guys are going to say something like 20k...) but 50kg is manageable for most I figure. That's one big sack of rice. getting rid of a keyboard mechanism made out of steel, all the keys ect that is going to drastically reduce the overall weight of the unit.
Another issue that then came up in the design was De-bouncing. Many early calculators have mechanical safeguards to prevent this but alas with today's keypads you don't have that luxury. getting the key to stop repeating itself in error was needed. So I got to learn all about the magic of de-bouncing circuits. (but I have seen keypads that did this for you... but again where is the fun in a black box?)
searching through the patents, you see that even simple things you take for granted in modern calculators took extensive research and development. Probably the most useful thing to come out of this so far is appreciating the years of work that went into making such a ubiquitous throwaway item what it is. For example, having a button that functioned both as an on and a off switch.
There are a whole bunch of other input options that I have yet to cover, but will come to soon enough.
Paper tape, punch card, magnetic card, magnetic tape, switches, rotary pots ect... all in good time.
-
Books/Journals/Papers:
:blah:
These are some of the books I found to be interesting useful.
Theory, calculation & digital computers
Basic Theory Of Digital Computers Vol1 Jan57 (https://archive.org/details/bitsavers_ibmsageBasmputersVol1Jan57_9329879)
Basic Theory Of Digital Computers Vol2 Jan57 (https://archive.org/details/bitsavers_ibmsageBasmputersVol2Jan57_13954615)
Basic Theory Of Digital Computers Vol3 Jan57 (https://archive.org/details/bitsavers_ibmsageBasmputersVol3Jan57_9609532)
These three volumes go into the theory behind many of the technologies that were available for computation at the time. It has some basic maths but it explains the 'wank words' that other books use as a barrier to entry. Its a great source of information, jammed packed full of worked clearly explained examples. This series of books takes a person from 0 to having a pretty good understanding of how the systems worked. It seems to have been created to bring personnel to the sage program up to speed as quickly as possible. But as I have said its biggest value is that it assumes the reader has started from square one and explains every point and concept clearly with examples. A lot of later books do not do this.
Project budgeting, logic design, project planning:
Thomas R. Blakeslee Digital Design with Standard MSI & LSI ( 1st.ed.) John Wiley & Sons 1975 (https://archive.org/details/DigitalDesignWithStandardMsiLsi)
A kick ass book, best of all its full of examples and advice.
Components
Diodes:
Sylvester P. Gentile Tunnel Diodes ( Basic theory & application of ) Van Nostrand 1962 (https://archive.org/details/TunnelDiodes)
Magnetic technologys
Ferrite core:
J. Smit & H.P.J. Wijn Ferrites Philips Technical Library 1959 (https://archive.org/details/Ferrites)
Radio
F. Langford-Smith Radio Designers Handbook Iliffe books 4th ed 1957 (https://archive.org/details/RadioDesignersHandbook)
Valve radio theory, 1500 pages of designs, articles and theory.
-
Logic/Function module Analysis
:blah:
Functions:
To make this project somewhat manageable, at first I just want to implement the four basic functions.
- Addition
- subtraction
- division
- multiplication
3 will be trivial to implement and division is going to be a bit of a pig, but its possible. I figure with the modular approach I can design the boards separately as I go and do the testing. building up slowly. Plus who knows perhaps the mysterious square root function could be added - but that's later.
(http://i62.tinypic.com/13yowfo.jpg)
Booth - Automatic Digital Calculators 2nd Ed.
suspicions confirmed...
Decimal point placement was a much bigger design consideration than what you would think. without the use of a calculator on a chip its actually quite a non-trivial problem having variable floating point placement. A number of solutions were proposed by early manufacturing designs
a. do not deal with fractions.
This was the KISS option, and has merit. But we are victims of today's expectations and to be honest I would feel I cheated myself if I couldn't use floating points.
b. have a fixed floating point placement
Probably the solution I will implement for the time being, this is much easier to implement than having a variable one. But this is already something I want to learn how to do properly. Given time I will implement the solution.
c. non-fixed floating point placement
mmmmmm.... nice but also hard. Various sources have warned that this is a major headache in early calculator models and represented significant development to get this to work. But on the other hand all the research has been done already so....
moving on..
leading zeros, who needs em?
It might come as a surprise (well it did to me..) that this was also something that came later in the development of the calculator. I like my machine to be vintage but not that vintage. I have a sharp patent here discussing the problem so following its approach I should be able to deal with this.
Clear function & checking for invalid register states.
The dreaded overflow, underflow, power on garbage...
I figure some signal hocus pocus should fix this. Something to work on in the later design phase.
-
I would just try and make it all digital. Forget about magnetic core memory. I would just use logic gates, transistors, or better yet, relays.
-
I would just try and make it all digital. Forget about magnetic core memory. I would just use logic gates, transistors, or better yet, relays.
Hey echen,
With relays its mostly a speed/cost/size issue. But they are interesting machines that's for sure. lol to think that zuse build one in his apartment, with the paper tape input. There is a photo someplace.
(http://www.weller.to/com/img/zuse_z1-bau2.jpg)
Tunnel diodes were always being touted in the books/journals of the 50s/60s as some sort of new wonder technology that was going to revolutionize logic. But where are they now lol... thats going to be an interesting branch of investigation.
(http://i58.tinypic.com/kbalw2.jpg)
In there turn I will look at them all, but as a first build I would like to just get something working then look at more complex solutions so doing the logic with Ics and a sprinkling of transistors/diodes would be the way to go i think.
-
Non Binary Memory Storage
:blah:
This could have been part of the memory module analysis but its so far out there I didn't want it messing up that post lol.
Binary took a while to catch on, a number of early electronic/electric machines were developed using other bases than 2. Perhaps you might come across one of these parts in your travels.
These sound like star trek tech - marketing gold for your next kickstarter.
Dekatron
A valve based technology, high power. For your government funded project with unlimited funding.
Advantage with these valves is you could actually see what was going on inside the register. Appear to have been available in multiple configurations/packages and number bases (8/10/12)
There is some confusion as the Ericsson data sheet seems to be unavailable for the referenced component GC10A, and on some websites I have seen it referenced as being octal number base, but the book seems to think it was base ten... hmmmm.....
(http://i58.tinypic.com/2h3z9ea.jpg)
ok straight off the bat, 350v rail...
(http://www.tnmoc.org/sites/default/files/styles/lightbox/public/7%20Chubb-Small.jpg?itok=NIbqKy-F)
This is the Harwell Dekatron.
STCtron
Looks to me to be a faster version of a dekatron.
Trochotron
Whenever a book tells you that something is so complicated the author does not want to attempt to explain its function probably a sign its not going to be a widely spread widget. (considering that Booth was one of the founders of modern computing, probably legit..)
But whenever someone tells you not to touch the big red button, there is a little voice that says "ah go on give it a go.."
Trochotron (https://www.youtube.com/watch?v=rmnFhOajfmw#ws)
Note from the video there were binary Trochotrons produced, my guess is the decimal Trochotron was of the ring type.
-
Dekatron
A valve based technology, high power. For your government funded project with unlimited funding.
Possibly quite extensively used in the East. At least there is surplus available on eBay, e.g.:
http://www.ebay.com/itm/260941919642 (http://www.ebay.com/itm/260941919642)
-
If you can find it, there was a Popular Electronics project back in the early 70's for building a calculator using neon bulbs. Input was a rotary telephone dial.
.
Not that you would build yours that way, but it might offer some insight into calculator design from a hobbiest viewpoint
That was around the time a 4 function calculator cost $400 US.
-
I remember that when I was at school in the mid 1970s (dates me a bit!) one of the electronics magazines ran a calculator project using NAND gates (7400 series chips I presume). (This wasn't the popular electronics magazine project referred to by PaulAm above as I think this one was in one of the UK electronics magazines).
It ran for several months and would have cost a fortune to build. Even at the time calculators were just starting to get much cheaper so I doubt anyone actually did build their own.
A bit of Googling and I found this scan of the articles (Practical Electronics 1972 running for 10 months!) :
http://bitsavers.trailing-edge.com/pdf/practicalElectronics/digi-cal/Digi-Cal_Jul72-May73.pdf (http://bitsavers.trailing-edge.com/pdf/practicalElectronics/digi-cal/Digi-Cal_Jul72-May73.pdf)
-
Possibly quite extensively used in the East. At least there is surplus available on eBay, e.g.:
http://www.ebay.com/itm/260941919642 (http://www.ebay.com/itm/260941919642)
Hi Uwe,
Its amazing how long valves lasted in the eastern bloc. It makes you wonder how much money an enterprising fella could have made going there in the 90s and buying up there surplus stock - what with the nixie boom amongst the maker market.
Its a shame there is not so much information about it in English. must have been fascinating just how much of the old technology was used/hybrid system.
If you can find it, there was a Popular Electronics project back in the early 70's for building a calculator using neon bulbs. Input was a rotary telephone dial.
.
Not that you would build yours that way, but it might offer some insight into calculator design from a hobbiest viewpoint
That was around the time a 4 function calculator cost $400 US.
Hi Paul,
I will take a trip to the National library here, I got a feeling they might have this in there archive. Thanks for the tip. I have seen pictures before of projects where guys used the rotary telephone dial. There was once a BBC computer show back in the 80s where they showed this massive valve machine with the telephone dial encoder on the front panel.
I remember that when I was at school in the mid 1970s (dates me a bit!) one of the electronics magazines ran a calculator project using NAND gates (7400 series chips I presume). (This wasn't the popular electronics magazine project referred to by PaulAm above as I think this one was in one of the UK electronics magazines).
It ran for several months and would have cost a fortune to build. Even at the time calculators were just starting to get much cheaper so I doubt anyone actually did build their own.
A bit of Googling and I found this scan of the articles (Practical Electronics 1972 running for 10 months!) :
http://bitsavers.trailing-edge.com/pdf/practicalElectronics/digi-cal/Digi-Cal_Jul72-May73.pdf (http://bitsavers.trailing-edge.com/pdf/practicalElectronics/digi-cal/Digi-Cal_Jul72-May73.pdf)
Hey jpd,
Great link! 10 parts, cant imagine the stress that would have caused having to wait for the next issue. If it was anything like where I grew up they only stocked a few copies of each issue of the electronic mags so if you were not there early the newsagent might have sold out then you had to go to the next village to find one. I had a quick look last night and its got the display multiplexing circuit. In fairness to my newbie eyes it seems quite an elegant solution for the time.
-
That's actually a more advanced calculator than the one I'm remembering, but an excellent article.
Note how this is a microcontroller with a stored program. Figure 9-7 is essentially the FSA diagram. The ROM is a diode matrix. Very cool.
-
Pretty sure I can source all of the parts for that project also... will figure out the total cost tonight and see if its practical.
Ics cost would be 2,347p so around 54 dollars, 72p for the resistors (say 2 dollars) not bad... not calculated the cost for the caps, transistors, edge connectors and bits and bobs but it looks like it would come in at around $100.
-
Pretty sure I can source all of the parts for that project also... will figure out the total cost tonight and see if its practical.
If you do decide to take it on, keep us posted. I don't have the time (or energy) to take on such a project myself but it would be fun seeing it done by someone else.
A bit like the building of Babbage's difference engine.
-
Would the LM317 be a good replacement for the MFC6030a? Seems it would work.
I thought about the LM7805 but the 5v circuit needs 2.5 amps but perhaps the LM7824 for the 20v circuit. And even better it would be 1/6th the price of a LM317. (But 2 LM317s wouldn't break the bank, still make a saving where its possible..)
(http://i58.tinypic.com/2vipilu.jpg)
Thinking the way forward with this is to simulate it on the computer, then see what tweaks could be made. am thinking the 74181 would simplify life a great deal (for the project overall, not the power board :P)
Like you I wonder how many people actually did this project. Considering the time this was done the work that went into these articles is stunning.
-
I'm not an expert on power supplies but there are lots of others on this forum who are. Looking at the circuit I think that the 2.5A (or perhaps it is 2-5A) shouldn't be a problem because transistor TR1 is the one taking the load not the regulator itself.
But if it was me, I'd leave the power supply until last and power it from a bench supply (assuming you have one) whilst building the rest. Of course, if you don't have a bench supply then building the power supply first makes much more sense.
-
Would the LM317 be a good replacement for the MFC6030a? Seems it would work.
I thought about the LM7805 but the 5v circuit needs 2.5 amps but perhaps the LM7824 for the 20v circuit. And even better it would be 1/6th the price of a LM317. (But 2 LM317s wouldn't break the bank, still make a saving where its possible..)
Perhaps you can deviate from the exact layout of the original project when it comes to the power supply. Personally I would suggest using a small switched-mode PSU for a computer. These normally offer 3.3V, 5V and 12V - from the 12V you could make voltage doubler for the 20V.
Alternatively there are three-terminal regulators from the 78xx series (in TO-3 package) available which can do 3A. Also the LM1085, LT1085 and LD1085 can deliver 3A if mounted on a sufficient heat sink (like the 2N3055 in the original schematics). LM323 also a 3A linear regulator.
LM2576 - simple switcher - a switch-mode step-down regulator which can replace a three-terminal linear regulator for the cost of an additional Schottky diode and an additional inductor.
-
There really isn't anything special in the power supply, although the current limit is a nice touch.
You could buy a couple of cheap switch mode LM2596 PS modules off of ebay for a couple bucks each. hang them off a transformer/bridge/cap and just set them to the voltages you need. They're good for 3A. Heck, somebody's selling 5 for $6 with free shipping.
If you haven't done a lot of construction, the PS is a good place to start.
-
Thanks for the advice and tips guys, will read up tonight and think of a plan. :-+
Am now looking at 145 1N5400s got a good deal 3p (6.6 cents) per diode at Newmark! O0 That's the best discount I have managed to get. On the down side I spent the afternoon exploring the Raon market lol. where does the time go? Checked out the Available Ics and around 75% are in-stock... Need to check next time if its just the LS series they don't have or if they are out of every variation. if I get desperate I might mix and match. Didn't get the ICs today. Wanted to compare online first. But as I was in the market already had a hunt around price checking and hunting for the missing Ics. This is where my time went today lol. Mind you its not all wasted - it seems that all the small stores in Raon get there stuff from 3 big suppliers that are actually in the market itself or brought back in some guys suitcase from mainland china. lol. I just followed the runners and it turns out that Newmark, Deeco are two - yet to track down the third.
One slightly frustrating thing is I did not find a 3x3 or 4x3 matrix anywhere physically in the market! On the other hand I have never seen so much audio and tv stuff.
Gave up and went to Deeco to have a look at there test equipment to see if its worth building the bench power supply or just buying one.
Much like ebay.ph there is a massive gap - example scopes, analog from the early 80s. $200, then the next model is a $1000 HP. nothing in-between. Bench supplies start at around $170 here as I was informed most people outside of big companies just build there own or bring them back in suitcases. Lots of high end second hand industrial test stuff available and I guess it must be a bargain because they were shipping them all over the place. But its all relative to what you need I guess. Counters, temp and pressure controllers, they even had the old style huge round Voltage and Amp displays. just shelves full of stuff. It was 5 floors of parts.
Marketing own goal they have the expensive bench supplies next to the $3 second hand transformers. lol. Its not going to be a huge cost to get a pretty good bench supply going. Transformers are dirt cheap, I just adapt a regulated supply they are just as cheap. But I figure I will learn more doing it myself. Found the HP guide on this forum so I will study this and probably pick the brains of you guys in the know.
One thing that seems to be very different to the UK at least, is enclosures are low cost, a 2U metal case, all painted and beautiful is 650p/$14! And that was one of the fanciest ones I could track down. the starting price was a few dollars for a meal case, plastic ones an instrument case will set you back a whopping $10! lol So at least the project will look good. (kind of gravitating towards having a few 2U cases in a rack with the keypad and output separate Wang360e style... need to think about this more)
Furthermore, imagine my surprise a huge section devoted to etching your own boards, every kind of configuration you could imagine, shelves of chemicals, baths. I must say I am not a big fan of etching at home I used to work with some guys that were into electronics a few years back and they all said the results were disappointing compared to what they could get from the factory. Got some of that protoboard stuff with the donuts holes for soldering. 40p (so just under a dollar for 14.5cmx9.5cm). and its got the edge tracks so I can mount it in a edge connector (the heavy duty type edge connectors here start at around 30p and reach 80p ($2) for 72pin).
Overall it was a really positive day, got all the resistors and diodes I will need, will get the caps and transistors next time.
-
Where are you located?
-
Where are you located?
Hey fella, Yea am in the Philippines, Manila.
Found a photo/illustration of the digi-cal project. Still unconfirmed if it was widely built - if at all. curious.
(http://www.epocalc.net/pages/mes_calc_liste/digi-cal.jpg)
http://www.epocalc.net/php/liste_epocalc.php?nom=D (http://www.epocalc.net/php/liste_epocalc.php?nom=D)
Another interesting note, I tried to find out what a "West Hyde type 'Red'" Diode is? I assume its a 1N5400?
I guess the other puzzle for today is tunnel diodes, when they are so fast it seems like a good choice for a switching component. So it makes me wonder why they are not used as such? lots of the early books talk about tunnel diodes like they were the next big thing. hmmmmm...
-
Tunnel Diodes - why can't I quit you?
Its strange the things that catch your interest in things like this, cast away comments in dusty books... it was a mystery how something that had so much hope just suddenly vanished into obscurity. Pick a publication, be it journal, textbook, thesis from the 50s/60s and tunnel diodes are widely reported as being the next big thing... but then they seem to vanish from the literature.. (with the exception of radar.)
Blistering speed, reliant what could possibly go wrong?
(http://i59.tinypic.com/2eujq7k.jpg)
Sometimes an idea keeps rattling around in your head, you can't explain why it keeps nagging at you... So here is a question, from a design perspective why not use them, they are fast, seem to have a fault tolerance and endurance rating like a prize fighter. I saw one negative that output voltage differential was not going to be more than +1volt but with today's precision ADC how is that a problem, especially with thermal compensation/management? Perhaps there was problems getting a tunnel diode thermal stability??? but I thought that was the advantage with these things that they had great thermal stability??? :palm: so confused... lol.
-
I spent many years working on GaAs research (monolithic microwave integrated circuits and devices). It was described as "the material of the future and always will be"! :)
The problem with more exotic materials and devices is that they are niche and expensive so if standard Si gets good enough for some of the applications (lower frequency for instance) then the niche gets smaller and the cost goes up.
GaAs has many advantages over Si - wider bandgap so more robust against radiation, very importantly it has a semi-insulating form that makes it good for low-loss transmission lines and GaAs FETs and HEMTs are good for much higher frequencies.
But GaAs is brittle and can only be processed on small wafers (when I was working in the field GaAs was on 2-3 inch wafers whereas Si was already up to 12 inch (a foot!) so it is intrinsically much more expensive. And all the time Si technology was getting better and going to higher frequencies. And of course SiGe came along.
-
Tektronix widely used tunnel diodes in trigger circuits in their scopes and pulse generators up to the 46x series, more or less. They're very dependable and work very well. My 453 will happily trigger at 150 MHz even thought the scope is rated for 50 MHz.
They always were a niche device and as capabilities of more mainstream components caught up, they faded away.
-
As a single device tunnel diodes are great, but you cannot integrate them to form more complex logic functions. A major problem with both Ge and GaAs has been the lack of a dependable native oxide as insulating layer between neighboring devices and as gate dielectric in MOSFETs. Nowadays we have ways around that, modern Si devices use different other oxides which perform better than SiO2 - but now the primary advantage of higher speed for Ge and GaAs is not worth the price.
-
Tektronix widely used tunnel diodes in trigger circuits in their scopes and pulse generators up to the 46x series, more or less. They're very dependable and work very well. My 453 will happily trigger at 150 MHz even thought the scope is rated for 50 MHz.
They always were a niche device and as capabilities of more mainstream components caught up, they faded away.
The tunnel diode triggering circuits in the sampling plug-ins (1S1 etc) were good beyond 1GHz. I have a "trigger countdown" tunnel diode oscillator unit here that will sync up to 5GHz.
-
I spent many years working on GaAs research (monolithic microwave integrated circuits and devices). It was described as "the material of the future and always will be"! :)
The problem with more exotic materials and devices is that they are niche and expensive so if standard Si gets good enough for some of the applications (lower frequency for instance) then the niche gets smaller and the cost goes up.
GaAs has many advantages over Si - wider bandgap so more robust against radiation, very importantly it has a semi-insulating form that makes it good for low-loss transmission lines and GaAs FETs and HEMTs are good for much higher frequencies.
But GaAs is brittle and can only be processed on small wafers (when I was working in the field GaAs was on 2-3 inch wafers whereas Si was already up to 12 inch (a foot!) so it is intrinsically much more expensive. And all the time Si technology was getting better and going to higher frequencies. And of course SiGe came along.
Thanks for the background jpb, I figured there had to be a reasion why they were not adopted more widely. So its the wafer structure that holds back development/lowcost tunnel diodes.
Tektronix widely used tunnel diodes in trigger circuits in their scopes and pulse generators up to the 46x series, more or less. They're very dependable and work very well. My 453 will happily trigger at 150 MHz even thought the scope is rated for 50 MHz.
They always were a niche device and as capabilities of more mainstream components caught up, they faded away.
Cheers for the info Paul, I get the impression that they are expensive components but when you need a feature/spec and not a price then it makes sence to use them.
As a single device tunnel diodes are great, but you cannot integrate them to form more complex logic functions. A major problem with both Ge and GaAs has been the lack of a dependable native oxide as insulating layer between neighboring devices and as gate dielectric in MOSFETs. Nowadays we have ways around that, modern Si devices use different other oxides which perform better than SiO2 - but now the primary advantage of higher speed for Ge and GaAs is not worth the price.
Hi Uwe, Thanks for the explanation. So its the oxide layer that holds the production yield back right? So perhaps there is a future for these components in a wider range of aplications after all. I guess its one of those situations where 'if you need it, you will know' for the moment.
The tunnel diode triggering circuits in the sampling plug-ins (1S1 etc) were good beyond 1GHz. I have a "trigger countdown" tunnel diode oscillator unit here that will sync up to 5GHz.
Hey GK - I forgot to thank you for the link your provided with your Analog computer project. The website had so many great books I am still reading through them now. It has been a major help. Regarding the above specs, That's performance porn.
Thanks everyone, this was exactly the thing I was looking for when I posted the project in this forum. Everyone has been so helpful and patent answering my questions and giving advice. Its helping me a whole bunch. :clap:
update: found multiple sources for surface mount components, rework stations locally. Also managed to track down some good sources for ics. bit of a quiet day finished the tunnel diode book. will try and get debian on a usb stick so I can get the prototyping/design tools this week. (I foolishly deleted windows xp and now can't make a live usb for linux.) :palm:
-
Mercury Delay Tank
When Looking at the magnetic ferrite cores, magnetic tape and delay lines sometimes the literature talks about its predecessor - the delay tank. I figured why not take a look. Delay lines were used in a number of early calculators but I am not aware of a calculator that used a delay tank. From a practical standpoint it would be difficult to find enough mercury to implement this today - even if you wanted to the health and safety guys would be all over this like white on rice. Added to this the difficulty with precision, size and cost. I can see why the magnetic delay lines were such a welcome development. But these beasts still represent a milestone in the development of memory technology. It seems sad that they are just forgotten when they proved such a crucial component in many early machines and applications.
Here is the UNIVAC I delay tank.
(http://i62.tinypic.com/3343fbp.jpg)
1951 UNIVAC I - Mercury delay line Memory for computers (https://www.youtube.com/watch?v=kignGE77l_I#)
I was watching a video today about a restoration project of an early computer in the uk,
EDSAC Project - Mercury delay lines (https://www.youtube.com/watch?v=xGEAPVCuwvY#ws)
they essentially said it was impossible to do a real restoration of the machine because of health and safety risks. That seems a cop out. If you were going to restore something and operate an exhibit do it properly or not at all. Its like re-creating the Babbage difference engine and implementing it on a FPGA because of the risk of the mill falling on someone... The presenter then reveals the true reason for not implementing the tanks and why the choice of even magnetic delay lines would prove a big problem.
Nobody was around that had any real hands on experience with these devices - there were some written documents but the practical know how about manufacture and implementation had been lost. They had to start from square one.
(http://i59.tinypic.com/2v3ndb7.jpg)
Here is a block diagram of how a delay tank would have operated.
Perhaps have a look at some of the issues that this technology had.
With sonic storage you match the impedance of the substance (in this case mercury) with two crystals (one sends data, one receives)- if they are mismatched you get weird reflected signal noise that will ruin the data. nothing is free and some energy in the signal is lost due to friction against the service of the container. The signal needs to be amplified and its wave shape corrected to account for the unavoidable changes in wave shape. Thus why there is a reshaper and amplifier stage.
Probably the biggest show stopper from a practical point of view is with the Mercury delay tank, your storage transfer rate is closely linked to ambient temperature. to within 1 degree Celsius +/- so you would have to build in a safety operating buffer so as not to have memory overflow conditions. I can see this must have been a big headache. It becomes a signal processing nightmare as any variation in temperature means you will be reading bits out of place in the stream.
The nice thing about archive is I was able to find the maintenance manual for UNIVAC 1. Its amazing the age we live in that documents I would never had had access to before and now available for free at the click of a mouse.
UNIVAC1 Maintenance Manual Jan58 (https://archive.org/details/bitsavers_univacunivceManualJan58_20360914)
Update: This next part probably should go in the delay line section but as Booth is talking about mercury and its uses figure I would drop it in here.
(http://i62.tinypic.com/351isyf.jpg)
-
How big does the mercury tank have to be? The word tank suggests large, but what about something only 1-2mm diameter, stainless tube or similar? Have to connect the transducers but could reduce the amount of mercury needed to just a few old tilt switches? The transducer must introduce transverse pressure waves in the mercury, so it isn't obvious why the actual delay line needs to be very large? Any ideas?
-
I'm at the Computer Lab where EDSAC was built - I received my diploma from Maurice Wilkes who designed and built the EDSAC - he was 93 at the time (2006) and sadly now has died.
He wrote an interesting autobiography "Memoirs of a Computer Pioneer" in 1985.
You might be interested in a more recent history (it's free to download the pdf). Chapter 4 covers the EDSACs - there are quite a lot of good photos.
http://www.cl.cam.ac.uk/ (http://www.cl.cam.ac.uk/)
-
I suspect if you used a glass tube, you would have problems with the wave travelling down both down the mercury and down the glass (at differing speeds) and reflecting at the end. I suppose you might be able to use something like a perspex (acrylic) block with a fine hole drilled through it, allowing the unwanted waves to disperse. I wonder how thin it would have to be to support mono-mode transmission, in the same way that fibre optics behave. That way, you would have lower losses and no dispersion.
Talking of glass though, how about using a 64uS glass delay line from a TV/VCR? They should be easy to obtain, and driving them should be fairly trivial in comparison to mercury.
-
How big does the mercury tank have to be? The word tank suggests large, but what about something only 1-2mm diameter, stainless tube or similar? Have to connect the transducers but could reduce the amount of mercury needed to just a few old tilt switches? The transducer must introduce transverse pressure waves in the mercury, so it isn't obvious why the actual delay line needs to be very large? Any ideas?
Hi woodchips, I have the measurements somewhere - in the case of UNIVAC it was large enough to store 10 words. but they had half registers so that would account for the two sizes of the units. Perhaps the large size of the unit was down to precision measurement/conversion issues at the time. Probably wrong here but wouldn't having a larger volume of mercury increase the overall stability of the liquid with regards to secondary waves and noise from the rest of the machine? I guess a greater amplitude could be applied to the transmission signal you put through this device, making filtering for noise and echos easier from the primary? Will have a think about this today and do some more reading.
I'm at the Computer Lab where EDSAC was built - I received my diploma from Maurice Wilkes who designed and built the EDSAC - he was 93 at the time (2006) and sadly now has died.
He wrote an interesting autobiography "Memoirs of a Computer Pioneer" in 1985.
You might be interested in a more recent history (it's free to download the pdf). Chapter 4 covers the EDSACs - there are quite a lot of good photos.
http://www.cl.cam.ac.uk/ (http://www.cl.cam.ac.uk/)
Hi jpd, Thanks for the link. I will have a read later. Its great he was able to preserve his memories for future generations, so many of these pioneers leave us without that.
I suspect if you used a glass tube, you would have problems with the wave traveling down both down the mercury and down the glass (at differing speeds) and reflecting at the end. I suppose you might be able to use something like a perspex (acrylic) block with a fine hole drilled through it, allowing the unwanted waves to disperse. I wonder how thin it would have to be to support mono-mode transmission, in the same way that fiber optics behave. That way, you would have lower losses and no dispersion.
Talking of glass though, how about using a 64uS glass delay line from a TV/VCR? They should be easy to obtain, and driving them should be fairly trivial in comparison to mercury.
Hey Zad, the fiber optic delay line is I think still used today in radar. and in applications where you have lots of noise from the rest of the circuit. I don't have the kind of equipment here to experiment with them but if you have it in your place its probably a great experiment. The fiber-optic systems seem very robust. you also get the benefit of your delay line being smaller than a magnetic one, not acting like a antenna and introducing a big noise problem to the circuit.
About the use of glass for the tube, I think the reshaper circuitry would be very involved. developed from testing and experience so if they developed the reshaper around the use of one material it would have been a pain to use another. I will need to look more closely at the Univac manual but my hunch is this will be the most complicated part of the whole delay tank circuit.
-
Project update:
Finally got a circuit simulator working on my pc! :phew: qucs will have a look at the tutorials.
still waiting for a gaggle of Ics to come into stock. In the mean time I have been going down a rabbit hole of dead technologies. Finished Booth's book on calculators/computers. The other night finally found a semi dependable smd source locally. One thing that is a bit of an issue is finding sub 0.6m solder. And that's weird as there is a lot of aftermarket cellphone repair going on here. So future research exploring is required. Am kinda gravitating towards using smd where I can on modules that can, through hole with the higher power stuff. That would give me a bunch of practice in both areas.
Power supply... well I found my unregulated universal AC-DC adapter. 1amp 5-12v so I plan to get some voltage regulators and make a poor mans bench power supply for my prototyping. well it says its 5-12v but in reality its 5-18v lol. On the plus side it was $4. I cannot for the life of me remember why I bought it but its been sitting in my draw for a year gathering dust.
(http://www.cdrking.com/apanel/modules/products/images/large/14566_7.jpg)
mmm it just radiates quality. :-DD
Later the next day....
Been looking to solutions around the patchy internet connection here, so have downloaded a fair number of the databooks on archive for reference. That should help figuring out the specs for old/obsolete reports/schematics/articles. Getting the more up to date data sheets is a bit more of a pain as I guess most manufacturers don't want you sucking down thousands of pdfs. Luckily fairchild seem to have built there site to enable me to make a mirror. Shame the others are not quite like that. But still its a start. Got the Jim Willams application notes also (i like the boat picture). And as many applications books as I could track down on archive.org. Beats a room full of dead trees.
I been trying to get as many schematics and books on calculators, electronics and early computers for reference and its surprising just how much is now available. I probably have more books on the computer than I will ever have time to read. But they are good to have none the less. It still blows my mind that anyone in the world has access to information that until even a few years ago would have been inaccessible to all but a few lucky people with access to research library in top universities or defense installations. With regards to calculators there is already a number of great websites that have tons of info that I am only just beginning to understand.
Now I have that circuit simulator I can begin to evaluate the various designs. So I am quite stoked about this! Been reading up on the 74181 & 74812 and the TI application notes are great. Its amazing how much good stuff is inside these documents - they tell you about gotcha's and pitfalls before you have a chance to make them. ^-^
Sorry if the progress is rather slow, but I figured it was better to do it right and read-up before committing to a design - I want to understand what it does first otherwise I have no way to judge if a circuit is better or worse in the overall design.
-
Update:
Debian is now installed and up and running, finally have a stable linux system!
Got the circuit sims and diagram software installed and it all works like a dream! even on my crappy pc.
once the temp goes bellow 34-35 degrees I will venture out and get the voltage regulator and a few lin pots. Will cobble together an analog circuit i found that can multiply, divide, square and square root. (with 1-2% precision) ok thats a bit over-generous for a error budget but in the end its a 4 component setup... ::)
2 x 250 ohm lin pot.
1 x 250,000 ohm lin pot.
1 x 10v voltage regulator.
3 x knobs.
1 x plywood board - front panel.
Wanted to get something physical on the desk as the little voice is saying *you are just doing book stuff, where is the practical?* I doubt I could mess up 3 pots and a regulator but who knows :-DD
Found a copy of Analog Circuits and world class designs. So am working my way though this. It seems to be one of these books where you will only get the benefit after reading it multiple times. Its full of maths terms and seems to have been designed from the outset to accessible only to someone with a maths degree. I still do not understand why they cannot write these books without resorting to another language to explain concepts - you lose 70% of the potential audience the moment they see complicated matrix functions ect. There has to be an easier way to learn the subject - it falls into the same trap as the MIT videos its just a tool for maths weenies. Will look for a more accessible book.
Now I said I was going to stick to digital but it quickly dawned on me without learning at least some of the analog side of life properly I was just setting myself up for a massive fall. Also getting seduced by digital differential analog... its a beautiful architecture/approach. I figure that one acceptable use of fpga - I can see it would be a very nice project to implement a DDA on one of these chips - no more analog drift, great precision, low power, no memory bottleneck.
-
The best laid plans of mice and men...
Temp outside was 36c... but I needed my components so I ventured out into the city.
Wanted to build the analog calculator I was talking about before - to give me more experience before tackling the digital project.
Got all the parts with the exception of the analog display for the time being I will use my multimeter but when I put the components in the case then sure I will shell out for a nice display (they run to about $6 for one of the round instrument panels.) Main cost was the bloody knobs! they cost more than the pots. ah well - things we do for presentation They do look nice appear to be off of some 70s hi-fi.
Could not find a 250ohm lin pot in deco - it was air-coned so forgive me for not going out and searching high and lo for a pot :palm: after a :scared: I remembered that the important thing was the ratio between the two pots and the third... so just got a 500ohm and a 50k ohm instead.
got a switch lucky bag and a heat sink and some LM7810s later I was ready to go! - Yea I could have done this with 1v supply, but that would have produced an horrendous error level of 10%.... I know its analog but thats just embarrassing. As its just a test went for 10v - this will give a 1% error level (the little voice inside is saying 'looser, should have gone for 100v' - that would have given 0.1% error level).
Finished up by getting a bag of electrolytic caps, a happy bag of leds and I was ready to go.
Confession, I wanted to include this into the power supply of the unit..
A self-made heat sink overheating indicator (https://www.youtube.com/watch?v=d7VdgYgOkTs#ws)
but i chickened out last moment as I forgot the part number of the transistor. mmmm classic point to point wiring going on in that video.
Will get the diagram done tonight of the system, still figuring out Kikad.
-
Power Supplys
(http://i58.tinypic.com/2na7a0y.jpg)
Somehow, somewhere along the line many of the modern text books seem to have forgotten that learning is supposed to be practical and fun... This is a beginners book from the 50s..
-
Been looking at how to do a "proper job" on voltage regulation. At first seems pretty straight forward. But that was me just being totally unprepared and flippant on what is in reality a very indepth subject. I guess it all comes down to how good do you need it to be. On the plus side, if the first module involves so much thinking and learning I can only imagine what the rest of the project will be like. This is a much better way to learn about electronics than just dry theory and maths for me.
Nice thing about actually prototyping stuff and breadboarding is it makes you confront a lack of skill and understanding in applying what you learn in the text books. Even getting the voltage regulator portion of a power supply to work right had me scratching my head. I am not the sharpest pencil in the tin lol. Its surprising how much you learn from making a dogs dinner of stuff though. I don't think you can realy know the subject just doing the theory and simulation. There needs to be a happy balance somewhere.
Some of the funny things that have happened,
Ferrite inductors.
Trying to find ferrite inductors, chokes has been "challenging" I found one that was 1500uh but its only 150ma... :'( guess I will just have to shelve the idea of rf filtering never mind. no Pi filtering for me. I was kind of surprised at this but then there appears to be very little RF hobby electronics here. Thats a real shame.
Capacitors.
Got my values mixed up and walked away with a 1mf tantalum bag of resistors instead of 1nf.. I lolled. anyway got a nice range of values now.
Heres a question before i go in and get the caps needed for the whole project, should i just bite the bullet and get tantalum or is there a better alternative? If I want to increase the life expectancy of the caps, would getting x2/x4 the voltage spec help out much with these?
Heat sinking of ICs:
my first thought was to have each of the ICs have there own individual heat sink, and then link them to the case so the case then acts as a large heatsink and should distribute the heat evenly. Here is one thing I don't understand if you use a fan, you get extra cooling on the board but what about the extra interference is it worth it?
-
It is very easy to get hooked up on infinite detail. Product design is a fractal! You know how when you zoom in on a photograph of a coastline, you get to see more and more detail in the line, and the measured length of that line gets longer and longer? That's engineering. You have to learn where to stop, otherwise you get involved in the intricate physics of the thing, and never get anywhere.
Does the product / module work? Is performance satisfactory? Yes? Great, put it down and move on.
Don't get drawn into using expensive low ESR caps with 105C spec. If you aren't going to use your machine in confined hot places for thousands of hours, then it doesn't matter. Dave quite rightly goes on at length about component quality in commercial equipment, but in cases like this it doesn't matter if the cap is 10,000uF or 7,200uF, with 4 Ohms ESR rather than 1 Ohm. The same goes for chokes. They are just there to take the sharp current (di/dt) edges off switching waveforms, absolute precision isn't needed. Just get a few dozen cheap 1A chokes. A lot of the work will be done by 100nF caps across the device power rails anyway.
Generally speaking, try not to use tantalum caps. they can be intolerant of high voltage spikes (even more so than bulk foil electrolytics) and burst into flames if mistreated - electros tend just to pop, steam and hiss. 2x working voltage should be fine. An engineer might order higher voltage ones, but that is probably because he doesn't know what voltage his future designs will use.
5V regulator putting out 5.1V or 4.9V? Moves around 0.2v under loads? No problem. You wouldn't want that in an audio amplifier, but this is a digital system, which is inherently resilient to fluctuations.
IC Heatsinks - Generally easier if each one has its own. Unless you are obsessing over power equalisation to stop the hot one hogging the power and getting thermal runaway.
Fans - Generally speaking, systems only use fans if they are getting pretty serious about things and are looking to package things up and run them for extended durations. I don't think I have ever shipped a final product which used a fan, except those with external power supplies supplied from third party companies. Heatsink smells funny and makes you go OUCH when you touch it? Use a bigger heatsink and higher power transistors. Still too hot? Look at redesigning the circuit to use less power, or stop driving it with so much voltage in the first place.
So to cut things short - Get something together and working. Or even not working - that way you can learn! Then build upon that foundation. Otherwise you will end up with a big box of components and nothing to see for it. I think every engineer has been there and done that!
-
It is very easy to get hooked up on infinite detail. Product design is a fractal! You know how when you zoom in on a photograph of a coastline, you get to see more and more detail in the line, and the measured length of that line gets longer and longer? That's engineering. You have to learn where to stop, otherwise you get involved in the intricate physics of the thing, and never get anywhere.
Does the product / module work? Is performance satisfactory? Yes? Great, put it down and move on.
Don't get drawn into using expensive low ESR caps with 105C spec. If you aren't going to use your machine in confined hot places for thousands of hours, then it doesn't matter. Dave quite rightly goes on at length about component quality in commercial equipment, but in cases like this it doesn't matter if the cap is 10,000uF or 7,200uF, with 4 Ohms ESR rather than 1 Ohm. The same goes for chokes. They are just there to take the sharp current (di/dt) edges off switching waveforms, absolute precision isn't needed. Just get a few dozen cheap 1A chokes. A lot of the work will be done by 100nF caps across the device power rails anyway.
;D Its like you are reading my mind. Yea to be honest I was getting hung up on making it the most perfect power solution, where people would weep with joy when they saw it. Like you say I lost focus on the requirements that were required and just went down the rabbit hole trying to make it beautiful.
Generally speaking, try not to use tantalum caps. they can be intolerant of high voltage spikes (even more so than bulk foil electrolytics) and burst into flames if mistreated - electros tend just to pop, steam and hiss. 2x working voltage should be fine. An engineer might order higher voltage ones, but that is probably because he doesn't know what voltage his future designs will use.
I went for standard electrolytic caps in the end but with generous heat and voltage specs. Nothing to fantastic.
5V regulator putting out 5.1V or 4.9V? Moves around 0.2v under loads? No problem. You wouldn't want that in an audio amplifier, but this is a digital system, which is inherently resilient to fluctuations.
IC Heatsinks - Generally easier if each one has its own. Unless you are obsessing over power equalisation to stop the hot one hogging the power and getting thermal runaway.
Fans - Generally speaking, systems only use fans if they are getting pretty serious about things and are looking to package things up and run them for extended durations. I don't think I have ever shipped a final product which used a fan, except those with external power supplies supplied from third party companies. Heatsink smells funny and makes you go OUCH when you touch it? Use a bigger heatsink and higher power transistors. Still too hot? Look at redesigning the circuit to use less power, or stop driving it with so much voltage in the first place.
So to cut things short - Get something together and working. Or even not working - that way you can learn! Then build upon that foundation. Otherwise you will end up with a big box of components and nothing to see for it. I think every engineer has been there and done that!
Result, added in a pair of electrolytic caps (330uf, 50v) and I got 10v! :-+
-
Was looking at the 10v voltage regulator again today, decided to see how it performed beyond just starting it up and checking the voltage...
Power on, check the voltmeter panel and its dead on 10v same as before. big smiles. lol then about a min into the run I start to see the needle flicker, then start to climb. I notice the 100 uf starting to bulge so I reach over to the wall socket.
Bang. :palm:
Glass half full - it was a regular electrolytic cap and not a tantalum one so at least it didn't burst into flames. :-+
luckily I shielded the monitor with my face lol. Sadly I am down a cap (5 cents). :-DD
time to start poking around.
Followup...
It appears that something has gone wrong with the chepo power supply... on further investigation it looks like the board has been fried. Shocking build quality. There was no fuse on this product. Have taken some pictures will upload later... sigh 4 dollars down the drain.
-
Here is the ac dc adapter in question.
Looks like a diode bridge of 1n4002, a 1000mf 25v electrolytic capacitor/condenser. two selector switches, and a transformer with multitaps. oh and of course a red led and a resistor.
appears that the diodes have gone to the big component bin in the sky.
Shielding is in the form of a paper cover on top of the transformer.
Enclosure was regular plastic, had no shielding and the clearance of the transformer to the top of the enclosure must have been touching. It used to run very hot and after opening it up I could see why. There was no space it had been jammed as tight as possible, you can see the way the heat has burned where the plug pins enter the enclosure (held in by glue...)
The transformer is fixed to the circuit board with a layer of glue. You can see there is no fuse - probably what lead to the calamity that ended the products life.. I had always assumed that all mains products would have been fused but I was wrong. Have become paranoid and have started checking the other devices in the house after this lol..
-
Update:
Got the AC/DC adapter replacement. Will see how it goes. Was going to just build one myself but for now will just go with the adaptor and a voltage regulator.
Been reviewing the magazine articles related to the calculator project & been trying to get my head round the cad/design programs to try and figure out what one would be the easiest to use.
Found some 0.3mm lead solder. for 70p! I couldn't believe my good fortune. I had previously been searching for fine solder here and I had given up on finding anything bellow 0.8mm.
Also enclosure wise found a keypad enclosure that will look very 70s/soviet. Also found some heavy duty keypads.
While I was waiting for some components to come into stock, I did my side project - previously I had found a kit for a telegraph straight key, I had a spare speaker lying around the house so I knocked together a sounder. Who would have thought so much fun can be had with a bunch of dits and dahs. ^-^ I have seen the paddle type keys on the net, will keep my eye out for one. But the straight key is just fine for me messing around.