On Wed, 13 Feb 2008, Charles H Dickman wrote: I have a 60+ year-old slide rule. -- David Griffith dgriffi at cs.csubak.edu A: Because it fouls the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing in e-mail?

On Feb 13, 2008 4:25 PM, Josef Chessor <josefcub at gmail.com> wrote: YOUR *MOM'S* A COMPUTER! OOOOH! Wait... It's like being a freshman all over again. John -- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn

My mom is 93 and she still has her Osborne 1 with a tan case. Bob On Wed, 13 Feb 2008 18:25:12 -0600, Josef Chessor wrote:

John Floren wrote: ...? Do adding machines count? No they just add. Runs and ducks. I would assume that means Digital computer, Orginal equipment with reasonable memory ( say 4K+ )

David Griffith wrote: Well my grandmother is 92 and still does her own taxes, so by an one definition of "computer" she beats your slide rule. :-) I am sure there is a museum with an ancient abacus or a board with some holes and stones, too. I was thinking about electronic, digital, stored program, Turing-complete computers. The PDP-1 at the CHM is the oldest I know of. I have no reason to believe it is the oldest though.

On Wed, 13 Feb 2008, Charles H Dickman wrote: Meaningless question, until you DEFINE: "Oldest" "computer" "operational"

Jason McBrien wrote: This is a tough group... :-) We can argue about that when we have some examples. Digital, electronic, stored program, Turing-complete. I could even be convinced to say just digital and electronic. If there is a Zuse machine or something out there, I am curious about it. Operational seems pretty clear to me, but a front panel wired to a light blinker does not count. Lets put the bar at the level of the PDP-1 at the CHM. -chuck

And thusly were the wise words spake by Jason McBrien So if one or more of its lights are burnt out, is it no longer operational? <ducks, runs and hides under the nearest flameproof and thermonuclearproof boulder> Cheers, Bryan

Tony Duell wrote: Pick the computer .... Granny #1,Granny #2, Granny #3 :)

Philipp Hachtmann wrote: I was thinking of the LGP30 when I asked the question. It is small. Something the size of a house is not going to survive. Sometimes is good enough for me... It was probably not much better when it was new.

Charles H Dickman wrote: I have a can of worms that can do math...

And thusly were the wise words spake by Charles H Dickman mewonders if the question could be.. "Has it been opened?" Cheers, Bryan

Why? What is preventing the use of vacuum tubes, or even transistors? 1900 had the facilities to build vacuum tubes sufficient for logic, had anyone known to try - and in fact someone did, just a few years later. I suspect 1900 probably had the facilities to build basic transistors (point-contact transistors aren't all that tough). It certainly had the ability to bootstrap shops that could make transistors; how much tools-to-build-the-tools is acceptable? /~\ The ASCII der Mouse \ / Ribbon Campaign X Against HTML mouse at rodents.montreal.qc.ca / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

On Wed, 13 Feb 2008, dwight elvey wrote: Is there any reason why reed relays didn't exist then? It's not difficult to make them by hand. Making them hermetically sealed, that might be tricky. -- David Griffith dgriffi at cs.csubak.edu A: Because it fouls the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing in e-mail?

dwight elvey wrote: Umm ... The inventor had no idea how they worked, so scrap that idea.

Tube technology: - DeForest was hacking around looking for a non-infringing (on other's patents) detector for radio waves, not an amplifier, when he developed the audion (not 'audiotron'). I'm using the word 'hacking' in the pejorative sense here: some of his earlier attempts involved trying to send currents through the flame of a bunsen burner. Even when he arrived at the idea of a three-electrode vacuum device (he disavowed any influence by the Fleming valve), his first attempts didn't even have the control electrode ('grid') between the cathode and anode (the control electrode was not in the electron stream). I don't think I'm out of line in saying he had no idea what he was doing at a theoretical level. - The ability of the audion to amplify would not be realised/implemented for yet another few years after 1906/7. - The laws of physics and chemistry haven't changed for billions of years. The knowledge of them most certainly has changed since 1900, even as they relate to the principles of vacuum tubes. Thomson's discovery of the electron was in just 1897, Rutherford's model of the atom as a +charge nucleus with orbitting -charge electrons wasn't until 1911. How the audion worked was not well understood at the time of it's development, even by a physicist. - The characteristics of a vacuum tube will vary with the remnant gas pressure, getting a high vacuum was still a problem in 1900. (re: Will's mention of lack of uniformity/variation between tubes in the early years). - Even if you coul get a good vacuum when the tube was initially built, the seals were an issue in the early years. IIRC, it wasn't until much later (the 20's?, where are my refs) that a metal alloy that had a thermal expansion coefficient matching glass would be developed. -- The 'need' for a 'computer': - It most certainly was there, well before 1900. Babbage's work was inspired/driven by need: to reliably calculate function tables (in part for navigation). -- The possibility of Magnetostrictive Delay Lines around 1930: - Need to look up when the magnetostrictive principle was discovered. It's a pretty subtle characteristic exhibited only by a few materials, it may not have been discovered till quite late. (Also needs sensitive reasonably low-noise amplifiers.)

I am sitting here, reading this thread, and am astounded by what people think tubes could do back in 1906. It would have been *impossible* to make any sort of computing device using the technology of the time. You would have better luck porting OS/2 to a 709. Tubes back then barely worked. When they did work, they barely amplified. And you were lucky to get 50 hours out of them. A try getting two that were vaguely similar electrically. And then try to find someone that actually knew how the things worked. Arguably, the best of the very early devices is the LRS Relay. They were very fragile devices, electrically and mechanically, and were something less than uniform in spec. In the US, the Audion was worse in every way. Even using three Audions in the same radio was so much of a handful that it was rarely tried. The only place tubes were used in signifcant numbers were some of the "high power" transmitters that basically used banks of tubes in parallel. In this configuration, there was some success, due to the inherent fault tolerant nature of the design. It was not until the Great War that tubes started to get reasonably uniform and "high" quality. The best example is probably the Western Electric VT-1, and that did not get good until 1917 or so. Even then, getting even a small number to work together was a chore. Signal Corps manuals of the time often include instructions telling how the radio operator should plow thru his allotment of twenty VT-1s to find three that would make the radio work properly. I have thought about how one would make a very early digital machine using tubes, and I have concluded that the very late 1920s is perhaps a good line to draw in the sand. The AC tubes were out in numbers and very consistent, high power rectifiers were also working well enough to eliminate the use of batteries, and even passive parts like resistors were finally coming out uniformly so every circuit did not have to be tweaked. Tubes and relays for motor control were well developed, and even phototubes were useful in a binary sort of way. The problem that always ends up being the show stopper is in RAM. Using a CRT for the memory was still off quite a few years, ferrite cores were not around for a few more, capacitors generally were too leaky to hold a charge long, magnetic recording was just not really there yet - so the only real option for online storage is an acoustic delay loop. Or paper, in the form of punch cards - which is exactly were the computer industry went. Cheap, easy, reliable, non-volatile. -- Will

In article <1203009683.32199.0.camel at elric>, Gordon JC Pearce <gordonjcp at gjcp.net> writes: The book "From Whirlwind to MITRE: The R&D Story of The SAGE Air Defense Computer" has quite a bit of coverage of the methods used on Whirlwind (and SAGE) to get repeatable quality components. Even in the late 1940s and early 1950s, the variance and reliability of components was not built in by the manufacturers. As a user of the devices, the MIT team developed methods for testing and characterizing components and rejecting the components most likely to fail early. This knowledge fed back to the manufacturers to the point where the manufacturers started doing their own testing and QA of their components. According to the book, this was the first systematic attempt at characterizing these sorts of components (tubes) in any meaningful way. -- "The Direct3D Graphics Pipeline" -- DirectX 9 draft available for download <http://www.xmission.com/~legalize/book/download/index.html> Legalize Adulthood! <http://blogs.xmission.com/legalize/>

In article <e1d20d630802141250o7d45a1e9o93b016a2b2345445 at mail.gmail.com>, "William Donzelli" <wdonzelli at gmail.com> writes: Its not to say that *no* electronic components of any variety were being produced with specific quality and reliability standards, its the components that were being used to fabricate Whirlwind. In particular, things like core memory were invented for Whirlwind, so there was no established quality/reliability metric. I can dig out the portion of the book and cite specifics if its really a huge point of contention. MIT/Whirlwind/SAGE didn't do this for things like wires, resistors, inductors, etc., but for the more advanced components that were newly invented. I know they did it for core memory, and I believe there were also some tubes mentioned. -- "The Direct3D Graphics Pipeline" -- DirectX 9 draft available for download <http://www.xmission.com/~legalize/book/download/index.html> Legalize Adulthood! <http://blogs.xmission.com/legalize/>

William Donzelli wrote: But NEW tubes for computers did not come out untill after about 1955 I guess. Looking at the 5965 datasheet you only get 1000 hours of tube life. Tubes became smaller was the trend. TV sure made a big difference in tube sales.

On Thu, 14 Feb 2008, William Donzelli wrote: Wrong, "computer tubes" were developped for computers and electronic calculating machines. That's why there was the Green Series, the Red Serien, the Blue Series and the Yellow Series (from Valvo of course) for commercial tubes. The "ordinary" tubes had only white paintings (BTW later on, may be in the sixties or seventies Valvo dropped the color codes). So: - Blue was for marine/sea applications (e.g. the PL21 thyratron is Blue) - Yellow was for communications/post ("Weitverkehrsr?hren" like the E88CC, the 18042 or 18046). - Green was for computers (e.g. the LGP-30 is full of Green Series tubes like the E91H, E180CC and E182CC) - Red was for industrial applications (like the E82CC) Christian

On Fri, 15 Feb 2008, William Donzelli wrote: Well, the types that I have listed are Noval/Miniature types and date from the fifties (the E180CC for example is the same as a 7062). These types were specifically made for computers and haven't existed during WW2 (where the development of Rimlock types had started at the end, but not Noval). According to your logic one has to call the E180CC an HiFi audio tube because everyone's using them as such nowadays. I think we have to differentiate between "computer tubes" as "tubes that were used in computers but not specifically made for computers" and "computer tubes" as "tubes specifically made for computers/digital applications". I am talking of the latter. Another example: The Zuse Z22 uses PL36 (or PL81?) for IIRC core memory drivers. This tube is actually a TV horizontal output tube and although extensively used in a computer, was definitely not a "computer tube". But the Z22 also uses logic modules with E92CC (7 pin dual triode, Green Series). Christian

On Fri, 15 Feb 2008, William Donzelli wrote: But there's nonetheless an important difference, namely the oxide cathode. Tubes in digital applications may be cut off for an extended period of time, i.e. there is no cathode current. Ordinary tubes (tubes not specifically made for computers) will create an intermediate layer (? don't know how this is called in English, "Zwischenschicht" in German) in the oxide layer which will render the tube unusable ("deaf"), i.e. with significantly decreased emission (think of a flip-flop that never changes its state). Another aspect is the guaranteed lifetime of the tube. All Green Series tubes are specified for at least 10,000 hours of operation. Christian

On Sun, 17 Feb 2008, William Donzelli wrote: Hm, I think you have to define the term "radar" for me. For me, a radar is an analog apparatus. And for most people a "radar tube" is just one kind of CRT. I can imagine radar transmitters and receivers and magnetrons. Maybe I'm just not following you, but I still don't see the direct relation to computer tubes or how computers used radar tubes. Then that's fine ;-) Christian

Hi, As the OP who started this discussion I appreciate all the comments. The original spec was not to invent new components, but to use things you could actually get in 1900. Just because I know how to make an IC now doesn't mean I could have made one in 1900. While I could have obtained Silicon, it's unlikely it would have been usable purity, and the imaging, etching and depositation would be simply impossible. The "was it available" we have looked hardest at is the transistor, and could have we made a point contact transistor then, even a FET (made before the transistor but didn't recognise it). While we might be able to do this, we're trying to limit our search to things we could get, rather than make, or materials we could get rather than make ie iron cores, as opposed to point contact transistors. ( We are, however keeping an eye on the 256 2n2222 project). Some of the components we have looked at are: Fluidics/pneumatics: possible, but way less fun, and we suspect manufacturing tolerances of the time would have made this really hard. Valves: Not really -- more like 1905/6 depending on what source. If it had to have gain then not. Cathode rays from John Joseph Thomson in 1897. Thermionic diodes: Yes -- from 1873. Solid State Diodes: Yes -- from 1874. Relays: Possible, but even in 1945 was very hard. Available by 1901 at the latest. Magnetic Logic: Certainly available, more later. Core Memory: No ferrite. :-( Probably not until 1947. Mechanical: Certainly, but we're electronics people, and that would be too little fun. Optical: Photo electric "cells" (not intentional) using selenium were around in 1873. Delay Lines: Possibly, piezo ceramic drive uncertain, but close. Magnetic drive -- yes. Both wire and mercury practical. Unfortunately the big book of everything we've looked at is at work, and I'm not right now. Our current research is based around magnetic logic, and there have been commercial magnetic computers made albeit much later. I'm not talking about the exotic cores with 19 windings on them, but basic 1/2/3 winding cores exploiting saturation and simple magnetic properties. It's early days yet, but we have found you need 4 things to make this work (1) simple cores, (2) diodes, (3) AC excitation and (4) some gain every now and again. 3 might be harder than it looks, but we think 20kHz should be enough for iron cores, scale probably 1-2kW. Right now I'm working on the "some gain" area using magnetic amplifiers in cores pushed near to saturation. Without some gain magnetic logic runs out of steam. We do have the schematics of the Univac magnetic logic computer using "Ferractors", but this gets deep into the "we made the material" area with a magnetic material with a square BH curve, and many, many transistors. Construction is likely to be bit serial, and the target is 500k - 1000k bits per second through the loop, maybe an instruction rate of 10% of that. I will say some of the replies on this group are realistic -- when I mentioned this at work one colleague (Windows Programmer) asked if it would run a web server. Sigh. -- Regards, Gavin Melville gavin.melville at acclipse.co.nz

Hi, The bias for the cores is the power hungry thing -- so far. The DC bias puts the AC bias to shame. My boss hopes for clock rates in the .5-1 MHz area, whereas the iron cores I'm currently playing with might top out at 20k. One of the silly little problems is that we can't find reference to clean sinusoidal AC sources of AC above 20k in 1900. The HF stuff in that era tended to involve spark gaps. My instant answer was motor - generator sets, but that needed 3 phase input, and brushless output ie. more like an alternator. I have visions of this thing generating the kW's of AC needed, arcs all over the place, ozone everywhere, and me trying to dynamically balance it at 6000rpm. 3 ceramic Tetrodes in a 3 phase amplifier would make it a cince..... (which is exactly what Univac did) There were 4CX1000's in 1900 weren't there ?

Gavin Melville wrote: How about a calculating music box? :) Do you have a instruction set one can look at yet?

Hi, Modern knowledge can help there of course -- I'm allowed to know why they weren't reliable. The trick they used in WW2 about setting them in wax inside a ceramic tube improved shock resistence and general stability greatly. It did, of course, also allow them to be changed quickly :-) Same thing here. Feedback can do wonders, even if the feedback is to the primary, and done with a magnetic amplifier. This is really a question of resources, and while finite, are substantial. This is why the interest in magnetic logic -- the idea works, well enough to build a mainframe with it and the material was available in 1900. The non-linear bit may, however be the killer, or in the case of magnetic logic, temperature may be the killer. We have basic gates working, but need some gain to cascade them, hence the current work on magnetic amplifiers.

This is an interesting topic. Without quoting anyone else's messages, but acknowledging that I am drawing on other people's ideas, here are some suggestions. Memory. I'd go for mechanical. Either decimal (there were mechanical calculators around at that date that had decimal and even duodecimal wheels for registers) or binary. I think the first binary mechanical memory was built by Konrad Zuse in the 1930s. They've reconstructed it, along with the rest of the Z1, at the Technical Museum in Berlin. It could easily have been done using Victorian precision engineering Logic. I'd still go for relays. I don't know whether there were relays as such in 1900 - my electrical engineering textbook of 1892 doesn't mention them - but the coils, cores, and components to make switches and actuators were all available. On the other hand, there's a lovely _programmable_ mechanical calculator, from about 2 years before Zuse, preserved at the IBM museum in Sindelfingen. This was just like any mechanical calculator of the day (decimal wheel registers), but with several accumulators. You could program nine steps of something resembling horizontal microcode on a plugboard somewhere; it then sucked in hundreds of punched cards and processed each according to the nine steps. There were conditionals in there - "if the card has a hole in position foo then add the total from accumulator 1 into accumulator 3" was I believe a possible instruction. This machine was made by Dehomag - the German Hollerith Machine Company - who were later bought out by IBM. Very fitting, since this machine was a typical IBM box - horrible processor but data throughput to die for... And despite the 1934 date, I expect it could have been built in a 1900 precision machine shop. Philip.

Stability is also a huge issue. Early tubes generally needed constant adjustment while operating. Screening Audions would help somewhat, but five minutes after the screening of new parts, you would end up with a bunch of mismatched tubes again. -- Will

Brent Hilpert wrote: My thinking was that the astronomical cost needed to get a device of unknown speed and reliability (which might not prove more efficient than the human workers it would replace) into production put a damper on most plans - people could likely see that a computer (in the electronic / electro-mechnical sense) would work, but not necessarily be able to judge how well... cheers Jules

Jules Richardson wrote: Umm you had 5 bit code not 8 bit [1] crap of today. [1] 7 or 8 bits depending just how you send data.

By 1901 relays were common items. Western Union did much of the perfection in the mid/late 1800s. I am not sure what the photocells of 1920 were up to, but by 1940 they were not half bad, with so many being used for FAX service. 8 KHz may be about right for off the shelf tubes. -- Will

I would say relays were definitely available in 1900. I have a book from 1895 [1] on electric bells, and while I can't remember if relays are explicitly mentioned (I think they are), the trembler bell was certainly well-known at that time. A trivial modification to a trembler bell turns it into a relay with one normally-closed contact (just remove the hammer and gong, and electrically disconnect the contact from the coil and bring them out seprately). It's then only a very small step to add a normally-open contact, and indeed other cotnacts. [1] My second-oldest book on electrical stuff. I have an 1892 book on telephone installations. -tony

John Foust wrote: Umm I can I still order stuff from them? I wonder how the prices compare with todays prices. PS: And I don't mean cheap chinese products, real *Made in America* stuff. You might have to back about 50 years to find those products.

Gavin Melville wrote: We are talking high speed of 5? Hz for relay computers. I am guessing a 1 HZ might practial for that era. PS. Doing a quick goggle on Zuse, I found this note. Telsa patented a relay AND gate in 1900.

I think you are perhaps just seeing bad, unmaintained relays in your elevator. A good, overengineered, maintained relay, even 100 years old, just will not break. The railroads are full of them (well, 90 years old). Just do not expect any sort of speed - maybe 1 second cycles. A 1900-era relay computer would be big, expensive, suck juice like crazy, and be stupidly slow, but I bet it would work. I can envision setting up batch jobs, running them, then seeing how many cycles were used, so between jobs workers could do preventative maintenance on the relays if some number of cycles was reached. -- Will

ard at p850ug1.demon.co.uk (Tony Duell) writes: Yes, they were -- my 1871 copy of Robert M. Ferguson's "Electricity" describes them in the selection about telegraphic applications: http://offog.org/stuff/ferguson-relay.jpeg -- Adam Sampson <ats at offog.org> <http://offog.org/>

No. No. No. No. If I were better at probabilities and statistics, I could probably *prove* it to be impossible, or close enough to it to not matter anymore. Take into account variations of the tubes and the time needed to tweak out the variations, and figure in the MTBF of the devices and the number of devices, and I think the result would be very near zero. I do not even think that the industry at the time could even make enough tubes to keep up with the demands for just one machine. If you have modern knowledge, and all that knowledge it built on, just make an Intel Core 2. People in 1906 did not know how a tube worked. And barely how to make them or use them. I can not think of anything in our modern world - or even in the near past - that was so much a "barely-there" technology. -- Will

William Donzelli wrote: So are you saying that by the time you've tested enough components to make the thing work, the first components tested have already started to fail (and by implication, that early tubes would fail very quickly even when not being actively used)? Or just that the early tubes would fail so often that replacement would be constant, with nothing left over for actual operational time? Given that we're talking digital logic here, I would have expected that it's reasonably easy to test whether a particular tube will perform in a digital environment, just not how long it'll stay operational. Like I say though, I'm not a tube person... Maybe... interesting point (although I suppose it doesn't affect the question of building a machine now using 1900-era components - just of whether it could have been done back then) Where's the fun in that? :-) I don't think the question was asked with an attempt to attain modern performance / reliability / ease of use / cost etc. in mind. But yes, we need the OP to give us a few more guidelines as to where the boundaries are, as "1900-era materials and modern knowledge" covers way too much ground. cheers Jules

William Donzelli wrote: Umm not really ... read the fine print that comes with your triode. The reverse is often true -- digital tubes make great audio tubes. See 5965 for a typical computer tube. Digital is the modern idea.

William Donzelli wrote: Where do you get that information from?

On Thu, 2008-02-14 at 12:56 -0500, William Donzelli wrote: I think one of us is missing the point. Whether it's you or me I can't tell yet. I read the very original post, about whether or not it would be possible to make a digital computer with the technology of the early 1900s as "Send a list member from 2008 back to 1900 and see if they can make a computer" - perhaps I misunderstood the question. If that's the correct premise, then I'd say yes, it's very likely that someone from 2008 could time-travel back to 1900 and (assuming money, tools and a workshop, and possibly some assistants) build a working digital computer. Now to test this, we're going to need a TARDIS and Tony Duell... Gordon

On Sun, 2008-02-17 at 09:32 -0800, Vincent Slyngstad wrote: Aha, but we already know that the TARDIS doesn't quite work properly anyway, so the plan is to leave Tony alone with it and a stack of paper until he draws the manuals himself. Gordon

Do we have any list members who work in IC fabrication :-) More seriously, there are a couple of books called 'The Voice of the Crystal' and 'Instruments of Amplficiations' written by a chap who made (AM) radio receivers using no commercial electronic components (he does use shop-bought enammeled copper wire). In the second book, he makes a few triode vlaves and both point-contact and (copper oxide) juction transistors. The devices have a pretty poor performance, and a very short lifetime. But they do work. And I suspect it's possible to do better, even in a lab of 1900. I'm darn sure it is. If you'll accept 'electromechancial' (relays, stepping switches, etc), then, although the machine would be large, and slow, I have no doubt it could be made. In 1900, while the _theory_ of electricity was not well understood, the practivcal operation of things like electromagnets seems to have been, You provide the first, I'll provide the second :-) -tony

On Feb 17, 2008, at 3:04 PM, Tony Duell wrote: I have both of these books; I bought them from Lindsay last summer. They're both fantastic! I've not yet built anything from the info contained therein, but I hope to. -Dave -- Dave McGuire Port Charlotte, FL

No ferrite. And not a chance that the tubes could drive the lines in a predictable way. -- Will

Hi I wish to make a correction. The tubes were called audions and were not made until 1906. Low gain made them hard to use for radio uses but that has nothing to do with digital uses. Dwight _________________________________________________________________ Shed those extra pounds with MSN and The Biggest Loser! http://biggestloser.msn.com/

Hi Low gain and high capacitance between grid and plate made them hard to use for RF. Audio amplification could be used after a diode detector. Dwight _________________________________________________________________ Climb to the top of the charts!?Play the word scramble challenge with star power. http://club.live.com/star_shuffle.aspx?icid=starshuffle_wlmailtextlink_jan

On Fri, 15 Feb 2008, dwight elvey wrote: The term "audion" describes a circuit for the demodulator stage in a radio set consisting of a tube (mostly called the audion tube like the common REN904). An audion may also be called a "grid rectifier". Sometimes the audion tube was also used as RF amplifier in a reflex circuit. But perhaps you're thinking of Lee DeForest's triode which he called Audion because of it's usage in a thereby called "audion circuit". But with the invention of the frame grid tube ("Spanngitterr?hre") and the PCC88 (called the "miracle tube" or "Wunderr?hre") it has been proved that triodes do much better in RF input stages like in UHF TV tuners which need to go up to around 800 MHz. The PCC88 (and similar tubes like the PCC189) are mostly used in cascode circuits with e.g. a PC93 as UHF oscillator. But we're completely off-topic now... (except perhaps with the frame grid tubes). Christian

I bet you could find the information in the Tyne book. I can not find mine right now. First off, the tubes from 1906 that we are thinking of are Audions, not Audiotrons. There were Audiotrons, but that was just another brand name in the whole nameverythingatron craze. Second, people that have Audions (or LRS relays) today never power them up. Most have broken leads and/or filaments anyway. Third, only a few Audions were used for telephone amplification before Western Electric secured the rights to make tubes. It took them quite a few years to get "repeater bulbs" decent. I do not have the values for a "typical" Audion transconductance, but it is so low it almost looks like a typo. The 1906 and the 1920s are worlds apart in terms of tube technology. The problem is that the basic operation of the tubes during their first ten years of the technology's life was shaky beyond belief. -- Will

Roger Ivie wrote: No idea why, but whenever I come across a system, I end up asking myself "how could this have been done* without electronics?" - I find mechanical analogues of electronic/electrical systems fascinating... * Not cheaper, faster, to the same scale etc. obviously - just whether it could be done at all. Anyway, my first thought on reading that previous message was how to do it using pure mechanical devices (and fluidic logic's a definite possibility). And now's probably a good time to mention the Heath Robinson Rube Goldberg computer again ;) cheers J.

dwight elvey wrote: It's possible, I'd think (there's a nice page about a marble-based binary adder at http://woodgears.ca/marbleadd/index.html if you haven't seen it before). The main problem with a marble machine is likely to be maintaining state, as any kind of 'read' within the system tends to be destructive. Doubtless there are ways around that (e.g. the presence of a marble presses a lever which results in a marble coming from some large hopper somewhere upon any kind of 'read' op, rather than the 'trigger' marble itself moving) I believe that the electronic computer was just an idea whose time had come; I don't think there's anything that would have prevented something similar (non-electronic). We've all heard about Babbage, military mechanical gun control computers, Meccano differential analysers, Victorian automatons etc. - it's just that prior to the "usual suspects" (Harvard Mk. 1 etc.) there was simply no perceived benefit in making an electronic, stored-program machine. cheers Jules

What about magnetostrictive delay line memory? Output could be a typewriter rigged up to solenoids, though I'm not sure when solenoids were invented - I've found companies that have been producing them since 1918. They're easy enough to build.

I think you need to be more specific. The laws of physics and chemistry have, AFAIK, not changed since 1900. The necessary raw materials to make transsitors and ICs were around then. So from what you said, you could just use a microprocessor.... So I think you have to consider just what you are, and are not, allowing with respect to making devices to make components, etc. Relays were certainly well known and understood at that time. I would propose a relay logic machine, therefore. Unfortunately my books on telegraphy are about 10-15 years too late. Some of the stuff in there could well be very applicable if it was known at the time you want. For example, and going a bit off the thread, one of my old books on electrical engineering describes a system used on the London Underground (underground railway, subway, call it what you will) to keep track of the trains. Each train had a 4 bit code, there was a drunm with 4 rows of pins whcih could be moved itn oone of 2 states to stoe the IDs of the trains. A set of solenoids moved the pins, the drum was then stepped to the next positon. A separate rotating set of contacts read out the states of a a row of pins and then moved on to the next one. It worked just like the classic circular queue with seaprate read and write pointers that we've all implemented in software at some point. The outputs of the contacts were decoded by a 'tree stucture' of relays to produce a 1-of-16 output. Now, as I said, I don;t know if that was around in 1900, but it certainly _could_ have been, The electric light bulb was known before 1900. The Edison Effect (diode valve) was, IIRC 1883. THe triode, was, alas, later (1907 I think), but maybe you'll stretch the point and allow it. -tony

In article <47B4490B.7040809 at gmail.com>, Jules Richardson <jules.richardson99 at gmail.com> writes: Its just one of those questions which crops up every once in a while that *could* be the beginning of an interesting discussion, if people didn't immediately devolve into smart-ass mode responding about slide rules and their grandmothers. -- "The Direct3D Graphics Pipeline" -- DirectX 9 draft available for download <http://www.xmission.com/~legalize/book/download/index.html> Legalize Adulthood! <http://blogs.xmission.com/legalize/>

A slight variation on the OP's original proposal/spec, but a fun project would be to re-design/implement Babbage's analytical engine using relays. This is something that could have been accomplished in 1900 - both the technology and the concept existed - if only somebody had put 2+2 together at the time (pun intended). It's interesting to speculate that if Babbage had been born or been working just a couple of decades later than when he was, or alternatively if electrical concepts had been just slightly more advanced in his time, he would have throught of implementing his ideas with electric relays and would likely have succeeded (although one might argue the same about the state of mechanical technology).

woodelf wrote: :-) I don't have a reference to the Meccano one, but the Lego flavour's here: http://acarol.woz.org

Charles H Dickman wrote: Another take on the original question: a functional reproduction of the ABC* from 1941, using period components, was produced in the 1990's. * the first electronic computer by one definition.

Charles H Dickman wrote: I assumed you were :-) I'm not sure if there are any parts in there which weren't available in the 1940s (can find out if needed), though - I think it's as close to 'original' as it's possible for a ground-up rebuild to get. Our Elliott 803 is a complete original* from 1960 (IIRC) - but doubtless someone can top that (I believe the CHM's PDP-1 is the same year?). * for suitable definitions of 'original' - 803 spares have been used to keep the machine running, and some of *those* are doubtless a year or two more recent! cheers Jules