Vintage> On Wed, 28 Jul 2004, Marvin Johnston wrote: Vintage> I'm thinking farther back. Was there any mini or mainframe Vintage> that had a real-time clock built in or as an add-on option? Vintage> It's come up in some litigation. The actual issue at hand Vintage> is that someone was able to overturn a patent by claiming Vintage> the IBM 650 had a real-time clock built in. They "proved" Vintage> this by submitting as evidence a printout that had the date Vintage> printed on it(!) I've checked the IBM 650 Manual of Vintage> Operation and it makes no mention whatsoever of a real-time Vintage> clock. I pretty much figured it wouldn't but I of course Vintage> had to do due diligence. Vintage> But it made me wonder: what was the first computer to have a Vintage> built-in real-time clock? I second the question -- defined how? For answering the litigation, you'd really need to know that... If the answer is: a device that maintains the time of day independent of the computer power, the oldest I know of is the DEC KW11-W (?not sure about the suffix). I don't remember when that came out, but it probably predates microcomputers. It wasn't a popular option. Another example, from roughly the same era, is the TOY clock in the Pro-350. That's a microcomputer chip, so presumably some PC type system offered it as well. (That may apply to the KW11-W as well... I don't know.) On the other hand, if the definition is a device that allows the computer to keep time while it is powered on -- which is all that you can deduce from the 650 example you quoted -- then certainly it goes as far back as the CDC 6600 (1964) and probably a lot further. Then again, that's after the IBM 650. By either definition we're talking about something that came out well over 20 years ago, so it would be "prior art" invalidating any current patent. paul

As I mentioned in another post, I think it was the KW11-C, and was a unibus-only option. There was a KWV11-C, but I don't know if they were associated. Early 80s. BTW - this board also had batteries, so the clock would maintain time across system power cycles (it charged the batteries when the mains were on) It was separate from the actual CPU chip... it might have been something like a Dallas chip, so probably someone else was using it as well. Mid 80s. Megan Gentry Former RT-11 Developer +--------------------------------+-------------------------------------+ | Megan Gentry, EMT/B, PP-ASEL,ST| email: mbg at world.std.com | | Member of Technical Staff | megan at savaje.com | | SavaJe Technologies, Inc. | (s/ at /@/) | | 100 Apollo Drive | URL: http://world.std.com/~mbg/ | | Chelmsford, MA 01824 | "pdp-11 programmer - some assembler | | (978) 256 6521 (DEC '77-'98) | required." - mbg KB1FCA | +--------------------------------+-------------------------------------+

At 04:49 AM 7/29/04 -0700, you wrote: The HP 9826 and 9836 had a battery backed RTC as part of the battery backup/power fail option but I'm not sure of the date. Hmm. Come to think of it, the HP-01 would probably count too. Joe Festival --- http://www.vintage.org ] http://marketplace.vintage.org ]

Andy Holt wrote: It goes further back than that. The 7094 modified for CTSS had a periodic interrupt. I assume Sellam means a device which will present the time and possible date maintained independently of the CPU. There I would agree I we are into the mini era if not the micro.... There was an add on RTC clock card for the PDP-11 (but Sellam specified "built-in" ;-) If it does not have to be built in I suspect you could find some earlier examples. ISTR hearing somewhere of units which attached to mainframes which would provide tame and date... -- HansP

I don't rememebr a clock card for the '11 -- there were line-time and programmable interrupt cards that gave a periodic 'heatbeat', and that's what DEC normally regarded as a real time clock... Looking back through my manuals, I've found a couple of somewhat interesting ones : There was one of the P851 (1978-ish) that was a battery-backed digital clock card (which either counted up to 100 days or 10000 hours in 1s increments). You could even hang an LED display off it if you wanted a human-readable clock in your machien room. I have schematics, but alas not the card (I do have a P851...) There was a real-time clock model (98035) for the HP9825 'calculator' (really a computer!). It actually contained 2 clocks. When the machine was powered up, it used an internal microcontroller clocked at a known rate. When the machine was powered down, there was a battery-backed chip (looks to be a repackaged chip from a digital watch) that kept the time, and which was read by the microocontroller at power-up (the interface to this chip was strange -- a couple of lines corresponding the the 'buttons' on an LED watch, 3 digit stobes, and the 8 lines for a 7 segment display (7 segments + annunciators for things like the leading '1' on a 12 hour/US date display, that's how they got away with only 3 digit lines). Almost certainly not the first RTCs on computers, though. If you go back to data loggers, the Solartron DTU (1960's vintage) has a clock card. This was a rack of mostly DTL logic that connected to a DVM or DFM and to an output device like a Facit 4070 punch or an ASR33 teletype. Even so often (the rate could be set using a switch on the clock card, or it could respond to an external contact-closure input), it would scan up to 20 channels (using relays in the DTU), connect each one to the input of the DVM, and record the result on the output device, along with the time. Thsi unit was sort-of a forerunner of lab minicomputers, in that it let you take results automatically and feed them (via paper tape or punched cards) into a mini or mainframe computer I also have a much later Fluke logger, controller by a pair of 4040s. It has a real time clock built in (periodic interrupt to the main processor IIRC) which goes up to 1000 days in second increments. Several of the early RTCs keep time in hours/minutes/seconds as you'd expect, but ject count days up from 0-99 or 0-999 or whatever. This gets round the stupid features of the calendar (and incidentally avoided the Y2K bug :-)) -tony

I know both of those cards, along with the DL11-W (which essentially has a KW11-L built-in), the MNCKW (MINC clock module, somwwhat similar to the KW11-P, the KW11-K (IIRC that's a dual clock card with some nice features), and so on. I think I have prints for all of them too. However, I don't think this is what Sellam meant by 'real time clock'. I think (and please correct me if I'm wrong) he was talking about a Time-of-Year clock type of thing. Ah, now that one I've not come across. -tony

Tom, I have to disagree with you completely! You are NOT being pedantic! An RTC is indeed any source of information which can be used to monitor the passage if time. It does not need to have any synchronization with "wall time" at all. The RTC may just provide an periodic norification that a certain amount of time has elapsed [16.666mS for a 60Hz LTC] or may have register(s) for accumulating a number of these events. If the processor contains built in timers that can divide a clock signal, then any of these can be considered an RTC. Once any type of RTC exists, a computer can be programmed with a value which matches the current "wall time" and use the information from the RTC to update the wall time. The biggest limitation with the minimal implementation of this is the duration which a a computer may be non-responsive [or even "off"] and still have the ability to correlate the information from an RTC with "wall time". In general LTC's are totally useless for this purpose [you have to manually enter the "base correlation" value at boot]. At the high end, the RTC may accumulate a significant number of "ticks" in hardware [usually with battery backup[ and have the ability to be "preset" in order to correlate with "wall time". In many of these cases the RTC also has decoding circuitary to represent the value in Seconds, Minutes, Hours, Days, etc... At this point the RTC has the functionallity of a "Clock/Calendar". If we are in agreement the the original question really needs to be broken down further... A) What was the first computer to make some use of a periodic signal to keep a time/date that was synchronized with "wall time" [either in hardware or in system software]? Or B) What was the first computer to make use of a device which could remain synchronized to "wall time" even if the computer was not operational?

On Sun, 2004-08-01 at 15:42, Vintage Computer Festival wrote: Ahh, specifically 650 -- I'll ask around. (Wouldn't a hardware manual and/or programers manual help here?)

On Mon, 2004-08-02 at 13:05, Vintage Computer Festival wrote: But in the spirit of the times, does any hardware facility have any time-related side effect that could have been exploited to extract time from? (Besides the obvious calibrated-subroutine hack.) Or does the feature have to be explicitly a "time-keeping" hardware facility?

On Wed, 2004-08-04 at 14:50, Paul Koning wrote: I agree with you generally, but disagree only in that an "OS" is hardly needed to take advantages of hardware facilities and side effects (in fact you're better off without one, they usually get in the way of "fun" :-) Agreed. I think it extremely unlikely that in a 650 an ordinary program would jump through the hoops necessary to extract a timebase from this particular machine. It's possible, however unlikely, that there's a manufacturers test port that counts drum revolutions or somesuch rot, but in hardware of that vintage there tends to NOT be unexploited hardware, it was painful and expensive enough as it is! Short of a peripheral device in a catalog that provides clock-time or real-time data, the answer is likely "no", 650's can't tell correct time (any more than you could tell it to WRITE 100 H1,M1,S1 WRITE 110 D1,M1,Y1 100 FORMAT (I2,I2,I2) 110 FORMAT (I2,I2,I4) I couldn't remember the details of the FORMAT statement, and none of my books are new enough to have alphanumeric formats in them!)

I think it would be interesting don't leave to go these books: http://cgi.es.ebay.com/ws/eBayISAPI.dll?ViewItem&item=5111921924&ss… STRK:MEWA:IT ... mostly having in mind the hp2100 simulator from SIMH, or simply for play with any kind of real HP2100. Greetings Sergio

On Thu, 29 Jul 2004, Andy Holt wrote: Yes. But I can't imagine there was not a real-time clock (i.e. as described above) as at least an option for an earlier computer system. -- Sellam Ismail Vintage Computer Festival ------------------------------------------------------------------------------ International Man of Intrigue and Danger http://www.vintage.org [ Old computing resources for business || Buy/Sell/Trade Vintage Computers ] [ and academia at www.VintageTech.com || at http://marketplace.vintage.org ]

On Thu, 29 Jul 2004, Tom Jennings wrote: I'm more interested in a clock/calendar: something that could be polled and produce the local time and date. I'll bet an interesting history lurks behind such a seemingly simple topic (real-time clock calendar deployment in computers). -- Sellam Ismail Vintage Computer Festival ------------------------------------------------------------------------------ International Man of Intrigue and Danger http://www.vintage.org [ Old computing resources for business || Buy/Sell/Trade Vintage Computers ] [ and academia at www.VintageTech.com || at http://marketplace.vintage.org ]

On Sun, 1 Aug 2004, Tom Jennings wrote: Hmm, okay. Then I guess what we'd be looking for is some device that keeps continuous time. I.E. you set it once, and through some manner of both hardware and software, produces the correct current date and time upon request. As far as need to keep the time during power off, I don't know that in this case it would matter too much. It would just have to be a combination of hardware and software that kept the current date in the IBM 650. The 650 was small enough that you could power it down during off hours, though I don't know what actual practice was (perhaps it was too much of a chore to boot up so it was kept on all the time). At any rate, as long as the 650 had some way to generate uniform time pulses, a real-time clock/calendar could be implemented, but this would probably have to be at a hardware level since the 650 was primarily a batch machine, or it would have to be programmed to store the date/time in some memory location that the other applications would not touch, and then some counter would have to be implemented that could update a memory location with the number of ticks since the last access or whatever. Of course, as I've been suggesting, there could have been a totally separate clock/calendar device (heck it could have even been mechanical) that kept the local time/date and then transferred it to the 650 upon request. Just like the clock in a modern PC. How fun. You'd have to shield it pretty heavily though or else your PC would be glowing at night (talk about a case mod!) Has anyone ever produced a random number generator that pulls it's "seed" (snow)? -- Sellam Ismail Vintage Computer Festival ------------------------------------------------------------------------------ International Man of Intrigue and Danger http://www.vintage.org [ Old computing resources for business || Buy/Sell/Trade Vintage Computers ] [ and academia at www.VintageTech.com || at http://marketplace.vintage.org ]

John> Computer-based ramndom number seed generation: John> I may at some point remember the machine/manufacturer, but I John> *do* recall tha one (perhaps Gummint-oriented?) device used a John> standard radio buried in it somewhere, tuned to a local John> station, the output of which was digitized and the seeds John> extracted from that... as I write this, IISTR that it was some John> kind of crypto-thingy... That would be quite a bad idea. I'm sure the spooks are smarter than that. Random number generators are built in to most crypto chips these days; they work starting from the noise generated by some simple electronic circuits. Then any biases are taken out by post-processing, for example by running the bitstream through a cryptographic hash. paul

On Mon, 2 Aug 2004, Paul Koning wrote: And just why would that be a bad idea? Remember this was back some years - your glib "these days" doesn't apply to 'them days'... in the 60's and 70's it was not trivial to develop 'real' random numbers, so various means were used to sample a physical source or 'randomness'. I have an audio whit noise generator (Grayson Stadler) that uses a 6CB4 tube with a small bar magnet in proximity to the envelope - this exacerbates the transit noise of the tube and produses a very nice 5 Htz to about 50 kHtz stream of chaos. For many years, going back into the late 40s', microwave noise sources (for testing and calibration) were made with a 2-foot flourescent lamp mounted transversly at an acute angle through a section of waveguide - when the microwaves passed thru the excited mercury vapor, the resultant harmonics went into the gigahertz range... Sheeshe! Kids nowadays! Why when *I* was your age,,, Cheers John

On Mon, 2 Aug 2004, Paul Koning wrote: Nits? Are we going to have a spirited discussion or are you going to pick nits? (While you're at it, might as well bag on my execrable spelling and egregious transposition errors, all firmly rooted in slowly advancing presbyopia and the need for (yet another) set of trifocals.) I have to force myself not to write cps, or Kc, or Mc... If sampled at 'occasional' intervals, and the resultant digitized samples are averaged and truncated, one approaches a fine approximation of statistical chaos. I'm not saying it's the most efficient, or even sorta practical, way to generate a random string of numbers, but it *was* interesting, and it bore upon the topic of this list, and the discussion at hand. Citation? URL? This is not familiar to me. All this is rapidly descending into Deceased Equine Flagellation... the points are evident that it is possible to feed a random number generation system with various chaotic signals derived from various well-known processes, and derive therefrom a suitably "random" series of as-unrelated-as-possible numeric values..... Cheerz John

On Mon, 2 Aug 2004, Vintage Computer Festival wrote: Damn, that's just lousy. Cheers John

At 01:53 PM 8/2/04 -0400, John wrote: Not if you eat them soon enough! Joe

This is pretty standard stuff in the military ECM world (I think the O-202/ALA-7 pulse generator works the same way, early B-52 era stuff). In fact, some relatively normal tubes were made for this trick. I found my box of phototube noise generators, pulled out of WW2 jammers. Cheap. Email if interested. William Donzelli aw288(a)osfn.org

I beleive the thermal noise in a resistor has been used as a source of random numbers, though. -tony

On Mon, 2 Aug 2004, Hans B PUFAL wrote: I didn't think of doing that! Thanks!! Yes, from what I've been told, the 7090/7094 had a clock feature of some sort. -- Sellam Ismail Vintage Computer Festival ------------------------------------------------------------------------------ International Man of Intrigue and Danger http://www.vintage.org [ Old computing resources for business || Buy/Sell/Trade Vintage Computers ] [ and academia at www.VintageTech.com || at http://marketplace.vintage.org ]

David> Regarding a message that appeared on this thread with the David> approximate quote: David> of over a minute per dat David> [sorry I it the delete key too quickly on the message to post David> the EXACT text] David> 1 Day = 86400 seconds * 0.01% [100 ppm] = 8.64 SECONDS per David> day. Oops. David> Not good enough for long term time keeping but much better David> than the previous poster indicated. Yup. Right conclusion, wrong numbers. David> I was involved in the development of som Military systems David> [1979-1983] that used a tempe0rature stabilized crystal with David> 0.5ppm stability. To the best of my knowledge this was a David> "state of the art" implementation of automomous time keeping David> for 1979. 0.5 ppm for boxes that get carried around in military trucks and bounced around in the field -- that's quite good. For something that's sitting in a reasonably controlled environment, that's not so great; I think that 10^-8 would be considered state of the art for OXCOs (1970s or not). For autonomous timekeeping independent of technology, the state of the art was a second per year or so (that's 10^-8, roughly) around the early 1900s. First with pendulum clocks (Shortt clock), then around the 1940s or so crystal clocks came in that could match this. And not too long after that there came the rubidium (10^-10) and cesium (10^-14) clocks. Some of that would be found in military gear, I think (Rb at least, Cs somewhat less likely). Consider GPS satellites, which have either or both built-in. paul

On Sun, 2004-08-01 at 15:32, Vintage Computer Festival wrote: re: computer 'subsystem' to produce clock time Are you looking exclusively at the IBM 650, or any computer? Does it have to be an 'automatic, electronic, digital, stored-program, computer'? Time-stamping of 'data' trivially dates to telegraph/teletype equipment -- I've seen stuff from the 1930's that does it. I know you don't want tty stuff but just to point out that the goal still seems vague to me. (Nahh, it could be a very small source, like a smoke detector. Not very dramatic though as a big scary lead coffin though :-) Oh I'm sure from the principal, yes, thermal noise would be a great source. Zener noise is easy too, not sure of it's distribution.

Vintage> On Mon, 2 Aug 2004, Tom Jennings wrote: Vintage> The overall discussion is very useful, but in my particular Vintage> case, it needs to pertain specifically to the IBM 650, Vintage> because the assertion at hand is that a document used to Vintage> discredit a certain scientist's claim to a certain Vintage> achievement has a date stamp on it that is claimed was Vintage> produce by the IBM 650 using its internal clock feature. Vintage> The 650 does not have such a feature inherently, so now I'm Vintage> trying to determine if there was ever an adjunct product Vintage> that IBM produced to give the 650 the ability to tell time. It sure doesn't look like that from the docs on Al Kossow's website. It seems that the 650 doesn't have interrupts, which would make a simple periodic tick type system pretty problematic. If the exercise is to attach the credibility of the document, it looks to me like you have enough data to do that. You can clearly say "there isn't any such thing as an 'internal clock feature' on the 650". That puts the ball in the other party's court, to try to demonstrate that there did exist some obscure add-on of which not a word appears in the dozen or so IBM docs that are on-line. And even then, another question becomes whether the time stamp is real or a forgery, independent of what machine generated it. paul

I was thinking the same thing. There were some programs I worked on during the early seventies where the computer did not have a clue about the date or time. The input "deck" contained this information which was basically just echoed on to the printouts. I remember running batch jobs [large decks] where the only cards that would get changed were a few parameters and the leading "identity" card....

BTW: The Sage system (tube based) predates the 709x systems (discrete component/transistor based) and was in use from 1956-63. Lyle On Monday 02 August 2004 19:54, Lyle Bickley wrote: -- Lyle Bickley Bickley Consulting West Inc. http://bickleywest.com "Black holes are where God is dividing by zero"

Bad Typo - thanks for the correction. The project was started at MIT in 1954, deployed in 1956 and the last one dismantled in 1983. Lyle On Monday 02 August 2004 20:15, William Donzelli wrote: -- Lyle Bickley Bickley Consulting West Inc. http://bickleywest.com "Black holes are where God is dividing by zero"

No, the streams were just digitized radar video. William Donzelli aw288(a)osfn.org

What I mean is that there is no time stamping, no complex format, no handshaking - just a bunch of bits. By the 1950s, the rotating yoke was mostly out of fashion, save a few World War 2 designs that refused to die. Even some World War 2 PPI scopes did away with them, doing the rotation electronically (SCR-584, for one). This is how the SAGE AN/FSA-10 gap-fillers worked - the scopes of up to six radars were directed at one radially scanned image tube. The result was a composite radar video stream. William Donzelli aw288(a)osfn.org

On Mon, 02 Aug 2004 22:30:30 -0400 "David V. Corbin" <dvcorbin(a)optonline.net> wrote: IBM PC-DOS 1.0, for those who have never booted it, presents the user with a Time and Date prompt that can not be bypassed by simply hitting carriage return. The user can enter fictitious time and date if so desired, but not null. This was changed with later versions of PC-DOS and the parallel MS-DOS versions. You could simply hit return and default time/date values were assumed. I'm not sure if PC-DOS 1.0 supported autoexec.bat, which bypasses the Time/Date prompt on later versions (i.e. the Time/Date prompts are invisibly bypassed and a call to a small program that accessed the 'clock' hardware was not included (did anybody else back then feel like their machine was stunted if there wasn't a clock card plugged into an ISA slot on the machine?)) My point in bringing this up is to note that even though it didn't have a real-timekeeping function built in, an original IBM PC running PC-DOS 1.0 did have an inherent 'timekeeping' fuction which was often somewhat accurate. It didn't cost any more or less to enter the real time/date than a fictitious one, since some input with the right format was forced on system bootup. Granted, not many people at all ran IBM PCs on DOS 1.0, which was crippled in many other ways, and wasn't really that viable an 'OS' to run for several further releases. But back 'with the thread' being discussed, when I was in a programming class in college where we had to submit our FORTRAN decks to a batch terminal, there was definitely a high quality time and datestamp placed on the resulting printouts. This was on CDC Cyber hardware in the late 70's.

On Mon, 2004-08-02 at 15:04, Paul Koning wrote: Absolutely -- a hardware clock is no less likely to be fudged than a software one.

William> I know that AT&T once marketted a true random bitstream chip William> using a bunch of unstable oscillators. The better Ethernet chips do likewise. Designing such a thing is *not* easy, especially if most of your ASIC design experience involves thinking about circuits as being digital (rather than analog, as they all are once you dig deep enough). paul

On Mon, 2 Aug 2004, Dan Veeneman wrote: Wow, very cool! Sure beats my idea for mounting the detector circuitry to a badge that you can then wear on your lapel in public to annoy smokers :) -- Sellam Ismail Vintage Computer Festival ------------------------------------------------------------------------------ International Man of Intrigue and Danger http://www.vintage.org [ Old computing resources for business || Buy/Sell/Trade Vintage Computers ] [ and academia at www.VintageTech.com || at http://marketplace.vintage.org ]