Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!cs.utexas.edu!rutgers!aramis.rutgers.edu!porthos.rutgers.edu!webber
From: webber@porthos.rutgers.edu (Bob Webber)
Newsgroups: comp.arch
Subject: Re: ENIAC Query
Message-ID: <Nov.14.05.57.52.1989.8685@porthos.rutgers.edu>
Date: 14 Nov 89 10:57:53 GMT
References: <38193@lll-winken.LLNL.GOV> <405@gvlv2.GVL.Unisys.COM>
Distribution: usa
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 55

In article <405@gvlv2.GVL.Unisys.COM>, kleonard@gvlv2.GVL.Unisys.COM (Ken Leonard) writes:
> In article <Nov.11.23.03.48.1989.3337@porthos.rutgers.edu> webber@porthos.rutgers.edu (Bob Webber) writes:
> * [re types of vacuum tubes used in ENIAC]
> *     4,200 6L6 Beam Power
> *     1,300 6V6 Beam Power
> well, for sure, neither if these is what one would call "low power consumption"
> but then...
> *       350 807 (Enlarged 6L6)
> I dimly recall ham transmitters that used a couple of these
> as final amplifier--at a plate power level of a couple hundred watts.
> Do I remember correctly, or do I have an advanced case of cranial decrepitude?
> What was ENIAC doing to need _that_much_ power in one stage of logic?
> Or did the builders include these just so they would have a place to fry
> their eggs in the morning?

Referring again to the Burk & Burk article (Annals of the History of Computing,
Volume 3, Number 4, October 1981, 310--398), a few comments might help clarify
this:

    1) We are talking about a machine that comprised 40 panels (each 3 feet
       deep by 2 feet wide by 8 feet high) consisting of 18,000 vacuum tubes,
       70,000 resistors, 10,000 capacitors, 1,500 relays, and 6,000 manual
       switches.  The machine operated at 78 fixed (direct current) voltage
       levels consuming a total of 140 kilowatts.

    2) Numbers and program signals were represented by a pulse of 50 volts
       with a sharp rise and fall and duration of 2.5 microseconds.  this
       pulse was buried in a 10 microsecond time period.  20 such units 
       formed the basic machine cycle (i.e., all operations started on
       the beginning of these 200 microsecond intervals).  Incorporated
       in such calculations was the usage of a saftey factor of 3 to 1
       in voltage and two to one in time (i.e., if 10 volts was needed
       within 10 microseconds, then the driving circuit was designed
       to do 30 volts in 5 microseconds).

    3) To increase reliability, 6.3 volt filaments were operated at
       5.7 volts and plate and screen power were limited to 25% of 
       their rated value.

    4) All signals were strong enough to prevent being destroyed by
       crosstalk from other wires.

It ran its first problem in December 1945, in 1946 was moved from the Moore
School at U Penn to BRL at Aberdeen Maryland where it continued to be
used until late 1955.

Referring to Brainerd and Sharpless (The ENIAC, Electrical Engineering, 
Feb 1948):
     There were 500,000 soldered joints in the computer.
     Over an 8,000 hour study, 400 tubes failed (one every 20 hours).
     A power shutdown results in 2 to 3 tubes failing, so as much as possible,
     it was left running.


---- BOB (webber@athos.rutgers.edu ; rutgers!athos.rutgers.edu!webber)