Schedule December 2006
return to main 1401 Restoration Page
go to Team Bios
Fri Dec 01 - Tape Team, Wed Dec 06 - general, Thu Dec 07 - Tape Team, Sat Dec 09 - 2nd Sat, Wed Dec 13 - general , Thu Dec 14 - Tape Team, Wed Dec 20 - general,
Fri Dec 01 - Tape Team,TAU & Tape Drive debug status: Dec 1, 2006 (Jeff. S and me)
Drive debug status:
- Pursued previously reported C-bit problem. (Was being written as 1, but always read as 0.) Discovered a loose screw lodged in the pin array of the drive's read amplifier card gate. It was lodged between the output pins of the C-bit's read head! Removed screw.
- The signal from the C-bit was now present, but weak. The output signal of the C-bit's read head was measured at 10 mV(p-p). All other bit signals were 30 mV(p-p). Discovered shredded tape residue on the inside track of the read portion of the read/write head. Cleaned head and the drive's tape path. C-bit's signal improved to 30 mV.
TAU debug status:
I suspect a bad signal level in the error detect circuitry.
- Tried to run the VRC diagnostic. (The one that originally detected the C-bit error.) The 1402 card reader would not read past the first card. The correct data was being loaded into locations 1 through 11 (the only columns used on first card), but the first instruction was not being executed. It seemed that the I-STAR was not having correct parity set in the thousands digit for location 1. The C-bit was set 0 rather than 1, all other bits were 0.
- Switched to hand code loops for debug.
- Wrote entire tape with 20-character standard (even) parity records. Read much of the tape back correctly. No errors.
- Tried to read back records using odd parity read instructions. Expected VRC errors on every record. Received VRC errors on some, but not all records (about 50%).
- Wrote entire tape with 20-character odd parity records. Read much of the tape back correctly using odd-parity read instructions. No errors.
- Tried to read back records using standard (even) parity read instructions. Expected VRC errors on every record. Received VRC errors on some, but not all records (about 50%). Whenever the "tape error" did not occur, the 1401's A-register indicated bad parity even though it contained correct parity and data.
and follow up e-mails
The 1402 problem I reported is not "gang punch" related. I believe the bug is in the 1401 CPU address generation circuitry. Regards, Bob --- Robert Garner wrote: > Bob, Jeff, Allen, et al, > > Thanks again for your great bug sleuthing and slaying > on the 729 and TAU circuits. > > Definitive progress towards a working tape drive! > > - Robert > > p.s. Sorry to hear about your 1402 reader problem. > While investigating why card bit locations are occasionally flipping > when they're being written into core memory (when cards have many holes/1's) > Ron, Bill, and Ed were fiddling with the 1402's 21V power supply on Weds...
Wed Dec 06 - general,Received from John Van Gardner - Dec 3, 2006
I just read your log entry about the 1402 problem when there are lots of holes in the card. This was a common problem with all card readers that used read brushes to flip cores. It first showed up on the 714 card reader after the Column Binary feature was installed to read binary cards.
The most common cause was low voltage getting to the contact roll. This could be caused by low voltage out of the power supply or anything causing high resistance from the supply to the roll. Some of the causes were worn or burned common brushes, loose screws in terminal blocks and frayed or broken wires. I wrote a detailed story about my first encounter with this problem in my story # 23 Chinese Binary.
The best way to see if this is your problem is to put one scope probe on one of the brush wires where it plugs into the connector block. Put the other scope probe on the output of the voltage supply and run some cards that have all the holes in the 4 row punched. When the 4 row feeds past the read brushes you will see a square wave and can measure the voltage on the roll. If it is much different than the supply you can move the second probe along the wiring path from the supply to the common brush and tell where the difference occurs.
Received from John Van Gardner - Dec 4, 2006
1401 Maint Manual page 95 I was looking at the 1401 Maintenance Manual 225-6487-3 pages 180 and 181 which show scope pictures and set up to check the read brushes and impulse CBs. There is a good diagram on page 95. I wish I knew some way to capture one page of a .PDF file and send it
The row bit cores are in the 1401 array and I marked with red where the -20 attaches. There is a chart on page 22.214.171.124 that shows the actual pin connections to the array. Also marked is a note that says, "Machines with serial numbers above 25000 have 1/2 write line removed. This allows the row bit core to set with full I from read brush".
ALD page 126.96.36.199 has a chart that show the connections between row bit cores and the read brushes. There is a note at the bottom of this chart that says, "All brush signal lines have a 180 ohm series resistor before 01A1 edge connector termination on the paddle card connector". I wondered if this resistor value is the same for all serial number 1401s. When they did away with the 1/2 write line how did they increase the current through the brush? Assuming no other resistance 21 volts through a 180 ohm resistor = 117ma. If you have all 80 columns punched in the same row that's 9.3 amps from the supply and through the transistors driving the common brush. Any resistance in the common circuit could explain why it fails when there are lots of holes punched.
Robert B Garner wrote: John, Thx for your note. A 9.3-amp pulse to our old 1402 21V power supply could be our issue. I'm not familiar with required core switching currents (in general, and wrt 1401). Is 117 mA already a marginal value?The only cores that I actually knew what the switching currents were in the 737 memory on the 704. We had to scope the terminator resistors for the X and Y array drive lines and look for a voltage level that represented 410ma +or- 10ma. These X and Y drive lines only went through the hole in 64 cores one time. We did not have any vacuum tubes that could supply this amount of current so that's why they used switch cores to drive them. As cores became smaller with less mass it took less current to flip them. The cores in the 7302 memory on the 7090 could pass through the hole in a 737 core. Then the 360 cores could pass through the hole in the 7302 cores.What does it mean to "remove 1/2 write line"? Why would they do that? (cost optimization?)If you look at the left hand red arrow on the attached diagram you will see it points to a line labeled 1/2 write * that goes from the power transistor up through a resistor then over to the right and down through the row bit core. This is the line that was removed on the 25000 serial number machines. This would certainly help reduce the cost of the array as these were all hand wired at that time. There is a second line from the power transistor through the read brush then through a resistor then to the row bit core. Notice that when this line goes through the core it is wrapped around the core several times then to the -20 volts. The extra turns help flip the core with less current.
Robert B Garner wrote: Van, Thanks for your advice! I suspect it will help us immensely again. > We had to scope the terminator resistors for the X and Y > array drive lines and look for a voltage level that represented 410ma +or- 10ma. I've been trying to encourage the guys to do exactly this to find out what's wrong with our extended core memory. > There is a second line from the power transistor through > the read brush then through a resistor then to the row bit core. It's amazing that the current directly from the read brushes is used to flip a core in main memory! (Imagine all the junky noise that could get on it via the ground loop formed between the 1401 and the 1402 in a noisy environment due to nearby elevators, etc.). - Robert
Wednesday finally came, and seven dwarfs went to "work" -
- Present were Ron Williams, Bob Erickson, Allen Palmer, Frank King, Glenn Lea, Chuck Kantman, Ed Thelen.
Allen Palmer arrived with his pickup loaded with re-furbished parts from Grant Savier's work shop. Here the truck is about half unloaded. So much weight (many motors and clutches) on such a little cart.
Frank King arrived with some cleaned card track parts. and we washed some more surface parts in a CHM vendor kitchen ;-)) Now to put the 083 sorter back together. Anyone remember how this thing came apart - and where are the screws? ;-))
- Bob Erickson and Glenn Lea continued cutting and fitting plastic shielding for the 077 collator. The general philosophy is to prevent accidental touching of moving parts, not to completely encase the 077.
Randy Neff came by and worried that if people reached around the shielding, bad things could happen.
Ron Williams was bug shooting a problem in the 1401, but occasionally he and Ed Thelen looked at the 1402 card reader problem. I were surprised to find:
- The roller the brushes contact is a hard rubberoid substance, not bronze.
- The roller is normally at -21 volts, rising to about 0 volts during the 12 pulses of the card read cycle - the 80 write core wires in the card reader core are normally at -21 volts so maybe this might reduce hazard of smoking the limiting resistors in case of some fault. See December 4th diagram above.
And a little study of the two read stations in the 1402. The read brush assembly of the second read station is removed for viewing.
- Ed talked some with Jim Sumers about the coming demo - Feb/March. He seemed pleased with the thoughts so far. Apparently the administration is contemplating demoing the operational machines (PDP-1 and projected IBM 1401) on alternate Saturdays - alternate being the PDP-1 one say even Saturdays and the 1401 on odd Saturdays - something like that?
Jim was concerned about my projected visitor training time to operate a key punch to key punch their name into a card - to be read into the 1401 and printed in BIG letters, etc and a description of the 1401. (We hope to have 3 operational keypunches by formal demo time.) I had projected 5 minutes to get someone familiar with feeding a card and successfully keying their name.
- Jim wondered what we would do if we had say 15 visitors which might overwhelm that procedure.
- Maybe there would have to be a plan be for say 5 or more people, maybe a docent feeds and registers the card and lets the visitor operate the keys, or maybe I am a worry wart, or we need to get more creative -
I showed Jim something like the following that now works with 1401 emulation under Ron Mak's ROPE :-)). This is currently coded in 1401 AutoCoder, could be recoded in the not so nice SPS format, the smaller less demanding 1401 assembler.
- The 700 character format table and the 700 character program are just not practical to keypunch in Object Card format. (I would really be great to get an assembler working, or a convenient method of getting object decks punched.)
There are plenty of test decks that can make:
- - the card reader pass cards,
- - the printer bang
- - a tape move with the tape drive chattering loudly,
*but* I think if a visitor
- 1) key punches his/her name into a card,
- 2) maybe sees the card go zinging through a 083 sorterto take home and show friends
- 3) sees the 1401 read the card,
- 4) listens to the printer banging out his/her name,
- 5) and gets a printout of his/her name in BIG BOLD PRINT
- 6) and gets the card he/she punched to take home
that is not only charming and memorable,
- - it is lightning and thunder !! :-))JJJJJJ UU UU DDDD YY YY JJJJJJ UU UU DDDD YY YY JJ UU UU DD DD YY YY JJ UU UU DD DD YY YY JJ UU UU DD DD YY YY JJ UU UU DD DD YY YY JJ UU UU DD DD YY YY JJ UU UU DD DD YY YY JJ UU UU DD DD YY JJ UU UU DD DD YY JJ JJ UU UU DD DD YY JJ JJ UU UU DD DD YY JJJJ UUUUUU DDDD YY JJJJ UUUUUU DDDD YY GGGGGG AAAAAA RRRRRRRR LL AAAAAA NN NN DDDD GGGGGG AAAAAA RRRRRRRR LL AAAAAA NN NN DDDD GG GG AA AA RR RR LL AA AA NN NN DD DD GG GG AA AA RR RR LL AA AA NN NN DD DD GG AA AA RR RR LL AA AA NNNN NN DD DD GG AA AA RR RR LL AA AA NNNN NN DD DD GG GGGGGG AA AA RRRRRRRR LL AA AA NN NN NN DD DD GG GGGGGG AA AA RRRRRRRR LL AA AA NN NN NN DD DD GG GG AAAAAAAAAA RR RR LL AAAAAAAAAA NN NNNN DD DD GG GG AAAAAAAAAA RR RR LL AAAAAAAAAA NN NNNN DD DD GG GG AA AA RR RR LL AA AA NN NN DD DD GG GG AA AA RR RR LL AA AA NN NN DD DD GGGGGGGG AA AA RR RR LLLLLLLLLL AA AA NN NN DDDD GGGGGGGG AA AA RR RR LLLLLLLLLL AA AA NN NN DDDD V V III SSSS III TTTTT EEEEE DD TTTTT H H EEEEE CCC OOO M M PPPP U U TTTTT EEEEE RRRR V V I S I T E D D T H H E C C O O MM MM P P U U T E R R V V I S I T E D D T H H E C O O M M M P P U U T E R R V V I SSS I T EEEE D D T HHHHH EEEE C O O M M M PPPP U U T EEEE RRRR V V I S I T E D D T H H E C O O M M P U U T E R R V V I S I T E D D T H H E C C O O M M P U U T E R R V III SSSS III T EEEEE DD T H H EEEEE CCC OOO M M P UUU T EEEEE R R H H III SSSS TTTTT OOO RRRR Y Y M M U U SSSS EEEEE U U M M H H I S T O O R R Y Y MM MM U U S E U U MM MM H H I S T O O R R Y Y M M M U U S E U U M M M HHHHH I SSS T O O RRRR Y Y M M M U U SSS EEEE U U M M M H H I S T O O R R Y M M U U S E U U M M H H I S T O O R R Y M M U U S E U U M M H H III SSSS T OOO R R Y M M UUU SSSS EEEEE UUU M M AAA PPPP RRRR III L 1 222 000 000 66 A A P P R R I L 11 2 2 0 0 0 0 6 A A P P R R I L 1 2 0 00 0 00 6 A A PPPP RRRR I L 1 2 0 0 0 0 0 0 6666 AAAAA P R R I L 1 ,, 22 00 0 00 0 6 6 A A P R R I L 1 , 2 0 0 0 0 6 6 A A P R R III LLLLL 111 , 22222 000 000 666 IBM 1401 (WORDS AND MUSIC BY ROBERT GARNER ;-)) THE IBM -1401 DATA PROCESSING SYSTEM-, INTRODUCED IN OCTOBER 1959, WAS THE -MODEL-T FORD OF THE COMPUTER INDUSTRY- AND -ONE OF THE MOST IMPORTANT AND SUCCESSFUL PRODUCTS IBM HAD EVER ANNOUNCED.- ABOUT 15,000 1401-FAMILY COMPUTERS WERE ...
Thu Dec 07 - Tape Team
TAU and Tape Drive Debug Status 12/7/06 - Ron W. & me729 Drive Debug:
- The drive did not want to load tape today when we powered up the system.
- Ron found that the mercury switch behind the take-up motor had fallen out of its clip.
- We also found that a wire going to the supply side capstan microswitch was broken. I probably broke the wire troubleshooting the mercury switch problem.
- The drive worked well the rest of the day.
- Troubleshot the missing "Tape Error" problem that I reported last week. Found two bad cards in the TAU. Now "Tape Error" seems to perform correctly.
- Ran the full set of tape diagnostics. Intermittent errors still occur, but only on about 1 in every 100 records. (Backspacing and retrying the operation seems to work 100% of the time.) Often errors are reported, but the data is transferred correctly. Other times it seems that we pick or drop the 8-bit (as in C-B-A-8-4-2-1). These may be real problems or just lack of proper adjustment. It's time to fine-tune the drive adjustments.
We had trouble reading the diagnostic decks. There were frequent "Reader Checks". Sometimes we had to read a deck several times to get the program to load correctly.
Sat Dec 09 - generalReport via Ron Williams
Present were: Ron Williams, Bob Erickson, Tim Coslet, Jeff Stutzman. Ron Crane came mid afternoon.
Star errors when reading or punching a card - false errors
Troubles writing/reading tape drive - 8 bit dead ? where?
Bob Feretich on vacation until mid Jan.
Wasn't a volunteer day? - no free food - Jim Somers owes us ;-))
Addendum to previous Thursday report- e-mail between Robert Garner and Bob Feretich
> > Ran the full set of tape diagnostics > > Fantastic! How many distinct tests was this? * 5000 Card to Tape * 5030 Backspace Skip * 5040 Move Tape * 5050 Load Tape * 5080 VRC Check * 5500 IRG Check - This test was executed, but I don't know how to interpret the results. The documentation did not cover the Model V drive. > > Can you summarize (for the web site) what they all do? > How many records total to you think were read/written > in the course of the day? Several thousand records. My debug hand loops wrote and read the entire tape several times. > Since Ron was with you, did he say whether they made any > progress on the 1402 gang read -> core problem on Weds? > We (incl Van Gardner back east) were very suspicious > of the 1402's 21V power supply. No. He did fix the addressing problem that I reported last week though. > Perhaps I should just bring in some ferrite beads? ;-) I would suggest sacrificing a goat. It always worked for me. Regards, Bob
Wed Dec 13 - General
- Present were Ron Williams, Bob Erickson, Frank King, Bill Flora, Ed Thelen.
- Robert Garner and Grant Saviers were in the building attending high level CHM meetings.
- Robert came by for a half hour at noon, and Grant dropped by about 5:00 PM
It was rumored that Robert was meeting with the museum's restoration committee :-|
- Like who are they??
- *THE PROJECT* for the day was to get the card reader working reliably. (It has been marginal and sub-marginal for the past 4 months.)
- There has been a lot of focus on the brush contact rolls and power supply for a month. Today Bill Flora discovered the brushes that control the voltage to the brush contact roll (not the read brushes) were installed in the wrong direction. We believe the machine was purchased in this condition. I believe this contact brush condition has been corrected.
While working on the above, Bill Flora noted that the feed rolls into and through the two read stations seemed to have detents. We all got our fingers in there and felt the very definite "detent" or negative lump once per revolution.
- The fiber rollers were removed, and the non-uniformity of diameter was measured to be about 0.004 inches. As the "detents" were cylindrical in form, and the micrometer lands are flat, it is possible the "detents" are larger. Folks estimated that reduction of radius of the "detented" fiber rollers by 0.005 inch would be good.
- Gears were removed from their shafts by punching out the tapered retaining pins.
Tapered pin removal is regarded as an art.
- Not enough impact on the drift (the clyndrical tool use to push on the pin) and the pin does not move, but starts to mushroom and become more difficult to remove.
- Too much impact on the drift, and you can bruise hole in the shaft.
- Bob Erickson has a jeweler's lathe, but it is not long enough to hold the roller shaft.
- Robert Garner called Grant on his cell phone - and Grant promised to come by later.
- Since Grant had not come by when we left (about 4:00), ??? took the shafts home.
- Grant came by about 5:00 and found no one and no shafts :-((
- There is the continuing question as to whether the machinery of 1402 card reader/punch should be completely overhauled, as was the case with the 729 tape units.
The second (Sellam's) keypunch works, but Bob Erickson was worried that by-passing to open hold coil on Relay #2 was over heating the low resistance pick coil of this relay. (The original fix suggested by "Chuck" at firstname.lastname@example.org. website)
The game was to reduce the current through the pick coil during the hold operation. (It requires a lot more magnetism to pull in a relay than to hold it in the pulled-in position. That is a reason many relays use two coils.) Our pick coil is 1.2 K ohm, excited by 160 volts. The calculated power dissipation is about 21 watts :-(( I imagine as the coil heats the resistance increases - but not by enough. The target resistance of the open hold coil was about 3.3 K ohms.
Lets not mention blowing the 2 amp slow blow power supply fuse while probing about. And the anguish of replacing it with the only practical fuse we had - a 5 amp fast blow.
In any case we decided to place a little electric lamp in series with the jumper "we" installed between the pick and hold coil, as suggested in an e-mail, to reduce the voltage on the coil during hold time. Now - what wattage lamp?? We measured the resistance of a 7 watt Christmas bulb, and watched the cold resistance of 183 ohms rise past 200 ohms as the slight current from the ohmmeter heated the tungsten filament. - Oh boy, OK time for experimentation as we had neither the resistance coefficients nor the inclination to try to calculate and guess several unknowns.
Here is the seven watt bulb, lit during the hold phase of Relay #2 (when a card is registered in the punch station.) The chassis on the floor at the right of the keypunch is a heavy duty resistor load on "long term loan" from the PDP-1 guys ;-))
The results were very positive :-))
- - The coil stayed quite cool, The surface maybe 5 degrees above ambient :-))
- - The lamp had about 100 volts across it - should last for years assuming failure by tungsten evaporation :-))
- rather than temperature cycling across the phase temperature of tungsten :-((
- - The hold in force of the relay armature was judged "just fine" by our panel of amateurs.
We also tested the above circuit with a 4 watt bulb - but the hold in force was judged "marginal".
- With the seven watt bulb, we seemed well within reliable, safe operating range.
"We" (I) decided that we had passed ninth grade basic electricity :-)) Oh Joy!! AT LAST!!
- Bob Erickson assumed responsibility for making a proper installation in the keypunch (and getting a proper fuse? and making proper documentation?, and clearing this with EPA and the environmental impact people?.)
And the final result is imaged to the left :-) Image supplied Jan 24, 2007
Frank King was testing the 083 card reader, which has been jamming - and got this -
It is interesting to watch experienced CEs removing jams. The first thing is to mark the cards with their estimated sequence position - so that when the cards are removed, (and de-wrinkled and maybe re-keypunched) the deck can be ordered as correctly as possible.
Bob Erickson continued with his project of encasing the 077 in plastic for visibility of the many moving parts, with minimal risk to fingers and neck ties.
- Much finer iron power (used in mag tape magnetic clutches) had been brought in by Allen Palmer last time. Ed Thelen wanted use this powder (in solution) to look at magnetic fields on the tape to try to localize the "Bit 8" problem. (It is currently unknown if bit 8 is being written and not read, or visa versa.) But 1401 power stayed off to permit safe investigation of the 1402 card reader. (A much higher priority)
And there also appears to be a time handling timing problem? Bob Feretich wrote a tape full of "V"s last week, and when different test was run, not only was bit 8 missing of the records written in this test,
- but there was an extraneous "V" at the beginning of every record as viewed in the 1401 memory and printer.
Somehow, the previous contents of the tape are not being properly erased, and are appearing in following reads.
- Oddly, there was always exactly one and only one extra "V". ????
- At least four of us plan to work/play with/on the 1401 system next Wednesday -
Thu Dec 14 - Tape Team
TAU Debug Status 12/14/06 It would have been nice if someone would have posted e-mail about the 1401 not being available to be powered up. Since I could not do hardware debug, I troubleshot a previously recorded software bug. The "Move Tape" diagnostic program would fall into an infinite loop repeating the first of about fourteen data records. The documentation in the Diagnostic manual gave a good hint on the program segment that was failing. Examination of the diagnostic deck revealed that a card had been mis-punched by Cardamation. All of the card's punches were shifted left 7 columns. A seven-byte instruction was thus omitted. The instruction shifted into that gap unconditionally branched, so the 1401 was able to execute through the gap without a program check. This was the second Cardmation mis-punched card found in the new batch of diagnostics. (The mis-punched card contained all of the data, but the punches were slightly misaligned.) I punched a replacement card, but I could not verify the fix due to the power situation. Next session: I will be doing some traveling and will not be available until 2007. See you all next year. If you need to contact me, e-mail is the best way. Merry/Happy Christmas/Chanukah/Kwanzaa/New_Year. Regards, Bob Feretich
Wed Dec 20 - general
- Present were: Ron Williams, Bill Flora, Allen Palmer, Bob Erickson, Ed Thelen, Robert Garner.
- Ron Williams had decreased the radius of one of the pair of idler rollers on a 1402 Reader shaft by 0.011 inch. This was installed, and the reader worked much better - read 4 passes of the Ripple Test with no problem. Much Better !!
- It works so well that we detected a card ( in the diagnostic function tests - punched and printed back east) with a printed column with no punches
- Allen Palmer is beginning to reassemble the second 729 Mod V tape drive. The power supply, checked for voltage and current output, has been reconnected.
- The project of enclosing the 077 collator ran into a snag. Gluing the complex shapes with what appears to be kid's airplane glue (as supplied by Tap Plastics) does not seem to work. A long section glued last time parted much too easily. And the results of the appearance do not seem promising.
- We are getting discouraged - maybe time to call in the Pros?
- Ron Williams suggests that we seem to be restricted to flat plates, shields, end plates, etc.
- Examples could be - a plastic on an end of the 1401,
- - plastic replacing the side doors of a 729 tape drive, etc.
- Ed Thelen bit flipped in a "test" he wrote that reads and prints cards. :-))
- (Ron Williams only had to correct a few bits as they were being toggled in :-))
Merry Christmas, Happy Holidays, :-))
Go to January 2007
go to Team Bios