Schedule October 2005

Attendees Ron Williams, Tim Coslet, Ed Thelen - later we kidnapped Jim Manley with his young eyes and steady hands for soldering.

Ron had a good day :-)) Got the printer controller to go through one print line cycle - unfortunately, you have to reset the 1401 to go through another print cycle. (No print hammers actually move.) The 1401 has an optional Printer Buffer installed, great for speed, (the 1401 can quickly load the Printer Buffer then start processing the next instruction without waiting for the print operation to complete) but "we" don't know how it works. Have to decode how it should work from the ALDs - time consuming!

When Ron finds a sick card, he can often identify which pin is not acting correctly, and can often identify say the input diode as open. He writes a note of his findings, bags the sick SMS card, and places it in a box for Tim Coslet to further check when Tim comes in on Saturdays.

	This is Tim Coslet's test jig and scope. He uses a small AC voltage in series with a resistor to non-destructively wiggle the various circuits in the card. Knowledge, experience, and replacement cards for comparison help identify the part that is not acting correctly.

Tim identifies the failed component and marks it with a green dot. Tim's hands shake, so it us usually convenient for a steadier handed person to do the actual replacement. The replaced component is marked with some red fingernail polish for tracking. The repaired card with updated tag is placed back in the plastic bag and placed in the "repaired" box.

Tim Coslet came in and identified a number of dead components on the defective SMS cards that Ron Williams has been identifying and swapping out of the machine. Ron finds about 2 bad cards/working day. Tim then marks the component to be replaced and hands the card to Ed for removal and replacement. After a while Jim Manley with young eyes and steady hands came in - and got chained to the bench - good slave :-))

Here are a few of the cards that got fixed today.

It is a little alarming that so many passive components (inductors, basically a coil of wire in plastic) and simple diodes fail. (In the glass diodes, corrosion seems to creep along the wire to glass contact, expanding and cracking the glass, and the diode fails by some mechanism.) To date we have replaced maybe 30 diodes and 10 inductors? Scary.

Attendees Ron Williams, Robert Garner, Ron Crane, Don Cull, Robert Feretich, Bill Flora, Ed Thelen. Hmmm, how to say this - politely? Today was not a winner. Ron Williams and Robert Feretich continued struggling to figure the desired operation of the 1401 Printer Buffer option from the prints as there seems no manual describing operation. This option has an added core memory buffer and drivers, and is not at all obvious what is supposed to go on in a detailed sense.

Bill F and Don C struggled with 1402 trying to read one card. Ed Thelen struggled with removing failed components from SMS cards so that he could replace them. Ron Crane struggled diagnosing failed cards. He found another transistor lead completely corroded through - like the three legged transistor was standing on 2 legs. Ron C seems much more adept than most folks at removing and replacing components on SMS cards. - We changed out about 12 transistors and 4 diodes. Just one 2N1302 or equivalent transistor left. Ed then struggled trying to find 2N1302 or equivalent germanium transistors on the Internet - and did not want to pay $5 each. Ron C pointed out that the "germanium" diodes that Ed had purchased from HSC Electronics were in fact silicon (a difference in forward current bias but will be OK in most 1401 circuits if required.) Yes, a lot of struggling, not much to show for it.

Those 25 2N1303 germanium transistors that Grant Saviers purchased on e-bay now look more like a miracle. A few day later Robert Garner - suggested that we replace transistors more on an "as needed" basis, - bought 5 plus a another group of 3 via e-bay - supplied the original to the thumbnail to the left (300 K bytes)

see added notes on 2N1302 transistors above -
Present were Ron Williams, Bill Flora, Don Cull, Robert Feretich, Robert Garner, Ed Thelen.

First Light - In building astronomical telescopes there comes a time when some important things are working, like the dome, the telescope mounting and drive, and the Great Mirror is figured, polished and mounted. It is time to see if all this work has promise. There is the night when you finally see (or photograph) a star. This is called "First Light" - there is a lot of work yet to be done, but some major things work well enough to see something. Like taking your kid to 1st grade. - The 1401 works well enough to read in the first two diagnostic tests from the card reader, and execute the tests correctly :-))

Time to party, pop a cork, ... October 19, 2005 is a great day, but the IBM tradition seems more restrained ;-((

	The 1402 Reader timing "Circuit Breakers", CBs have been cleaned and adjusted, the read brushes checked, the card path transport works, two broken wires found and repaired, relays checked, timing light struggled with, belts replaced, bearings checked and lubricated, on and on, ...
	In the 1401 processor, the multiple power supplies checked and working, some 70 faults in the 1401, including about 50 cards have been replaced/fixed, the clocks and lower memory works, instruction fetch and execute cycles work, about half of the instructions work when manually inserted via switches, ... . And at long last, the card reader and processor work well enough to read and execute the first two diagnostic decks :-)) Here are the 1402 Reader/Punch and 1401 Processor team members present at First Light :-))
	And the original Diagnostic (and other decks) were read in using a borrowed card reader to CD ROM and other storage. Diagnostic deck information Here are the Diagnostic Decks that still yet to be performed correctly. Note that decks one and two are missing - the machine read in and performed these two tests correctly :-))
A 3.2 megabyte .wmv file	Here is a 30 second movie of the first attempt to read a large deck (of blank cards) into the 1401 via the card reader - note that there is work yet to be done ;-|

see added notes on 2N1302 transistors above -
Present were Ron Williams, Robert Feretich, Bob Erickson, Tim Coslet, and (briefly) Ed Thelen.

Robert Garner requested that I (Grant Saviers) update the team on our 729 reel clutch rebuild sub-project.

Allen Palmer & I have spent in excess of ten man-days working on the rebuild of the six clutches for the 729 V his team initially restored to tape moving status. This report provides a summary of what has been accomplished. An expanded version will be posted of the 1401 site with additional data, graphs, and pictures. Why rebuild? The initial reason after Allen got the 729 moving tape was that the right reel clutch assembly started to make seriously bad noises. Disassembly of this three clutch assembly revealed four problems: Interference between the clutch back cover and driven disk. In a prior report (see pictures on the 1401 site) it was believed that the clutch may not have been manufactured correctly. This still may be true, however we now know that improper re-assembly of the three clutches on the common shaft can cause this interference. Interference and cover wear was also discovered in one of the left reel clutches, strengthening the improper re-assembly as the likely cause. Contamination of the clutch powder by rear cover aluminum debris and steel wear particles from the driven disk. Worn out felt seals between the bearings and the clutch powder. Damaged clutch bearings – worn, lumpy, and dry grease. Initial investigations: The three right reel clutches were disassembled and photographed, see prior report. The left reel clutches were also disassembled and in this process we created a problem by pulling a clutch coil loose. We believe we will be able to solder the leads back to the slip rings although it is a touchy proposition. A new bearing press out fixture has been designed but not built that allows removal of the bearings from the other side, eliminating the stress on the coil assembly. Electrical measurements: The six coils averaged about 104 ohm resistance and about 0.24 henry inductance. Repacement bearings: Multiple searches and discussions with bearing makers were unsuccessful. IBM had a special bearing made, it was last produced in 1972. We can get them made, 1000pc minimum order, about $50 each. An alternative using standard bearings will be used. (see the web site for data and chart regarding axial angularity stiffness tests). Clutch powder: A replacement powder was obtained, however the IBM specs are unknown and little is known about the magnetic circuit properties of the clutch. Tests were undertaken with the new powder to see if the performance of the IBM powder could be duplicated. On the advice of the powder supplier, we adjusted the weight of the powder charge. It was also learned that the clutch powder (Placid Industries) is a 410L stainless steel alloy of 325 mesh balls. A trade secret powder is added to reduce squeal. The IBM powder is considerably finer and .75x the density than the new powder. However, coarser than 325 mesh 410L balls are used in currently produced larger clutches. We measured the sustained torque vs current characteristics for: - 22gms of IBM powder per IBM specifications - 29gms of new powder (same volume as IBM spec) – too much torque/amp - 17gms of new powder (guess-equal magnetic moment?) – too little torque/amp - 20gms of new powder – looks good! See the web site for the data and chart, also pictures of the test setup. Felt washers: No commercially available internal seal washers could be found so raw felt material was procured in F1 and F3 (high density and medium density resp.) grades in 1/8 and 3/16 thickness. The IBM CE manuals comment on two piece felt washers but none of the clutches so far have contained such a part. The medium density felt is intended as a springs for the high density felt as a seal (high density felt will be next to the bearing). Grant machined a four-piece die set that can punch the ID and OD of the washers concentrically and within a few thousandths’ tolerance. Enough washers have been made for the 6 clutches. The existing external felt seal washers will be used as they are much less worn. A die set has been designed but not built to make new external seals in F1 x 1/8” thick felt. Next steps: Allen is having the damaged threads on one shaft chased (restored). 12 new bearing retainer nuts (spares) are ordered with 50 new locking rings for all 30 clutches. 50 pcs of what I think is the right size bearing spacer (0.006”) have been received. Another man day or at worst two should have both left and right reel assemblies ready to re-install in the 729. Written by Grant Saviers 10-24-2005

Robert Garner and others are securing germanium transistors for the project.
Present were Ron Williams, Robert Feretich, Bill Flora, Frank King and Ed Thelen.

• Bill Flora worked on sense switches and logic in the 1402 card punch path.
• Ron Williams and Robert Feretich worked on printer control logic - it was hang in a state that a subsequent print command could not start.
• Frank King worked on 1403 printer control logic and lamp displays.
• Grant Saviers sent a further report about the 729 reel clutch rebuild sub-project.
  This work was performed at Grant's San Jose machine shop :-))

  	The clutch assembly (right reel) on the 729. The slip rings can be seen. The left two clutches are the tape up and tape down w.r.t. the columns and the front geared clutch is the brake and load/unload clutch.
  	The left reel assembly with the series wound motor used for high speed rewind. The other clutch functions are the same as 00249
  	Allen Palmer with version 1 of the torque measuring setup to test IBM vs Placid clutch powders.
  	Rev 2 of the powder torque vs current setup. Chatillon force gauges are connected to the aluminum swing arm to apply a calibrated force at three different radii. The DVM measures the coil current.
  	Grant Saviers using the hand press to make new internal felt seal washers with the dies designed for this purpose
  	The .xls spread sheet "clutch powder" and chart shows the measured torque values for different clutch powders and different amounts of the Placid powder. 20gms of Placid powder was selected as the best amount.
  	The .xls spread sheet "bearing deflections" shows the angular displacement vs load for three (none, one and two) 0.0039" shims between the two radial ball bearings to provide preload. The original MRC duplex bearing was also measured. As a result a compromise shim value of 0.006" will be used in the final assembly.

And some diagnostic e-mail - I guess I wish we had a blog -
On Thu, 2005-10-27 at 20:40 -0700, Bob Feretich wrote:

Van, The 1402 card reader seems to be mostly working, but we need a diagnostic to help us make the reading more reliable. Normally, we could run card-to-print and see which columns need to be adjusted, but the printer isn't working yet. (Ron Williams and I have been focusing on printing for the last three weeks.) Can you write a diagnostic that can validate reader reliability? I'm thinking that we could start with a card that was punched... 9876543210-&9876543210-&9876543210-& then rotate the card's punches 1 position for each subsequent card... 876543210-&9876543210-&9876543210-&9 76543210-&9876543210-&9876543210-&98 6543210-&9876543210-&9876543210-&987 and so forth. Since we may be scoping, the deck of data cards would need to last at least a minute.? If we had about 400 data cards we could move cards from the reader's output hopper to its input tray and keep it going indefinitely. The diagnostic program could read the first card, save it, and check following cards against the first one using the appropriate rotation permutation. That way we could use other multiple punch per column data cards in the future if we thought we needed them. What do you think? Can you provide such a diagnostic program? If you can, I can get Cardamation to punch up the cards. Regards, Bob Feretich

And Van Snyder responded

Bob: I think 1040A does what you want.? Run it with SS D and E on, and C off. SS D on doesn't try to chain-load the next program after reading 60 detail cards.? Instead, it just tries to read another 60 detail cards. SS E on halts instead of printing on errors.? SS C off suppresses printing correct results.? Also make sure not to put "1" in 1252 (S52) as that would turn on header printing. 1040A has the desired pattern built in, instead of getting it from the first detail card. Have you sent 1040A.cd and 1040A.in to Cardamation?? 1040A.cd produces the same core image as 1040-2object.txt, but it's 19 cards + detail instead of 61 cards + detail. The first card of 1040A.in is actually part of the program (as is card 62 of 1040-2object.txt), but in a format that Autocoder can't produce, and the headers are in the program. 1040Anew.cd is entirely in Autocoder format, but the first *two* cards of the detail deck (1040Anew.in) are header cards (instead of being part of the program in a goofy format). I've attached a gnuzip-compressed tar file with 1040A.s, 1040A.cd, 1040A.in, 1040A.lst, 1040Anew.s, 1040Anew.cd, 1040Anew.in, 1040Anew.lst, and 1040Abig.in (the goofy last card of the program plus ten copies -- 600 lines -- of the detail), all in Cardamation's ASCII-to-BCD code. WinZip knows how to read gnuzip-compressed tar files. If you want to patch the program so it doesn't halt on error, change the halt at 916 in 1040A to a NOP.? Insert the following card before the last one to make that patch: ....5...10...15...20...25...30...35...40...45...50...55...60...65...70...75 N? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? L001916B068 ? ? ? ? ? ? ? ? 1040 In 1040Anew, the error halt is at 593.? Insert the following card before the last one to make that patch: ....5...10...15...20...25...30...35...40...45...50...55...60...65...70...75 N? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? L001519B068 ? ? ? ? ? ? ? ? 1040 If you use 1040-2object.txt, insert the following card somewhere between 48 and 62 (sequence numbers are in cc 78-80): ....5...10...15...20...25...30...35...40...45...50...55...60...65...70...75 L0129161001N Let me know if 1040A won't do what you want. Van and presented a different way - with out what is ever causing the question marks > Bob: > > I think 1040A does what you want. Run it with SS D and E on, and C > off. > SS D on doesn't try to chain-load the next program after reading 60 > detail cards. Instead, it just tries to read another 60 detail cards. > SS E on halts instead of printing on errors. SS C off suppresses > printing correct results. Also make sure not to put "1" in 1252 (S52) > as that would turn on header printing. > > 1040A has the desired pattern built in, instead of getting it from the > first detail card. > > Have you sent 1040A.cd and 1040A.in to Cardamation? 1040A.cd produces > the same core image as 1040-2object.txt, but it's 19 cards + detail > instead of 61 cards + detail. > > The first card of 1040A.in is actually part of the program (as is card > 62 of 1040-2object.txt), but in a format that Autocoder can't produce, > and the headers are in the program. > > 1040Anew.cd is entirely in Autocoder format, but the first *two* cards > of the detail deck (1040Anew.in) are header cards (instead of being > part > of the program in a goofy format). > > I've attached a gnuzip-compressed tar file with 1040A.s, 1040A.cd, > 1040A.in, 1040A.lst, 1040Anew.s, 1040Anew.cd, 1040Anew.in, > 1040Anew.lst, > and 1040Abig.in (the goofy last card of the program plus ten copies -- > 600 lines -- of the detail), all in Cardamation's ASCII-to-BCD code. > WinZip knows how to read gnuzip-compressed tar files. > > If you want to patch the program so it doesn't halt on error, > change the > halt at 916 in 1040A to a NOP. Insert the following card before the > last one to make that patch: > > .... > 5...10...15...20...25...30...35...40...45...50...55...60...65...70...7 > 5 > N L001916B068 > 1040 > > In 1040Anew, the error halt is at 593. Insert the following card > before > the last one to make that patch: > > .... > 5...10...15...20...25...30...35...40...45...50...55...60...65...70...7 > 5 > N L001519B068 > 1040 > > If you use 1040-2object.txt, insert the following card somewhere > between > 48 and 62 (sequence numbers are in cc 78-80): > > .... > 5...10...15...20...25...30...35...40...45...50...55...60...65...70...7 > 5 > L0129161001N > > Let me know if 1040A won't do what you want. > > Van

and some printer stuff from Ron Williams - who hopefully understands it??

OK - we have the interesting challenge of printing a line of about 130 characters. In this printer, the characters are print slugs whizzing horizontally behind the ribbon, behind the paper, in front of the hammers. To print the line we have to fire the 130 hammers when the correct print slug is in front of the appropriate hammer. Interesting enough if we are printing one line a second, but we are printing at least 10 lines per second, and allowing time for the paper to move up a line and stop in between print times. This with relatively slow logic, like 500 nanoseconds switch time and 11.5 microseconds to fetch a character from memory. Frankly I have no clue - very curious!! Apparently there are several counters that operate in trick ways??? And there is also an (extra cost) option which allows the 1401 to continue executing instructions before the print cycle is complete.