Friday, October 13, 2023

The tricky installation of the tab and carrier return cords, as well as challenges loosening the rotate spring tension

CARRIER RETURN AND TAB CORDS POWER LEFT AND RIGHT MOVEMENT

There are two string like cords in a Selectric typewriter mechanism. One hooks to the left side of the carrier, able to pull the carrier to the left. The other cord hooks to the right side of the carrier and pulls it towards the right. Teeth fitting into racks hold the carrier in position against the rightward pull, allowing it to move one step during a space or to move freely during a tab until it strikes a tab stop that is set at some column. 

COUNTERROTATING DRUMS TAKING UP AND RELEASING THE CORDS

The carrier return cord is routed leftwards from the carrier and over a pair of pulleys on the left which reverse the direction of the cord and direct it behind the racks. It is wound around a drum at the rear of the machine - that drum is on a common shaft with a mainspring and a drum that sits forward of the spring. The rear drum winds up the carrier return cord as the carrier is moved to the left, feeding the cord out when the carrier is pulled to the left so that the return cord remains under steady tension. 

The tab cord is routed to the right side of the machine, changes direction over a special spring loaded pulley and is then wound around the drum in front of the mainspring. The mainspring will be wound up by the process of a carrier return, storing up energy necessary to pull the carrier rightward. The drum winds up the cord as the carrier is pulled rightward during spacing and tab operations, feeding out the cord during leftward (backspace and carrier return) operations. 

MAINSPRING POWERING RIGHTWARD MOVEMENT - SPACE AND TAB

The source of power to move the carrier to the right is the release of energy stored by winding up the mainspring. This winds as the common shaft rotates in one direction and seeks to rotate the shaft in the other direction unless the carrier is held in position, for example with a tooth in the escapement rack. The spring gains energy while the carrier is pulled leftwards during a return and gives it up to move the carrier rightward. IBM refers to this as negative energy for some reason but it is just the potential energy in the wound up spring that powers rightward movement. 

MOTOR POWERS LEFTWARD MOVEMENT - CARRIER RETURN AND BACKSPACE

The motor of a Selectric is always running, spinning the operational shaft but that shaft isn't coupled to anything else unless a clutch is activated. This is what IBM refers to as positive power in their training and documentation. In order to pull the carrier left in a return operation, the clutch connects the operational shaft to the common shaft with the cord drums and mainspring, turning it to simultaneously feed out tab cord, pull in carrier return cord, and wind up the spring. The power comes from the motor to make the movement and to restore energy into the mainspring. 

SPRING LOADED PULLEY MAINTAINS CORD TENSION

Since the cords can stretch over time, a spring loaded pulley is built into the right side of the Selectric. The pulley can move left or right but has strong spring tension rightward. This holds the cords in the proper tension and automatically adjusts to stretching, within the limits of its range of motion. 

INSTALLATION THEORY

The simple sounding theory for installing the cords is much more challenging to put into practice, the issue being getting the cords to be the right length and on the drums correctly, in addition to which setting the mainspring at its proper tension. 

The carrier return cord is hooked to the rear drum, threaded around the pulleys and attached to the left side of the carrier, while the carrier is sitting in its rightmost position. 

The carrier return mechanism is activated so that as you turn the operational shaft manually the rear drum takes up the cord and pulls the carrier leftward. Proceeding carefully so as to keep the cord winding on the correct spiral grooves of the drum, you move the carrier all the way to the left. 

Next one should attach the tab cord to the front drum, thread it around the spring loaded pulley and attach it to the right side of the carrier. Here there are challenges in real life. The drum is usually turned so that the attachment point is downward and away, not convenient to hook up a cord. The spring tension of the pulley will make it almost impossible to hook the cord on, so you really leave the cord off that pulley, hook to the carrier and then force the pulley leftward while pulling the cord over it. Alas, the cord may be too long so that the spring loaded pulley can't move rightward enough to put on tension. 

If the cord is hooked to the front drum and engaged in the spring loaded pulley with proper tension, one then has to wind up the mainspring to appropriate tension. If you do all this, the carrier is ready to move left and right, handling the carrier return, space, backspace and tab operations properly.

When I have the carrier full left but the front drum mount point is inaccessible, I need to move the carrier a bit to the right until the drum rotates. Unless you are very careful looking both at the front and back of the machine, you can loosen the carrier return cord making a jumble on the rear drum. 

I had to tie several knots in the cord next to the hook that connects to the right side of the carrier, to shorten the cord appropriately. I must then tension the mainspring by winding it nine or ten turns. If all went well, I should be able to space the carrier all the way to the right. During this movement I have to watch that the tab cord winds properly around the front drum. 

I have failed multiple times just as I am moving the carrier a bit to the right or trying to connect the tab cord to the front drum. This is very tricky, as moving the shaft the wrong way loosens cords which then scramble up on a drum. 

ROTATE SPRING TENSION IS TOO HIGH, BUT RELIEVING IT IS HARD

Inside the carrier there is a vertical rotating shaft that the type ball spins on. It has a spring underneath that provides power so that when the metal rotate cord is loosened, the rotate spring takes up that length by crisply turning the shaft. It has to move fast enough to get into position during typing. However, if it is too tight, then it restricts the range of rotation. This is the cause, I believe, of the failure of this typewriter to rotate to either the +5 or the -5 positions. 

This is checked using an IBM spring scale that pushes against the rotate tape pulley on the right side. One sets up a +5 character, hand cycles the machine halfway through a print operation so the rotate tape is appropriately tensioned, then push on the pulley and see 2 to 2.5 pounds of force. My spring is offscale beyond 3 pounds right now 

Adding tension to the rotation spring is easy - there are notches around its perimeter that you can see and reach from the top of the machine. Pushing it one way will turn it to click into the next notch, adding tension. However, releasing some pressure requires simultaneously holding it steady by the notch and also pulling on a detent spring to let it rotate the other way. I can't even see this detent, much less figure out a way to get a tool in to move it.

Assuming I find a way that I can pull on the detent while also having another tool control a slow rotation of the spring to release some energy, I can remove tension until the scale shows me the target force of 2 to 2.5 pounds. This is challenging which is why I have spent quite a few hours on the 1053 just trying to deal with the rotate spring; then there is the time fighting to get the tab and carrier return cords properly attached. 

Thursday, October 12, 2023

Cover in place - could fine tune the mounting hardware but acceptable already

COVER INSTALLED ONTO REAR OF 1130


The mechanism has a groove on the bottom mounts, allowing the cover to be placed onto the raised ridge inside the IBM 1130 display box. On the two sides are mounts with a rotating lever to either hold the cover behind the raised side ridge or allowing the cover to tilt out if the lever is oriented vertically. 

The lever has detents that lock hold it either vertically or sideways to extend behind the ridge. A spring holds the side screw heads outward, which pulls the lever against its stop plate inside so the lever locks into one of two orientations, vertical or sideways. To operate the cover, the screw is pushed forward against the cover, moving the lever out of its lock, rotated ninety degrees to its intended orientation and released. 

The change I anticipate making is to swap the screws on the sides with longer versions that allow me to install a locking nut on the inside end. The lever has a nut epoxied to the part but the screw therefore not only rotates the part but loosens or tightens. Using a second nut to jam against the first will keep the nut purely rotating with the screw head rather than having the screw threads move in or out.

Wednesday, October 11, 2023

Put together two designs and sent the boards off to be fabbed; challenge removing typeball from 1053 for more adjustments

REPLACEMENT FOR IBM 1130 CONSOLE LOADER PROJECT

The quick way I implemented the console loader was using two eight relay boards and one four relay board connected to an Arduino. The Console Entry Switches (CES) on the front panel of the typewriter printer are wired to the relays and the other side goes through a resistor to ground. Then there are two buttons, Prog Start and Load IAR, which are wired to three more relays. 

By activating one relay the power is removed from the existing Prog Start switch. The user turns the rotary mode control switch on the 1130 to Load and then uses a terminal program to interact with the Arduino. The text in the file then causes the Arduino to set the CES and push either Load IAR or Prog Start in order to set memory addresses and store contents. 

The issue is one of space and elegance. The relay boards take room under the cover of the 1130, in front of the keyboard, plus we need an additional box elsewhere to hold the Arduino. I designed a board to use optoisolators to replace the relays, that board mounts atop the Arduino and we only need to find a space for that smaller stack inside the console. I sent this off to the fab as well, since I already have the optoisolators and the Arduino. 

MUSEUM DISPLAY CONTROLLER

The Cape Canaveral Space Force Museum has several locations, one of which is outside the security perimeter thus readily accessible to the general public. It has a number of small LCD displays that run various videos discussing activities. These are older units and are behaving very badly. I suspect some kind of corruption to the files as they have repeatedly lost power over the years. 

I decided to whip up a Raspberry Pi based solution that can drive any HDMI monitor as a replacement. It makes use of the Pi Presents open source application to animate such displays. There are a few challenges that aren't properly addressed by the Pi hardware and application. 

Any time power fails, the Linux system running on the Raspberry Pi can leave the file system in a corrupted state since it didn't shut down gracefully. The Pi Presents software can schedule automatic start and shutdown of the displays, but it needs a persistent clock or an internet connection to know the current time. The building does not have internet connectivity. Finally, the staff needs an easy way to start these up for an unscheduled event and to shut them down when leaving at the end of that event. 

I whipped up a circuit that will be built as a 'PI Hat', a board that fits above the Raspberry Pi and plugs into the 40 pin connector. It will provide a real time clock with battery protection so that the system always knows the correct time. It will feature a supercapacitor based system that provides 90 seconds or more of power when the input power drops, plus sends a shutdown signal to Linux, allowing the system to gracefully close all file systems. It has an autostart feature when power comes back on, a start button for manual startup, and a shutdown button that triggers a Linux shutdown. 

This was completed and sent off to the fab for production. It mostly uses surface mount components for compactness. The supercapacitor system that charges the capacitors and then uses them to produce 5V for the machine during power outages is built around a MAX38889 chip. The real time clock is built around a DS3231 chip and a CR2032 battery. The battery needs to be replaced every two years and the capacitors should be replaced every eight years. 

IBM SELECTRIC 1 TYPEBALL BROKEN LEVERS

When IBM first introduced the Selectric typewriter, the typeballs had a plastic lever and cap to remove them from the typewriter, but they soon switched over to a design that had a knurled metal cylinder that the lever was molded onto. The newer construction produced balls that tend to break so that the lever snaps and the element won't detach from the typewriter. The metal cylinder is left on the cap but can't be turned without a lever.

Inside the cap is a strong metal spring that has two arms which snap shut into a groove on the spindle of the typewriter. The metal cylinder has ramped ends that will push the two arms apart so that the ball can be pulled up and off. The lever would cause the cylinder to rotate, that is until it snapped off. This is a very very common failure of these early type elements. IBM eventually went back to all plastic mechanisms which don't break. 

All of the 1053 console printers and most other I/O Selectric devices use the typeballs with metal cylinders, meaning on most of them the ball is stuck on the unit. This doesn't cause a problem in operation but it would be nice to remove the element to clean off built up ink and it is important to make certain adjustments with the ball out of the way. 

The space through the plastic cap to the ends of  the spring arms is very tight, thus ordinary tools like snap ring pliers can't fit in. I will work out a way with my existing tools and perhaps the hardware store to remove the ball for the current restoration, but I want to accomplish two things longer term. First, I want to modify the balls to have a working removal lever. Second, I want to fabricate a tool to easily remove these elements with broken levers. 


Friday, October 6, 2023

Still working on 1053 character selection after a busy day at Cape Canaveral

ASKED TO JUMP IN AND HELP AT SPACE FORCE MUSEUM TODAY

A tour van was scheduled to visit Hangar C and the main museum at LC26, but the original staff who would have met them were unable to work it. I was called to head over, open up the buildings and deal with the visitors. I left around 8:30 this morning and wrapped up the tour support by lunchtime.

ATLAS V LAUNCH SUPPORT IMMEDIATELY AFTERWARDS

United Launch Alliance had an Atlas V rocket scheduled to launch the first two Kuiper internet satellites to allow them to be verified in space. Kuiper is the constellation of birds that Jeff Bezo's Blue Origin plans to put into orbit in order to offer a competitive satellite internet service.  It was scheduled to launch at 2PM, so I had time to stop by the base cafe for a quick lunch and then police the visitors watching the launch from ITL road. All in all, I didn't bet back to the shop until after 3:30PM because of all the above. 

REFINING DOOR MOUNTING HARDWARE

Having finished both sets of mounts and pivot mechanisms for the rear cover of the 1130 display pedestal, I did a test fit and wasn't happy with the operation of the latch. I switched to a longer bolt that gives more room for the spring loaded movement to release and turn the pivot.  

1053 ROTATION STILL NOT SATISFACTORY

The 1053 console printer is having issues moving the typeball to the -5 position. The ball stops turning around -4, when the detents lock the ball into its incorrect position for typing. Something is maladjusted but I am not yet certain of the culprit. 

When this is combined with the failure of the positive rotation latches to return under the restoring bail, I suspect there are multiple problems. I can't imagine how a Selectric mechanism got this far out of adjustment, unless someone had tried to work on it in the past and randomly changed all sorts of adjustments. 

Bottom line is that I don't understand yet why these two problems exist. I do have both my own 1053 and the parts donor Selectric to compare against, so sooner or later I will spot the critical difference(s) and then they can be corrected. 


Confession - being too reliant on black box components, my uTracer Vacuum Tube Curve Tracer heater regulation board project didn't work

UTRACER AND HEATER REGULATOR BOARD

A while back I built a vacuum tube curve tracer designed by Ronald Dekker which measures tubes and produces curves to compare to the spec sheets to assess tube health and performance. One weakness of this design was in its ability to drive the filaments, which is overcome by a heater regulator board design from Stephan Lafferty. 

REGULATOR BOARD BUILT

The regulator board made use of a very accurate 2.5V reference source to precisely control the voltage of the filament supply through an opamp feedback loop, with extreme precision resistors used to set the voltage points to generate 5V, 6.3V and 12.6V output levels at up to 5A. 

It also is a four wire delivery system, meaning that there is a separate sense wire for both ground and + voltage levels. Thus any voltage drop on the supply wires to the tube socket is detected because the sense voltage is lower than the target. The heater regulator board design is controlled to drive whatever voltage is necessary to produce the exact filament voltage as seen by the sense wire. 

The buck converter used to drop the 19V incoming supply voltage to the desired filament voltage was a board sourced from eBay that was built around the XL4005 IC. This board was manufactured in China and provided 5A buck voltage regulation with either constant voltage or constant current control, but at an extremely low price, thus it was not worthwhile for Stephan to design his own converter. 


The part specified in the design of the heater regulator board was no longer available on eBay or Amazon. I found some similar boards, based on the XL4015 chip, whose specs and layout appeared to match. I chose to buy those but when I tested the overall heater regulator board, it didn't work. I lazily assumed that XL4015 was a more modern and potentially slightly cheaper chip that the makers had switched to. 

I finally switched on my brain recently and downloaded the spec sheets of the two chips, XL4005 and XL4015. Two differences were apparent - switching frequency and the set point for the regulation. 

The original chip operated at 300KHz while the newer regulator chip operated at a lower 180KHz. This might introduce some remnants into the uTracer which weren't filtered by the component values on the heater regulator board. 

The fatal difference, however, was in the set point of the chip. The XL4005 used 0.8V as the feedback level, while the newer XL4015 dropped that to 0.2V. The entire design of the heater regulator board was oriented around monitoring the filament output voltage and selecting a divider network to produce 0.8V exactly when it was on target, that being driven through the opamp to control the XL4005 board. 

I dug around once I realized this and found some different buck regulator boards that did used the XL4005. The first order I placed on Amazon disappointed as the boards came with XL4015 and the vendor had sneakily updated the item description on Amazon so that it now claimed to use XL4015. Luckily it wasn't a very high cost item so I simply dumped them in the trash.

The next try was successful, the boards were indeed built on XL4005. The modification for the heater regulator board opens the onboard feedback from a small potentiometer and instead connects the feedback pin to the opamp output. The board works perfectly. 

Thursday, October 5, 2023

Success with 3D printing of door hardware; continuing on 1053 typewriter restoration

THE MAGIC WORD IS Z OFFSET

I discovered an important setting for the printer, in addition to bed temperature, extrusion speed and extruder temperature. There is a parameter called Z Offset, which determines how high the extruder nozzle actually sits above the bed. Once the leveling is complete and the auto-leveling takes place as printing starts, the difference between the bed level determined by the leveling sensor and the tip of the nozzle is the offset.

The printer moves left and right on the X axis, forwards and back on the Y axis, while it moves up and down on the Z axis. After the auto-level completes it has a mesh of offsets from the presumed home level of Z=0, since it tests sixteen points across the bed to compensate for tilting and warping of the bed. To that we have to add the Z Offset to place the nozzle the ideal height above the bed. That is a tad less than the thickness of a sheet of printer paper, to produce the best results.

I experimentally determined the offset, set it in the software and immediately achieved successful printing of both the bottom mount and the pivoting side parts. I am delighted with the quality of the parts and can't wait to print the second set of parts and install it on the rear cover. No issues at all with adhesion, nothing fancy needed. 

Side pivot parts as they come off the printer

TYPEWRITER WRESTLING CONTINUES

I discovered a broken part on the side of the 1053, the cord self-tensioner. Two cord are used, to haul the carrier to the right and to wind the mainspring when the carrier does a carriage return to the left. The cord that pulls to the right is routed around a pulley with a strong spring, called the self tensioner. The purpose of this part is to absorb slack as the cords stretch with age and usage. 

The self-tensioner part can be made with plastic, but those parts are famous for cracking with age (and abuse as it is sticking out on the right side). 

Damaged tensioner from 1053

Luckily the parts donor machine I bought had a good tensioner - one that was made with metal instead of the plastic used in the 1053s. It is stronger than the original part, therefore.

Donor part now installed on the 1053

Getting the two cords installed on the machine is trickier than I expected. I need to end up with the cords wound over the drums on the mainspring axle with just enough tension to partially compress the self-tensioner. My first attempt resulted in cords that were too short to even reach the tensioner.

This cord fit on the 1053 when it arrived, presumably, so there is some trick to installing it that I haven't yet found in the several IBM manuals covering the printer. I will keep at it until the cords are installed with a proper position on the self-tensioner. Once that is done, I can wind up the mainspring and that part of the restoration will be complete. 

I am still looking into the failure to fully reset the latches for the rotate selection at the end of a print cycle. My objective is to ensure reliable printing of characters which depends on all the settings for tilt, rotate and print shaft timing. I don't want latches that are left over from a prior print operation to cause the printing of the wrong character. 


Wednesday, October 4, 2023

3D printing mounting hardware for door, continued tweaking typewriter (1053 Console)

NEW HARDWARE FOR MOUNTING REAR COVER ON DISPLAY PEDESTAL

The main issue I had with the first attempt was that the exterior parts of the mounting hardware were large and clunky. Rather than using four T handles to secure the cover, I designed a new part for the lower edge that will sit on the bottom lip edge of the pedestal display cabinet. This new design means we only have a single visible screw head on the outside of the cover behind each of the two lower edge parts. . 

For the two side mount mechanisms that will latch the cover in place or release it, I kept the T handle for the outside but developed a set of rotating parts, spring loaded, so that the user pushes in and turns the handle 90 degrees to release the latch. Other than the 3D printed parts I will use a spring, two T handles two screws and four nuts with epoxy securing the nuts to my interior parts. 

The arm part on the left faces down onto the holder part on the right, so that the bulge in the arm will nestle into the gaps on the holder in one of two rotary positions. The holder is attached to the rear of the cover with epoxy and sits so that the edge that is closest to the hole is next to the lip edge of the display pedestal enclosure. 

Thus the arm, when the bulge is fitted into the gap opposite the lip edge, will extend over the rear of the lip edge and keep the cover from opening. Rotating so the arm bulge fits in the other gap, the arm will be rotated parallel to the lip edge and thus allowing the cover to pivot out. 

A spring sits between the T handle head of a bolt and the cover, so that it tends to pull the bolt outwards. The arm is epoxied to the nut on the end of the bolt. Thus the spring keeps the arm pulled into the holder so the bulge and gap blocks rotation. Pushing forward on the T handle against the spring allows the arm to move away from the holder and be rotated. 

TRIED OUT MY NEW 3D PRINTER TO BUILD THE PARTS

I had picked up a 3D printer from a fellow computer restorer - so it is used because he owned it before but it is new to me. I brought my mounting hardware files to the shop, adjusted the printer and gave it a try printing these in ABS plastic. 

I am still having difficulty getting the printer to start the first layer reliably, but I work on it every day so eventually . . . 

TYPEWRITER MECHANISMS CLEANED AND ADJUSTED

There are so very many tiny parts, pivots, levers, springs and the like on a Selectric typewriter, all of which have the potential to be gummed up from stale lubricants. A good working estimate is 3,000 parts. Now that I have addressed all the larger and more obvious components, it was time to focus down on the tiny details that remain stuck.

The ribbon advance mechanism is a case in point, As the typeball is thrown forward to type a character, it should slide a plate forward which has pawls that would turn one of the two ribbon axles. The pawls as well as the plate were stuck with sludge, but with some reoiling and careful manipulation they are ready to go. 

Character selection to rotate the typeball involves four levers that are pulled down in specific patterns to cause the rotate tape to be pulled out or released inwards a given amount, ultimately turning the ball within a 180 degree arc to one of 11 positions. Simultaneously two other levers are pulled down in patterns to tilt the ball up or down to four rings of characters. 

I found that the mechanism for the R5 selection was sticky and working erratically. That is, we can either see the typeball rotate as it should in the positive direction which is what R5 selects, or by its absence cause a negative direction spurious rotation. With the same selections on the other levers, I can see either a -5 rotation or the ball remain in the idle position, but the sticky mechanism causes the R5 selection to be missed in many cases. 

I had to locate the sticking part, a tiny lever pivot deep inside the machine that is the latch which keeps R5 from activating unless the solenoid releases the latch due to an R5 selection. The latch is not being pulled back to block the R5 lever movement in spite of the spring that is attached, because of gumming in the pivot. Oil loosened the old lubricants and my tools gently rotated the pivot until it was moving freely. 

Having the rotate and tilt tapes properly installed it was time to work through the adjustments to be sure it will print well. It was going along reasonably well until I noticed that the character selection latches aren't relatching correctly. This leaves the typeball partly rotated, not back at the home position. 

This may not happen under power and only while hand cycling the machine, because I do see that the next print cycle completes the resetting before the new character is latched in. I compared it to behavior of the 1053 that came on my 1130. My typewriter does fully reset when hand cycling. 

I reattached the cords that wind up the mainspring as the carrier is pulled to the left margin by a carriage return operation, then pulls the carrier to the right for space and tab operations using the power in the spring. These are two cords that wind around drums on the same shaft as the mainspring, one side letting out the same amount as is wound up by the other drum. I have to figure out how to wind the spring and tension the cords in order to complete this task.

I also discovered that some levers used to trigger the carriage return were gummed up. After I freed them up the CR button would reliably shift the mechanism into power return. This activates a clutch to turn the mainspring shaft using the constantly rotating operational shaft, drawing the carrier to the left. When the carrier strikes the left margin setting it pushes the margin bar leftward to release the return mechanism. 

This typewriter uses is a high speed rewind feature to reduce the time to return across the line. To decelerate the carrier, there is an air piston that stops the carrier from bouncing as it comes to a stop. 

I am still hunting for suitable tape to create a replacement color shifting tape that selects between the upper and lower half of the ribbon, which is a bicolor cartridge with red and black ink. This is 46 inches long,  0.105" wide and 0.003" thick. IBM used a plastic for the tape but I suppose as long as whatever I find is flexible enough and strong enough, it will do.