1971-1974 Change

Analog to Digital Electronics

Post Boost Propulsion System work came to an end mid 1970 when the production phase began. Out of a job I pondered what was needed next. “Think tank” people were concerned about our ability to launch through radiation in the event of an attack,.steps were being taken to “harden” launch sites. I was convinced our analog controls electronics would have a better chance surviving radiation if converted to digital on vs off. I suggested work should be initiated on how to digitalize the current analog controls electronics. They agreed it was a good idea and suggested I look into it. I said but I’m a mechanical engineer, I’ve only dabbled with electronics – and turned to walk out the door. My friend responded, since when did that stop you? I paused and considered the alternative? Only 6 of 75 former Post Boost personnel remained. I needed to find a new job as there was no work in my prior organization. I’d been taking night classes taught by the Art Hammond, Chief engineer of the new Micro-Electronics division, electronics technology was changing rapidly. Perhaps I should look into what’s possible?

Missile X, follow on to MM 3

At the request of management I’d been working with TRW, advisors to the AF, on concepts for a follow on to MM 3. Lou Purpura, who arranged these meetings, and I called it a Missile X or MX, and others began using the same term. Some time later the AF issued a contract for an MX missile, with Autonetics on contract to build the Inertial Guidance system. I was asked to work with Dr K. O’kief and others at TRW, who took an interest in the work I was doing to digitalizing flight control electronics.

Gimbaled Nozzle thrust vector control for MX

Solid propellant motor contractors were being funded to develop a gimbaled nozzle method for thrust vector control. This method required separate Roll Control.

Change to flat missile raceway – and few wire data bus

Side winds on the protruding raceway at launch caused the missile to roll, and guidance system wiring, like a turkeys neck, could not tolerate a rolling twist. This incentive to flattening the raceway prompted the idea of “data bus” communication with fewer wires – another incentive for conversion to digital electronics.

My plan

Design and build a demonstrator with upstage commands via a few wire data bus to downstage nozzle controls operating in all digital servo control mode. I didn’t have a clue on how to do this so my first task was to learn about digital electronics and methods.

Special Hardened Minuteman electronics

Electronic devices had to be compatible with a pending “hardened TTL Minuteman parts, which would be emerging new Transistor to Transistor Logic devices. I found I could obtain new device samples for experimental tests. Microprocessor methods then evolving could not be used as they were more vulnerable to radiation.

In Over my Head

Problems became more complex – too late I found myself in over my head – unable to just quite when others had stuck their neck out in my support. In the end I succeeded, working myself out of a job -- the concepts became part of a front end digital Flight Control computer.

MM 3 flat packs to Etched Circuits to TTL logic devices

The technology was moving from MSI (Medium Scale Integration) to (LSI) large scale integration of elements.

Process

Micro Electronics Chief Engineer Art Hammond’s hand out above describing chip fabrication

Paid $60 for calculator at left. My notes from after work class’s Paid $20 for one at right 4 yrs later

Growing Crystals: Silicon Crystals are grown in a solution and look like long icicles, Monsanto Chemical was one of the first to provide these in a 3 inch diameter size, to specified quality.

Cut into Wafers: The silicon crystals were then cut into wafers to begin their passage through a photo-etch-deposit-photo-etch-deposit process required to “build” semi conductors. By 2000 wafer size had increased to 12 inches in diameter, and automated factories redesigned, a 12 inch diameter stack of glass became quite heavy. A typical production plant costs 2 billion dollars.

Photo-Etch Wafers were coated with “photo-resist” sensitive to light. When exposed, the image is developed as if a photo leaving some areas protected and some expose to be chemically etched, then placed in glass tube ovens and exposed to chemical vapors which deposit specific material on exposed areas. The process is repeated to “build” devices.

P or N deposits Special semi-conduction material, seasoned with P or N types of rare earth elements, are vapor deposited on the exposed areas, and the process repeated until a transistor has been built. Electrons flow from a P junction and to an N junction.

Aluminum Inter Connect “wiring” The transistor arrays are interconnected by depositing aluminum on photo created circuits. A powerful microscope is require to see them.

Postage Stamp array of chips: Multiple devices, as if a field of postage stamps are created during the process. After the process has been completed, these are “cut” into “chips”, as if making tiny panes of window glass.

Packaged chips The chips are then placed in a carrier, and connected, initially by tiny gold “bond wires” to the main package leads.

Yield “Production” is measured by yield, the % that are good, in 1972 a 10% yield was considered good. At after work classes I learned Autonetics was supplying Japan so many hand calculator chips they were selling them as if produced in Japan. The division was a primary supplier of phone modem chips until spun off from the aerospace industry.

Reliability The Minuteman Reliability program contributed significantly to advances in the semi conductor industry, paving he way for microprocessor development. During MM1 just moving an electronics device from one bench to another would cause it to fail or to work. Microscopic contaminate, inside a sealed package, could short tiny circuits. Absolute clean room operations imposed by Minuteman standards were adopted by the industry to increase yield and profitability – making micro processors and desk top computers possible.

Logic Devices

After WW II interest increased interest in making logic devices. IBM, and others, made Logic cards and semi-conductor firms make families of integrated circuit devices as: RL (resistor logic), DL (diode logic) and RTL (resistor transistor logic) and Emitter coupled logic. These were marketed in the Dual Inline packages.

IBM Flip Flop on a plug in card

Logic Elements A logic device is a gate. Below is a TTL NAND gate with Truth Table.

TTL NAND gate Diagrammatically it looks like this invert NAND and you have AND:

When A and B are both High or 1's the output for this device of A+B is low or 0. Since it requires A and B to be l for an output 0 the AND logic produces 0 the inverse of 1 or Not 1 so the device is called a Not AND or NAND. The addition of a small circle on the nose of the symbol indicates the AND output is the inverted to a NAND, A line above a letter indicates the inverse; the line above A+B defines it a “not”.

These became widely used standard terminologies, with Truth Tables defining the possible conditions.

Boolean Addition, Multiplication and complementation

A Flip Flop is a single bit programmable memory.

Clocked S-R (set reset) Flip Flop made from logic devices

O for Output was called Q, to not confuse with zero. The inverse of Q is Not Q (a line above Q)

Counter Connecting FF’s as shown creates a binary counter, where the output of one clocks the next. To start all FF's can be Reset to zero. The binary count is read at A, B, C and D as 0's or 1's like an odometer on a car. Counters can be caused to count up or down; start with preloaded values and caused to reset after a specified count, etc.

Shift Register By connecting FF’s as shown creates a shift register; whatever is applied at the input will be shift fallowing each clock pulse. By connecting the output to input a ring counter is created in which any combination of 0's and 1's can be shifted in a circle. Shift registers can be caused to shift left or right, be preloaded or cleared.

Latch connecting FF’s as shown creates a latch; when the clock is strobed each input at D is stored in the FF and can be read at Q. A latch can thus read a counter or shift register by strobing its clock whenever the device is to be read.

Decoders route an input to a binary coded address, a 4 bit binary to 1 of 16 is shown.

1 to 16 4 bit binary decoder Decoder with Latch Dual 1 of 4 Selector

Display Decoder a Read Only Memory is used as a Binary to 7 segment decoder

Counters, decoders and displays can be set up to program a specific sequence of events.

Clock Prior to semi conductors Clocks were created by using crystals to create oscillators. I was stumped until I found the clock circuit below in an RCA handbook; these were very easy to make and functioned flawlessly.

Clock binary decoders (prior) data selector above

Programming the system. The clock connected to a sequence of counters with their outputs into 4 bit binary into 1 of 16 decoder I could create a fixed program – I could hard wire a routine of choice. It worked very well when refined to provide solid data transfers.

The problem, how do you close a servo loop with digital electronics? This required receiving a digital command, subtract a digital feed back command and produce a digital error command for issue to the servo valve.

Up-Down counter as digital summing junction I decided to use an Up-Down counter as a Summing Junction. An input command would be parallel preloaded into an up down counter then have the feed back up or down count and end up with an error signal for specified time increment.

Digital Command: To simulate a computer command, for the demonstrator, used a voltage controlled multi-vibrator to drive two 8 bit parallel output counters to produce a 16 bit binary command signal in a predefined “cycle time”.

The output of a voltage controlled multi vibrator can be controlled over it’s linear range

Joystick command simulator In my home shop I mounted a variable resistor (potentiometer) referred to as a “pot” on a piece of aluminum angle for pitch, then made another for yaw. One angle was fixed and the other attached to the shaft of the fixed pot, and a joy stick aluminum tube, flattened at one end, to the shaft of the other pot. It took a while to work out but they worked like a charm.

Joystick motion controlled separate pitch and yaw output of the voltage controlled multi-vibrator with a timed burst captured the 16 bit command signal. This command was periodically stored in a “shipping dock” register pending it’s timed transfer to a 16 bit parallel load shift register.

Digital Feedback: A Clock provided excitation frequency and level detector determines sign and enables up or down count. Demodulated dc voltage feedback commanded a voltage controlled multi-vibrator frequency output to summing junction counter.

Digital Feedback patent award: My boss asked me to submit a patent disclosure on an idea I’d abandon. The AF assigned a patent attorney and proceeded with the patent. Some 10 years later I received an award of $1000 for submitting the concept. I never felt right about that because I never proved the concept worked. When hired we signed an agreement the US Gov owned all patentable ideas originating while they paid the bill. Photo below, receiving patent check.

Servo Valve Driver: Highly reliable analog servo valves had been designed for 8 ma control signals produced by vacuum tubes. How do you command an analog valve digitally? Emil Kohler said why not use an “H” switch, a switch in each leg and coil in the middle.

Transmission Gate “H” switch made with Transmission gates 4 Transmission gates on one IC

Logic device drives servo valve. Experienced engineers believed this was impossible until demonstrated. An ON or OFF semiconductor switch does not heat – thus a “logic” device could switch 8 milliamps.

I derived much personal pleasure, observing their surprised smiles.

Binary Number to Valve Command: I struggled with how to convert a binary number to a valve command. A large command was full on, I only and to convert lower bits to a ramped command. I decided to use a 16 bit shift register to cycle a % on vs off time, 1 on, 0 off. From prior experience I knew the valve coil would “integrate the signal” and apply the averaged torque command. A Johns Hopkins “accelerating switching valve” we tested during the Navaho days cycled at 128 cps, the system could respond and shook the system. I decided to clock the shift register using the feedback excitation frequency at 5 khz, commands would average and not dither. The difficult problem was how to convert a binary number to a ramp of 16 one bit steps until full on saturation. I spent days trying to design logic to make the conversion, too many chips were required. Then I came across a new 128 bit programmable read only memory, you “scratched” the pull up resistor from the cell to make it a 0 otherwise it was a one. I took it to near by Microelectronics research lab to scratch a “ramp”. They cut the links with a laser and a photo of what they did. The highly enlarged image revealed they had blasted craters in the chip – it was non functional. I replaced that with a new programmable memory and moved on.

Extend or Retract I needed a way to determine a sign bit to extend or retract the actuator. I devised a way to store the sign bit in a set-reset flip flop where the command was reset to Plus at the beginning of each cycle and set to Minus if the Summing junction counter passed through zero, from all zeros to all ones.

The above shows how binary error was looked up as a “% on time” with a look up table and the sign bit enabled an extend or retract command. This Digital to “analog” conversion solved one of the most difficult problems. Engineers in the aerospace industry were encouraged by AF, Navy and their home company to attend meetings to exchange ideas and present papers. It was essential to have access to these meetings to keep abreast. I was repeatedly to describe how the above was done, it was perhaps patentable at the time.

Cutting edge knowledge was often within the industry before it was taken up by colleges.

Nozzle control actuators I chose to use MM 3 post boost axial engine electric servo actuator to move the simulated engine nozzles in sync with the joysticks at the top. At home I made a gimble joint from a square nut and aluminum tube segments with a bracket added to attach actuators. We stuck plastic funnels on the tube as if rocket engine nozzles. The improvisation drew smiles of approval.

The above card was used to drive the electrical actuator motors.

Upstage and Down stage circuit boards I made an aluminum frame on which we mounted circuit cards for holding IC sockets, IC’s on top wiring on bottom.

Wiring was done by ladies who wired missile guidance electronics, they strived to make the wiring neat, as if neat was a requirement.

Wiring Instructions I provided instructions as show above on what pin to connect to what pin.

Data Buss Demonstrator: an electronic rack served as a “missile” with up stage electronics at the top, down stage electronics at the bottom connected with a five wire data bus: +5 volts , ground, clock, data out and data in. Two 4 bit shift registers made an 8 bit word, shipping first a stage identification code, then pitch then yaw binary command to the down stage servo controller, and returned position feedback.

Power and system clock was supplied to down stage electronics, which permitted locking the two sets of electronics together. The shift registers were parallel loading and unloading. Data loading latches held data ready to ship and data unloading latches held arrived data; as if loading and unloading shipping and receiving docks. An Address latch read the address code and enabled or disabled that stage from loading or unloading data. Line Drivers and receivers were at each end of the data bus.

System Demonstration: After the system was built I demonstrated it to others. Chief Engineer Tom Shuler, whom I’d known since 1955, came by and I showed him how wiggling the top Joy sticks caused the “nozzles” below to track. We were pleased to see Tom’s face change from scowl to smile when he wiggled and watched.

Data Buss Study: The data bus idea had generated sufficient interest, in fact TRW requested and received AF funding to study it’s merit – in presenting the idea to TRW and AF I was hoping we would receive funding for such a study.

In House Design Review: An in house review of the concept was favorable except for one thing – the design could not be modified to perform filtering and shaping functions.

Motor contractors study funds: AF funded motor contractor servo actuator studies for gimbaling booster nozzles. The study of hydraulic servos was excluded as that was a refined proven method. I learned motor contractors were using 100 hp gas generators to powered clutched servo actuators. It was obvious to experienced flight control people that such a systems would require signal shaping and processing to maintain stability. A high speed data processor would be required.

Fiberoptics: I was constantly in search of ideas and early one morning watched an educational program sponsored by Monsanto advertising a product line of light emitting diodes and sent off for sales literature. About that same time someone sent a couple of fiber optic cable samples. If signals could be sent through these they would be immune to electromagnetic effects. I sent for send and receive devices from Monsanto and looked up local suppliers of ornamental displays I’d seen sending light from a source to a fanned array of small fibers. I found a source and purchased a five foot length of glass fiber. I then proceeded to set up a driver and receiver to send signals through the fiber from one work bench to another. I soon realized our lab was not set up to make such things as it required polishing the ends and focusing send and receive signals. When I expressed my idea I was told such could not be used for our missiles because radiation clouds the optics and destroys the signal.

Earths atmosphere has limited capacity. At the time of WW 2 Wave Bands were controlled by the government – they were limited. Later improved quality and frequency of transmissions permitted more allocations.

Each Optic cable becomes an independent atmosphere. Optic fibers soon relieved the wave length limit. Initially there were problems on how to Re-amplify optic signals – there were huge losses when a cable was broken and spliced. Then a means was found to make a loop in the cable and by some means re-boost the signal. Today we accept the benefits of cable as if it’s always been possible.