HD-70-SemiCondEvol

Part II 1955-1980 Semi conductors to Microprocessors.

Introduction

From Tubes to Transistors to Etched Circuits to packaged Chip

Components

Welcome to a walk through my experiences with semiconductors, I was beginning to understand vacuum tubes when transistors arrived. Transistors were first used as substitutes for vacuum tubes. Though logic functions had long been done with switches, it was in these applications that transistors really took off. They were ideal for making logic devices on a card where wiring was applied by dipping cards in solder; IBM specialized in making huge computers with replaceable plug in cards. The big jump came when it was discovered how to print-etch-deposit multiple layers on silicon wafers; the term micro electronics was coined and put electronics evolution on the fast track. This era will not repeated, electronics will remain the same but your challenges will be different.

Systems

Following Minuteman missiles I, II & III there was an incentive to look ahead; there was a need for better nuclear event survivability and reduced roll control demands. Conversion from linear transistor amplifiers (analog) to binary numbers (digital on vs off) logic devices offered better nuclear survivability and fewer flattened wire between stages would reduce roll disturbance. After pointing this out, it was suggested I look into what might be done – a big leap from hydraulics and rocket engines. With management support the following system concepts were evolved using Minutemen missile booster control as the model. The technology was rapidly changing and each system concept was an improvement on the other. Digital electronics was new even to most skilled electronics engineers.

System #1 Time share signal wires (serial data buss) to booster nozzle control units, used binary counters to compare guidance position commands with feedback (actual position) and command servo-actuators to correct booster thrust vector. (parallel arithmetic was not yet an option)

System #2 Time share signal wires (serial data buss) to booster nozzle control units, used 4 bit Arithmetic Logic to compare guidance position commands with feedback (actual position) and command servo-actuators to correct booster thrust vector. (parallel arithmetic had arrived)

System #3 Move electronics from each booster stage to front and time share digital control via dedicated wires to each servo actuator to correct booster thrust vector.

Each of the systems were designed, built and demonstrated. System #3 became the concept used for the MX missile, incorporating the servo control electronics in the Flight Control Computer. That concluded my conceptual experiments and I was assigned to help restart the B-1B program.

!953 X-10 Navaho Missile tube electronics plug in module, front and back side

Basics

Rather than starting with the big picture and work down, lets start with discrete parts and work up.

Semi conductors entered the existing world of tube electronics – the symbols and terms changed.

Vacuum Tubes: The standard vacuum tube has a heater element that heats a cathode which boils off electrons that float out into the vacuum inside a glass tube. There is a grid element between the cathode and plate element. The plate is connected to a high plus voltage to attract electrons from the low negative voltage cathode. A signal voltage applied to the grid will attract or reject the flow of electrons from the cathode to the plate. A low level signal applied to the grid can become a high level signal from the plate – tubes are signal amplifiers.

Bi-Polar Transistor These semi conductors have three parts a collector, base and emitter. They are either P or N kinds, the one shown is a P kind where electrons flow from the emitter; for this case the collector would be plus and the emitter minus. Current to the base (sourceing) will turn the transistor on, current taken from the base (sinking) will turn the transistor off. A small variable signal to the base can control a large signal through the transistor – the transistor is a signal amplifier – base current must flow to enable flow of collector current.

MOS FET Transistors These Metal Oxide Silicon Field Effect Transistors have three parts: source, gate and drain. They also come in either P or N kinds. These operate on a Field effect where a field on the gate will enable or disable flow through the device, a plus or minus field turns the transistor on or off depending on if it is a P or N kind. Current does not flow from the gate, the charge on the gate controls flow.

CMOS This Complementary MOS device connects a P kind with an N kind such that a signal on the gate will turn one off and the other on, thus they are called complementary. These devices made digital microprocessors possible. The output is either full on or full off, there is no power drain, except for the brief instance when they change state. Desk top computers are equipped with CMOS memory chips to hold set up information – a small battery sustains their state while main power is off.

Transmission Gate By connecting a P and N transistor in parallel, with one side commanded by the inverse of the other, flow can be turned on or off. These, used in conjunction with CMOS, enable the selection of multiple devices connected in parallel. They permit “tri state” logic: On, Off or Open.

Potentiometers This is a resistor with one end connected to plus the other to minus and a wiper that can read a selected location. These are used to measure the position of a device that moves the wiper. This was the primary kind of position measuring device during WW II. They could not be used on missiles undergoing severe vibration as the wiper would lose contact.

Variable Resistor Similar to a potentiometer except that the amount of resistance is determined by the wiper, which can go from low to high resistance. Volume controls on radio and TV work on this principle.

Transistor (analog mode) A transistor can be used as a variable resistor. This is how they were initially used, as substitutes for amplifier tubes, to vary flow through a load.

Analog transistors get hot: When transistors behave as variable resistors, they get hot and require the use of heat sinks. Much innovation has been applied on how to heat sink power devices such as audio amplifiers.

Digital transistors stay cool: When transistor is OFF there is no current flow, thus no heat load. When a transistor is full on there is no resistance, no voltage drop, no heat load; watts = I x R where I = current and R = resistance.

An H switch is like a goal post where the coil is the cross bar and the switches are in the upper and lower legs. When diagonally opposite switches are turned on and off, current can be directed to flow one way vs the other.

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

I’ve jumped ahead, lets go back to early transistors:

Above and below is from an early Sylvania handbook, an aid to hobbyist

This worked very well in my CA garage, pulling in signals from Mt Wilson, it wouldn’t pull in a thing in NW KS

This circuit, from a transformer supplier, permitted experimenters to make radios patterned after tube sets.

I assembled the about 1956, it didn’t work, later I found the IF transformers coils were not wired per the label, I reconnected according to wiring and it worked -- tenacity prevailed.

1958 Hi Fi Cabinet: Tape deck, Tuner & Hi Fi Amplifier, Phonograph Heath Kit Tube HiFi Amp ran very hot

The cabinet was made from one 4x8 ft plywood sheet, I spent hours selecting matching patterns. The Heath kit wiring under the chassis was very crowded. It worked well but ran VERY hot – I should have incorporated a cooling fan. The left drawer include a Tape Deck, the right a multiple rpm phonograph turn table. The center section housed a Hi Fi Tuner and the Heath Kit amplifier, then the Home made all transistor HiFi stereo amplifier. The transistor set worked beautifully for many years until too many little hands twisted the knobs disconnecting preamp wires. These are now in the Oberlin KS museum.

This all transistor dual channel HiFi was built about 1958, with encouragement from Bob Kelley and use of assembled kits for: Vacuum Tube Voltmeter, Oscilloscope & Signal Generator. Except for input power, it did not use transformers as required for tube sets. The key to HiFi was quality biasing of amplifiers.

Prior photos were taken on this card table

Made from 9 kinds of wood from our yard, coffee table contains some 350 pieces, 3/8 x 1 x 3 inches.

Reset – back to Basics

Resistors: Impede the flow of electricity and come in a variety of sizes, their resistance value color coded.

Capacitors: Capacitors store electrons as a balloon stores air. Their primary purpose is to smooth signals or power by absorbing spikes and filling in dips. There are of many kinds, and types, almost always a part of any circuit. When a transistor cuts off flow to a capacitor, the capacitor retains it’s charge until leaked off.

Inductors: Inductors are similar to capacitors in that they can absorb spikes and fill dips, but they work on a different principle – when current increases it builds a magnetic flux field, when current is cut off the field collapses and generates an emf (electromotive force), pushing current. When a transistors cut off flow through an inductor, they require “kick back” diodes to relieve the induced voltage.

Diodes: (above right) permit flow in one direction only.

Diode threshold Diodes have a small voltage threshold that must be overcome, they must be replaced with demodulators which use commanded transistors for rectification of low level signals.

Zener Diodes: permit reverse flow at a specified voltage, serving as a relief valve.

Phase Shift: Capacitance causes current to lag voltage and Inductance causes current to lead voltage, the combination of which creates a phase shift when present. The amount of phase shift is measured in degrees. This is important when working with alternation current. It’s necessary for power plants to try and balance phase shift loads. The effects are present in semiconductor circuits but not as obvious.

From Hausmann & Slack 1939 physics book

AC to DC power supply: Below the input transformer dropped household voltage to the desired DC level and the diodes provide full wave rectifiers; the diodes are a huge improvement over the use of vacuum tubes or selenium plates as rectifiers. Diodes made it possible to replace car DC generators with simpler AC alternators, diodes permitted simple low cost rectification into DC.

Silicon Controlled Rectifier power supply for home work bench. This worked quite well and was used for several years.

Silicon Controlled Rectifiers: It was found that diodes could be controlled, making it possible to select at what point rectification would occur. This provided significant opportunities and options. The above power supply is an example. Though widely used in commercial applications we did not use them for military purposed as they were vulnerable to the effects of nuclear radiation.

Logic Devices

Switches had long been used to perform logic functions, but not thought of as logic. After WW II there was increased interest in making logic devices. When sending for literature and reading trade magazines I encountered reference to various ways of creating logic devices. IBM, and others, were making 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. Users and experimenters could readily obtain catalogs defining these devices.

LOGIC ELEMENTS

The basic element of any logic system is a gate. A TTL NAND gate is shown below.

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

Packaging took on various forms, with plastic encased dual inline the most common with ceramic packages considered better. The most common were 14 pin then 16 pin packages, with more pins for complex functions. Packages had it’s own terminology. Pins numbering started with zero. The terminologies were similar but different depending on the kind. Bi-polar power pins were always V_cc and Grd, CMOS power pins were V_dd and V_ss ; as shown above.

Motorola Mc MOS INTEGRATED CIRCUITS 1974

MC14000Al Dual 3 Input NOR Gate/Inverter 1.22 MC14001Al Quad 2 Input NOR Gate 1.22 MC14002Al Dual 4 Input NOR Gate 1.22 MC14006Al 18 Bit Static Shift Register 5.39 MC14007Al Dual Pair and Inverter 1.22 MC14008Al 4-Bit Full Adder 5.59 MC14011Al Quad 2 Input NAND Gale 1.22 MC14012Al Dual 4"nput NAND Gate 1.22 Mc14013Al Quad Type D Flip.Flop 2.26 MC14014Al 8.Bit Static Shift Register 5.87 MC14015Al Dual 4 Bit Static Shift Register 5.87 MC14016Al Quad Analog Sw/Quad Multi 2.26 MC14017Al Decade Counter/Divider 5.87 MC14020Al 14.Bit Binary Counter 6.52 MC14021Al 8-B,t Static Shift Register 5.87 MC14022Al Octal Counter/Divider 5.68 MC14023Al Triple.3 Input NANO Gate 1.22 MC14024Al Seven-Stage Ripple Counter 4.22 MC14025Al Triple 3 Input NOR Gate 1.12 MC14027Al Dual J-K Flip-Flop 3.50 MC14028Al BCO-To Decimal Decoder 5.09 MC14032Al Triple Serial Adder (positive) 5.96 MC14034Al 8.Bit Universal Bus Register 12.55 MC14035Al 4 Stage Shift Register 5.83 MC14038Al Triple Serial Adder (negative) 5.96 MC14040Al 12.Bit Binary Counter 6.20 MC14042Al Quad Latch 4.70 MC14531Al 12 Bit Party Tree 4.81 MC14532Al 8 Bit Prlority encoder 6.20 MC14534Al Reat Time 5.Decade Cnter 27.69 MC14536Al Programmable Timer 11.89

MC14046 Phase Locked Loop 4th Qtr 74 MC14049AL Hex Inverter/Buffer 2.24 MC14050AL Hex Buffer 2.24 MC14051 8 Ch Analog Multiplexer 4thQtr 74 MC14052 Differl 4 Ch Analog Multi 4th Qtr 74 MC14053 Triple 2Ch Analog Multir 4^th Qtr 74 MCI4071AL Quad 2.lnput OR Gate 1.22 MC14076 Quad D FF 3 State Out 4th Qtr 74 MC14081AL Quad 2.lnput ANO Gate 1.22 MC14501AL Triple Gate l.65 MC14502AL Strobed Hex Invr/Buffer 3.70 MCI4506AL Dual Expandable Gate 1.65 MC14507AL Quad Excl OR Gate 2.22 MC14508AL Dual 4-Bit Latch 13.68 MC14510AL BCD Up/Down Coter 6.94 MC14511AL BCD to 7 Seg Latch/Decr 7.39 MC14512AL 8 Channel Data Selector 4.15 MCI4514AL 4/16 Line Decoder (Hi) 14.96 MC14515AL 4/16 Line Decoder (Lo) 14.96 MC14516AL Binary Up/Down Counter 6.00 MC14517AL Dual 64-BitStalicShReg 20.66 MC14518AL Dual BCD Up Counter 7.39 MCI4519AL 4.Bit ANO/OR Selector 2.22 MC14520AL Dual Binary Up Counter 7.39 MC14521AL 24 State Frequency Dr 8.31 MC14522AL BCD Divide by-N Cnter 6.46 MC14526AL Binary Divide by N Cnter 6.46 MC14527AL BCD Rate Multiplier 6.46 MCI4528AL Dual Monostable Multiv 4.15 MCI4529AL Dual 4.Channel Multiplex 5.55 MCI4530AL Dual 5 In Maj Logic Gate 2.50

MC14539Al Dual 4 Channel Digital Mix 4.l5 MC14541 Oscillator Timer 4^th Qtr 74 MC14543Al BCD to 7 Seg latch Decod Dr 7.39 MC14549Al Successive Approx Reg 11.35 MCM14552 64x4 RAM 4^th Qtr 74 MC14553Al Three Digit BCD Counter 15.49 MC14554Al 2x2 Bit Parallel Binary Multi 4.61 MC14555Al Dual Binary 1 of 4 Decoder 2.76 MC14556Al Dual Bin 1 of 4 Decoder (INV) 2.76 MC14557Al 1 to 64 bit Shift Register 1.30 MC14558 BCD to 7 Segment Decoder 4^th Qtr 74 MC14559Al Successive Approx Reg 11.35 MC14560Al NBCD Adder 7.077 MC14561Al 9s Complementary 2.17 MC14562Al 128 Bit Static Shift Register 18.60 MC14566Al Industrial Time Base Gen 4.61 MC14572Al Hex Gate 1.48 MC14580Al 4x4 Multiport Register 21.18 MC14581Al 4 Bit Arithmetic Logic Unit 11.35 MC14582Al Look Ahead Carry Block 11.15 MC14583Al Dual Schmitt Trigger 2.76 MC14585Al 4.Bit Magnitude Comparator 5.55 MCM14505Al 64Bit Static RAM 20.66 MCM14524 1024 Bit ROM Consult Factory MCM14537Al 256.Bit Static RAM 41.32 MC14410 2 of 8 Tone Encoder 4^th Qtr 74 MC14411 8it Rate Frequency Gen 4^th Qtr 74 MC14415 Quad Precision Timer Driver 4^th Qtr 74 MC14435 3 1/2 Digit A/D logic Subsystem 4^th Qtr 74 MC14490 Hex Contact bounce eliminator 4^th Qtr 74

The above listing is typical of what was becoming available in TTL and CMOS in 1974. At the time I was focused on using commercial TTL parts of the kind intended to be “hardened” for military applications.

From Art Hammond lesson hand out describing fabrication process.

My sketch from after hours class by Art Hammond head of Autonetics Micro-Electronics manufacturing

Ladies in clean rooms bond chips to frames with gold wires

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. The industry increased the wafer size and by 2000 they were 12 inches in diameter -- the automated processes had to be redesigned, a stack of them became quite heavy; a single production plant can cost 2 billion dollars.

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

P or N deposits Special semi-conduction material, seasoned with P or N types of chemical 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 Each device had to be tested, the slightest flaw would cause failure. “Production” was measured based on yield, like what percent were good products. I was taken through the Micro Electronics plant at Autonetics, at that time, 10% was a good yield. They found it necessary to move the plant because of contaminates in local smog.

At one time Autonetic Micro Electronics Division was supplying the Japanese calculator market, in such quantities that the Japanese were selling the calculator chips back into the US market as if produced in Japan.

Eventually the division was spun off as it never reached profitability as part of the defense industry.

Reliability It became common knowledge in the industry that the Minuteman program contributed significantly to advances in the semi conductor industry. It came about with the need for very reliable parts. During Minuteman I it was found that just moving an electronics box from one bench to another would cause it to fail or work. The problem was eventually traced to flaking of contaminates inside the package, which shorted out the tiny wire circuits; remove the flake and it worked. The interconnecting wiring on a chip, only visible with a powerful microscope, permit particles, too small to be seen without magnification, to cause a malfunction -- malfunction of one chip, failure of a missile. A very comprehensive program was established for clean rooms, processes and training of personnel. Suppliers soon found their yields greatly improved and the super clean measures were applied to commercial lines. This kind of quality would lead to the ability to manufacture micro processors – and the desk top computer.

From Motorola literature

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

FLIP FLOP ELEMENTS

Now let’s have a look at another basic building block, the Flip Flop (FF). A single FF is shown schematically as follows:

Discrete parts Flip Flop logic Flip Flop

The type D and J-K Flip Flops are widely used as memory cell building block, single or in combinations.

A Flip Flop is a single bit programmable memory.

Making Counters, Shift Registers and Latches from Flip Flops

Lets take a look at what can be done with FF’s

Connected as a Counter

Counter By connecting FF’s as shown above we create a binary counter where the output of one clocks the next. Before starting a counter 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, caused to reset after a specified count, etc.

Connected as a Shift Register

Shift Register By connecting FF’s as shown above we create 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.

Connected as a Latch

Latch When connecting FF’s as shown above we create 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.

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

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

3 bit binary to select 1 of 8 in to a single out. Read Only Memory used as a Binary to 7 segment decoder

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

RAM Random Access Memory 4 x 4 bit read write memory – they were small in 1974

CLOCKS

Prior to semi conductors Clocks were created by using crystals to create oscillators. I felt stumped until I found a very simple clock circuit as show below in an RCA catalog; these were very easy to make and functioned flawlessly.

Clock voltage controlled multi-vibrator

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

System # 1 The Data Buss link to Servo-control System

Minuteman Raceway Problem: Autonetics and TRW were presenting ideas to the AF at Norton Air Base with regard to a Minuteman III follow on; TRW had come up with a huge Roll control requirement based on what MM stage I could do rather than what was needed; far in excess of what Lockheed, with a flat raceway was using. I discussed this with our flight simulation people who said if we could change our 6 inch high raceway to a flat one, lowering wind caused roll disturbance we too could reduce roll control requirements. Thus in incentive to flatten our raceway.

Party Line Control System I decided that a single data link, emulating a rural telephone party lines system, would work on a missile with to & from digital serial data packet. Like farmers each stage could have it’s own “ring”, a coded address. The idea was great but the Party Line name was a poor choice, others thought political not rural party.

Developing the building blocks

Before proceeding it was necessary to determine how to do the various pieces as: Servo valve driver, Summing Junction, Feedback conversion, Command simulator, Timing of a system, Data buss method, Conversion of binary error command to valve commands. It was necessary to work out each of these before they could be incorporated as a system.

Driver used for MM III Post Boost P-106 electro mechanical servo in first system demonstrator

Digital Valve Driver: Thoughts were focused on how to command an analog hydraulic servo valve with a digital command. This required an On-OFF valve driver if done with digital electronics. From MM I experiments I knew from work Bob Kelley did, that a valve coil integrates high frequency signals, signals too fast for it to respond; the valve torque motor would average out high speed bits. I could achieve the equivalent of an analog signal by controlling the % ON vs % OFF time. The task would be to ramp a binary signal from low level % ONn to full 100% ON at saturated command. I could cycle % ON vs OFF as 1’s vs 0’s in a shift register and apply these to the servo torque motor. I could figure out how to do that after I had a digital valve driver.

For our demonstrator we needed a driver for a MM III P-106 post boost electro-mechanical actuator that could be commanded as if it were a hydraulic servo valve -- above is what we used.

When T1 and T3 are ON current flows through the motor coil to extend the actuator. Current flow is reversed with T2 and T4 ON. D1, D2, D3 and D4 protect the respective transistors from induced voltage when the motor field collapses. When extend commands are removed for example the coil sucks in current through D4 and shoves it through D2 into the supply thus precluding breakdown spikes to occur at the transistors.

T1 and T2 are Flip Flop like switches. T1 is a slave to T3 and T2 is a slave to T4. When T3 shorts to ground current can then flow through R1 which opens T1. When T3 goes OFF R1 cannot flow current and T1 goes OFF. Thus by commanding one transistor, two are switched. Note that T1 & T2 are “N” transistors while T3 & T4 are “P” kinds of transistors. T5 and T6 drive T3 and T4 commands are provided by Integrated Circuit (IC) parts that interface with the digital logic system. This design was later replaced with a single IC – applicable to hydraulic servo valves.

Digital Summing Junction: I had been trying to figure out some way to compare a command with a feed back to find the error difference. One day driving to work my speedometer-odometer began acting up, in need of lubrication -- all of a sudden it came to me. A digital up/down counter is like an odometer that can count up miles or reverse miles – a negative feed back signal to a counter could subtract from a digital command. I could parallel load a command in a counter, then subtract the feedback from it by down counting to arrive at the magnitude error definable as plus or minus error. Such a counter is the digital equivalent of an analog summing junction.

Analog Summing Junction Digital Up/Down Counter Summing Junction

Digital Feedback: Converting the feedback signal to digital was the subject of much study, and the subject of a patent disclosure. I no longer have a diagram of the method used for this first system, so will use these diagrams to describe what was done.

Clock provided excitation, level detector determines sign and enables up or down count. Demodulated voltage feedback commands voltage controlled multi-vibrator frequency output to summing junction counter.

Binary Error to Valve Command” We used a 16 bit shift register to cycle a % on vs off time, 1 on, 0 off; knowing the valve coil would integrate the signal and apply the averaged torque. I recalled that the Johns Hopkins accelerating switching valve cycled at 128 cps and the system could respond, shaking the system; so I cycled the shift register using the feedback excitation frequency at 5 khz. 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 bi-polar PROM (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 see if they could help. Yeah, they helped, they cut the links with a laser, then provided me 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 something and we moved on. I stepping through the process with pencil and paper I realized I needed a way to determine the sign bit. (later systems using arithmetic logic units, 2’s complement numbers did this – but I was proceeding from scratch.

Sign Determination: When the Up (command) count is larger than the Down (feedback) count we have a number left which we will call a + error meaning we want a 1 stored in the latch sign bit. However if the Down count is larger than the Up count our error is - and we want a 0 stored in the latch. We connect an 8 bit NAND gate to the outputs of our two U/D counters and a Flip Flop to the NAND Gate. As the down count progresses backwards through zero the zeros revert to ones. For example:

When the 8 bit NAND gate inputs momentarily go all l's, its output is zero. A zero to the Set terminal of our FF makes Q go to a 1 and “not Q” to a zero. Thus we connect the “not Q” to the latch sign bit so that when we strobe the latch, at the end of our count, we insert a zero into the sign bit. By connecting the FF Reset to the clear line to the counter we restore the FF to a + status, where it remains unless we count through zero again. A simple solution for this unexpected problem.

Joystick command simulator A command signal generator was needed for pitch and yaw. The dual channel voltage controlled multi-vibrator flip flop chip was ideal for the task. 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.

These were connected to a pitch counter and a yaw counter in timed bursts to define binary commands.

Down stage gimbaled Nozzle: Jim Anderson was able to provide PBPS P-106 axial engine gimbal actuators made by Autonetics for simulating down stage Pitch and Yaw control of a nozzle. In my home work shop I built a gimbal using a large Square Nut to which I attached split pieces of aluminum tubing as In and Out of a universal joint drive. One end was fixed to a plate and the other tube extended to represent a thrust nozzle under control. I made an actuator attachment piece from a block of aluminum to which we could connect the P-106 actuator shaft. In the final demonstration we used two sets of joysticks at the top and two movable nozzles at the bottom -- to demonstrate handling multiple devices at the same time.

Timing of the system: I anticipated timing of functions would be difficult, however once I knew how to make a clock, to operate counters connected to binary decoders, I became pleasantly aware I could specify both the length and read write time of any function – what ever I needed, I could build.

System Timing Diagram: It was important to make a system timing diagram for upstage, downstage and data buss functions. The clock frequency was set high, then down counted by linking counters, then converted to time segments using decoders. It was important to work out the details of read, write, apply data, strobe data, enable and disable data, clear data. Each device had it’s own protocol built in. Some clocked changes on high vs low, some on change from low to high or high to low. Items in parallel had to be inhibited or enable in their turn. At first it seemed a daunting challenge, but it was in fact easy to achieve once the system plan was worked out.

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.

Typical card layout, connection were from IC # to pin # First system used soldered wires, second used wire wrap.

Open Collector Logic: TTL devices are normally equipped with their own “pull up” resistor, When multiple sources of TTL data are placed on a parallel data buss it was necessary to use “open collector” devices to the buss and place a “pull up” resistor at the end of the parallel path.

Tri State Logic: CMOS devices used Transmission Gates to enable or inhibit a connection to a parallel data buss. When tri-state logic was used the data line floated and there was no inner-connect unless intended.

Grounding: Literature I was reading harped over and over how it was necessary to maintain proper grounding as TTL was sensitive. I found the devices to be very rugged, even functioning when one pin rested against another; showing only as a slight blip on the oscilloscope, it still did it’s logic function duty. This applies only to devices on the same circuit card and has nothing to do with vehicle grounding to prevent “ground loops” cause by multiple independent vehicle grounds.

Inter-stage Serial Data Buss

Digital Data Buss For the demonstrator I placed up stage electronics at the top of an electronics rack and a down stage electronics at the bottom and connected the two with five wires, plus 5 volts, ground, clock, data out and data in.

I used two 4 bit shift registers to make an 8 bit word, shipping first a stage identification code, then pitch then yaw command to the down stage servo controller, and returned position feedback.

Power and system clock was sent down stage, 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 this system was built I demonstrated it to others. Tom Shuler, our Chief Engineer came and had a look at it. I showed Tom how you could wiggle the top Joy sticks and watch the “nozzles” below track your motions. We were all pleased to see Tom’s face change from scowl to smile when he wiggled and watched.

Data Buss Study: This created considerable interest, in house and with TRW. As usual we provided TRW with the results of my data buss study and not too long after TRW was given and AF contract to study the benefits of a serial data buss for communications between stages.

In House Design Review: A design review was held regarding the merits of this design concept. By and large it was favorably received – except for one thing – the design could not be modified to perform filtering and shaping functions.

Motor contractors study funds: The AF funded motor contractor servo actuator studies for gimbaling booster nozzles for a possible new vehicle. The study of hydraulic servos was excluded as that was a refined art; therefore the options came down to gas driven servo actuators. We learned that 100 hp gas generators would be clutching servo actuators for attitude control. Without making studies it was obvious that much energy could require much signal shaping to maintain stability. The Up-Down counter summing junction concept would not do the job, the loop closure would require a high speed data processor.

System #2 Loop closure down stage using an Arithmetic Logic Unit.

Arithmetic Logic Units: The new 4 bit ALU was a godsend to the computing business. Prior to it’s arrival arithmetic was done by shifting “A” numbers and “B” numbers through a two bit adder and accumulate them in a “C” shift register. Fortunately ALU’s created a big stir, trade magazines included articles on how to mechanize multiply operations. Soon there was a companion Look-ahead carry chip that permitted automation of the carry process. The ALUs could be connected to make an 8 bit parallel processor. A big hurdle was choosing and handling 2’s complement numbers.

Binary Number Systems: The 2's complement number system is one of the three common notations for representing both positive and negative numbers; the others are 1's complement and signed binary numbers. The unique properties of the 2's complement system have made it the most .commonly used notation in general-purpose computers. There is a single unique code for each number and all numbers are treated alike in arithmetic operations and in D-A and A-D conversions, regardless of sign. Arithmetic operations are implemented through a single processing path, providing a greater computational speed than is possible with other notations.

Definition of 2's Complement Numbers Positive binary numbers are known as true binary. To represent negative numbers a sign bit is sometimes added; these are known as signed binary numbers. However, complement binary numbers, either l's or 2;s complement, are more useful representations for positive and negative numbers.

In the 2's complement system the negative of a number, A, is defined by 2ⁿ -A, where N is the number of significant bits required for the full range of numbers in the problem.

The following table: shows sets of 5-bit (four bits plus sign) signed binary 2's complement, and 1's complement numbers. Note that for these number systems a negative number is indicated by a “1" in the MSB of sign bit position. Also, as indicated by the definitions, the 2's complement is always one greater than 1's complement. The 2’s complement set could be generated by presetting a counter to the state 10000, and then counting up. Thirty-one counts brings the counter to the upper level of the set. and the thirty-second count returns it to the starting point. This demonstrates the reentrant property of the set. The 1's complement set also has the reentrant property. There two representations for a zero in the 1’s complement set; in arithmetic operations either form of zero may result. depending on the computational sequence.

Most significant binary bits at left, sign bit at far left.

Hexidecimal	Decimal	Signed Binary.	2’s Complement	1’s Complement
F E D C B A 9 8 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -A -B -C -D -E -F	15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15	01111 01110 01101 01100 01011 01010 01001 01000 00111 00110 00101 00100 00011 00010 00001 00000 10001 10010 10011 10100 10101 10110 10111 11000 11001 11010 11011 11100 11101 11110 11111	01111 01110 01101 01100 01011 01010 01001 01000 00111 00110 00101 00100 00011 ()0010 00001 00000 11111 11110 11101 11100 11011 11010 11001 11000 10111 10110 10101 10100 10011 10010 10001	01111 01110 01101 01100 01011 01010 01001 01000 00111 00110 00101 00100 00011 00010 00001 00000 11111 11110 11101 11100 11011 11010 11001 11000 10111 10110 10101 10100 10011 10010 10001 10000	15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15

The generated command and feedback numbers are read as signed binary. The numbers are the same for: plus signed binary, 1’ and 2’ complement numbers. For a negative signed binary invert the magnitude value to get 1’s complement then add 1 to get 2”s complement magnitude number.

4 bit Arithmetic Logic Unit

Left: Block diagram of 2's complement multiplier unit

Algorithm

1. Line up the MSB of the multiplicand register with the MSB of the accumulator.

2. Starting with the LSB, examine successive bits of the multiplier. For 0 to 1 transitions, subtract the multiplicand from the accumulator. For 1 to 0 transitions, add the multiplicand to the accumulator. If there is no change (ie, 0 to 0 or 1 to 1 ), leave the accumulator unchanged.

3. Shift the contents of the accumulator to the right by one bit, but do not change the state of the MSB.

4. Repeat steps 2 and 3 for all bits of the multiplier, but do not shift after operating on the last bit (the MSB).

Booth’s Algorithm for Binary Multiplication.

Multiplier bits:

01 Add multiplicand X, then shift partial product right 1 bit.

10 Subtract multiplicand X, then shift partial product right 1 bit

00 Shift partial product right 1 bit.

11 Shift partial product right 1 bit.

From Slide rule to had calculator

Digital Computer Design Book: About 1973 I came across the above book published in 1963; I bought it and poured myself into learning how to do arithmetic computations. I learned you must precondition binary numbers to be either in 1’s Compliment or 2’s Compliment before or after doing an arithmetic operation, only the very end of the book referred to parallel processing; there were no “cookbook” designs to follow but I was able to extrapolate. Booths Algorithm, was helpful as the logic used associated with look ahead for multiplications.

Multiply by repeated addition: You multiply by performing repeated addition. The Arithmetic Logic Unit could add or subtract but could not do multiply or divide. We didn’t need to divide but needed multiplication.

First ALU processor layout SAR device with associated circuitry

First Processor The above left diagram shows my first attempt to apply the new devices to perform multiplication functions. After 35 years I needed more information than shown to reconstruct exactly how it worked.

Successive Approximation Register for Feedback conversion: A micro electronics research engineer told me he was working on a Successive Approximation Register and explained roughly how it worked. A short time later I found such a device, with an appropriate application circuit in an AMD sales booklet. I ordered one and built it into our system. (Yes this is the same AMD now giving Intel a rough time in the CPU market for PC’s.) The SAR compares it’s “guess” with the unknown signal to home in on the unknown. On a scale from 0 to 100 it will first guess 50, then if the unknown is lower, next guess 25 and so forth until it has resolved it’s lowest bit. It was very fast and I was delighted to incorporate it into our system.

Hydraulic Servo nil filtering – Turbine Gas Servo much filtering

Digital Processor Requirements: I had assumed we would be using proven hydraulic servos and asked Mal Johnson to provided a worst case requirement for such a system. Though we knew motor contractors were looking into systems other than hydraulic we didn’t take that too seriously until Mal told me he’d been asked to look into what it would take to control such gas generator powered servos; so I asked for him to provided a definition of what would be required. This was a much different animal. Mal expressed the requirement in standard LaPlace terminology as used for an analog system. I was stuck, I didn’t have a clue on how to do LaPlace transform functions digitally.

Dr Blare Bona’s “Rosetta Stone”: Differential Equations > LaPlace > Z Transforms > and Sample Data methods

Sample Data Systems: I was having a terrible time, reading many books, becoming bogged down trying to determine how to perform signal conditioning function with a digital processor. I was becoming as irritated with the books and my own ineptness. Dr Blair Bona came to my rescue. Blair would often come in to visit with Mal and observed me struggling with my face in a text book. We’d become acquainted on such things as how to rig up switches to turn on/off garage door lights from house and garage. Blair said forget the stuff in those books, this is what you need. He proceeded to my black board and wrote a lines worth of differential equations, which I could follow. He then expressed the same information in Laplace transforms used for analog systems, which I didn’t understand but could track to a degree. He also showed Z Transforms then converted those to Sample Data representation. I didn’t understand right away how it worked, but I could certainly understand the data processing method. It was as if someone had written translations on a Rosetta Stone – this immediately illuminated the way to do things. I said don’t erase anything until I can write that down. Thank goodness he happened to take an interest; it lifted me out of a pit, I was lost. Using combinations of information I began the design of a Digital Signal Processor.

Digital Processing required for One Gas Turbine Servo Actuator

With this Sample Data loop closure diagram I knew what I had to do.

Test Equipment: we were using an analog electronics test lab, and Jessie, the lead man provided us with a dual trace scope and some old hand calculator parts -- from when he worked for near by Microelectronics. I needed something to use for test equipment and after studying the had calculator parts for a while decided to use the keyboard and display, then Input/Output and memory electronics on a board under the keyboard. We could issue a command and capture the response. 128 bit RAM chips had just come out so I ordered some and made a design. It took Lloyd Gardner about a week to “wire wrap” the parts and have it working.

This “Test Equipment” is now in a show case at the Oberlin KS museum; it worked great.

Arithmetic Logic Units set up to perform 8 bit computations for servo loop closure using Sample Data mechanization..

It took time before I could figure out how to do this for a single multiply process, and more time to handle a full set of servos and their shaping networks. We knew we were on the right track.

System #3 Loop closure upstage using an time shared Arithmetic Logic Unit.

Who would provide the Servo Actuator Electronics? We had meetings with TRW G&C on what would be needed to control the new motor contractor gas powered servos -- they asked what would happen if the electronics was given to the motor contractors along with the servo actuators. I said the responsibility for control and stability should remain with Autonetics, sighting prior circumstance when it became obvious the propulsion people did not understand flight control or electronics. TRW G&C agreed but TRW Propulsion thought they could and should also do the electronics. At that time many were locked on to having a data bus with servo electronics down stage. TRW G&C wanted Autonetics to be responsible for testing the adequacy of any system – but who would design and build electronics for the vehicle was left up in the air. Developmental electronics was being provided by the competing gas turbine servo actuator contractors.

Abandon Data Bus – Use Direct Wire: I made layout studies on the idea of doing away with a data bus and moving all electronics up stage. I found that this would fit well as a flat cable – it was a doable concept. I had another visit with Lou Purpura, saying I believed we should move all downstage electronics up stage – that way we can be assured of keeping that part of the business, and perhaps find a way to time share the electronics from one stage to the next. That it had to handle two stages at a time for a brief period during staging. Lou agreed – adding we can probably make it a part of the Flight Control Computer. Even after I’d worked the details on how to do this, selling the idea encountered the momentum built up for a data bus; what had been sold had to be unsold.

Dr Ken O’Kief Lou Purpura called and said Terri Miwa of TRW called to see if their new employee O’Kief could come visit with me so I could bring him up to speed on missiles. Terri was a good friend of ours since MM I, and then head of their flight simulation group. I told Lou it was OK with me if it was OK with Shuler. Lou said Tom’s already approved the idea – he wants us to keep in contact with TRW as they look into new concepts. This began a fascinating relationship where Ken and I would meet every other week, at Autoneics and at TRW Redondo. After our first meeting I prepared four reports for Ken, covering various options on doing flight control of a Missile X; a term Lou and I had been using. Ken took this up, and helped spreading the term; it was no surprise that the new missile was indeed called MX.

O’Kief had been preparing his reports and recommendations based on information I’d giving him – thus TRW Redondo and Norton were warming up to the idea of using digital electronics using a time shared digital processor up stage.

Programmable Coefficients: It now became necessary to accommodate storing different filter coefficients for each stage; plus the need to program specific processing for each stage as they could be different.

Pulse Width Lookup Feedback select, demodulate, Analog to Digital convert, 16 bit value ship – pre SAR method

Autonetics Anaheim CA

MM I 202

MM II 203

MM III 69 > 235

R&D 1 271

R&D 2 231

R&D 3 231 & 235

B-1B 222

252 new “bomb shelter”

836 final flight control

242 Micro-Electronics

People & Location Changes: When Flight Control moved the Anaheim plant during MM I, we were in bldg 202. When reorganized for MM II we moved to bldg 203. When I was assigned to help start MM III I was sent to bldg 69 and a year later to bldg 235 under Bellamy. Our Post Boost organization went from 75 to 6 people placed under Dale McLeod of Navigation Systems in bldg 271; I was placed on Bellamy’s staff. Other parts of Autonetics were also cutting back, my desk was on the second floor of 271 with some 20 other people put on a “do not lay off” list by Tom Shuler, who became alarmed at loosing key people, only later did I learn of this list. This was a sound call, skilled people were retained and migrated into other assignments. With Post Boost program dead I’d suggested to Lou Purpura that someone should look into how to convert analog to digital electronics. He said why don’t you look into that, I did and distributed a letter on my findings. I was due to that letter I was provided access to a Navigations Systems lab and Emil Kohler, previously on loan to Rocketdyne, was assigned to help. We built system #1 in bldg 271. Circuit cards for that first design were wired by Navigation systems lab personnel then installed by Emil and I working as technicians building the system ourselves. Walt Evans, who had invented and written a book on “Root Locus” method of controls analysis had been brought back from lay off and assigned to look at what we were doing. I had known who Walt was when he car pooled with George Keller by initial boss. Walt was a mathematician and felt out of place and soon found other work, I felt sorry for him in a way, his world had been analog electronics and this digital stuff didn’t mean a thing to him – and me being a neophyte was not much help. About a year later Walt suffered a stroke and could no longer speak. As previously told we demonstrated System #1 to many people including Shuler (chief engineer) and held a design review attended by Niese (chief scientist) & McLeod (department head) and others whereupon it was decided we needed the ability to do signal filtering – thus the need for a digital data processor.

We reorganized again and I was placed on George Anderson’s, (Navigation system Group leader) staff and shared an office with Mal Johnson in building 232 and given access to a Navigation systems lab in bldg 235. Lloyd Gardner technician under Frank Philips of Flight Control was assigned to work with me. Emil Kohler took retirement and Karl Loefgren a new engineer with masters degree was also assigned to work with me.

I had been asked by Carl Boody to attend a meeting with him at Norton Air Base, for a TRW presentation of their “data bus” studies at which time I found the criteria they were using came from reports I had written on the concept. Boody knew this and asked me to join him. It was after this I became convinced we should move the nozzle control electronics up stage.

In another shake up we were assigned space in a Frank Phelps Flight Control lab in bldg 235, Lloyd Gardner, Karl Loefgren continued working with me and we built system #3 in that lab. We needed a “missile” mockup so I took them with me in my car to look at some white plastic patio tables – after looking them over we decided we could stack them and make a three stage booster mockup with a stage 4 on top.. I bought the tables and we brought them into the lab – you’d have though we were kids with new Christmas toys.

Princeton Algorithm Mechanization Princeton Algorithm Multi Coefficient Processing

Speed Requirement I checked with Mal Johnson on how fast the data had to be processed; after some discussion Mal said we should close each servo loop every 2 milliseconds. Since a final design would be built with nuclear hardened parts; I assumed a down grade of 50% in speed as compared to commercial bi-polar parts. I made speed calculations and found our design could not do seven multiplies on four servos, the design was simply not fast enough and something would have to be changed. The transition between stage I and II would be the most critical because staging was done under the influence of atmospheric pressure – when the missile is still flying with the CG aft of the center of pressure – aerodynamically flying “feather” end first. Again Dr Blair Bona came to my rescue.

Princeton Algorithm: Blair said what you need to do is use the Princeton Algorithm. This permits setting up coefficients in such a way that a multiplication can be done much faster. That was in 1975 and now in 2007 I don’t recall exactly how that worked but it took advantage of the fact that shifting a binary number does an instant multiply by shifting the binary “decimal point”. I modified the design to handle arithmetic in this way – the computations could be done fast enough but it was necessary to modify the way data was handled. I presented these findings to Dr O’Kief of TRW, he wrote a report on it and this put the subject to bed about digital being fast enough for the high speed servo loops.

Our “missile” at the back of the lab looked quite impressive. We set it up with three stages, with pitch and yaw servos of the kind we’d used for system #1, only this time we stuck on a plastic funnel on the gimbaled stick to look like a rocket nozzle.

Karl Loefgren had worked for Bell Labs before coming to work for Autonetics and knew about new microprocessor systems and had been in contact with Earl Hicks head of the Micro Processor Lab. Earl provided Karl a Motorola 6800 microprocessor system for us to use. We set this up as “launch control” on a work bench and strung an umbilical cable from it to our near by missile. We also added an Autonetics built AIM computer system to aid in setting up the signal processor program.

Programmable Executive Program

Karl showed me how to use the new microprocessor system to and we used it to check our missile control circuits that were time shared in their turn for missile nozzle control. We were confident we were on the right track.

I’ve already told you that: When I learned Karl was going to be gone on vacation I decided I’d better learn how to run the 6800 processor system and began trying to set it up – I got stuck and asked Karl how to do a certain thing. Karl said, but I’ve already shown you how to do that. I looked at Karl and said I’m aware that you have almost perfect recall but I do not, you showed me a lot of stuff two weeks ago and I need your help in refreshing my memory. We had a good chuckle and moved on. There was a way to display a memories content in hexadecimal format, and I didn’t recall the digital coding required to cause it to happen. This was long before such things as a BASIC language for entering commands.

Dr O’Kief’s Appraisal: We were at TRW Redondo and had been going over the concepts for what I had always called a digital signal processor, or a “Digital P-92” the name of our analog processor; when Ken leaned back and said, “this is fantastic – but it’s not a processor, it’s a programmable computer; it has all the attributes of a full up computer.” Since I didn’t know anything about computers, I’d never thought of it that way, I was only making use of the new devices to do necessary signal processing; going another step each time to solve a problem.

It was then Ken the told me his Phd was in computer science. Flabbergasted I asked, why didn’t you tell me? He said you were so far ahead, using these new TTL parts, from what we were doing in school -- I was too embarrassed to tell you, especially when you kept telling me you didn’t know what you were doing. We discussed how this could be, as I constantly had a feeling of being behind trying to catch up, always feeling dumb, each day faced with something I didn’t know how to do. It was the availability of the new parts that made it possible, new ones were coming out every week. I told Ken it was thanks to Blair Bona, I had learned about how to handle sample data and about the Princeton Algorithm to perform fast data processing.

The DNA Molecule when out to lunch O’Kief would tell me about his night class on Mexican history and I’d tell him about my night class on biology. He once called me up, asking if I’d join him in writing a book on the DNA molecule, saying we could do better than an existing book by Asimoff. I declined as I was reading a book by Watson, one of the co-discoverers of DNA, on Molecular Biology and was overwhelmed by his excellent book on the subject.

Self Replicating Computers: About that time after returning from lunch Ken asked” “do you think it would be possible to make a computer that could replicate itself? His question caught me off guard, I’d never given automation any thought. I pondered his question, imagining setting up an automated production line to make such a thing as the digital signal processor we’d been talking about. I could imagine automation handling everything except the supply of parts – my answer was, “no, we could not handle the supply problem”. Ken smiled and said, “I’m looking at one.” It took an instant to change gears and recognize what he’d said. We humans are computers that replicate ! For days and days there after I pondered on how does nature handled the supply problem. About a month later, one Sunday afternoon, looking at particles drift in the back yard pool, it came to me. I knew from Molecular Biology how atoms can assemble themselves to a provided a code – and that this always takes place in a fluid environment. Random thermal motion of a liquid can transport chemical parts for assembly to a pattern that rejects all but those that fit the code key. Life began in the water and all living things need water to transports the chemicals of life.

I lost contact with Ken, his wife was in foreign service and was to be posted in Germany, Ken took a crash course learning German and went there with her, he got a job doing computer programming for some firm in Germany while she was posted there.

Nancy, Shuler’s Secretary saw me at the coffee machine, I’d known her since I hired in. She asked what I was doing now, so I took her to the lab and showed her our missile. She said Tom’s got to see this. In about 15 minutes Tom was there asking what I was up to, so I showed him. Some six months prior I had shown Elliott Buxton what we were doing when still in bldg 232, and had feedback that Buck was quite impressed. Tom looked over what we had, and without lingering too long said I’m going to send Bob Niese (chief scientist) and Tom Gunkle (heard of computer department) to see this.

Dr Niese and Dr Gunkle Review: Bob & Tom arrived at a time when Karl Lofgren was off on weeks leave with his wife for Olympic rowing tryouts. (they were quite good but that was the year President Carter cancelled US participation.) Lloyd Gardner was also gone so I played host to Bob and Tom for almost a full day. Tom sat on a tall stool, watched and listened while I explained and Bob asked questions. I don’t have a clue what they really thought. They were professionals in this field and I was an amateur. They were both very polite and listened to my full tale of why I was doing what. But it really didn’t matter because a very short time later it was announced there would be an MX missile and that Flight Control for boosters would be done by Autonetics by electronics incorporated as part of the Flight Computer’s I/O section. It became a part of Tom Gunkle’s organization and our Flight Control Electronics was out of the loop -- I had helped sell the idea that put me out of a job.

Gas Turbine Servo-Actuators fail when tested by Autonetics: There were two kinds of actuators competing for use, one where the output of a gas generator drove a gas motor which drove a flex shaft to power a mechanically clutched servo. The other incorporated it’s own turbine and was powered direct from a gas generator. We tested the one with the flex drive. I had been invited to look at the servo actuator and made the comment that the unit would fail where the flex shaft connected to the actuator via a brass bushing bearing. They laughed an said but it’s not been tested yet. It was tested and failed where I predicted it would. Naturally I was asked why I predicted it would fail there. I told them I had once used a rotating stiff wire brush driven by a flex shaft when working at a tire shop, in order to buffer tire rubber. That when I would start to use the buffer it would almost yank the part I was holding out of my grasp. I knew that making a 90 degree turn with a flex shaft places a huge load on the bearing – and their system was harnessing 100 horsepower of hot gas energy, not the fractional hp I tried to hold.

Buxton called in – TRW G&C choose Hydraulics: I was convinced the use of the intended kind of gas powered servo actuators would be a disaster for the MX program and visited with other flight control fellows as Mal Johnson, Jim Anderson, Lou Purpura and others about what we could do about it. I suggested we tell Buxton about our concerns and see if he could talk with his old TRW G&C flight control friends about looking into the use of hydraulic servos. We spoke with Buck and I explained that when TRW-AF funded the motor contractors, they excluded the use of proven hydraulic servos, as a way of seeing if there might be a better way. I said the propulsion people have become enamored with their experimental system. Jim Anderson had been in contact with Moog and they were ready with a hydraulic servo concept designed for the application. Buck talked to the TRW G&C people as did Moog. TRW placed responsibility for the servo controls back in G&C hands and they called for the use of hydraulic servos. The need for 7 th order digital filtering and the Princeton algorithm went away. The Buxton’s “feat” was remarkable as he was only working part time -- as he was recovering from a stroke that left him partially paralyzed on his right side – his solid judgment was recognized as functioning and valid. That left some people quite unhappy, but it was the better choice.

Rockwell 6502 Computer & Commodore Computers: My recollection of events becomes muddy as we were reorganized more than once in a short time, I had an office on the second floor of bldg 231 then we were moved to the second floor of bldg 235 as MX got underway. That’s where I experimented with newly purchased Commodore Computers at home and at work plus a Rockwell 6502 operating system at work. I had in mind using these machines as test equipment for testing systems built in the lab. To demonstrate I wrote a program for a set of servo controls using a new small Commodore and storing the program on a cassette tape and gave it to Tom Gunkle to show how these desk top machines could be used. I also experimented with using an Autonetics built computer system called the “Rockwell 6502”, which had to be programmed in machine code.

Patent Award: while sill in bldg 231, listening during a George Anderson Staff meeting, I was asked to write up one of my ideas and submit it as a patent disclosure – my “free” support had a price. Reluctantly I did and the AF decided to sponsor it for patent by the AF. (Employees sign an agreement when hired that patents belong to the company or AF depending on who was paying the bill.) Some six years later I received an award of $1000, for having submitted an idea that was patented.

B-1B Reactivated: President Regan reactivated the B-1B shortly after he came into office; John Cox (ex B-29 pilot) and I were sent to help restart the B-1B program. I took two Commodore Computers, belonging to Autonetics for our work there, these were the first desk top machines they had ever seen there. We used a word processor program I’d written in machine code, that would to text with arithmetic, and we used this on our machines for B-1B work. I was assigned to work on Hydraulics and Pneumatics. My boss George Anderson referred to me as a generalist – a specialist in multiple fields. I was repeatedly caused to adapt to and learn a new fields of endeavor as contract demands changed needs. It would be some 8 years before I become involved in electronics design again, to solve a problem on a C-130H Gun Ship design.

Epilog on Digital Processor: I happened to return to Autonetics while working on the B-1B and met Terry Miwa of TRW who was arriving for an MX meeting regarding booster servo controls. Terry had been present at a meeting when I presented concepts for how to use digital in lieu of analog electronics. After the usual greeting and hand shakes we started to part, then Terry said, they should have stuck with your design. I asked how’s that? He said someone removed the programmable RAM from your design and used fixed coefficient values which now have to be changed. I was pleased that Terry remembered.

Semiconductors Changed the world in Many Ways

Train Pulling Power Before ending this I’d like to tell how semiconductors permitted train engines to apply full power at any speed. Coal powered engines had switched to Diesel powered engines, where Diesels powered an electric generator which powered electric motors. Semi conductors made it possible for the operator to cause the motors electro magnets to always be applying maximum pull at any speed, even pull in reverse.

Most household electric motors run at a fixed synchronous speed, their magnetic poles rotating at the 60 cycle rate set by the power source.

Semi conductors permit a train operator to cause the magnetic poles to “move” in sync with the motion of the train. When at dead stop the magnetic pole moves like a carrot on a pole in front, always just ahead – unless put in reverse to stop. This feature greatly increases the efficiency of our infrastructure, though it’s out of sight out of mind.