On this page, scroll to:

1.  Introduction [left]
2.  1624 [next]
3.  Computer Types: UDT | NIKE | RISC | CP-2044 | UYQ-70 | CLC 
4.  Special Purpose Computers by John Alton
5.  Special Purpose Computer notes by Wayne Olson

The 1624 was designed to emulate the Digital Equipment Corporation (DEC) PDP-24 machine using Micro Program Controller techniques.  These units were packaged into commercial 19" racks and used for an A.F training program.  Bill Bauer was the project engineer for these units. [lab] 

3. Computer Types

3.1 The CP-788 [UNIVAC type # 1215, Universal Digital Trainer (UDT)] was built using printed circuit cards from the 1218 computer, with many of the same salient features but 15 bits instead of 18 - a lesser memory addressing capacity.  The memory speed was 8 usec.  This was also called the Universal Digital Trainer.  Don Mager was the lead logic designer for this development. First unit delivered July 5th, 1962. [lab]  

3.2 NIKE by Ron Schroeder -

     When I first went to work at Remington Rand Univac, I worked on the Nike project. There were two computers, the TIC and MAR I (GPDC) machines built circa 1960 that were quite advanced.  They were followed by a prototype CLC machine that I believe never went into production.  Much of that work went to CDC when Joe Bloomfield [sp?] left the company.
     George Gray does provide some info in his Sperry Military Computers paper.  John Alton's career summary also touches on this computer type.

3.3 Reduced Instruction Set Computer (RISC):
     An example of specialized engineering was the RISC project.  In the mid '80s a Sperry Univac design team proposed and was awarded a contract by the Central Intelligence Agency (CIA) to develop a radiation-hardened MIPS RISC microprocessor chip design.  Lowell Benson was named as the Technical Program Manager.  Don Bennett was the architect for the design concept.  John Porter was the lead engineer doing the design and simulation.  Dr. Vic Wells was the Semi-Conductor facility leader for chip processing.  A paper design with a logic simulation demonstration was delivered for Phase I.  RCA and Westinghouse had a parallel contracts.  We and RCA were awarded a Phase II contracts to build and demonstrate chip set operation - a basic processor and a floating point processor. Don Bennett and I presented our design at a classified conference held on Hilton Head island off of the South Carolina coast.
     After Burroughs bought out Sperry and we became Unisys, the new management decided to close our Eagan semi-conductor facility which was supposed to be the builder of this Phase II radiation hardened chip set.  Between Engineering's Jim Stewart and PM's Lowell Benson, 15 'sand factories' were visited to ascertain their technical capabilities.  Jim and I then prepared a Statement of Work and Request For Quote (RFQ) document which was sent to six of these fifteen facilities that we had determined were capable of building the pre-requisite 2 micron devices using our design files.  Three of the six responded to the RFQ.  After using a Kepner-Tregoe evaluation method, we selected United Technologies in Colorado Springs to build from our design files.  We did an update to our design files to adapt the design to their processing design rules.  They built, we demonstrated, our extensive simulation led to first pass operational silicon chips during phase II.  {Editor's Note:  We ended up with a few spare parts from the assembly process.  As program manager, I kept one to use in presentations, shown at the right.} 

This was the world's first radiation hardened 32-bit microprocessor on a chip. 

     In spite of having to change to a new manufacturing facility, we were within budget on this Cost Plus Fixed Fee contract.  We were only a month behind the original proposed schedule.  According to our CIA technical contact, RCA was four months behind and over spent.
     The CIA opened the competition to four companies for Phase III.  Neil Hahn brought in a 'systems engineer' from Harris [I don't recall his name] to lead our proposal effort.  WE LOST, a minor moral victory was that the CIA directed the winning bidder to use United Technologies as the primary manufacturer of their chip set. [lab]

3.4 CP-2044:
The CP-2044 used three 68040 microprocessor chips to effect an emulation of the CP-901 ISA as a next generation replacement for the core processor in the Lockheed P3C ASW aircraft for NAVAIR. [lab]
     Today the P-3C’s anti-surface warfare improvement program (AIP) mission system is used on 98% of the world’s maritime surveillance aircraft (MSA). Since the early 1960’s we have contributed to more than five million successful MSA flight hours.

4.0 Special Purpose Computers

June 2008 - Recollections from 40+ Years Ago - John Alton [with help]
4.1. Introduction
Following is an unclassified summary of special purpose computers built under US Navy contract by Sperry Univac during the period from the mid 1950s through 1967. These computers in development sequence were called: Blueplate I, Blueplate II, Thornhill, Blondell, Belter I, Belter II, and Seacrest.

The application of these machines involved solving very large computing problems. This was accomplished by using an exceptionally large quantity of logic and memory circuits and by implementing most of what would normally be done by software with hardware circuits instead. These machines had roughly ten times the number of logic circuits as our general purpose machines at the time.

The following descriptions focus primarily on Thornhill because it was the first in the series that used new technology followed by other machines which used essentially the same basic technology.

4.2. Thornhill Cabinet Structure

a) General
The Thornhill machine was housed in a large cabinet 35 ft long X 8 ft high and 7 ½ ft. wide. The cabinet contained three major sections, the electronics, the air handler portion of the cooling system, and the power supplies. Photo 1 [right] shows the cabinet under construction. On the left are the power supplies. In the center is the air handler portion of the cooling system. On the right is the electronics end of the cabinet.

b) Electronics

• The electronics end of the cabinet has two banks for electronics with a walkway between them for access to the back side of the chassis. Photo 2 [below left] shows an end view where the chassis will be installed. There are five bays on one side and 4 bays on the other. Each bay contains 15 chassis; 3 wide and 5 high for a total of 135 chassis.
• The chassis were made up of a stack of alternating printed circuit back panels and insulating layers topped by a ¼ inch thick aluminum ground plate. The back panels were made up of many [up to 20] 15” X 12” two layer printed circuit boards with plated through holes. The back panel boards and the ground plate were interconnected with electro-less nickel and gold plated roll pins that also formed part of the sockets which accepted the logic and memory circuit cards. Photo 3 [below right]shows the ground plate with white plastic insulators that were molded into the ground plate. It shows ground bushings being inserted into the ground plate to connect the circuit cards to the ground plate, two bushings per connector area. The roll pins that interconnect the back panel boards will terminate at the surface of the white insulators, forming the socket for the circuit cards. The circuit cards, which were made with discreet components, were unique in that the connector pins were not straight, but waved such that there were multiple opposed pressure points where they mated with the back panel roll pins. This provided a highly reliable connection. Single width cards were 1” wide X 4” high. Cards were also made in multiple widths. Photo 4 [below left] shows a partially populated chassis. 
• Printed circuit routing for back panel artwork was generated by automated design programs developed for the project. These programs also drove a large flat-bed plotter which rendered the artwork used for fabrication. Photo 5 [below right] shows artwork, photo mask for etching, and a finished back panel board. 
• Clocking the Thornhill circuits at 10 MHz was a major challenge. It was solved by synchronizing the clock with fine tuning the timing circuits to nanosecond accuracy at each of the 135 chassis in the machine.

c) Power Supply and Distribution

•  The two large power supply cabinets and power supplies were developed by the Bogue Electric Corp under contract from the Thornhill project. Five voltages were provided; among them +5V, +15V, and a negative 5V. Two thousand amps were delivered by the +5V supply. Lesser current was delivered by the other power supplies. 
• The power was distributed to the chassis by a series of bus bars. The main bus bars traversed horizontally along the cabinet bottom and top of each side of the cabinet from the power supply to every bay of chassis. Photo 6 [below left] shows one of the four bus bar sets. Smaller vertical bus bars carried power to the chassis stacked 5 high and 3 wide in each bay. The main busses were made up of 2 copper bars, each ½ inch thick and 4 inches high, separated by a very thin dielectric. This design was chosen to maximize the capacitance and minimize the bus inductance. 
• Photo 7 [below right] shows (between the power supplies and the air handler) the bus bar terminations that will eventually be connected to the power supply. 
• A 150 horsepower motor-generator set provided ac power to the power supplies.

 

 

d) Cooling system

• The cooling system provided 35 tons of air conditioning, built as a conventional refrigerant evaporator – condensing unit system with an added 100 Kw reheat unit. This allowed the air to be cooled below system requirements to dry the air by condensation then reheated to the correct cooling temperature for the electronics. 
• The evaporator unit is housed in the air handler shown on photo 1.  The air handler is driven by a 20 horsepower blower.  The compressor and make-up water tank are located elsewhere and are shown in photo 8.  [below]



 

The condensing coil was in the cooling tower located on the roof of the building.


• Photo 6 also shows plenum air outlets for each column of chassis on one side of the cabinet. Return air would be captured by similar boots above each column of chassis. The same design applies to the other side of the cabinet.

 

 

 


 

 

 

4.3  Blondell
The same technology was used on Blondell as was used on Thornhill.

 

4.4  Belters I and II Cabinet Structure

a) General
The Belter cabinets; housing the electronics, air handler and power supplies were 32 ft long, 8 ft high and 7 ½ ft wide, three feet shorter then the Thornhill. The system also included a 1230 Computer, 1532 I/O Console, 1240 [2 handler] Magnetic Tape Unit, and an Emergency Control Unit for switching the two Belter machines.
b) Electronics
The logic and memory circuits were packaged as in Thornhill, using many of the same card types. The memory circuits were done with integrated circuits.

 

4.5. Other Data
a) Circuits and Packaging


Machine	Logic	Memory	Packaging
Blueplate I	Transfer Circuits [like USQ-17]	Magnetic Core	USQ-17 [Note 1]
Blueplate II	Transfer Circuits [like USQ-17]	Magnetic Core	USQ-17 [Note 1]
Thornhill	Transistor-Transistor Logic (TTL)	Magnetic Core Thin Film NDRO Thin Film DRO	Described in par 4.2 b), bullet 2 above
Blondell	As in Thornhill	As in Thornhill except no NDRO	Ditto
Belter	As in Thornhill	Integrated Circuits	Ditto
Seacrest	As in Thornhill	Integrated Circuits	Ditto

            Note 1: Discrete component PC cards and wire wrap back panels were used on the Blueplate machines.


b) Miscellaneous


Machine	Input Data Means	Design Area	Build Area
Blueplate I	Punched Paper Tape	Building 3 Basement	Plant 3
Blueplate II	Punched Paper Tape	Building 3 Basement	Plant 3
Thornhill	Punched Paper Tape	Building 6*	Building 6
Blondell	1230 Computer	Building 6	Building 6
Belter	1230 Computer	Building 6	Building 6
Seacrest	1230 Computer	Building 6	Building 6

    *Building 6 was located just east of the main Plant 2 office building on Minnehaha Ave. in St. Paul.
4.6. People Consulted
Curt Bute, Joe Kimlinger, Gordon LaValley [Provided significant data on hardware design], Wayne Olson, Jack Pollack [Provided significant data on Belter and Seacrest], and Norm Petrowski.

5. Special Purpose by Wayne Olson

VIP Club page 58 updated on Wednesday, January 20, 2010