THE KWR37 ON-LINE CRYPTO RECEIVER by Jerry Proc VE3FAB 68 Braywin Dr., Etobicoke, Ontario. M9P 2P3 NOTE - This article contains no classified information and the material which is presented has been cleared by the National Security Agency. To many outsiders, cryptography is a fascinating subject and detailed material is not easily obtainable. At the turn of the twentieth century, messages containing confidential information were encrypted using code books and continued this way up to end of World War 1. In the 1920's, mechanical, rotor-based machines were developed in Europe and the United States for the purposes of encrypting commercial business traffic. By the late 30's, the German military had adapted a three rotor crypto machine that we now know as the Enigma. Rotor based machines continued to be refined until replaced by vacuum tube technology and finally, by the computer based technology of today. One of the most critical applications for secure communications was the military naval environment and the story which follows, features the KWR37 on-line crypto receiver, as applied to a naval role. GENERAL DESCRIPTION ------------------- During World War 2, most 'low level' traffic from surface ships at sea was manually coded and decoded using code books. In the post war period, the amount of 'broadcast' traffic continued to increase to the point where it required some level of automation as manual decoding could not keep up with the volume of traffic. Automation made its appearance in the form of a teleprinter based receiving system. The sole purpose of the KWR-37 was to automatically decipher the encoded fleet broadcasts which was sent to military ships at sea and other applications where it was required to decode a steady stream of classified traffic. On the input side, the '37 was connected to the 60 millamp current loop output of a frequency shift converter. The output side of the '37 was connected to a Teletype page printer. The interconnection of the KWR-37 is expressed by the following diagram: ANTENNA-->--RADIO RECEIVER-->--FREQUENCY-->--CRYPTO-->--TELE- SHIFT RECEIVER PRINTER CONVERTER The shore station started each day's broadcast at 0000 Zulu and transmitted without interruption for 23 hours and 55 minutes each day. On shore, the encryption device such as a KWT-37 was synchronized with a time signal station (CHU or WWV) and the originating station sent an automatic 'start' signal followed by a continuous stream of encrypted, non-repeating traffic throughout the day. The decoding 'key' which was similar to an IBM style punch card had a pattern of randomly punched holes, and had to be changed daily, prior to the start of the next day's broadcast. Encryption keys were changed by unlocking a front door on the KWR-37, removing the existing card, and installing the card that was designated for the next day. These cards were inserted behind a small door in the front of the KWR-37 using built-in, alignment pins. The door closed against a block of small, spring-loaded steel pins. Where a pin touched the paper card, no signal passed; where a pin poked through a hole in the card and touched a silver-plated metallic track, a circuit was made. Each card held enough keys to cover 14 years of usage before the key repeated itself. Used cards were destroyed on a periodic basis. John Dill of Kingsville Texas, was a crypto mechanic in the USN in the 1960's and 70's, and kindly documented his experiences with the KWR-37. "The holes in the punched cards directed the key stream to a series of bistable multivibrators (flip-flops) which were wired on thirteen printed circuit boards located on the left side of the machine when one opened the equipment drawer. All the flip-flops plugged into a motherboard which was positioned horizontally. The active devices in these circuits were sub- miniature, type 6088 sharp cutoff pentodes with wire leads and they were made by Raytheon or General Electric. These tubes were 5/16 inch in diameter and 1 1/4 inches long and anchored by metal clips on each circuit board. The 6088 pentode was also known as type CK522AX. Depending on circuit design, the 6088 could be driven to produce as much as 10.5 mw of power at the high end or as little as 1.2 mw at the low end! One multivibrator stage consisted of two 6088 pentodes for the flip-flop and one 6814 sub-miniature triode amplifier, a vacuum tube originally designed for late generation tube computers. All stages had to be perfectly balanced, hence the use of resistors with 1% tolerance. Typically, the pentodes ran at 67.5 volts B+ and the triodes at 100 volts. There were four flip- flops per board and the entire unit contained approximately 500 tubes! Mechanically, the '37 was 19 inches wide, around 24 inches deep and 8 to 9 inches high with a case finished in navy cabinet grey. It could either be rack mounted or positioned on a equipment shelf. With a weight approaching 100 pounds, it was definitely a two man lift when being installed. At the rear of the unit there was only three sets of connections: current loop in, current loop out and power. In addition to the operational key cards, there were also cards used strictly for testing. Each card in the test deck, checked a different KWR-37 function. Two of the cards, produced a distinctive pattern of beeps to indicate proper operation and the technician had to listen attentively. With care, the test cards could last for years. OPERATION --------- Transmissions began at 0000Z and continued without pause or repetition for 23 hours, 55 minutes each day. Whether any messages were being sent or not, the 'customers' KWR-37's were on-line, in sync and receiving the transmitted key stream. In the event of a power loss or if the unit went out of sync, the operator would have to initiate a restart. When the sending station stopped transmitting, all receiving units worldwide would be prepared to receive transmissions for the next day. If radio conditions were normal, the transmitting station's Auto Start signal would automatically start the machine. If Auto Start was missed due to atmospherics, the operator had to late start the unit. This procedure is discussed further in the text. On the front panel of the unit, there was a control composed of two concentric dials; the outer for hours and the inner one for minutes. Above that, were three miniature switches marked Start, Reset and Sync. Two small, orange lamps tagged Mark and Space flashed alternately in time with the incoming signal. Re- synchronization of the KWR-37 required that the machine be reset, then run it forward in time at high speed to catch up to, then slightly pass, the transmitting station's key stream. The operator would set the Hours/Minutes dials to the difference between 0000Z and the current Zulu time. The Hours dial was marked in 1 hour increments up to 23. Similarly, the Minutes dial was marked in 5 minute increments up to 55. The Reset switch would then be pressed. This would reset the flip-flops in the Key Generator and the Internal Clock and ensure that all these circuits started up from a desired, known, pre-set value. Internally, the reset signal was routed to the flip-flop stages through the Key Card, thus changing the initial 'set' state of the Key Generator. Pressing the Start switch would enable and start the clock which began to drive the flip-flop stages thus producing the key stream. Activating the Sync switch would give approximately 15 seconds worth of high clock speed, akin to a fast forward function. If for example, the KWR-37 had dropped off-line at 14 hours into the broadcast day due to loss of ships mains power, and restoral took 15 minutes, the operator would set 14 hours, 15 minutes on the dials and hit the Start button. The machine would run in high speed for several minutes until the clock had advanced the key stream 14 hours and 15 minutes, at which time it would drop back down to normal speed and start searching for sync. This process forced the KWR-37 in constantly comparing it's own internal timing to that which was being sent on the broadcast. If a clock comparison was unsuccessful, the clock would delete a pulse, effectively dropping it back in time by a small amount. Each time this pulse deletion occurred, an audible beep was sounded through a panel mounted speaker. As the beep rate slowed, it told the operator that synchronization was approaching. After several seconds of silence the Sync light would illuminate and the teleprinter (ie Teletype machine) attached to the '37 would start printing. If the search for synchronization ran over several minutes duration, the '37 would alarm again with a steady, irritating, much-hated tone from the speaker along with the dreaded red Alarm light. Standard procedure called for resetting the machine and trying again. Since no two KWR-37's were exactly alike, the presence of the alarm did not mean that the machine stopped searching for sync. The alarm simply meant that the allotted amount of time had elapsed, during which, synchronization should have been attained. In many cases, the '37 achieved sync with the alarm sounding and the SYNC light on. At this moment, the operator would silence the speaker and everything would run normally. This was the official procedure for achieving synchronization. In practice, however, it was an entirely different world. An operator would generally attempt the formal procedure. If this did not achieve results, a whole series of 'homebrew' remedies could be applied. Among these miracle cures for lack of sync were: a) Pounding the front panel briskly just prior to pressing the START switch. b) Opening the equipment drawer and hitting the tops of the circuit boards with some hard object such as a mallet or cleaning brush. c) Opening the front door; removing the key card and cleaning the conductive tracks in the rear of the front door with a rubber eraser. This practice removed the plated silver on the tracks and was frowned upon. d) Cleaning the conductive tracks with Teletype paper or paper money. Since Teletype paper contained trace amounts of oil to assist with lubrication, this practice was highly discouraged. e) Rapid and vigorous spinning of the time-delay dials, followed by many shots on the RESET button. f) Uttering foul, abusive language at the machine in order to let it know who was in charge. MAINTENANCE ----------- The KWR-37 was very old, tired and well past it's design life in 1968 and did not improve with age. Many technicians only had a modicum of training in the art of soldering. For the '37 family, this was a disaster as the most frequently performed corrective maintenance involved the replacement of wire lead vacuum tubes. One can only imagine the damage that was done to the printed circuit boards after 20 years of mediocre maintenance. To ensure the highest reliability, crypto mechanics tried to turn out a machine capable of operating normally with only 1 volt of filament voltage to all the 6088 pentodes. The standard setting was 1.25 volts and was indicated by a front panel meter. Each pentode had a filament draw of 20 ma. If the unit ran properly at a reduced filament voltage, that meant that the tubes had strong emission and the unit would run reliably. As emission decreased, the operator could increment the filament voltage to restore normal operation. When the machine became unreliable at a setting of 1.25 volts, it was turned back to the maintenance depot. Checking for operation at a reduced filament voltage became known as 'margining'. The 6814 triodes which used indirectly heated 6.3 volt filaments were not margined. Later and unofficially, an extender board was developed which allowed individual circuit boards to be margined. Once each board ran reliably at one volt filament voltage, the filament supply to the entire machine was reduced. If it worked, it was considered ready for use. Testing each board individually improved the quality of the troubleshooting process. The majority of maintenance problems in the '37 originated in three areas of the machine: the 'S' circuit cards, (the ones containing the key stream flip-flops); the 'T' cards which combined the 'S' card outputs and the 'U' or alarm cards. Next, were the cards which allowed the '37 to run at high speed. The modified card extender was invaluable in finding these circuit faults and eventually won official approval. A maintenance bulletin was circulated among all KWR-37 holders documenting the modified extender, the construction details and stock numbers of the parts required". John goes on to comment about his worst KWR-37 repair job. "A navy technician had the '37 drawer open for maintenance. Innocently, a brand new Ensign, who was the ships Communications Officer noticed the activity and came over for a look. He must have been having a hard time at sea because of the large bottle of Maalox (stomach antacid) in his shirt pocket. As the Ensign leaned over to peek at the '37, the bottle fell out and broke on the top edge of the equipment drawer. Needless to say, the Maalox spilled throughout the machine and a large blue flash ensued as the power supply shorted out. Flames and smoke began issuing from the drawer. The tech had been sitting on the deck in front of the '37 cross-legged with his legs underneath the extended drawer. His burning trousers were quickly extinguished by the remainder of the Maalox running out of the equipment. In his haste to escape, the tech placed his full weight on the card rack and broke the motherboard in several places. The '37 was eventually repaired but the cost to repair, likely exceeded the value of the machine". CONCLUSION ---------- During it's in-service life, the security of KWR-37 system was compromised on at least two different occasions. When the USS Pueblo was captured in 1968, the north Koreans acquired fully working KWR-37's along with active key lists. Naturally, there was a mad scramble to quickly change all of the key lists held by KWR- 37 'customers'. In the mid 1980's, it was discovered that the infamous 'Walker spy ring' was selling active key lists (ie the actual IBM style punched cards) to the Communists. Once again, the key lists had to be quickly changed. It's important to note that simply possessing a machine was insufficient to copy the traffic. Any adversary had to be in possession of the active key lists as well. By the early 1990's, any remaining KWR-37 crypto receivers were taken out of service and destroyed. This sounds like a sad ending, but such is life in the world of cryptography. I extend my sincere thanks to John Dill for providing the source material which made this article possible. --EOF--