top of page

Overview of Toll Switching Systems


This section provides a broad perspective of toll switching systems as elements
in the North American domestic telephone network. Call routing falls into two functional categories: local and tandem. Some of the materials in this introduction are derived from [Engineering]. 

Local switching systems (central offices) connect customer loops directly to other local subscribers or to trunks leading to a remote office. A trunk is a talking path (and digit forwarding) link between two offices (wires, coaxial, radio). 

A central office may support two or more 3-digit central office codes. The last four digits of a telephone number provide up to 10,000 line numbers within each central office code. By example, for 552-1677, 552 is the office code and 1677 identifies a subscriber in that (San Francisco) office.  In the hierarchy of switching systems, a local office is designated as a Class 5 office. 

The term tandem (acting in series) is used generically for any switching system that connects trunks to trunks, where a trunk is a path (2 or 4 wire) between offices of any type. Tandem switching systems connect local switching systems to each other or other tandem offices.

As an aside, voice trunks were analog until 1962 with the introduction of T-1 digital carrier by Bell. It needed a repeater every 6,000 feet and could carry 24 voice channels over twisted pair copper wire. The J-1 analog carrier system from Bell (~1937) could carry 12 voice channels over a twisted pair. The L-1 analog carrier system (1941) supported 600 voice channels over coaxial cable. These analog carrier systems greatly improved on the "one voice channel" per pair methods. However, the more voice channels carried over any wire type, the more often the signal needed repeating to maintain integrity.   

The need for toll switching

Tandem offices are classified according to Fig 1.  The terms local and toll reflect the tariff distinction between local and toll traffic. In addition, some switching systems perform both local and tandem switching functions. Fig 1 relates the toll Class to the exchange type that supports it. The exchange types (XBT, #4,..) are explained later in this article.

Fig 1, Toll Exchanges and Classes of Service

Why 5 classes of switching? Well, the most efficient way to route calls uses a hierarchical topology. A flat configuration requires too many connecting trunks. A hierarchy of switching nodes provides the most efficient routing. Each pass through an office reduces call quality somewhat (more on this below) so minimizing the number of hops is paramount. See the article on Traffic Engineering, Appendix C, for more on efficient trunking concepts. 


Fig 2 shows classes 1 and 2 toll switching on a route map in 1965 for NA.  Class 3, 4, 5 offices are omitted else the map would be a dense tangled web of trunks. We can climb up/down the hierarchical ladder with a simple example. Using Fig 2, start in Portland Oregon and trace out the path to Jackson Mississippi. Five offices are crossed not including the local, Class 5, offices at each end.

Fig 2, Map of Class 1 and 2 toll switching centers with trunks in 1965 (Bell) [Fundamentals]

See the Fig 3 table on toll switching usage in North America. It’s surprising that the oldest crossbar toll switch, the XBT, was in 2nd place (126 systems installed) for electromechanical systems in 1983. The “ESS” rows are for the Electronic Switching System, and we should expect superior performance, compared to the “slow” electromechanical systems. The number of total "terminations" (5.75 million) implies that there were ~2.9 million trunks in 1983. 

Fig 3, Census of Bell System Toll Switching Systems, January 1983

Fig 4 provides a good perspective on the number of Class 5 local offices compared to the Fig 3 total for Toll offices. There were 718 toll offices (classed 1-4) in 1983 compared to 9,780 Class 5 local offices. The roughly pyramidal hierarchical nature of toll offices is dwarfed by local offices. Interestingly, 65% of all toll offices were electromechanical as late as 1983. See Endnote 2. 

Fig 4, Census of Local Switching Systems (approx.) January 1983

Toll Call Routing Example

This section assumes the reader has a basic understanding of call switching operations for a local crossbar office. In comparison, a toll office lacks any attached subscribers, has no need for dial or ringing tones, no ringing power and other simplifications.  Toll switching is a trunk-to- trunk router. However, selecting the ideal outbound trunk, maybe out of several thousand, requires complex operations discussed below. The crossbar switch is fundamental to toll switching, learn more about the switch basics here

Fig 5 provides an overview of subscriber (A) in Oakland dial calling her grandma (B) in Cleveland and then secondly a person (C) in Canton, Ohio. Let’s look at the first A-B case.


  • The called number is 216-622-1234 in Cleveland. The local office forwards all 10 digits to a toll office (possibly a Class 2 in Sacramento based on Fig 2) that it knows can reach area code 216 “somehow”.  This first toll office cannot reach Cleveland directly over a trunk so decides to forward all 10 digits to an intermediate office (maybe Class 1, 2, 3) in Chicago. 

  • This office has a direct trunk to a Cleveland toll office in area code 216. So, it “skips” (drops) the three 216 digits and “spills” (forwards) the seven 622-1234 digits only.

  • The Cleveland toll office has direct trunks to the local Main-2 office, office code 622. The toll equipment spills the 622 prefix and forwards only the 1234 digits, the called subscriber ID. The phone rings and all is well.   

  • Little did the caller know the routing gymnastics and thousands of relay operations needed to reach grandma. The second call from sub A to C uses similar logic. The forwarded digits may also include other code digits, not shown here, to assist in call control at the next office. 

Fig 5, Calling number processing operations for a toll call from Oakland to Cleveland. [Fundamentals, Fig 11-9], (redrawn)

The switching plan for nationwide dialing restricted the maximum number of links for transmission stability, uniformity, and efficiency. With manual toll switching, the maximum was four toll switching centers in series, but the loss could be as high as 20 decibels. With direct dial, as many as nine toll switching centers could be in tandem. This requires intelligent routing for coast-to-coast calls especially when one or both subscribers are located in distant small towns [Duffy], [Fundamental].


There were many standards set in place by Bell System engineers to ensure end-to-end call quality. The key "talking path" drivers are attenuation, frequency response, noise, crosstalk, distortion and echo. Each of these has mitigation strategies. See Appendix A for more on call quality including audio examples of the induced aberrations.  

Toll Office Operations

The addition of a toll office at Gotham Tandem in New York City in 1941 permitted operators in New York State and northern New Jersey as well as distant operators to dial or key pulse directly into the tandem equipment. This allowed for completion of 7-digit calls to approximately 350 connected central offices (by 1947 due to WWII war delays) [Adam]. This was before 10- digit Direct Distance Dialing was introduced in 1951.The value of even one toll office in the NYC area had a huge impact on efficient call routing between local offices.

Next, let’s look at the workings of a toll office. Fig 1 correlates the exchange types that can implement a given toll class. The Crossbar Tandem (XBT) toll exchange works across classes 1-4. Using the XBT as a pattern for all toll offices provides good insights into overall operations. Fig 6 is a bare-bones diagram of the XBT office. 


Receiving and then translating the digits into routing instructions is the most challenging aspect of toll routing. See Endnote 1 for more on this.  

Fig 6, Crossbar tandem switching configuration [Fundamentals, Fig 7-20]

Route Control 

The Trunk Link and Office Link Frames (Fig 6) perform the switching between the input and output trunks.  The Frames are crossbar switch fabrics. The in-to-out path is locked in place by common equipment invoked for a brief duration. How does this occur?  

Senders and markers are the major common control circuits. They are attached, temporarily as needed, by their respective “connectors”. The sender's function is to register the 10 digits of the called (incoming) number, and transmit the top 6 of these to the marker.  

After the marker receives the 6 digits from the sender, it parses this into two 3-digit groups. The first is the area code and the marker records this by engaging one of its “area relays.” The operation of the area relay causes an associated Foreign Area Translator to be called and it returns the outbound trunk ID based on the 3-digit office code. The marker selects this trunk (if busy, uses an alternative) and closes the crossbar links to connect the input trunk to the output trunk. 

The marker passes the digit out-pulsing instructions to the sender and it performs the out-pulsing (or uses multi-frequency tones) of digits directed to the next office over the outbound trunk. After out-pulsing is completed, the t
alking path is cut through. Note: the order of operations is simplified here and actual operations are more involved.  


Over the years from 1941 until the 1990’s many upgrades were made to toll offices to keep pace with the ever-growing call rates, billing and maintenance needs. Long distance calling put pressure on telephone companies to maintain acceptable received speech quality over ever longer distances. See the article on transmission impairments and the seven deadly sins of speech quality over toll networks. 

See the section on this site, The Crossbar Exchange Family Tree for more insights into toll, local and PBX crossbar exchanges, with pictures. 

Importantly, for nearly 50 years crossbar-based toll domestic offices switched countless billions of calls. In 1954, ~10.4 million toll calls were made per day with 50% dialed without operator intervention. This required ~70,000 dial-type domestic toll trunks [Cozine]. Naturally, amounts would grow until there were ~50 million toll calls per day in 1980 [Engineering, pg 445].  See Fig 3 for the state of some toll metrics in 1983. 

The Western Electric #4ESS started routing toll telephone calls in Chicago in early 1976. Slowly, this electronic system and others, began replacing electromechanical systems due to ESS’s many efficiencies. 

Fig 7, First number 4 Crossbar on cutover day [Myers]

Caption: Taken at the time of the #4 Crossbar toll cutover in Philadelphia on August 22, 1943, this photograph shows, kneeling, J. E. Murdoch (left), Chief Engineer, Eastern Area, Bell of Pa., and A. B. Clark, Director of Systems Development, Bell Laboratories; standing, left to right, F. J. Chesterman, Vice-President, Operations, Bell of Pa.; C. H. McCandless, Switching Development, Bell Laboratories; and C. R. Freehafer, Vice-President and General Manager, Eastern Area, Bell of Pa.

Endnote 1: The nationwide numbering and routing plans required the ability of toll offices to receive 10 digits and interpret the three-digit area code and three office code digits.  In general, this real-time interpretation requires the simultaneous "translation" of six digits into up to ~1,000,000 trunk selection and forwarding decisions. 

During the 1940’s relay-based translation (7 digits, no area code) met most toll routing needs. However, a new apparatus in the art of translation was required for 10-digit nationwide direct dialing.  Bell Laboratories engineers invented an ingenious electromechanical device, the card translator. Its inner workings are not covered here but it was an add on to Fig 6 [Hampton], in principle. It was only used in 4A tandem offices and required a large cast of supporting equipment. 

The card translator is the "seeing eye" of the 4A common control equipment. It translates the code digits registered in the incoming sender into information used by the marker and sender to switch calls. The card translator replaced the limited functionality of route and area relays in the marker and FAT translators in the XBT system. Image below from [Joel]. 

card translator for 4A tandem office_edi

Endnote 2: All Local exchanges, of any switching type, needed to interconnect to Bell and non-Bell systems (Step-by-Step switching mostly).  GTE Corporation, formerly General Telephone & Electronics Corporation, was the largest independent telephone company in the United States during the days of the Bell System from 1926 until 2000. In 1955 there were 311 telephone companies operating 2,398,521 telephones in Texas alone. So, this required cooperation between all competitors for local-to-local and long distance calling. 

                          Appendix A

                    Mechanical Amplifiers and 2-Way Repeaters


Thomas Edison had his fingers in many pies. During his 84 years, he was granted 1,084 US patents. It wasn’t until 2015 when Lowell Wood surpassed Edison with 1,085 US patents. Edison had 23 patents with telephone in the title and five with the words telephone repeater in the title.  In 1879, Bell purchased some of Edison’s telephone patents (via Western Union), and this helped consolidate Bell’s power.  


Of interest here is Edison’s work on signal repeaters (amplifiers) for telephone transmission. As mentioned in the main body of this article, the first repeaters were mechanical in nature. In 1875, Edison was granted US patent 158,787, Improvement in Telegraph Apparatus. The focus was on repeating telegraph signals. However, the concepts would soon be applied to analog telephone circuits. He called his invention the “electric motograph.”


The idea is best illustrated using a different Edison patent, US221,957, Improvement in Telephones, granted in 1879. Its central element was the electric motograph. The motograph is a mechanical amplifier requiring an electrolyte impregnated, hand cranked, rotating chalk cylinder. The triode vacuum tube was almost 30 years distant. 

Mechanical amps and 2-way repeaters

Fig A1, Edison patent figures illustrating the electric motograph “amplifier”

Fig A1 is from the patent and shows a telephone (microphone, receiver, bell, motograph in red box). The concept is to amplify the received signal for improved listening.  The focus here is on the red box and its rear view (right side) not all the workings of the telephone.  The description to follow is derived from the patent. Like most patents, a little patience is needed to follow the (very clever) reasoning. Although not shown, there is a battery inside the telephone set.


This detailed explanation shows Edison’s creativity to devise a smart, albeit complex, way to boost audio signals in 1879. The prototypes worked well. In one case, distant speech and songs were clearly heard by over five thousand people [Dyer]. Edison's efforts to amplify electricity were ground breaking.


A Rube Goldberg Amplifier


In the figure, a’ is a cylinder of compressed chalk soaked in an electrolytic solution, such as a caustic alkali a conductor of electricity, mixed with “a salt of mercury” (!). This cylinder is secured to a shaft, b', and the whole is rotated by the operator (hand crank f) by means of the toothed wheels c' and d’.  


Resting upon the cylinder is a flat spring g connected to the (earpiece) diaphragm, d2. This is the essential element of the receiver earpiece. T is the carbon transmitter. This spring is pressed upon the chalk with a pressure of several pounds by means of the wire k and spring h. The telephone line-wire is connected to the spring, g, while the telephone earth-wire, ground, is connected to the shaft.


The incoming speech signal level modulates the friction (by electrolyte reaction) between spring g and the chalk cylinder. The friction is directly related to the signal strength. So, arm g vibrates due to this changing friction and the listener hears the variations via the attached diaphragm, d2. This is the key to the device’s ability to amplify.


In addition, there is a vessel with water, n’, and a sponge, r, operated by lever r’, that should be used (weekly) to wet the chalk cylinder, a’, to keep the electrolyte active. The subscriber had to turn the crank (the power source) to hear anything.  Rube Goldberg, step aside.


These were baby steps in mechanical amplifier development. Edison’s electric motograph was sidelined due to its complexity, need for cylinder rotation and frequent maintenance. Still, it pointed a way forward for others who would invent more practical mechanical means to repeat signals.


The Mechanical Repeater


Edison saw the need for a speech repeater, especially for calls outside the local area. In 1886, he was granted US patent 340,707, Telephone Repeater. This device was not part of a telephone but a standalone device for telephone offices and “repeater shacks.” He leveraged the electric motograph as the amplifying element.


The invention is fiendishly clever and used only one device to amplify both sides of the conversation without feedback and howling (in principle). Fig A2 shows a related version (circa 1912) that is quite similar to the Edison method but more intuitive for understanding.  


So how does it work? The amp output current was divided equally between the East and West sides. If the signals were alike, they would cancel, and nothing from that side would be returned to the amplifier input. So, no feedback! It required a so-called hybrid (output) transformer pictured in Fig A2. To achieve perfect cancellation required perfect East/West "line-characteristics balancing" and this was nearly impossible for the changing real-world conditions.

Fig A2, Block Diagram of Type 21 repeater [Fagen]

Fig A3 shows Edison’s version, from his patent.

                    Fig A3, From Edison’s patent 340,707, telephone repeater

Looking at the pieces, coils C and D form the equivalent (not identical) of the hybrid transformer in Fig A2. Note the adjustable resistors R and R’. These were tuned to balance each A and B line to improve the delicate balance and reduce any “singing” from amplifier feedback. Edison likely knew little about the complex inductive (L) and capacitance (C) aspects of transmission lines so using only a resistor to equalize for perfect speech cancellation was far from ideal. Future solutions would use complex R/L/C networks to achieve the needed cancellation.


How did it amplify? E was the receiver portion (input) using an electric motograph (amplifier) attached to a diaphragm. F was the microphone portion (output) that “listened” to the vibrations of the diaphragm. So, this combo functioned as a standalone audio amplifier. But be quiet, its receiver could pick up ambient noise. The motograph was turned using a dedicated motor not shown. It's likely, ~20 years later, that H.E. Shreeve (Bell Labs) was influenced by Edison's combination of a receiver/transmitter as a "speaker-listener" device when he invented the 1A/3A mechanical amplifiers.   

As an aside, A.G. Bell’s first microphone used the variable resistance of the meniscus of enclosed water.  The granular carbon microphone was concurrently and independently developed by David Edward Hughes in England, Emile Berliner and Thomas Edison in the US. Edison was the first to patent the idea (222,390) in 1879 with working models demonstrated a few years before by all three inventors. Bell purchased this patent, and all future telephones used the carbon concept until circa 1980. So, element F in the patent was a carbon microphone. 


Final words

Dr. George Campbell (Bell Labs) was likely influenced by Edison’s single amplifier repeater ideas (Type-21). He knew of Oliver Heavyside’s work on transmission line theory and H.E. Shreeve’s work on practical mechanical amplifiers. So, Campbell leveraged the work of these giants to improve repeater technology for long-distance lines. Campbell’s invention of the 2-wire Type-22 repeater stand as a testament to the combined ingenuity of the many innovators before him that helped contribute to this magnificent invention.



Adam, A. O., Crossbar Tandem as a Long-Distance Switching System, The Bell System Technical Journal, Vol. 35, January 1956

Brady, P. T., Effects of Transmission Delay on Conversational Behavior on Echo-Free Telephone Circuits, Bell System Technical Journal , January 1971


Chapman, A.G., Open-Wire Crosstalk, Bell System Technical Journal, April-June, 1934


Clark, A.B., H.S. Osborne, Long Distance Telephone Circuits in Cable,  Bell System Technical Journal, Oct, 1932.

Cozine, J.J., Two–Train Switching in Toll Crossbar Offices, Bell Laboratories Record, October 1955

Duffy, F,P. et al, Echo Performance of Toll Telephone Connections in the United States, Bell Laboratories Record, Feb 1975.

Dyer, F.L., Edison his Life and Inventions, Volume II, 1910.


Engineering and Operations in the Bell System, AT&T Bell Laboratories, Ed 2, 1982


Fagen, M.D., A History of Engineering and Science in the Bell System (1875-1925), Early Years, 1975


Fundamentals of Telephone Communications Systems, Revised Edition,1969, Western Electric Co.

Hampton, L. N. and Newsom, J. B., The Card Translator for Nationwide Dialing, Bell System Technical Journal, September 1953, p. 1037

Joel, A. E., Jr, A History of Engineering and Science in the Bell System (1925-1975), Switching Technology, 1982

Korn, F. A.,  A Crossbar Tandem Office, Bell Laboratories Record, Aug 1942

Myers, O., Markers for the Crossbar Toll System, Bell Laboratories Record, Aug, 1944

Survey of Telephone Switching, compiled by staff of Pacific Telephone Co, 1956, Chapter 8, 4A toll switching.

Swenson, P.W,, Contacts, Bell Laboratories Record, Feb 1949

Toll: Technical Developments Underlying the Toll Services of the Bell System, Bell System Technical Journal, Oct-Dec, 1936.

bottom of page