The Crossbar Switch
Highlights
Switch category: 2-D, XY cross point
​
Inventors: Mostly, G.A. Betulander and Nils Palmgren (Sweden); J. N. Reynolds and C. L. Goodrum (Bell Laboratories)
Important dates: Formation of Nay Autotelefon Betulander company in Sweden, 1910. Reynolds US patent 1,139,722 in 1915 for the coordinate switch. Ericsson company involvement starting 1919. Bell restarts crossbar effort with Ericsson’s help in 1930 leading to #1 crossbar exchange in 1938.
Legacy: In 1975 there were 6,549 crossbar offices in North America. Crossbar switch usage spread worldwide by L M Ericsson and others and it became the metro exchange switch of choice for 45+ years.
Background
​
Electromechanical switch design is an art and a science. From an early date engineers desired a scalable switch with (1) minimum of movements, (2) minimization of selecting electromagnets, and (3) a large number of relay-like contacts at the switch contact points. The motor driven panel switch design did not advance these goals. So, designers started looking in new directions.
​
The coordinate or crossbar switch design went a long way to meeting these goals. Don’t confuse the crossbar switch with its namesake, the crossbar exchange. Next, let’s briefly see how a simple crossbar works.
​
Crossbar basics
​
A crossbar switch is a type of 2-dimensional, X-Y, switch. It includes the worthy feature of enabling A x B switch points with A + B electromagnets. It has electrical contacts (cross points) at every intersection of its horizontal and vertical bars. The cross-point contacts are very similar to the contacts of a relay. Generally, small crossbar switches (100, 200 points) needed less maintenance than panel switches and some older Strowger-type switches.
​
Below is a 9-point switch with 3-row and 3-column electromagnets. Here, up to three independent connections can be made at once (ex: 1, 5, and 9). More on this limitation later.
​
One popular operational mode of the switch is as follows. To close switch point 5, energize H2 first then V2. Then release H2. V2 holds the cross point closed for as long as needed. While point 5 is closed, the Y2 input (talking path) is fed to the X2 output.
Fig1, Crossbar basics diagram
All crossbars work in a similar fashion. Now that the basic idea is covered, some historical development follows.
A long winding path
In 1910, along with colleague Nils Palmgren, G.A. Betulander started the Swedish company Nay Autotelefon Betulander. The company specialized in products for automatic exchanges. By 1919 they had developed a crossbar switch design that would become the master prototype of many versions to follow. See Fig 2.
This switch must have intrigued the L M Ericsson company (Sweden) because they purchased the small company the same year. Betulander returned to the Telegraph Agency, his previous employer, and continued to develop a crossbar switch version for them.
​
L M Ericsson further enhanced this design and manufactured it for many years. See Fig 3 [Xbar] with eight switches highlighted. By 1971 L M Ericsson had produced one million switches, continuing their dominant position in crossbar development and sales worldwide [Ericsson]. See the article on the Pentaconta switch for insights on a derivative device.
Fig 2, 100-point crossbar switch, L M Ericsson, Sweden, 1921
source: London Science Museum
Fig 3, World's first major crossbar exchange, Sundsvall, Sweden, 1926
Meanwhile, work at Bell Telephone Laboratories was underway to create a very large coordinate switch. The first crossbar patent was issued to J. N. Reynolds of Western Electric in 1915. Some authors believe this patent may have influenced Palmgren and Betulander’s initial work but the two mechanisms were quite different.
​
After some false starts, Reynolds and colleague C. L. Goodrum built several working prototypes during 1921-1925. One model was 8.5’ tall and 4.5’ wide and consisted of 60 vertical rods and 45 to 70 horizontal rods. See Fig 4. It was a behemoth with up to 4,200 available cross points. Interestingly, it had 60 vertical selector rods, the same as a panel selector switch. It seems to have been designed as a stand-in, in some manner, for a panel switch.
Fig 4, Reynolds & Goodrum coordinate switch, ~1923, from [Fagan]
​
According to [Fagan], “The cost studies completed in the following year showed that the system did not produce sufficient savings over the panel or step-by-step systems (Strowger) to warrant its commercial introduction, and work upon the coordinate development was discontinued.”
​
Work restarted in 1930 when W. H. Mathies of Bell visited Sweden to see Ericsson’s crossbar in action at small telephone exchanges. Apparently, he was very impressed. Samples were obtained and after structural improvements for repair and manufacturability were made, a Bell version was born. See Fig 5. Some authors question if Bell bought patent rights to the L M Ericsson design or it was developed in stealth mode.
​
This result looks quite like Fig 2 from 1921 by Ericsson. The Bell switch was first used in the No. 1 Crossbar system (1938) that became the large-city successor to panel.
Examples of modern crossbar switches
​
The crossbar became a workhorse in the Bell system for 45+ years. Fig 5 shows a 10x10 crossbar made in 1970 by Western Electric. The model weighs about 30 pounds (13.6 Kg) and is made mostly of steel. It’s about 21” (53 cm) long and 9” (23 cm) tall. Source: Wikimedia Commons
Fig 5, 1970 version, Bell 10x10 crossbar switch
​
Notice the 10 vertical bars and 5 horizontal bars. Why not 10 horizontal bars? Cleverly, each H-bar can rock (tilt) up or down slightly to assist in cross point selection. Notice that each H-bar is connected to two electromagnets (rock up, rock down). A video to follow will demo how crossbars rock!
​
Figure 6 shows a partial perspective of a 100-point crossbar switch. Note the following points:
-
10 electromagnets for controlling the 5 horizontal selecting bars. Each bar is tilted up/down by using two electromagnets. One to tilt the arm up and one to tilt it down. The corresponding rod’s selecting fingers are moved slightly up or down accordingly.
-
Only two “holding bar” (vertical) electromagnets are shown, numbers 0 and 9, but there are 10 in a fully configured 100pt switch.
-
There are 50 hanging selecting fingers. These assist in closing the desired cross point when activated by a “holding bar.” Each finger can have any of 3 positions; neutral (no op), up slightly (H-magnet 1 operates), slightly down (H-magnet 0 operates). The hanging selecting finger may be the idea of G.A. Betulander since it appears on switches built in 1921. See the Appendix for some Betulander legacy pictures.
Fig 6, Partial Perspective 100pt Crossbar Switch,1937 [Reynolds]
Fig 7, detail of switch point with contacts and selecting finger [Scudder]
Fig 7 shows details of two cross points (also, 2 make contacts for each point) out of 100 in a typical switch. A selecting finger can choose either the upper or lower contacts. The finger can be moved slightly up or down to select one of the two contacts. The vertical holding bars are not shown in this view.
This may be difficult to visualize so check out a short explainer video to see how a crossbar closes a desired set of contacts with the help of the selecting fingers and electromagnets.
​
​
​
​
​
​
​
​
​
​
​
​
Many engineers in telephone labs worldwide attempted to create the "perfect switch". It's a hard problem and some inventions were less than practical. Here are a few examples of amusing switch designs.
​
Exchange notes
​
In a typical Crossbar-based exchange, a single switch is cloned repeatedly to create a larger, aggregated, switching fabric. For a #5 Crossbar exchange, roughly 825*, 200pt, crossbar switches would be installed to support 10K subscriber lines. Fabrics are fascinating -- check out the section on crossbar fabrics.
​
The Bell System had several million crossbar switches installed in NA exchanges in 1975. In addition, more than a million by other manufacturers outside of NA. Worldwide, for at least 40 years, licensed and copy-cat companies built countless versions of the Ericsson and Bell switches.
If there was a poll for its importance in exchange history it would likely rank near the very top, contending for first place alongside the Strowger switch. If you have an opinion on switch rankings, please send me a note. Here is more information on how the switch integrates with its namesake exchange.
​
--------------------------
*This amount comes from adding the number of Line Link Frames plus Trunk Link Frames (20 switches per frame) plus other required miscellaneous switches.
​
References
​
Ericsson Review, Vol 48, #4, page 162, 1971
​
Fagan, M. D., History of Engineering and Science in the Bell System -- The Early Years (1875-1925), Bell Telephone Laboratories, 1975
​
Reynolds, J.N.; The Crossbar Switch, Bell Laboratories Record, July 1937
​
Scudder, F.J. & Reynolds, J.N., Crossbar Dial Telephone Switching System, Bell System Technical Journal, 1939a.0
​
Walker, Peter, Telecommunications Heritage Group, UK, Crossbar Systems talk, Annual General Meeting, 2000, thg.org.uk
​
Xbar exchange picture, Swedish National Museum of Science and Technology.
Appendix
G.A. Betulander Legacy Pictures
(Tekniska Museet Stockholm)
Autotelefon Betulander workshop. Test room for all-relay exchange proposal for Stockholm
Telephone system 1915. Betulander was an early adopter of the all-relay concept. No complex mechanical switches needed.
Telemuseum (Stockholm) exhibit 1962 showing Betulander’s crossbar. With Mrs. Betulander and telephone company executive Klas Weman
GA Betulander at his desk, Telegraph Works (1891-1910, 1920-1936), Stockholm.
​