In other words, where did this value came from? Why 20 Mhz and not 10 Mhz?
Some oscilloscopes did use a 5 or 10 MHz bandwidth limit function instead of 20 MHz.
In some cases this was simply because their maximum bandwidth was not much higher than 20 MHz so a selectable 20 MHz bandwidth limit did not make much sense and a lower one was included. The 60 MHz Tektronix 22xx models with their 10 MHz bandwidth limit are an example of this.
Another reason a bandwidth limit lower than 20 MHz would be used is to support a sensitivity higher than the typical 2 or 5 mV/div on an otherwise higher bandwidth oscilloscope. The 50 MHz Tektronix 2225 series combine their 5 MHz bandwidth limit with their x10 vertical magnification to achieve 500uV/div. The 7000 series 80 MHz 7A15A does the same thing. The 7000 series 100 MHz 7A13 vertical amplifier has a 5 MHz bandwidth limit to make maximum use of its 1mV/div and high common mode 10mV/div sensitivity. (1)
Later oscilloscopes with much higher bandwidths seem to have settled on supporting both 20 and 100 MHz bandwidth limits which I think points to the reason Dave cited. 20 MHz become the lowest common standard and the informal standard for power supply noise measurements. Later 100 MHz did as well although to a lessor extent simply because 100 MHz oscilloscopes are so common.
We received a nice Tek 7000 series CRO which had one vacant plugin position without a cover.
(And No!,the supplied plugins didn't,have a 20 MHz limit switch)
Some of the 7000 series vertical plug-ins had a switchable bandwidth limit. I prefer the 200 MHz 7A26 (and single channel 7A16/7A16A) over the 75 MHz 7A18 even in a slower mainframe simply because of their 20 MHz bandwidth limit and not because of their higher bandwidth.
Compared to the older instrument, the trace was a lot thicker,& we grizzled about it.
A cover was obtained, & the noise level was reduced to an acceptable level.
Depending on what oscilloscope you replaced, this may not have been do to higher bandwidth. Like most new oscilloscopes at that time, the 7000 series used scan expansion trading better deflection sensitivity for worse spot size. I have heard a story about one TV studio that replaced their out of production 50 MHz 547s with 7000 series mainframes and then switched back because of this.
A real-time scope with 1 GHz bandwidth is REALLY hard to do. You need a power amplifier that can deliver GHz signals at several hundred Volts amplitude. You also need segmented deflection plates with delay lines between segments, so the signal follows the electron beam as it progresses toward the screen. The Tek 7104 had that capability.
It is 10s of volts for the 1 GHz 7104. The 7104 CRT itself has about 3 times the deflection sensitivity as the 500 MHz 7904 CRT so a lower voltage higher frequency process could be used for the vertical output amplifier.
(1) The 7A13 is about 5 times noisier than a typical 100 MHz input simply because of the design necessary to support differential inputs with a wide common mode range although it is still less noisy than some modern DSOs. Despite this it supports a higher input sensitivity then its contemporary 100 MHz amplifiers making a bandwidth limit function unusually useful.