Tektronix pushed the limits of electronics and devices to deliver leading-edge instruments. The 7612 digitizer introduced in 1980 needed a high speed ADC (analog to digital converter) to achieve its bandwidth of 80 MHz with a sample rate of 200 Msamples. No semiconductor devices could achieve this performance so Tektronix designed their own utilizing the T7610, a hybrid CRT-semiconductor converter tube. More information on the construction is on our T7610 Electron Beam Analog to Digital Converter page.
The T7610 semiconductor target was arranged in 8 columns and 256 rows, forming an 8-bit Gray code output in the columns depending on which of the 256 rows was hit by the ribbon beam. Here is an example of a 4-column, 16 row Gray code:
There is only one transition from 0 to 1 in the most significant column, so as the ribbon beam is swept up and down the target by the input signal, the most significant Gray code bit will change at the same frequency as the input signal. But there are two transitions in the next most significant bit, so it will have to toggle at twice the frequency of the input. Likewise, the third most significant bit has 4 transitions and will toggle at 4 times the input frequency, the fourth most significant bit will toggle at 8 times the input frequency, etc. For a full 8-bit Gray code and an 80 MHz input frequency, the least significant bit will toggle at 128 * 80 MHz, or over 10 GHz!
While the electron beam bombarding the target could actually generate current at that frequency, the readout electronics of the day were nowhere near up to the task of following a 10 GHz waveform. With a fast input, the least significant bits would read out either as random (due to noise in the readout electronics) or stuck at a 1 or 0 (due to offset in the readout electronics).
At low input frequencies, the T7610 produced a clean 8-bit output code. But as the input frequency increased, the lesser significant bits would “drop out” in turn as their toggle frequency exceeded the readout electronics bandwidth, no longer representing the instantaneous input signal level. So the T7610 was clearly not still an 8-bit converter at high input frequency. The drop-out did not occur at a precise frequency, however. Rather, each bit in turn became gradually noisier or more frequently stuck in one state as the input frequency rose.
As the story goes, Lyle Ochs was looking for some way to accurately describe this loss in converter resolution at higher input frequencies, and hit upon the idea of mathematically fitting (by least-mean-square-error) a sine function to the captured record of an input sine wave. He defined the “effective bits” (sometimes called Effective Number of Bits or ENOB) of the converter at any given input frequency as the number of bits an ideal quantizer would have to have to create the same mean-square-error (from quantizing) as the real converter. Even before it dropped out entirely, as a bit got noisier it contributed more to the mean-square-error of the acquired record, and in turn became less “effective” at determining the instantaneous signal level.
This generic method of rating the loss in dynamic accuracy of an ADC at higher input frequencies caught on in the high-speed digitizer market, and was adopted as an industry standard by the IEEE TC-10 committee in Standard 1057 (first published for trial use in 1989). It is still in common use today as one of the banner specifications of a high-speed ADC, though few know the history of its origin in describing the rather unique Tektronix 7612.