Machine Unit

From TheAnalogThing
Revision as of 14:02, 28 August 2021 by Tfischer (talk | contribs)

THAT can compute values in the range from -1 to 1. In most applications, this requires users to *scale* programs to fit into this range. THAT, in turn, represents this unit scale with voltages ranging from -10V to 10V. This voltage interval is called the machine unit.

To understand the need for these scale conversions, consider that users want to model values at any scale while THAT is limited to its modest physical boundaries. Modeling the dynamics of the global human population (currently around 7.9 Billion) should obviously not require 7.9 Billion volts. The population value needs to be scaled. As a convention, analog computer models are scaled to the model unit [-1,+1]. The global population might, for example, be represented as 0.79. In its machine unit, THAT would then represent this value as 7.9V.

Good scaling results in programs that use broad parts of the machine unit ​​without exceeding it. Using only small ranges of the machine unit reduces computational precision - much in the same way as weighing small amounts of cooking ingredients on a scale intended for people would not be very precise. Exceeding the machine unit leads to meaningless results - much in the same way as weighing a person on a kitchen scale would.

Using the -10V to +10V range as the machine unit is an engineering choice and a convention. It happens to be a very good choice because it allows for easy conversions in the decimal number system, it can be handled very precisely using affordable electronic components, and it is perfectly safe for humans.

When any value in a program exceeds the machine unit, the red OL (overload) LED lights up. This is not dangerous for the device. However, in this condition, THAT will likely "clip" any excess voltages, and the values in question will not be computationally correct.

Output signals available via the RCA (or "chinch") jacks on the device's backside are not in the machine unit. Instead, they are attenuated to smaller audio signal levels to conveniently be read into software oscilloscopes via sound card interfaces.