|
High-speed RTL? Errrr... XD
I made these RTL gates to investigate various details about how they worked. While adjusting resistor values I saw a correlation between current draw and propagation delay, so I wondered, "Can I make high-speed RTL?" I tried. I think I could, but every transistor would need to be a high-current transistor with a heatsink! :)
With a 300ns delay, these gates are only half as fast as the default EveryCircuit gates, but draw nearly 5mA to output a logic zero. Can you imagine a computer built from these? The maximum input frequency these 4 inverters can handle is about 200kHz. (They might take 250kHz, but I wouldn't risk it.) The maximum frequency a counter can handle, and thus the maximum clock frequency of a CPU, is much lower than what each gate can. If I remember right it's about 1/4, yielding a 50kHz CPU. Oh joy!!!
What about this hypothetical CPU's power draw? Well, I'm guessing a bit, but I think I could make each 'bit' of a full adder in 7 basic gates. An 8-bit adder with carry input could draw as much as 280mA. I guess we're looking at over 1A, more than the capacity of a 7805, for the CPU alone. At only 50kHz, it's not good. :)
Going back to speed, I think the adder would be sufficient with ripple carry (the simplest type, and thus in RTL the lowest-power). It looks like the carry would take almost 5μs to propagate through the adder, while a clock tick is 20μs.
I've got a feeling I've missed something. Maybe I should build more RTL circuits. There's no chance of exceeding EveryCircuit's simulation speed limits! :)
To think they used this sort of stuff in the Apollo guidance computer...
|
