Talk:Register
|
|
Moved this out of the article, it needs to be looked at.
One of the most significant developments of computer architecture was the discovery that binary numbers were the least expensive form for electronic storage.
Two different approaches to registers have been pursued.
The first was to have an electronically-controlled switch for each possible state of each digit of the register. Any one switch, when on, turns off all the other switches. This system is stable, and is standard engineering practice.
The other method is to map several states to the charge in one capacitor. This approach can sometimes be used to save money. For example, most digital cameras use a charge coupled device as their sensor.
Well, I explained the logic, and put the text back in. It's the truth, really. Ray Van De Walker
- It looks better now, thanks for rewording it. -- mike dill
I think the stuff about mapping several states to the charge in a capacitor is really more relevant for an article on memory technologies, myself. AFAIK (and I'm not a hardware expert, admittedly), registers have been implemented using switch logic for at least the microprocessor era and probably longer. --Robert Merkel
This is not my experience. There's a dead give-away in the spec sheet. If the clock tolerance goes to zero Hz, then the unit has no data storage in capacitors. It seems like complex CPUs often have a minimum speed. User:Ray Van De Walker
Capacitors for registers? Sounds fishy to me too. Capacitors were/are used in DRAM (main memory, the kind you buy in chunks of 64 or 128 megabytes), but when I think registers, I think SRAM, i.e. switches.
Also, some oooold machines did have decimal registers. Your math sounds spotty though. Yes, with ten switches I can do a single decimal digit, 0-9. I can also do it with four switches (in binary, and just overflow after 1001). I'm not certain which method was actually used, however: In binary, with ten switches I can count from 0 to 1023, with four, from 0 to 15. So using binary was either a hundred-fold increase in efficency, or a 50% increase in efficency. 100-fold increases are not common in engineering, so I'm leaning towards the four-switch-as-original-digit method. Regardless, the whole thing, as I said, sounds very suspicious. I left it in since I'm not certain, but I'm going to go find out, because this sounds wrong.
-Mudlock
I was about to summarily execute the paragraph about binary being least expensive, but I'm going to be nice and give it a chance to be saved. There are two problems. First, a register is a blob of storage; binary, BCD, etc is irrelevant. Second, the ternary logic freaks will be all over this - they can show that 3-state representation is more efficient, and the circuitry isn't really any harder either. So it's not that binary is more efficient, it's that like QWERTY for keyboards, it's hard to change now.
Easiest thing is to lose the paragraph - better would be to move it to a discussion of binary representation. Stan Shebs 08:13 Feb 16, 2003 (UTC)