the Echo Timer
The last month I saw Charles Moore and talked to him about Forth and the F21 chip and how he had advised me long ago to not promote the chip until it was working. I reminded him how I had asked him not to talk about certain things on the chip because there were just almost unbelieveable, like the echo timer. Because this circuit is more than one hundred times faster than circuits that one can design with conservative design tools in this process it is simply unbelievable to many people.
I told him how I had published the basic specs that he gave me on the circuit in the documentation and as a result had had the chip dismissed by one expert at MIT. He was almost ready to believe the other stuff until he got to the description of the echo timer on the chip and said that at that point he knew it was impossible. It was exactly the reason I had asked Chuck not to talk about it when he talked about the chip.
I said that a few people including myself may have guessed how it works and we can think of lots of useful things that we could use it to experiment with. However I worried that Chuck's original stated use for this circuit was to construct a keyboard with one wire and set of switches and I didn't want him getting press for another quirky keyboard and have people miss the fact that this circuit is truely amazing.
I said that I wanted to try to explain how it worked and run my explanation past him for review. He added that, "The speed of light used to seem fast to me, and now it doesn't." I knew what he meant.
I have said before that part of Chuck's style of constructing hardware or software involves getting from point A to point B to solve a problem. He will sometimes encounter an obstacle on the original planned path from A to B. Whenever Chuck encountered an obstacle he didn't let it stop him for long, he found his way around it. Then after he was past it he would look back at it, now from both sides to try to figure out a way to take advantage of it in the future.
I said that I knew that he had to deal with the actual behavior of the signals on the chip in terms of properly sensing intended true and false signal levels and that echos sometimes give false signals so that the amount of drive of level of sense comparison must be adjusted to compensate for signal echos. My guess was that after dealing with this for a while Chuck decided to try to take advantage of it by designing a circuit to time an intentional echo on a line as fast as he could in .8u technology.
Chuck confirmed that this was the case. I remind you that the literature from MOSIS about the .8 micron HP process fab line states this technology is sufficiently fast to contruct a 75Mhz flip-flop. Chuck built a circuit that can generate a signal, and time the amount of time until the echo comes back from the end of the wire. A 75Mhz interval of time makes light seem pretty fast, light will travel about 13 feet in free space in each click of a 75Mhz clock. However Chuck is talking about a timer running at 10 gigahertz in .8u and timing a signal on and off chip at this rate. Most people do not have access to technology with 100ps timers. That gives you measurements to about 1/2 inch at the speed of light.
When the pin is turned on the timer begins counting at 10 gigahertz. The current comes pouring out into the wire and the voltage level stays low as the current flows like water into an empty pipe under low pressure. The longer the wire the longer before the current reaches the switch at the end and splashes up against it. At this point a wave or "echo" heads back toward the source of the current. After the same amount of time the echo returns to the source and the voltage level goes up quickly to the a to detect the echo. In each .1ns interval light can only move about 1/2 inch out and back.
Microcontrollers used in toy robots do this with sound. We can do it with light or radar or current in wires on F21. I am sure that scientists, hobbyists, and kids could figure out imaginitive things to do with this tiny little timing circuit that can make the speed of light seem slow. Chuck wanted to make a keyboard by measuring the distance to each unique key at the speed of light. Other people have come up with some other interesting ideas for what they could do with a 100ps timer. I didn't want Chuck to talk about it because he got so much press for his three key keyboard ten years ago and his original idea for this circuit was a one wire keyboard.
A ten gigahertz real time circuit in a cheap microprocessor is so unheard of that some experts have been willing to dismiss the whole chip as impossible because things like this or the ten gigahertz ramp encoders in the analog I/O subsystem just look like a documentation error to most technical experts. Most people try to compare the specs on our analog I/O coprocessor to a PC sound card or the analog unit in a common garden variety microcontroller. One can do this but they must realize that they are looking at lots more zeros in the numbers because this chip is thousands of times faster and can do many more things than very slow cheap microcontrollers or slow expensive PCs.
Jeff Fox 4/8/99.