These explanations, have explained a number of puzzles I've had over a long period of time, as to why certain things couldn't/didn't exist in practice. I've sometimes seen these amazing looking transistors (a while ago, many seemed to have 4 leads coming out at 90 degrees between the leads). These days, especially (ignoring component shortages), mouthwatering specifications, such as 35GHz, yet not too expensive, perhaps $0.40 to $1.25 or so, depending on the exact part, and the quantity, bought.
So, why the heck can't we go and make cheap/small, discrete part, 2GHz NE555 timers, comparators, and so forth. But as I can see from your post, and the other explanations. Some circuits, can practicably made, such as microwave amplifiers, radar detectors (as used in cheap person detectors, e.g. burglar alarms), and certain other microwave things. But, the other things I listed, (presumably/intuitively now I understand it better), can't realistically be made at such high frequencies.
So, it is like going into an electronics lab, and them saying diagnose this very high frequency circuit for us. Feel free to use the 40GHz Oscilloscope on the bench, but we can only let you use this old/battered 50MHz probe with it, sorry.
Keep in mind, what's possible is far broader than what's available. What's available, is possible, AND has been developed into a product, AND is a commercially successful product (at least, more or less, depending on how much investment or loss-leading the manufacturer might be using with that product/line..). If you want just the first one, pick up some time at a research lab, and bright a hearty budget; the second, move it out of the research lab into a commercial development lab; only the last one is available from suppliers, and at that, not necessarily at any kind of affordable cost still. There's a lot of hurdles to pass before something winds up on Digi-Key!
So, 2GHz 555s are perfectly feasible, I'm pretty sure -- but no one wants them. You can get close right now, with an equivalent made from ECL or what have you -- check out what kinds of timers, one-shots, etc. are available. I don't know much about ECL families; I'm guessing there are some?
I know, it was just an example of course. But, that said, it can be interesting to do a design experiment like that -- follow through and see how close you can actually get, off-the-shelf, and what the performance specs would be. (For my part, I've designed a current-limiting switch using discrete transistors, with about 3 times lower power consumption than any building-block solution -- trouble is, the sheer parts count means it costs more. It's not that such a function is impossible in IC, it would trivially do 10x better than my version; there just doesn't happen to be a commercial offering of it!)
And there's lots of reasons besides profitability:
- Such a fast timer is a liability, not an asset, in most applications -- you need to manage the sub-100ps risetime, somehow not letting it radiate, while also hopefully getting all your timing components close enough to the poor chip that stray inductances (transmission line lengths, really) don't completely smash the response, making it some mixture of LC oscillator rather than just the RC that was intended.
- The operating voltage must be quite low, or the power dissipation will be extraordinary (I'd guess most such circuits run at 2.5V or below, if CMOS, and ECL a bit higher). If you wanted say 5V of output swing, or even 15V, at 2GHz -- you'll be paying for it with a whole chain of power amplifiers!
- And again, it's about transmission lines -- their use is an unavoidable necessity. You can't drive a net and wait for it to settle, like you can with CMOS or TTL. Practical nets are some cm long, while the 1/4 wave is some mm (depending on whether you're counting just the edge (harmonics), or the whole waveform as such). It's always waves in flight. So you can't make an assumption like, "this pin will draw some current during the edge, but it'll eventually settle down, so we don't need to waste much space on power dissipation for these output transistors" -- instead, those waves need to dump into a (termination) resistor eventually, lest they come back around and dirty up the signal.
- And who needs an RC timer? Most applications up there are radio-related, so, very precise frequency control is required. (Not all of them; spread spectrum might be a virtue for CPU clocking for example!) You're only going to do that with a quartz reference and PLL multiplier.
But outside board-level components -- there are applications on-chip. Most PLLs, AFAIK, are some form of this. Greatly simplified, since they don't need the versatile trigger/threshold function, but in the general sense that it's an RC circuit, yeah. Typical example is a ring oscillator, where the resistance is some combination of device transconductance and Rds(on), and capacitance is node / body C. By varying the bias current into the amp stages / inverters, the frequency is controlled. Hook that to an error amp and phase detector, and you've got yourself a PLL.
So, in a sense, you're not wrong, about thinking about something like that -- it's just not something you'll see in catalogs, but very much out of sight, tossed in a boring corner of many ICs, just another piece of invisible fabric supporting a vastly more complex whole.
Regarding clock speeds in general -- fT can be seen as an upper limit for edge rates or propagation delays; only the simplest circuits can run at some fraction of fT, or indeed above it (usually using distributed amplifiers). CPUs have to run a small fraction of this rate, due to the complexity of the gate chains, and bus lengths, used to implement them. Only very simple CPUs could run that fast (sub-100GHz say), and they'd be utterly useless for any kind of real work (an 8051 running this fast, is absolutely no competition against a modern 64-bit machine at 4GHz).
Kind of a small example of this, and heh, a bit less useless, actually a practical example. Compare PIC to AVR for example: PIC is an ancient 8-bit (give or take extensions) architecture, based on a 4-clock (mostly?) instruction cycle. AVR is (mostly) single-cycle. A lot of PICs run up to 64MHz, for 16MIPS performance; a lot of AVRs run up to 20MHz, for ~20MIPS performance. Both are 8-bit, both have various peripherals, memory (and give or take penalties for accessing certain memory spaces, or competing with DMA bus access, etc.), and, they're fairly close, considering. Mind, these are 3.3-5V fabbed parts.
Or compare further to non-pipelined ARMs (Cortex M0 something or other). Typically topping out at 30-60MHz, 3.3V operation. A bit finer fab scale, but much wider buses, far more powerful as a result. (Adding in pipelines and caches, operation blasts up to 200MHz or so; although some/most of these may be internal low-voltage core designs, I'm not sure.)
Or taking things in yet another direction -- there was a fairly reasonable op-amp I was looking at some years ago, I forget what part, something Analog Devices. Somewhat over 20MHz GBW, precision inputs, RRIO, 16V supply limit or something like that, I think -- competitive price too. The hook is this: supply current is just a bit lower than everyone else's, oh and by the way they're using their proprietary SiGe process to do it (XFCB I think?). Bwha!? Yes indeed, using seemingly RF-geared tech just to get a higher figure of merit (namely GBW/Iq: in most things, opamps, comparators, etc., you pay for bandwidth with bias current) than its conventional relatives. And unlike CMOS which has to run at lower voltages for greater speed (there are indeed many fantastic <=5V or <=3.3V amps, comparators, etc. out there in CMOS), HBTs (SiGe) are perfectly comfortable mixed with conventional BiCMOS designs.
So they can sneak in that high technology, right under your nose, in other ways too.
So on the one hand, I'd be excited by the 40GHz Oscilloscope, but the 50MHz scope probe, with its (highly likely), very significant probe capacitance, would mean that I would have a job, using the scope beyond, 20MHz or so (ignoring trick question cheats).
So these 35GHz (or whatever they are these days) transistors are so tempting, yet would probably struggle to even reach the speeds of a normal (relatively slow) bjt, in most normal circuits. Apart from where those circuits which do allow extreme performance from them are used, along with the appropriate construction techniques, to go with it.
Yeah -- the probe itself wouldn't be necessarily a problem for sheer frequency response, but flatness along with it, good luck. That's why e.g. 500MHz probes have smaller ground leads and tips, and pretty much anything above that you need either a special probe, or just route the damn signal into a coax connector and view it directly.
And, again about feasibility: 40GHz scopes do indeed exist; they're very complex to make, and though there isn't much demand for them -- that's paid for by the serious need customers have for such equipment. And yeah, you're definitely using the right hardware to hook up one of those beasts, probes are irrelevant up there.
Also, oh good, mawyatt's joined the thread. Definitely take his real IC experience over my mere suppositions of the market.
Tim