0
\$\begingroup\$

I am trying to make a 30 GHz 32 (or 64 bit) counter on flip flops.

Ready counters do not satisfy me due to max frequency and price, so I decided to make it on my own from flip-flops.

Ready flip-flops also do not satisfy for the same reason, so I decided to make flip-flops from ready NAND elements.

NAND elements also are pretty slow, so I decided to own NAND elements from transistors, diodes and resitiors.

There are resistors with satisfying operating frequency, for example, this one with 110 GHz.

30 GHz has about 33 ps period, however the fastest diodes I found have only 1 ns. I typed “switching”, “fast”, “ultrafast diodes”, but the minimum one is 1 ns.

I probably can use transistor as diode, but maybe there are, perhaps even more faster diodes, and I just searching wrong?

\$\endgroup\$
12
  • \$\begingroup\$ PIN diodes maybe? \$\endgroup\$
    – Andy aka
    Jul 22 at 20:11
  • \$\begingroup\$ @Andyaka thanks, I will search for it \$\endgroup\$
    – Stdugnd4ikbd
    Jul 22 at 20:13
  • 1
    \$\begingroup\$ For reference - in 33 ps light travels in vacuum by 1 cm. I think in metal it can drop to 5 mm. I would imagine you would need to fit the circuit into 1 mm region even with ideal diodes/transistors (for real life ones you need to take into account capacitance of them). \$\endgroup\$
    – Maciej Piechotka
    Jul 22 at 20:31
  • 3
    \$\begingroup\$ I think you might be underestimating the sheer complexity of what you want to do. \$\endgroup\$
    – Hearth
    Jul 22 at 21:43
  • 1
    \$\begingroup\$ Incidentally, your 110 GHz transistor is only going to be 110 GHz if you keep it from saturating. This part is meant to be used in linear mode as an RF amplifier, not for digital stuff. \$\endgroup\$
    – Hearth
    Jul 22 at 23:37

2 Answers 2

Reset to default
2
\$\begingroup\$

There are no board-level components usable for such a function, in this frequency range. You are ignoring stray capacitance and inductance, or real delay for that matter since you're talking frequencies where multiple wavelengths may extend between components. When you consider these limitations, you will see this is not a viable approach.

The only meaningful build-your-own route is to get fab space on a wide bandgap semiconductor prototype or university teaching fab process. This might be just possible on SiGe, else InP or GaN something, probably. Needless to say, this will cost exponentially more than ready-made parts.

You can save some component cost by using a locked oscillator, or harmonic mixer, but these methods carry various downsides or compromises, which I'm guessing you had intended to avoid by using a more general-purpose (logic) counter.

\$\endgroup\$
7
  • \$\begingroup\$ Perhaps You know existing counters solutions with such frequencies? \$\endgroup\$
    – Stdugnd4ikbd
    Jul 22 at 20:10
  • 2
    \$\begingroup\$ Afraid I do not, but I would expect to pay dearly for them, relative to more common components anyway. Such frequencies are not trivial to work with. \$\endgroup\$
    – Tim Williams
    Jul 22 at 20:13
  • \$\begingroup\$ Okay, just a bit dumb off topic theory question: do I understand correctly, that in synchronous counter max frequency does not depends on flip flops number? \$\endgroup\$
    – Stdugnd4ikbd
    Jul 22 at 20:15
  • \$\begingroup\$ Generally no, but mind that combinatorial complexity may depend on it. You may wish to ask another question on that. But beware if it isn't well focused, it will probably get closed. \$\endgroup\$
    – Tim Williams
    Jul 22 at 20:21
  • 1
    \$\begingroup\$ To respond to that point, the operating frequency may depend on the number of flip-flops depending on the design; in a simple design, transitions from 01…1 to 10…0 require the state-change to cascade along the word, while more complex designs use multi-input gates to ‘look ahead’ and generate these transitions in a single step and so extending the word length has little effect on operating frequency but at the expense of significantly increased complexity. \$\endgroup\$
    – Frog
    Jul 24 at 19:43
2
\$\begingroup\$

Small (physically and current capability) Schottky diodes are usable at those frequencies (eg. Skyworks), but I think your chances of success are not very high.

You can get a 10-26GHz ÷ 4 for less than $100, off the shelf, from AD.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks, I will search for it, although, as I remember, I already searched through Shottky diodes \$\endgroup\$
    – Stdugnd4ikbd
    Jul 22 at 20:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.