I found some relevant info in TI's app note Understanding and Interpreting Standard-Logic Data Sheets, SZZA036B.

Under the Propagation Delay Time explanation on page 57:

A common misconception about logic devices is that the maximum data-signaling rate (or maximum frequency, as it is commonly misnamed) is equal to the inverse of the propagation delay. The maximum data rate on buffers is dependent on several factors, such as propagation delay matching, input sensitivity, and output edge rates. A device can have a high maximum signaling rate if the propagation delays from low-to-high and high-to-low are matched, the input is fast enough to respond to the fast data rate, and the output edge rate does not interfere with the low and high-level steady states. Clocked devices behave in the same manner, but now the set-up and hold times must be taken into account

I can't find anything else about input sensitivity in this app note. That's probably part what I need. If my question was easy to answer, TI would probably do it here.

If input sensitivity is hard to find, and a small factor compared to the propagation delay, maybe the loosey-goosey method I've been using (invert Tpd and include a margin for error) isn't so bad at approximating the max frequency.

Anyway, I think I'll be incommunicado for the holiday weekend. Any additional responses are appreciated, but don't expect them from me. Thanks for everything so far.

 

There is nothing special about free running clocks. In fact I'm unsure of what the alternative is. That said if you want to distribute such a clock you use specially designed clock buffers that have output stages which can rapidly charge and discharge a capacitative load. You would not typically waste time trying to get an ordinary gate to do the job. In an FPGA or an ASIC you would not use ordinary gates to buffer a clock. When it comes down to it you probably should be using an FPGA in place of discrete packages if you are interested in speeds higher than say 50 MHz.

 

I'm sorry but if you want to be that specific you must be that specific and not waffle generalities.
Otherwise you are wasting everyone else's time.

So post a specific circuit and ask a specific question about it.

 

I'm sorry but if you want to be that specific you must be that specific and not waffle generalities.
Otherwise you are wasting everyone else's time.

So post a specific circuit and ask a specific question about it.

Ultimately, I’m not trying to be specific. I want to know how to determine the max freq for a digital IC when the datasheet doesn’t say. If I knew how, I’d be ready in the future when a problem comes up. The common case I’m focusing on is when the datasheet has Tpd.

If a specific circuit helps, here: A 74LCX86 has one input held low, and at the other input a square wave with 1ps rise and fall times. The 74LCX86 is powered by 3.3v, the square wave is from 0 to 3.3v, ambient temp 85C, load of 500 ohm and 50pF (if one doesn’t like these numbers, one can change them). What is the maximum frequency of the input square wave for which there is an output of equal frequency (regardless of the output amplitude)?

If one doesn’t know, one need not reply. I’ve had enough responses attempting to change the question to something with a known answer, though I’ve appreciated these thus far. I understand that all responses have been attempts at help. Also, I'm sure many are not responding because they don't know the answer.

I’m getting the idea that one can’t know the max freq from a datasheet that has Tpd instead. As WBahn suggested, the way to find out may be to ask the manufacturer of the part.

My recurring problem is that I work with short deadlines. I may not have time to wait for a manufacturer’s reply. I may not have time to find a part with a sufficient quoted max freq (which might not exist). I was trying to fix this by asking how to determine max freq, because I don’t have a deadline now so it’s a good time to learn. Perhaps the better thing to do is try to foresee my future needs, and accordingly search for parts and ask manufacturers questions now when I have the time.

 

If a specific circuit helps, here: A 74LCX86 has one input held low, and at the other input a square wave with 1ps rise and fall times. The 74LCX86 is powered by 3.3v, the square wave is from 0 to 3.3v, ambient temp 85C, load of 500 ohm and 50pF (if one doesn’t like these numbers, one can change them). What is the maximum frequency of the input square wave for which there is an output of equal frequency (regardless of the output amplitude)?

If you truly want the answer to this, then you must be prepared to conduct some testing yourself.
Any design engineer should know that this is an ill posed statistical question that requires a statistical answer.
He should also understand the statistics involved.

Any manufacturing outfit that attempts production without this information is heading for a major product failure.

You don't do engineering by dictat, but by the rules of physics and engineering common sense.

That's my opinion

 

The answer to you original question is that you cannot determine a single maximum frequency. The best that you or the manufacturer can do is estimate the maximum frequency by testing a sample of the population. You can make heuristic or rigorous arguments about the efficacy of your methods but ultimately you either subscribe to the proposition that the sample mean and variance apply to the population or they don't. Then you need to deal with the consequences of that decision.

Let me make another thing clear. Datasheets do not contain guarantees. They contain statements for which the likelihood of an instance of a specification violation is extremely small. The absence of such a statement in a particular datasheet is an indication that they choose not to make such a statement, both because they don't want to and they don't have to.