Old 5v TTL: What does a NAND gate looped back on itself do at the analog level?

I no longer have access to a scope; I've been wondering about this off and on.
Back in college I pondered a NAND gate with one input held high, and
the other input being tied to the output.
Basically wondering about the oscilation, and just how "clean" it might be.
Now I guessing that it's not going to look anything like a clean square wave, because the transition voltages would be met mighty fast. But are the transitions monotonic? If not, does any noise goof up which direction its transitioning when switching that fast?
I suppose what I'm asking is: Is there an odd place where the transition voltages from true to false (and back) are not "quite" enough there to have the NAND output enter an completely undefined state? At first I was assuming that the there would just be a tight cycling around the transition but I'm not sure now what noise and any need for the analog circuits to be "left alone" might cause.
Thanks!
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Depends. What is the rise time in comparison to the delay? What's the loop gain? If the loop gain is high and the rise time is fast (in comparison to the delay) it'll oscillate as you suggest. The faster the edge rate the "squarer" the oscillation. If the device (delay) is slow or the loop gain is low, it'll settle to a mid point, much as an op-amp. With few exceptions, this isn't recommended though.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Sure, understood. The assumption I'm making is that the gate internals are slower than the electron propagation :) , so I'd have to say my initial assumptions were the settling toward the mid-point.
But I'm I'm still not entirely convinced though that there wouldn't be some odd shape to this "curve", noise directed perhaps, that wouldn't be what we're expecting.
Imagine this for 1 example: The voltage rises near transition, get's stuck in a mini-curve down (very small due to noise) and is then brought back up. If that NAND loop back happens along the time domain more narrowly than the noise, it might be an interesting settling point that doesn't, well, seem obvious. OR that the gate then tries to shift states before it's supposed to and then we see a beat frequency between the internals and the noise or something else goofy.
Though I'm sure it's nothing more than the settling like you suggest . I just find it intriguing.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

What's electron propagation have to do with it? Again, it'll settle to the midpoint if the loop gain is <1 at 1/delay frequency.

If it settles in the linear region, sure, noise will be amplified.

If the loop gain is high it will want to oscillate and noise will push it into oscillation. If the loop gain is <1 noise will be amplified but it will stabilize back to the midpoint.

It *is* an amplifier, just not a very good one.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Fri, 22 May 2009 10:03:10 -0700 (PDT), " snipped-for-privacy@gmail.com"

All digital logic families have voltage gain above 1; if it didn't, logic signals would fade away in digital systems. Whether it oscillates or not is a control-theory issue; gates with one gain stage are 1st order systems and tend to not oscillate; gates with multiple gain stages are higher-order gadgets and tend to oscillate.
Old 74-series TTL and 10K ECL had gate voltage gains in the single digits and 1st order time response (single dominant delay node inside) so tended to be DC stable if you connected an inverter input to its output.
A gadget can have very high gain here and still be stable; an opamp for example.
John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Fri, 22 May 2009 07:12:03 -0700 (PDT), "Thomas G. Marshall"

Old slow 74-type TTL had essentially one low-gain stage, 1st order response, so a simple gate would usually stabilize at about +1.2 volts DC. Later gen faster parts, like AS and F will probably oscillate. HC-type CMOS parts have multiple internal stages, higher-order delay, so usually oscillate. Early ECL (10K, 10KH) would DC stabilize too; faster, higher gain parts (Eclips, EL, EP) will oscillate.
10K type ECL was often used as a fast amplifier. A 7400 gate can be used as an amplifier, too.
John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On May 22, 1:21pm, John Larkin

Great responses you two! Thanks!
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

with a direct connection it might not even oscillate. put a twin-tee or other phase inversion network there and it's likely to though.

by adding negative feedback people have used biased gates as low power audio amplifiers in the past.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.