74HCT04 oscillator... is this expected?

[havard]

So I decided to build Grant Searle's simple 6502 machine, and started with the oscillator.  Rather than getting a clean 1.8MHz clock, I had a not-quite-square wave that was somewhere between 40 - 50MHz, depending on tongue angle, regardless of load (into a 6502's clock input) or termination (tried 100 ohm and 10k pull up and pull down).  I figured my crystal was a dud, so I dropped in a 32kHz watch crystal soldered onto some header pins and still had a 40 - 50MHz not-quite-square wave with a 32kHz modulation.  Assuming I seriously screwed up, I began pulling out components until I got to the point where there was a resistor across the input and output of two gates.  I no longer had a 40-50MHz not-quite-square wave.  I had a 75 - 100MHz sine wave.  I chalked it up to gremlins in the cheap breadboards I had at my disposal.

A few weeks go by.  After watching eevblog #568, the thought hit me... what would happen if I soldered resistors directly across the leads and held the thing in the air?  A few more weeks go by.  I finally gave it a shot tonight.  Grabbed a few 100 ohm 1206s, soldered them across the appropriate leads and sure enough, I have a 75 - 100MHz sine wave.

What is going on here?  Is this expected?  A bad 74hct04?  Or most likely, did I miss some basic fundamentals?

edit: the oscillator originally started out as the one from http://searle.hostei.com/grant/6502/Simple6502.html

[valentinc]

      Try building this: http://www.electro-tech-online.com/attachments/cmos_oscillator_761-png.7103/   instead... The topology you have used relies on the propagation delay of gates only, which vary significantly with temperature... So it's not a stable circuit... The one in this schematic relies on an external capacitors and 2 resistors... And the propagation delay doesn't count much at all...

       If you want an even better one, with very sharp edges, I recommend you build one with a Schmitt trigger:

[T3sl4co1l]

Needs HCU04.

Unless otherwise noted, all HC series chips are input and output buffered.  An HC04 is not a pair of MOSFETs, but a chain of three inverters.  This gives so much gain and phase shift in the threshold region that it cannot be stabilized with simple resistor bias.  Read up on 'ring oscillators'.

[havard]

Needs HCU04.

Unless otherwise noted, all HC series chips are input and output buffered.  An HC04 is not a pair of MOSFETs, but a chain of three inverters.  This gives so much gain and phase shift in the threshold region that it cannot be stabilized with simple resistor bias.  Read up on 'ring oscillators'.

Thank you!  It all makes sense now. 

I actually read up on ring oscillation when I first stripped down the crystal circuit into what I have now, so I figured it would be something like this.  It's working as it should, just not as I would expect!  That will teach me to read and reread the datasheets, even for something as "obvious" as a hex inverter.

[Gabb0]

There is also an episode of The Signal Path Blog that explains ring oscillators and shows some measurements.