Pointless, yet intriguing.

[NinjaKitten]

I've pondered this question for a long time now and I've not come up with any concrete answers.

What happens if you take a TTL/CMOS device such as a NAND gate and loop the output back onto one of its inputs, then set the other input high? Theoretically it should oscillate at as high a frequency that its switching speed allows. I shall demonstrate:

Initial State:
A: 0
B: 1
O: 1

Set A high:
A: 0 -> 1
B: 1 -> 0 -> 1 -> 0 -> 1
O: 1 -> 0 -> 1 -> 0 -> 1 (oscillation)

But in practice I'm not so sure this would occur.

I'm mainly used to digital electronics, so I don't tend to venture into how the actual devices work internally - I just find the datasheet for the device so I know the pinouts and the device description, then off I go.

Would it really oscillate in reality? If it did oscillate, would it give out a huge amount of heat and burn itself out? Or would it happily oscillate at a very high frequency forever?

I'm tempted to use one of my old 7400N NAND chips, a 9V battery and my oscilloscope to test this out.

[exurbia]

Go for it, pointless research, isn't really.
I'm sure i'm not the only one who'd be interested for a moment, in the answer.

[lineup]

i think this is used
to create digital oscillators, vibrators

and if you add one cap + resistor, an RC delay filter
then you can define the osc. frequency more or less precise

[jcx]

google: logic gate ring oscillator

[NinjaKitten]

Originally posted by lineup
i think this is used
to create digital oscillators, vibrators

and if you add one cap + resistor, an RC delay filter
then you can define the osc. frequency more or less precise

I get that they're used in oscillators, I use RC/NAND oscillators all the time when I don't need a really precise clock. I'm just intrigued as to what would happen without the RC part.

I'll dig out an old 7400 tonight and let you all know.

[NinjaKitten]

Done.

The chip puts out quite a bit of heat and oscillates at a frequency too high for my scope to detect it properly. After 45 seconds or so of operation, the 7400 died and I got a short across the chip between the inputs and output of the gate I was using. However, I must mention that this is a very old chip (circa 1999) and may not work to modern standards.

According to the datasheet, the switching time is around 9.5ns between states. This means that the frequency should be 1/(2 * (9.5x10^-9)) as it takes two switches for one cycle - a theoretical 52MHz.

Would this be lower in reality due to the capacitance of the chip's leads and the wiring between them? If so, how much lower?

[dmills]

You can in fact use some cmos parts as amplifiers, and at least one of the 4000 series inverters was available as a 'unbuffered' part for exactly this use.

They do not make very good amplifiers you understand, but for things like input stages for variable reluctance pickups (small signals of a few dozen mV or so) to something roughly logic level, they were good enough and were cheap (and you got 6 in a 14 pin package as I recall).

Incidentally, most cmos will pull a lot of current if the input is too far from either rail as both output fets conduct, hence the reason to tie unused cmos inputs either high or low.

Regards, Dan.

[NinjaKitten]

Forgive me my ignorance, but I don't see the relevance.

[dmills]

Gain falls with frequency, so an inverter wired back on itself will probably end up with its input swinging slightly around 0.5vcc (or whatever threshold is).

You cannot consider these parts to be digital components when abused in this manner, they have to be treated as non linear inverting amplifiers for the behaviour to make sense.

I was merely pointing out that there were 'digital' parts explicitly designed to be used in this manner (and that some were designed to actually be stable when operated this way).

Regards, Dan.

[lineup]

How about using one schmitt trigger inverter?
Will this behave differently?