Can any one solve the Multisim DigitalCounter sample is now broken puzzle?

[Euler's_Identity]

Setting to zero starts the counter at 11.

[Euler's_Identity]

Bingo!

My idea worked.  Force the oscillator into a known state, and then she runs like a champ.  (Note that I sub'd a Nand for a Nor, but it shouldn't matter.)

Frequency is approximately 737Hz.

Now the question is how much setup time does it need until the CMOS simulation fails?

 

[Euler's_Identity]

I'm seeing the holdoff time must be approximately 170ns or longer in order for the oscillator to run properly.


Here's my theory about why the oscillator circuits in the past didn't work correctly.  In real life this oscillator begins because a gate changes state.  It changes state because one of the joined legs runs slower (or faster) than the other.  In the Multisim world, the legs run at the same rate, but the circuit isn't in error, so the Multisim algorithm just keeps cranking away looking for a solution to a problem that, in real life, would've already been solved by one of the legs tripping the gate into a known state.  Hence, my idea was to tell Multisim how to make up its mind.

It kinda reminds me of the Star Trek episode where Spock tied up the computer with an unsolvable problem, put it in a loop.  Later Kirk asks Spock what he did.  Spock tells him that he told the computer to go calculate Pi.

What I see was the problem was Multisim's algorithm was essentially told to calculate Pi when it came to CMOS oscillators.  It can't decide which leg is first until it hits the CMOS time threshold, which, somehow, causes that time to become the driver instead of the RC design's time.

Yes?

   

[lacy]

I can confirm your findings. Maybe you have found something here.. I know that most oscillators need a kick start with Spice, but why does it have to be with an arrangement like this. In Multilsim, you can set an intial voltage on any net. Looks like that would be sufficient to get it started.

I am going to try this same circuit with inverters to see if it still works. Most CMOS on-chip oscillators are of the inverter variiety. I will let you know if it still works or not.

Message Edited by lacy on 08-02-2008 02:54 PM

[lacy]

Just tried this with your circuit containing only inverters. Inverters don't have 2 legs to apply your theory, but what I did was to tie the capacitor to just one leg of the inverter and the other end to ground or VDD. That worked beautifully. I think you may have solved the CMOS situation and that is the best news. I had tried everything I could think of and just got fustrated and quit. If I designed a circuit I would just use a clock source to replace a cmos oscillator and then add that later in the real circuit.

The only thing is that with the inverter oscillator that I just tested is the fact that the capacitor does effect the frequency due to where it is connected in the circuit. but this can be tuned out with the values.

Thanks for this as it will add something new to Multisim that for me wasn't there before.

[Euler's_Identity]

Ahh!  I just reloaded the circuit and found out that, even though my "set to zero" was unchanged, my "real" setting had reverted back to "ideal" some how.  This means I didn't have realistic voltage levels on my CMOS.  Changing this setting causes my oscillator to develop a combination of the design frequency and the limiting frequency.


Hmm...

[Euler's_Identity]

Ok, try this one.  Technically, the diode is necessary to keep the leg from raising above the supply voltage on the start of the oscillator.  As we were causing the inner-ic clamp diode to conduct, I decided to protect the leg, and it seems to have worked.

 

[Euler's_Identity]

Here's the stand-alone CMOS oscillator circuit.

[Euler's_Identity]

Also, I get approximately 752 Hz.


Ok, the fun's over.  Now I have to go back to my original design, as I'm days behind. 😞


Whoohoo!  Multisim's CMOS simulation  works!  [insert Snoopy dance]