Waveforms in a 74LS04 ring oscillator.

[joeqsmith]

Setup a quick test changing from the whiteboard to perf board.    Using same 3 gate ring and buffer.  Added termination.  Didn't have any 450 ohms, so used a couple of 1K 1206s in parallel.   Pulled up the unused gate rather than grounded as shown.

Still a mess, but should speed up the oscillator.   

[SeanB]

I think that wet tantalum is rather more expensive than the rest of the board though......... Definitely overkill using that 1990's vintage part, though it will still have a low ESR and will still pass all it's military certification tests.

[T3sl4co1l]

Gotta love the nice sharp pull-up of a CMOS gate, compared to the old weak TTL version...

[SeanB]

Now you need to get some ECL devices and try them.

[T3sl4co1l]

Anyone want to start a donation fund for purchasing some Hittite gates?

"The scope can't tell if it's running... let's point the camera at it and see if it's emitting light instead?"...

Tim

[SeanB]

Anyone want to start a donation fund for purchasing some Hittite gates?

"The scope can't tell if it's running... let's point the camera at it and see if it's emitting light instead?"...

Tim

It only does 28GHz...............

https://www.hittite.com/products/view.html/view/HMC842LC4B

And might go as high as 45GHz.

Don't think dead bud style will work here, and as it is in a ceramic carrier only the price might be a little on the high side. wonder if you can ask AD for a sample device.

[joeqsmith]

Sorry, I'll have to leave the ECL test to someone else.  My antique stash of TH parts is pretty limited. 

Swapped out the Wavecrapper.  First plot showing the Motorola SN74LS04N, second the RCA CD74HC04E. 

The third plot is an old CMOS 4000 series, Motorola MC14069UBAL.  27.8ns   Way better than I would have guessed.....

[SeanB]

Unbuffered CMOS is always faster, even if it has an output that will barely qualify it as digital. Supply it with 18V and it will be even faster, though it might also run quite hot as well. RCA CMOS was qualified to 20V, as opposed to the 18V of the other suppliers. The rise and fall time is very dependant on supply voltage as well, which can be an issue. 2V supply and most would not work over a few kHz.

[joeqsmith]

Remember when....  databooks?   Finding the book was faster than searching the internet.   

[joeqsmith]

Top is running the MC14069UBAL at 3.3 followed by 10, 15 & 18 volts.   Those wide thresholds made it idea for some applications. 

[albert22]

Here is a similar ring oscillator, by varying the supply voltage they turned it into a VCO. 
What can be expected of something like it?

[T3sl4co1l]

Seems unlikely that it would be so linear, but it's worth noting a similar thing is used for traditional PLLs and DLLs such as you find in embedded CMOS (MCU / FPGA / radio / etc.) stuff.  What's different is, the gates themselves are controlled from current mirrors (with a cascode circuit), thus varying the slew rate on the switching node (which has a ~constant capacitance), so you'd expect reasonable linearity.

Tim

[JoeN]

What if this circuit was built out of ECL instead of TTL?  Would that work?  What approximate frequency would you expect to see on the output?  Example IC for implementation:  http://www.onsemi.com/pub_link/Collateral/MC10EL04-D.PDF

[joeqsmith]

Here is a similar ring oscillator, by varying the supply voltage they turned it into a VCO. 
What can be expected of something like it?

Not thinking this would be a real good source but depends what you need.  Even if you ran it closed loop with a little PI controller, hardly seems worth it.

It was easy enough to automate this test with Labview. The data should give you some idea of the open loop non-linearity.  I used the same 3 stage oscillator used for the previous tests.  The power supply voltage was swept from 1.0 to 5.2 volts in increments of 0.1.   To get some decent numbers, a counter running off a GPS was used to measure the output frequency.      I ran all of the same devices.   That old RCA part again looks very good for linearity.   

[T3sl4co1l]

Wow, the HC keep going below 1V... (at room temperature, anyway!).  I've heard of them recommended for micropower apps at say 2V, where they're slower than CD4000 at 5V... didn't think they'd keep going that low though.

Tim

[joeqsmith]

I have updated the voltage vs frequency graph to include the three new parts. 

Notice the Fairchild F part tops out over 100MHz!  I don't remember anything faster than the Fs back then (in the 7400 series).   Also seem to remember them being more sensitive to ESD.

[joeqsmith]

Link to the data sheet:
https://www.fairchildsemi.com/datasheets/74/74F04.pdf

Attached to the LeCroy 8500A with the semi rigid as before.  Sub nS fall times now.  Expected the weaker drive.   

[tggzzz]

Now you need to get some ECL devices and try them.

... and stop measuring time in Siemens. Most people prefer time to be measured in seconds.

[tggzzz]

Wow, the HC keep going below 1V... (at room temperature, anyway!).  I've heard of them recommended for micropower apps at say 2V, where they're slower than CD4000 at 5V... didn't think they'd keep going that low though.

I was more surprised at the voltage at which the LS device ran.

[joeqsmith]

Wow, the HC keep going below 1V... (at room temperature, anyway!).  I've heard of them recommended for micropower apps at say 2V, where they're slower than CD4000 at 5V... didn't think they'd keep going that low though.

Tim

At least the two parts from Motorola and National did.  The RCA fell a little short.   

Plot showing the Fast part running at 5.4 volts stuck to a Peltier then cooling it.     With a little higher voltage and a little more cooling, it may break 110MHz.   

 

 

[joeqsmith]

What if this circuit was built out of ECL instead of TTL?  Would that work?  What approximate frequency would you expect to see on the output?  Example IC for implementation:  http://www.onsemi.com/pub_link/Collateral/MC10EL04-D.PDF

That seems like a good experiment for you to try and then post your findings.   

I was trying to stick with the same pinout and package that the OP was using.       

[SeanB]

I remember some boards that used a 74F 138 as a logic decoder, just so they could get those extra few ns of propagation delay down, so that they could use a slower and much cheaper SRAM device. Important if you were making a 128k bank switched ram board and wanted the slower RAM battery backed, the difference in leakage current for the slow parts as compared to the faster ones was actually quite significant. Still ran at sub 4MHz though, using a Z80 processor. There was a note in the service manual that you had to use a F device in those address decoder locations, and use LS or S chips in the rest of the decode logic as well. Supply fail was detected with a simple transistor and zener diode fed from the 8V logic supply, so you would write protect the RAM array when the 8V bus fell to 6V. The transistor drove a CMOS inverter that simply deselected the alternate chip select, putting the RAM to low power mode.

Those days when a 6116LP3 chip was considered both a large chip and a low power one as well.

[joeqsmith]

After the hour warmup, I collected data for an hour.    The frequency drifted almost 250KHz.  Lower graph is showing the power supply drift, about 2mV.   Keep in mind from previous data, frequency went up with voltage, not down.   Suspect this is more temperature than anything.   

[joeqsmith]

From the data sheet, the F part will handle up to 5.5 volts.   To show the effects on the frequency with voltage, I stepped it up few times.  Notice how it jumps in frequency then slowly drops off.     

[joeqsmith]

I pushed the supply up higher, then took a cotton swap with alcohol on it, at dabbed it onto the IC.  You can clearly see how the temperature effects the frequency.   To see if the part would break 110MHz, I raised the voltage to 5.500.  This was not enough even with the alcohol.   

Placed the Peltier on a fair sized heat sink and laid it over the IC.  To the right, you can see the oscillator speeding up well beyond 110MHz as it is cooled.  Final temp was +12C.