If you are comfortable with coil oscillators at lower frequencies, How about a ring
oscillator?
I personally don't have experience with discrete components at that high of a frequency, but I have
put together IC layout using a ring oscillator method to obtain Ghz frequencies.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Basically a ring oscillator is a series of inverters ( always an odd number) with the final inverters output
connecting to the first inverters input. This creates a phase shift relationship between each inverter
stage.
Suppose you have 3 inverters and at each input/output stage you have a LC or RC tuned for 1MHz or
a 1uS delay. (in reality propagation delay of the inverter also needs to be factored into the delay). If
you take a "tap" at every inverter you will see a 120 deg phase shift at each inverter output.
By combining the phased outputs you can effectively synthesize 3MHz although your clock is only running
at 1MHz.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Here is a very crude example of a ring oscillator using 1MHz crystals. You could just as easily
use a LC "tank" for your desired frequency in their place. To combine the outputs, you need
a "one-shot" for each phase that results in a pulse shorter than the output of each phase.
Ideally this would be equal width to your target frequency after "OR"ing the phases together.
...So for 3MHz you want the width to be about 166nS.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
I knew there was somthing I was not catching, I was puting inverters in a row in my head and couldnt relate the phase shift. A tank or crystal between eack one is how ?
Yeah, it's a little difficult to get my head around myself, but it does work. You must think in parallel, and not
on an individual inverter crystal/LC section. Consider a single inverter with a 1MHz crystal tied across the
input and output. With the exception of a small propagation delay within the inverter, your frequency is 1MHz.
So the crystal is acting as a delay line of 1uS between the input and output. Now take the same idea, and
place it in a chain of odd number inverters rather than just a single inverter. When you first power up, all
outputs will be in phase minus the propagation delay from the inverter. As the ring oscillator runs it will "seek"
balance and eventually stabilize in the correct phase due to electrical stress of either pushing or pulling the
crystal away from it's properly tuned value and the switching threshold of the inverter gate. This balancing
does not work properly if the inverter to crystal ratio is not a 1:1 correlation.
In Silicon, the above is much easier to accomplish because you base your phase delay on the actual
propagation delay of the inverter. This value is empirically calculated based on the size and type of transistor
used in the inverter delay line and you lay out your chain accordingly. Using discrete components, you must
create you own or a "known" delay value to ensure proper timing this way.
Just a thought:
Another way to generate multiple phases is to have a single LC coil with various tap locations along the coil.
By adjusting C, you could compensate for any offset or loading from the taps.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Short
Posts: 26
Comments
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"When all think alike, no one is thinking very much.' - Walter Lippmann (1889-1974)
······································································ Warm regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan
I was thinking about 800 mhz I cam make coil oscs but I am not sure about that high.
Thanks SHORT
oscillator?
I personally don't have experience with discrete components at that high of a frequency, but I have
put together IC layout using a ring oscillator method to obtain Ghz frequencies.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Thats interesting, how is that a different diameter of ring to vary freqency !
Thank You Short.
connecting to the first inverters input. This creates a phase shift relationship between each inverter
stage.
Suppose you have 3 inverters and at each input/output stage you have a LC or RC tuned for 1MHz or
a 1uS delay. (in reality propagation delay of the inverter also needs to be factored into the delay). If
you take a "tap" at every inverter you will see a 120 deg phase shift at each inverter output.
By combining the phased outputs you can effectively synthesize 3MHz although your clock is only running
at 1MHz.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Here is a very crude example of a ring oscillator using 1MHz crystals. You could just as easily
use a LC "tank" for your desired frequency in their place. To combine the outputs, you need
a "one-shot" for each phase that results in a pulse shorter than the output of each phase.
Ideally this would be equal width to your target frequency after "OR"ing the phases together.
...So for 3MHz you want the width to be about 166nS.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
I knew there was somthing I was not catching, I was puting inverters in a row in my head and couldnt relate the phase shift. A tank or crystal between eack one is how ?
Thank you
Short
Yeah, it's a little difficult to get my head around myself, but it does work. You must think in parallel, and not
on an individual inverter crystal/LC section. Consider a single inverter with a 1MHz crystal tied across the
input and output. With the exception of a small propagation delay within the inverter, your frequency is 1MHz.
So the crystal is acting as a delay line of 1uS between the input and output. Now take the same idea, and
place it in a chain of odd number inverters rather than just a single inverter. When you first power up, all
outputs will be in phase minus the propagation delay from the inverter. As the ring oscillator runs it will "seek"
balance and eventually stabilize in the correct phase due to electrical stress of either pushing or pulling the
crystal away from it's properly tuned value and the switching threshold of the inverter gate. This balancing
does not work properly if the inverter to crystal ratio is not a 1:1 correlation.
In Silicon, the above is much easier to accomplish because you base your phase delay on the actual
propagation delay of the inverter. This value is empirically calculated based on the size and type of transistor
used in the inverter delay line and you lay out your chain accordingly. Using discrete components, you must
create you own or a "known" delay value to ensure proper timing this way.
Just a thought:
Another way to generate multiple phases is to have a single LC coil with various tap locations along the coil.
By adjusting C, you could compensate for any offset or loading from the taps.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.