lunren wrote on Dec 14th, 2011, 2:13pm:loose-electron wrote on Dec 14th, 2011, 11:27am:I have used the Maneatis device with good success many times. Generally redesign the bias circuitry.
Consider running the ring oscillator at lower frequencies and using multiple phases from the ring to get the desired clocking rate.
Hi Jerry,
Do you think 1.5V supply for a 2.4GHz Maneatis device is doable? Can you give me some hints or reference on how to use low frequency multiple phases to get higher frequency clock?
Perhaps, but it depends on the threshold voltages
of the devices. It all scales down, but have to look at the speciifics
of the foundry process.
The poly phase method is pretty straightforward, think of it like this:
(illustrating using 9 delay cells)
Oscillator runs at 1 MHz
There are 9 1MHz square waves in your ring oscillator
Phase seperation of the 9 phases = 360 degrees/9 = 40 degrees
Take 9 clocks with 40 degrees phase difference
You have edge rates at 9MHz
Use some digital logic to combine the 9 signals to create
a signal which is 9X faster than the VCO frequency.
Thats the concept. I used this first way back in 0.5 micron CMOS
to get a SerDes system running at 2GB/sec.
We used 20 phases to get that to work with the slower transistors.