Holy wasted opamps Batman.

How do you do that? If you start an oscillator with a -random phase with respect to the crystal oscillator, then lock the oscillator to the crystal, the phase has to shift and you lose the initial phase. How do you keep the start/stop oscillator edges time-coherent to the trigger?

There are a number of excellent papers on bang-bang pll loops. Here are some:

(unfortunately, my link to the zip executable no longer works. I could upload it if you like)

Verification of Bit-Error Rate in Bang-Bang Clock and Data Recovery Circuits Ken Kundert

Richard Walker has a number of papers on bang-bang loops. Search "bang- bang"

Many more references online.

There are a few techniques. The Hewlett Packard 5359A Time Synthesizer used a triggered ECL delay-line oscillator which was synchronized to a crystal oscillator using a heterodyne phaselock technique. The manuals are available online. Several people copied that idea. My trick is to have an ADC that's clocked by the XO and that samples/observes the waveform of the triggered oscillator. An FPGA looks at the ADC data stream, figures out the phase relationship, and locks that in, with a DAC/varicap on the triggered oscillator. Done carefully, the error is low picoseconds.

I'm not doing a b-b loop now, but I'll save those in case I do again.

Thanks. That helps a lot.

It looks like I got snookered. Apparently there are several PLLSim programs with the same name, and the one I have is for a conventional charge pump PLL. It looks useful in it's own right, but it's not a bang-bank.

I' trying to track down the authors of the original program and will update if successful.

How does positive feedback make the filter oscillate and not latch?

There are two methods. The one I like is to use an auxiliary Schmitt trigger to pull the input a millivolt or so off the null. That gives you a nice triangle sweep and works with any sort of loop filter. (I posted a circuit like that upthread.)

The other one is to use a Wien-bridge style network. That doesn't need the extra amp, but it complicates the filter design. Since an LM358 is better than good enough for the Schmitt, it seems like a no-brainer.

Cheers

Phil Hobbs