Digital Frequency/Phase Comparator (DFPC) - better than 4046 PH.C. II ?

ething

code in

time

using

ang-bang PD,

roblem.

if you look at the timing spec for a xilinx cmos output the numbers for data-to-pad and tristate-to-pad is exactly the same, so whether you drive the data or the tristate pin the output transistor must be doing the same

you get the same "slam to the rail"

you could do the same with two outputs and no diodes

something like this:

assign pc2_p = up ? 1'b1 : 1'bz; assign pc2_n = down ? 1'b0 : 1'bz;

-Lasse

type

in

using

bang-bang PD,

problem.

Unless you are pulling up and down at the same time.

You'd just connect the outputs?

a 4046-

ething

th code in

time

of using

a bang-bang PD,

a problem.

ve

with a single high-low-tristate output that would be impossible

yes, of course if you do up and down at the same time they will fight each other but so will two push-pull outputs with diodes unless I missed how you'd connect them

-Lasse

type

something

code in

time

using

bang-bang PD,

problem.

Most of my recent PFD's have controlled current outputs. ...Jim Thompson

he 4046 type

ust a 4046-

something

with code in

the time

ion of using

y.

as a bang-bang PD,

't a problem.

er..

nsive

odes

tate

ed)

of

I guess xilinx cmos outputs are also current controlled (ish) at least they are described that way, being programmable with

-Lasse

type

something

code in

time

using

bang-bang PD,

problem.

This is one version:

The resistors on the left define the currents into the integrator.

This drives a VCXO. In non-PLL mode, we can turn on U4, turn the integrator into a linear amp, and make the UP/DOWN signals into a delta-sigma DAC, to open-loop tune the XO frequency.

he 4046 type

ust a 4046-

something

with code in

the time

ion of using

y.

as a bang-bang PD,

't a problem.

er..

nsive

odes

tate

ed)

of

I think the diodes are redundant if you made the two outputs push/ tristate - pull/tristate and there would be diode voltage more for the resistors to set the current

is it really needed, wouldn't pwm/delta-sigma into the integrator work just as good?

if it works for pll-mode it should work for non-pll mode, every thing is the same it is just what controls the pins that changes

-Lasse

a 4046-

something

with code in

time

of using

bang-bang PD,

problem.

expensive

But the integrator would integrate! Then we couldn't set the open-loop VCXO frequency. In delta-sigma mode we turn the integrator into a lowpass filter.

We do the dual-mode thing for instruments that can optionally lock to an external 10 MHz reference. When they aren't locked to a reference in PLL mode, we can set the VCXO frequency using a saved cal factor and the delta-sigma thing.

pins.

burn

of

expensive

are

Nice. For many things i am contemplating i am looking for something like an old 22v10. Just checked, they are still around, cool. It can do several PFD2s and replace several packages of tiny logic at the same time.

?-)

type

code in

using

bang-bang PD,

problem.

What sort of current noise do you get? I'm always suspicious of IC-generated current sources--IME they tend to be many times the shot noise, whereas a nice CMOS switch to a regulated power supply just looks like a resistor.

Cheers

Phil Hobbs

type

something

code in

using

bang-bang PD,

problem.

I don't know off the top of my head. I'll have to ask my clients what they've measured. My applications are considerably different from yours... frequency synthesizers where the VCO is tank-type with high-Q, or grotesques DLL's where I'm synchronizing digital clocks on multiple chips. ...Jim Thompson