Which Phase Detector to choose ?

[Jadeit]

Let's have a clock signal frequency 5-10MHz , positive pulse width always 30ns, 0/5V.
A signal derived from it, which is delayed by 5 to 15ns.
We are interested in what the signal delay is.
We have an MCU any as STM32L available and we are interested in what the signal delay is.
I will probably use a phase detector, direct its output to the RC cell and measure the voltage of the ADC 12bit.
The question is, what type of phase detector is most suitable?
1. Exclusive OR gate 74HC86
2. Edge triggered JK flip flop
3. Dual D type phase comparator

4. High speed Analog comparator
5. Some other

I wonder what type of detectors do you consider the most suitable?
For an analog comparator, how fast must it be?
All you need is a high speed comparator in the MCU propagation delay typical 55ns or I have to use any as rel High speed comparator for exmple 4,5ns **broken link removed**

 

[barry]

The XOR might be a problem since you don't have 50% duty cycle, unless your reference clock has the same duty cycle. Also, the HC86 is way too slow, maybe look at an LVC86. I'm assuming you're charging a capacitor through a resistor(I'm guessing that's what "the RC cell" is), and measuring the voltage with the ADC?

I would suggest you simulate your proposed circuit and see if the performance meets your requirements (which you haven't stated).

I don't understand your last sentence at all.

 

[KlausST]

Hi,

I don´t see the need for 50% duty cycle.
The XOR output gives the rising delay as well as the falling delay.
An LPF gives the ratio: (rising_delay + falling_delay) / period_time. As long as both duty cycle are within 15% ... 85%.
(without having tested it)

But here I see a problem: since your frequency is 5MHz...10MHz you get varying LPF filter output.

Klaus

 

[Jadeit]

But here I see a problem: since your frequency is 5MHz...10MHz you get varying LPF filter output.

for clarification. The frequency does not change Now we are testing the thing with 5MHz, but clever books say that the frequency should be from 10MHz higher.The reason is the thermal stability of the physical process that we measure.
I think it won't be a problem, but a version with a 10MHz clock will be created for verification.

 

[dick_freebird]

There are, I believe, "time to digital converters" but
whether these have the resolution & range you are
after, at a price you're willing to pay, I couldn't say.

I do not believe the phase detectors will be super
accurate when you attempt to measure time as an
analog voltage. Phase detectors are often used within
a feedback loop which buries their offset by loop
gain and loop filter.

If you had a range of (say) 0 - 50nS and wanted a
resolution of 0.5nS (1% resolution) that would fit in
a 7-bit space. If you had a counter that could run
from (say) a 2GHz clock then you could get decent
resolution by a gated clock, gated reset digital
scheme.

Then again I don't see why you wouldn't just use
a decent 'scope and low capacitance probes and
use the cursors, if it's a one-and-done measurement.
Of course if you wanted to datalog it over time, you
might prefer an "adjunct probe" such as one of these
phase detector or time-counter kludges. But many
'scopes would let you accumulate crossings and
histogram the built-in delay math function.

 

[KlausST]

I do not believe the phase detectors will be super
accurate

Not very accurare, one will need to do some offset correction, because of tr to tf (and their delay) mismatch of the XOR..and output resistance mismatch.
But I expect the precision to be rather good .... but still one has to maintain constant temperature and constant VCC.

Klaus