Jump to content
beerandmusic

CLOCKS, what should we look for in next generation

Rate this topic

Recommended Posts

50 minutes ago, mansr said:

The Leeson effect has nothing to do with this. I'm talking about a direct mathematical consequence of phase noise in the clock on the reconstructed signal in a D/A converter.

 

Ok, we discussed that in another thread. 

 

Phase noise at 1Hz is an error of 22 MHz +/- 1 Hz and the level of the corresponding frequency variation is this level x f/22 MHz . Phase noise at 10Hz offset is roughly f/2 MHz ...

 

So how much below -120 dB is audible? (That’s roughly the frequency error if the phase noise is 0dB!!! at 10 Hz)

 

The error for >10 Hz for reasonable good clocks is already very low.

Share this post


Link to post
Share on other sites
56 minutes ago, mansr said:

Maybe clock phase noise at 1 Hz isn't actually audible despite the widespread rumours to the contrary.

If you look at the frequency vs phase noise curves there is a flat baseline perhaps -140 - 170 dBc/Hz, as the frequency approaches 0 Hz the phase error rises logarithmically. 

 

So the phase noise > 100Hz shouldn’t be audible, if anything is audible it is the close in noise. 

 

The point is that the effect of phase noise goes down linearly as clock frequency goes up, but the performance of the crystal at close in frequency offsets goes down logarithmically so as sampling goes up with clock frequency there is a net reduction in the effect of phase noise.

Share this post


Link to post
Share on other sites
On 2/12/2019 at 3:44 AM, jabbr said:

The digital FIR but the subsequent analog filter should diminish the 22 MHz +/- component to the point where - 100 dBc/Hz phase error becomes ??? voltage variation ... below any reasonable noise floor ... no?

 

You are now talking about something that is not time-invariant? Or are you still talking about clock line delay for conversion latch?

 

For static delay, you can think about moving some of the unity-weighted FIR taps sub-sample amount in or out. But that amount doesn't change over time, it is the same all the time (because PCB trace length doesn't change). You can model this in digital domain with suitable amount of oversampling if you like. Effect is change in the filter frequency response (if we disregard the other more beneficial effects). So no effect in pass-band, but some effect on transition band.

 

Share this post


Link to post
Share on other sites
38 minutes ago, Miska said:

You are now talking about something that is not time-invariant? Or are you still talking about clock line delay for conversion latch?

No here I’m talking about the +/- 1  Hz component which is at -100 dBc/Hz (phase noise level) ie the error signal compared with the center 22 MHz. That error ripple will itself be attenuated by the analog LPF. 

Share this post


Link to post
Share on other sites
2 minutes ago, jabbr said:

No here I’m talking about the +/- 1  Hz component which is at -100 dBc/Hz (phase noise level) ie the error signal compared with the center 22 MHz. That error ripple will itself be attenuated by the analog LPF. 

 

Noise components are transfered down in the ratio of frequencies as discussed earlier, and the analog LPF cannot fix it. Converting same sample multiple times (through AFIR or similar) helps reducing it though because it affects transformation of the clock to analog domain. The boundary where conversion happens is the critical point where the clock becomes embedded into the analog signal.

 

Share this post


Link to post
Share on other sites

Ultimately, phase noise defines how well energy of a single discrete frequency in digital domain is concentrated into actual analog frequency component. So this time, instead of talking about time-domain blur, we can talk about frequency domain blur... Of course the two are also related, but often discussed separately.

 

 

Share this post


Link to post
Share on other sites
4 minutes ago, Miska said:

Ultimately, phase noise defines how well energy of a single discrete frequency in digital domain is concentrated into actual analog frequency component. So this time, instead of talking about time-domain blur, we can talk about frequency domain blur... Of course the two are also related, but often discussed separately.

 

 

Right, so perhaps my attempt at a time domain analysis isn’t helpful.

 

Frequency domain is easier and 1/22e-6 @  > -100 dBc is a very very small number

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×