Driscoll and Differential oscillators have almost the same phase noise at all frequencies, so there are no other parameters to influence the choice.
The only way could be listening to both in the same system and then make the choice according to personal taste.
I suggest to place each oscillator in a separate aluminum box to shield them as best as possible, so the SMA connectors and cables are essential.
As I said several times these are state of the art oscillators therefore it would be desirable to implement them in the best possible way.
As you all know I totally disagree with the implementation method suggested by others on this forum such as stacking devices or keep them close to each other, but of course everyone is free to chose his own way.
What I have to say for professional correctness is that there are rules to be followed based on precise physical principles, although these rules are often more complicated and expensive.
Simulators like Spectre RF can simulate phase noise, as long as it isn't dominated by obscure unmodelled effects inside the crystal (which probably excludes Andrea's oscillators, as he optimized out everything else).
With all due respect to Cadence I believe that it is not possible to adequately simulate a crystal oscillator, there are too non-linear parameters that the simulator cannot take in the right account.
I don't know Spectre RF, but I wonder how it could include in the simulation the impurities on the surfaces of the crystal. These are not "obscure unmodelled effects inside the crystal", this is the real world.
These impurities dominate the random walk of the crystal and therefore the 1/F3 phase noise of the oscillator (and in part even the 1/F noise).
The Rubiola.org website contains a lot of literature on these aspects.
Some time ago I asked Enrico Rubiola how the phase noise could be improved once the performance I posted was reached.
His answer was lapidary: working on the oscillator circuit is a losing battle, once certain performances are reached the close in phase noise is dominated exclusively by the crystal.
SMA plug connectors and RG400 semi-rigid cable like in the attached picture.
You can buy them from RF Supplier
Custom Cable Assemblies | No Minimum Order
Maybe you also could find them on eBay.
Attachments
No selection criteria and both exist because when we started developing we had no idea about the final results so we have built both.
Finally we can say that the crystal makes the difference, but to draw this conclusion it took years of development and a lot of iterations to get the best results.
Without building both, it was not possible to determine which would be the best performer.
Andrea, what converter IC is on these boards in the picture? Or is that a secret.
Yes, that's more or less what I meant.
The usual way to model a crystal, which was perfectly adequate for all crystal oscillators I ever designed (mostly fast starting oscillators for ISM transceivers and a few for broadcast radio and television), is one or several LRC series circuits in parallel with a capacitor that models the electrode capacitance. As long as you don't design for the kind of close-in phase noise levels you design for, the oscillator circuit rather than the crystal usually dominates the phase noise and then you can simulate it perfectly well with Spectre RF using this kind of model. The only thing that's a pain is the effect of drive level dependence on start up, but that has nothing to do with phase noise.
As long as no crystal manufacturer or anyone else comes up with a suitable simulation model for the effects of impurities you write about, I think it is fair to describe them as obscure unmodelled effects inside the crystal. At least they are unmodelled effects inside the crystal, obscure or not. When they dominate, any simulation without them is obviously useless.
Last edited:
So the members can blame me in near future if they wouldn't be satisfied with prices?
BTW i will order
Hello Andrea,
I have been regularly following this thread with increasing interest.
Congratulations with the performance level you achieved for the Driscoll and differential style oscillators! You indeed reach absolutely state-of-the-art performance now.
At what drive level are the crystals driven now in the 5.6448MHz versions?
I have been regularly following this thread with increasing interest.
Congratulations with the performance level you achieved for the Driscoll and differential style oscillators! You indeed reach absolutely state-of-the-art performance now.
At what drive level are the crystals driven now in the 5.6448MHz versions?
Hello Andrea,
I have been regularly following this thread with increasing interest. Congratulations with the performance level you achieved for the Driscoll -and differential style oscillators! You have reached absolutely a state-of-the-art performance now with these oscillators.
At what drive levels are the crystals driven in the 5.6448MHz oscillators?
I have been regularly following this thread with increasing interest. Congratulations with the performance level you achieved for the Driscoll -and differential style oscillators! You have reached absolutely a state-of-the-art performance now with these oscillators.
At what drive levels are the crystals driven in the 5.6448MHz oscillators?
in post 8 under
Clock Generator for Cambridge Audio DISCMAGIC1/CDT - Adjustment Instruction for VR1 ?
I had upload some years ago the schematic of LC technology LClock "XO".
Maybe of interest.
Clock Generator for Cambridge Audio DISCMAGIC1/CDT - Adjustment Instruction for VR1 ?
I had upload some years ago the schematic of LC technology LClock "XO".
Maybe of interest.
Hi Joel,
difficult to say, I have not yet compared them in a listening session.
It might seem curious but at this moment I don't have an audio system available to compare them, too busy with these projects.
In Italy we could say "the cobbler with broken shoes".
Joking aside I expect they perform almost the same, only slight differences, maybe small nuances based on personal taste.
- Status
- Not open for further replies.