The wave-shape looks nice, Andrea. But what about phase-noise, have you been able to measure that , yet?
As I said several times, as soon as all the oscillators will be ready (Clapp, Driscoll, Butler and now Pierce too) I will go to a university lab to measure their phase noise, accessing an Agilent E5052.
So.... very long time.
***********************************
Hey brother, how're you doin...
Look, it is probably just me....but your question just doesn´t make any sense to me...
What is the objective of your design? what are you tryin a achieve?
Last edited:
************************************
Hej Jesper...
The higher frequencey clocks, (24.576) as apposed to 11.2896 will have higher phase noise.
It would not be comparatively correct to measure two oscillators with different frequencies or xtals.
But if you need to modifiy the RutgerS to accomdate a 24.576 MHz xtal....just send me an email...
cheers
Last edited:
**************************
Hello Andrea.
Unfortunately this is an incorrect assumption.
Exceeding the maximum DL is a BAD idea.
The max DL is the Mechanical-Elasticity-Limit of the quartz blank. It is there for a reason.
An exceedingly high DL, will lead to small surface-fractures.
It will also lead to partial evaporization of the surface-etched electrodes, and elevated levels
of gas-contamination inside the holder.
These mechanisms WILL in-time damage the substrate.
There is an optimal operating range for the xtal unit, defined as Normal operating Range (see pic belw)
The motional resistance (R) will have a fairly constant behavior within this region (at least, this should be important to you.)
Exceeding the max DL will lead to increased R, increased heat dissipation in the substrate, and in some oscillators,
even spurious freq jumps, replacing the main selective-mode.
Last edited:
The objective is to improve the close in phase noise of a higher frequency clock using a lower (eg 1/2 or 1/4 ) frequency clock.
***************************
Hello Herbert and I hope everything is fine with you.
74HC inverters do not provide good isolation, that is the reason why you NEVER see them in precision
oscillators, specially as buffers or isolators.
Their additive PN is not easy to measure, however it is easily audible (you can easily hear it in your audio system).
Measurements of the RutgerS Clock show comparison of LT1016 vs 9 stages of inverters inside the 3 cascaded 74HC04:s....
...just take a look at the internal circuitry of LT1016 and you'll understand why the 74HC04 measured better...
Skewing of the clock flanks (meaning using slower devices) will lead to higher jitter-content in the
clock signal, presented to the next stage, not necessarily to the clock signal itself.
Why?? read-on...
The final jitter reaching your DAC IC is much more dependent on the Input-Referred-Noise-Density and
the related Noise-Bandwidth of the DAC's clock-input stage.
These parameters are made far worse by the physical leakage mechanisms of other active, on-chip,
functional-blocks (unavoidable in IC's).
This means you do not necessarily need to make the clock flanks faster than necessary.
I think the best way to measure the effects of PN of your favorite clock on the audio signal would be to perform measurements on
the analog signal on the output of the DAC chip.
RutgerS is already an excellent clock for audio.
I am certain it will measure even better with a properly designed buffer, and correctly biased resonator (decreasing the DL).
--------------------------------
Keep up the good work, Gentlemen.
Last edited:
Thanks for this explanation, Alexiss.
One question: You mention measuring the PN effects after the DAC. Would that imply a standardization of the DAC (and it's implementation) with which the effects are to be measured?
Cheers,
Edwin
Hi esgigt...
...I am not sure if I understand your question correctly, but I don't think that it is necessary to standardize
DACs and measurements.
Just observing the sidebands generated on the reproduced single-tone signals on the DAC output will serve
as a better indicator of the jitter performance of the entire chain (clock signal inside the IC + DAC), as this
will be system-dependent.
As audio guys, I think this should be more important to you than going all-out trying to reduce SSB-Noise figures
on the oscillator output (these are already very good indeed in the RutgerS).
...I am not sure if I understand your question correctly, but I don't think that it is necessary to standardize
DACs and measurements.
Just observing the sidebands generated on the reproduced single-tone signals on the DAC output will serve
as a better indicator of the jitter performance of the entire chain (clock signal inside the IC + DAC), as this
will be system-dependent.
As audio guys, I think this should be more important to you than going all-out trying to reduce SSB-Noise figures
on the oscillator output (these are already very good indeed in the RutgerS).
Last edited:
I guess you did understand my question as I intended it to be...
Actually I have no idea how much jitter and other junk a DAC adds to the audio signal... I also assume that different DAC's have their individual contribution to distortion products. That's why I asked.
True... the end-result is what matters most.
Last edited:
Good point. I stated this several times in other threads, but people continues to see the audio chain as separate devices, not as a system.
I would think to a system, from the source to the DAC output. I would feed the DAC directly by the master clock slaving the source. But this implies to develope an entire digital system. Someone skilled in uC and FPGA could develope such system, rather than design separate devices (USB to I2S, FIFO, R2R DAC and so on).
BTW, a good master clock is a must also in the above way.
@Andrea:
Hi Andrea ... Interesting thought that I do much agree with. On a smaller scale - and not optimized - I've experimented with this using the Amanero Combo 384 where I removed the original oscillators from the Amanero board and instead fed the clock that comes from the oscillators feeding the DAC. This worked and to my ears has some quite unusual sound qualities.
However, IMHO there's a current "undesirable" state of things in that the audio systems use both 22.*** MHz and 24.*** MHz clocks. Not least because the frequencies are quite close and I reckon that - with the possible exception that one of the frequencies might be more consonant e.g. with current Western World tone tuning - the difference is negligible. Thus one frequency IMHO would be desirable.
As it is now the ability to make a quality switch between two close-by frequencies is necessary.
Cheers,
Jesper
Hi Andrea ... Interesting thought that I do much agree with. On a smaller scale - and not optimized - I've experimented with this using the Amanero Combo 384 where I removed the original oscillators from the Amanero board and instead fed the clock that comes from the oscillators feeding the DAC. This worked and to my ears has some quite unusual sound qualities.
However, IMHO there's a current "undesirable" state of things in that the audio systems use both 22.*** MHz and 24.*** MHz clocks. Not least because the frequencies are quite close and I reckon that - with the possible exception that one of the frequencies might be more consonant e.g. with current Western World tone tuning - the difference is negligible. Thus one frequency IMHO would be desirable.
As it is now the ability to make a quality switch between two close-by frequencies is necessary.
Cheers,
Jesper
**************************************
Jesper, this is really not a problem that cant be solved.
You can have two low jitter clocks (RutgerS 22...MHz & RutgerS 24... MHz) running on a common platform with minimal interference.
The question is: why would you want to?
Cheers....
Last edited:
***********************************
This is exactly in accordance with my own observations.
***********************************
This is where I get lost....I cannot understand nor explain the realtionship...
I'm not sure I fully understand this claim but it seems to be that random frequency modulation in the 1 Hz to 100 Hz (?) is audible at what are very low levels. I think its fair to say these variations are as much as 100 dB smaller that the wow and flutter of a conventional turntable or tape transport. This is why I think a real effort needs to be made to test and quantify audibility.
This modulation should be pretty easy to spot as sidebands on a carrier in the analog output. Possibly a low frequency carrier (100 Hz) would show them more but they should be present on a 1 KHz carrier. Another way would be to look at the time interval variation on a 1 KHz tone. This would be more sensitive to cycle to cycle variations.
However I may have completely misunderstood this.
This modulation should be pretty easy to spot as sidebands on a carrier in the analog output. Possibly a low frequency carrier (100 Hz) would show them more but they should be present on a 1 KHz carrier. Another way would be to look at the time interval variation on a 1 KHz tone. This would be more sensitive to cycle to cycle variations.
However I may have completely misunderstood this.
*********************************
well, Herbert is saying that an inferior oscillator at these low offsets produce more natural sound!!!
Look at the above-mentioned text... maybe he means the opposite??
Phase Noise Predictions Poll
Allright gentlemen.
As you know by now, Andrea is planning on performing comparative
measurements on 4 oscillator configurations.
These are:
1. RutgerS
2. Driscoll
3. Pierce
4. Butler (aka common-base)
Thanks to the input from everyone here, this thread is quickly becoming one the best sources
of collected information on audio-related oscillator design (many thanks to Andrea).
In the true spirit that started this thread, let us gather our knowledge and predict which oscillator-design
you think would come out as the victor and why.
So far it seems that the close-in phase noise is the most important parameter for digital audio
reproduction quality.
RutgerS oscillator has excellent close-in phase noise measurement.
However, I will place my bet on the Driscoll as number 1, with the Pierce coming to a very close second.
Why??
I believe the Driscoll produces the highest Loaded Q than any other circuit and a 11.2896 MHz AT-cut Driscoll
will be the winner.
As mentioned earlier, the Pierce mechanism, along with its class-A operation, and its way of using the resonator, allows
for tighter control of the resonance frequency than the Clapps-type, hence leading to better PN performance.
What are your ideas??
.
Allright gentlemen.
As you know by now, Andrea is planning on performing comparative
measurements on 4 oscillator configurations.
These are:
1. RutgerS
2. Driscoll
3. Pierce
4. Butler (aka common-base)
Thanks to the input from everyone here, this thread is quickly becoming one the best sources
of collected information on audio-related oscillator design (many thanks to Andrea).
In the true spirit that started this thread, let us gather our knowledge and predict which oscillator-design
you think would come out as the victor and why.
So far it seems that the close-in phase noise is the most important parameter for digital audio
reproduction quality.
RutgerS oscillator has excellent close-in phase noise measurement.
However, I will place my bet on the Driscoll as number 1, with the Pierce coming to a very close second.
Why??
I believe the Driscoll produces the highest Loaded Q than any other circuit and a 11.2896 MHz AT-cut Driscoll
will be the winner.
As mentioned earlier, the Pierce mechanism, along with its class-A operation, and its way of using the resonator, allows
for tighter control of the resonance frequency than the Clapps-type, hence leading to better PN performance.
What are your ideas??
.
Last edited:
- Status
- Not open for further replies.