The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator

[bson]

Are you looking at 150k for R5 on the TWTMC-C? According to both schematics I have as well as the BOM R5 is listed as 180k. 220k/180k gives me a 53% duty cycle according to my Agilent 53131A (referenced off a Trimble Thunderbolt, although it does have the opt 010 ocxo). 11.2896MHz, 15V supply, 3.3V LVC04 output, single U.FL socket in the corner position. I haven't gone back the last 20 pages or so, so if you're talking about something else just ignore me. 🙂

BTW, this thing (TWTMC-C) is super stable; drift is about 0.05Hz over 30-60s, a slow steady wander. No other short term drift that I can see (basically the counter ticks at 0.01Hz (~1ppt) up and down, likely only a display artifact). Tempco is a much bigger factor (as is microphonics). Warming it or making noise at it makes it jump around up to 10-20Hz (~1-2ppm). Which in absolute terms for audio use is practically an ideal XO. Absolute accuracy for that particular 11.2896MHz xtal is only about 200ppm (0.02%); it's centered at 11287375.30Hz.

 

[TNT]

bson, can you do also spectral analysis? Would be interesting to se the impact from different "noise"... even loud music.

//

 

[gentlevoice]

@andrea, gerhard, 1audio or bson? ...

I have a possibly quick question: The CLK input gates on e.g. D/A or A/D converters - are they significantly different from the inputs of an inverter like the 74HC04? My question really is if the first part of the TWTMC couldn't just a well be connected directly to the D/A CLK gate and then the last part (the inverters) could be saved if the TWTMC is close to the converter. I have a vague recollection that I've seen this elsewhere ...

Might there be any comments/insights on this?

Cheers,

Jesper

 

[bson]

TNT, depends on what you mean by spectrum - the spectrum of the shaped output, the oscillator itself, or the phase/frequency noise. I guess they're all related. I can easily get the output spectrum using my VNA. The oscillator itself, that would be harder and would probably require soldering a 10M/100M/1G resistor to the board (tombstone) and a compensation network on the input (which will need some measure of calibration against a well-known reference).

The counter has a default gating time of about 1/2s, which means I can't observe variation shorter than this. Going to a shorter gating time gives up digits of precision and I wouldn't really expect this to show anything useful. The default/auto setting seems to be the sweet spot where drift begins to show...

Phase noise is really not easy to measure.

 

[bson]

@andrea, gerhard, 1audio or bson? ...

I have a possibly quick question: The CLK input gates on e.g. D/A or A/D converters - are they significantly different from the inputs of an inverter like the 74HC04? My question really is if the first part of the TWTMC couldn't just a well be connected directly to the D/A CLK gate and then the last part (the inverters) could be saved if the TWTMC is close to the converter. I have a vague recollection that I've seen this elsewhere ...

You could, especially if the inputs are CMOS, however you don't want to distribute the XO directly over coax. The cable will load it. Even when on the same board and in close proximity to a CMOS clock input it might still be worth at least prototyping with the inverter so you can easily probe it. Then when it works reliably, try jumpering the inverter pads of an unpopulated footprint; if that works consider a board run with the inverter removed.

 

[1audio]

You need some good isolation even with CMOS. Any change in loading (and the internals of a DAC change a lot reflecting back to the inputs) will affect the oscillator. You will get lower phase noise with a good buffer. A 74AC buffer will give excellent results with added phase noise that will be unmeasureable in this context.

If the clock is any good measuring with anything less than specialized intruments will show nothing useful. I put together this note on using an FM tuner for measring jitter/phase noise http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html Its not real easy but its probably the most cost effective solution. I never could calibrate it but I'm confident that you can spot a problem really quickly with this trick. With two tuners and correlation software you could probably get pretty decent results.

 

[gentlevoice]

@bson:

You could, especially if the inputs are CMOS, however you don't want to distribute the XO directly over coax. The cable will load it. Even when on the same board and in close proximity to a CMOS clock input it might still be worth at least prototyping with the inverter so you can easily probe it. Then when it works reliably, try jumpering the inverter pads of an unpopulated footprint; if that works consider a board run with the inverter removed.

Thanks for your comments & ideas. Also seeing 1audio's input I'll just consider what may be the most feasible way to progress.

@1audio: Hmmm ... I'm somewhat amazed at how many aspects of audio you seem to have touched upon over time ... 🙂 ... Thanks again for your input, Demian.

I read the blog post you linked to and if I understand it correctly it could be a quite accessible way to get an impression of what a jitter spectrum may look like. While reading it a question popped up, though, which is related to how useful it is in practice ...? If calibration is not really possible will it then be usable e.g. in relation to the TWTMC which already (most likely) is very low jitter ....

Cheers to you both ;-)

Jesper

 

[gentlevoice]

A brief follow-up to my comment in post #782:

4. Using the 150k & 220k resistor options for configurable duty cycle I get about 25%/75% duty cycle from visually reading the oscilloscope ... (C6 is 100 nF//680uF; PSU Cap values higher - shouldn't make a difference here...?)

Replacing the 680 uF with a 1 uF returns duty cycle to appr. 53 - 54%. So it is as it should be. I suppose the leakage current in the 680uF is higher and skews the DC point of the resistor voltage divider.

Hope this caused no inconvenience :-L

Jesper

 

[1audio]

RE calibrating the FM tuner for phase noise. The challenge is that the tuner output is relative to deviation and phase noise is measured as amplitude wrt to the carrier. There are several hoops to jump through to get levels and then translate them into phase noise and then jitter. I tried to figure out the relationships and how to check them but I could not get something that shows the actual relationships in a way I could work with. the ratios were always too big with too many steps to trust. I'm sure someone on the forums could do it and quickly, just not me.

 

[HpW]

You need some good isolation even with CMOS. Any change in loading (and the internals of a DAC change a lot reflecting back to the inputs) will affect the oscillator. You will get lower phase noise with a good buffer. A 74AC buffer will give excellent results with added phase noise that will be unmeasureable in this context.

The other (different) old engineering solution would be to isolate the oscillator from any long ground loops using:

- ECL based oscillator (negative power) with symmetric output drivers

- feed the signal using a symmetric twisted pair (transmission line)

- convert the it back to CMOS level using a logic or transformer

just my

hp

 

[Alexiss]

Hi Alex,

you are right, the drive level of the Clapp oscillator is strong, but not so high to damage the crystal, at least for frequencies up to 49 MHz, where the crystal ESR is 5 to 13 ohm. The Laptech AT-cut crystals are specified for 500 uW dissipation. There could be a problem wit 90/98 MHz crystal, where the ESR is greater than 30 ohm, but for these frequencies, probably, I will move to Driscoll circuit. Obviously, high drive level influences the aging of the crystal, but in audio devices we don't care about long term stability. Moreover, typically, higher drive level means lower phase noise.
At 11 MHz, considering the crystal ESR around 8 ohm, the RMS current flowing in the crystal should be around 7 mA, and so the dissipation less than 500 uW. At 22 MHz the crystal ESR is around 5 ohm, so the current is around 8 mA and the dissipation around 350 uW. At 45 MHz, where the crystal ESR is around 14 ohm, the current will be less than 7 mA, and the dissipation around 600 uW. At 90 MHz the ESR of the crystal is around 40 ohm, so the RMS current is around 7 mA, and the dissipation around 2 mW, that is a very high drive level.

The Driscoll oscillator drive level will be much lower, since the crystal is located at the lowest power point in the circuit (emitter circuit). The power dissipated in a crystal with ESR around 40 ohm will be less than 300 uW.
Work in progress around the Driscoll oscillator ...

Andrea

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Hello Andrea and I hope everything is fine with you.
Thank u for your quick response.
I am very busy with work and unfortunately I don't get to check your very interesting thread as often as I'd like to :-(

Anyways:
A quick check revealed an I-peak DL of close to 15mAac (30mAac pp) through the resonator in your Clapp oscillator.
As you correctly pointed out, the higher DL will lead to improved phase noise figures, but this will be short-lived and ill-advised.

Remember that the maximum allowable dissipation in the resonator-element is a function of the oscillation frequency. The higher the f, the lower the DL.

As you know, that is (one of the reasons) why all precision oscillators have ALCs. The Gate diode is a bad solution as it seriously degrades the In-circuit-Q and thus should be avoided if low "jitter" is a primary design objective.

I just feel that your efforts and the Laptech resonator are too valuable to be compromised by petty details.

At the risk of sounding a tad critical, I feel the DL and the I(ds) are set too high. Maybe a revision is in order? ;-)

Best regards: alexiss

 

[Alexiss]

1audio comments in the previous posts

Going quickly through the ever-growing number of posts in this wonderfull thread, I did notice that 1audio has pointed this out (drive level, remedies, and associated evils) several times in the previous posts.

How come no-one is taking notice??😕

...Kudos to 1audio BTW.
cheers...

 

[gentlevoice]

@alexiss:

Going quickly through the ever-growing number of posts in this wonderfull thread, I did notice that 1audio has pointed this out (drive level, remedies, and associated evils) several times in the previous posts.

How come no-one is taking notice??

Well, not that I may be the most obvious one to reply to this but for my part I can say that it is outside of my area of knowledge and so I wouldn't have much to add in this context ...

And also not that I want to imply that "because you have noted this you may also have some answers to this" but if you have I'd be positively interested in hearing them ...

1audio:

RE calibrating the FM tuner for phase noise. The challenge is that the tuner output is relative to deviation and phase noise is measured as amplitude wrt to the carrier. There are several hoops to jump through to get levels and then translate them into phase noise and then jitter. I tried to figure out the relationships and how to check them but I could not get something that shows the actual relationships in a way I could work with. the ratios were always too big with too many steps to trust. I'm sure someone on the forums could do it and quickly, just not me.

Hi again ... My impression is that at least currently this is beyond me so I will leave it here for now - but have it in mind should I need it onwards.

Cheers to you both,

Jesper

 

[gentlevoice]

Hi ... just in case it happened that some (more) people gain interest in a 6.144 MHz crystal I've added myself to the list.

Jesper

Interest List SC-cut Crystals from Laptech
- * andrea_mori: 3 x 11.2896 MHz + 2 x 22.5792 MHz + 2 x 24.576 MHz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- Acko: 1 x 45.1584 MHz + 1 x 49.152MHz
- iancanada: 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- Clsidxxl: 1 x 45.1584 MHz + 1 x 49.152MHz
- lindamar: 1 x 11.2896 MHz 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- * tagheuer: 1 x 45.1584 MHz + 1 x 49.152MHz
- palmito: 1 x 45.1584 MHz + 1 x 49.152MHz
- damohpi: 2x 11.2896 Mhz
- Naimster: 1 x 22.5792 MHz + 1 x 24.576 MHz
- Phi: 1 x 22.5792 MHz + 1 x 24.576 MHz (Added title )
- ambrosia168: 3 x 11.2896 MHz + 3 x 45.1584 Mhz + 2 x 49.152 MHz
- * flowerpot: 1 x 45.1584 MHz + 1 x 49.152MHz + 2 x PCB + 1 x daughter board + 2 x DIL
- jdlvfr: 3 x 11.2896 Mhz
- tods: 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- wlowes: 1 x 45.1584 MHz + 1 x 49.152MHz
- * fralippo: 1 x 45.1584 MHz + 1 x 49.152MHz + PCBs
- * duster1: 1 x 45.1584 MHz + 1 x 49.152MHz + 1 x 90.3168 MHz + 1 x 98.304MHz + PCBs
- deanoUK: 1 x 45.1584 MHz + 1 x 49.152MHz
- * BDL: 1 x 22.5792 MHz
- * vita : 1 x 45.1584 Mhz + 1 x 49.152 Mhz
- Malvin: 1 x 45.1584 MHz + 1 x 49.152MHz
- rlim: 1 x 22.5792MHz + 1 x 24.576MHz
- Canvas: 1 x 11.2896 MHz + 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- * mcluxun: 1 x 45.1584 Mhz + 1 x 49.152 MHz
- grzegrzol: 2x 11.2896 2 x 22.5792 MHz + 2 x 24.576 MHz + 2 x 45.1584 MHz + 2 x 49.152 MHz
- * jims: 1 x 11.2696 Mhz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- * madkid 1 x 45.1584 Mhz + 1 x 49.152 Mhz
- * tbrowne 2 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152 MHz + 1 x 90.3168 MHz
- * jborden: 2 x 45.1584 Mhz + 2 x 49.152 Mhz
- * Stijn001: 1 x 45.1584 MHz, + 1 x 49.152MHz

Second batch AT-cut Crystals from Laptech
- * andrea_mori: 2 x 6.1440 MHz + 2 x 12.2880 MHz + 1 x 16.9344 MHz
- Miklos: 1 x 16.9344 MHz
- Marlowe: 3 x 16.9344Mhz + 3 x 25MHz
- carz 1x 27MHz
- 1audio : 1X6.1440MHz
- * mravinsky : 2 x 16.9344Mhz + (2 x TWCMC-C XO board +1 x daughter board) if possible
- * noizas : 2 x 16.9344 MHz
- randytsuch: 2x 25.0000 MHz
- casshan: 1 x 11.2896 MHz + 1 x 12.288 MHz + 2 x PCB + 1 x daughter board
- Fabian85: 1x 16.9344 MHz, 2x PCB + 1x daughter board
- * duster1: 2 x 16.9344 MHz, 1 x 45.1584 MHz, 2 x PCB + 2 x daughter board
- * RollE2k: 1 x 45.1584 MHz + 1 x 49.152MHz + 2x PCB + 2x daughter board
- * shendrik: 2 x 22.5792 MHz + 2 x 24.576 MHz
- babaudio : 4 x 22.5792 MHz + 4 x 24.576 MHz
- Justubes : 1 X 25mhz
- grzegrzol : 2 x 16.9344 MHz + 2 x 22.5792 MHz + 2 x 24.574 MHz + 2 x 25 MHz 2 x 45.1584 MHz + 2 x 49.152MHz
- * jborden: 2 x 25.000 Mhz + 2 x PCB
- smanz:1 x 45.1584 MHz, + 1 x 49.152MHz, 1x TWTMC-D&D, + 2x TWTMC-C
- * jims: 1 x 6.1440 Mhz + 2 x 16.9344 MHz + 6x TWTMC-C +3x TWTMC-D&D + 6 x TWTMC-DIL
- * madkid 1 x 90.3168 MHz + 1 x 98.304MHz + 4 x PCBs
- * tbrowne 5 x 22.5792 MHz + 3 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152 MHz
- gentlevoice: 1 * 6.144 MHz

 

[Gunders]

Added 5 * 24.576 MHz AT-cut + 2*TWTMC-C
------------------------------------------

Interest List SC-cut Crystals from Laptech
- * andrea_mori: 3 x 11.2896 MHz + 2 x 22.5792 MHz + 2 x 24.576 MHz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- Acko: 1 x 45.1584 MHz + 1 x 49.152MHz
- iancanada: 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- Clsidxxl: 1 x 45.1584 MHz + 1 x 49.152MHz
- lindamar: 1 x 11.2896 MHz 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- * tagheuer: 1 x 45.1584 MHz + 1 x 49.152MHz
- palmito: 1 x 45.1584 MHz + 1 x 49.152MHz
- damohpi: 2x 11.2896 Mhz
- Naimster: 1 x 22.5792 MHz + 1 x 24.576 MHz
- Phi: 1 x 22.5792 MHz + 1 x 24.576 MHz (Added title )
- ambrosia168: 3 x 11.2896 MHz + 3 x 45.1584 Mhz + 2 x 49.152 MHz
- * flowerpot: 1 x 45.1584 MHz + 1 x 49.152MHz + 2 x PCB + 1 x daughter board + 2 x DIL
- jdlvfr: 3 x 11.2896 Mhz
- tods: 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- wlowes: 1 x 45.1584 MHz + 1 x 49.152MHz
- * fralippo: 1 x 45.1584 MHz + 1 x 49.152MHz + PCBs
- * duster1: 1 x 45.1584 MHz + 1 x 49.152MHz + 1 x 90.3168 MHz + 1 x 98.304MHz + PCBs
- deanoUK: 1 x 45.1584 MHz + 1 x 49.152MHz
- * BDL: 1 x 22.5792 MHz
- * vita : 1 x 45.1584 Mhz + 1 x 49.152 Mhz
- Malvin: 1 x 45.1584 MHz + 1 x 49.152MHz
- rlim: 1 x 22.5792MHz + 1 x 24.576MHz
- Canvas: 1 x 11.2896 MHz + 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- * mcluxun: 1 x 45.1584 Mhz + 1 x 49.152 MHz
- grzegrzol: 2x 11.2896 2 x 22.5792 MHz + 2 x 24.576 MHz + 2 x 45.1584 MHz + 2 x 49.152 MHz
- * jims: 1 x 11.2696 Mhz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- * madkid 1 x 45.1584 Mhz + 1 x 49.152 Mhz
- * tbrowne 2 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152 MHz + 1 x 90.3168 MHz
- * jborden: 2 x 45.1584 Mhz + 2 x 49.152 Mhz
- * Stijn001: 1 x 45.1584 MHz, + 1 x 49.152MHz

Second batch AT-cut Crystals from Laptech
- * andrea_mori: 2 x 6.1440 MHz + 2 x 12.2880 MHz + 1 x 16.9344 MHz
- Miklos: 1 x 16.9344 MHz
- Marlowe: 3 x 16.9344Mhz + 3 x 25MHz
- carz 1x 27MHz
- 1audio : 1X6.1440MHz
- * mravinsky : 2 x 16.9344Mhz + (2 x TWCMC-C XO board +1 x daughter board) if possible
- * noizas : 2 x 16.9344 MHz
- randytsuch: 2x 25.0000 MHz
- casshan: 1 x 11.2896 MHz + 1 x 12.288 MHz + 2 x PCB + 1 x daughter board
- Fabian85: 1x 16.9344 MHz, 2x PCB + 1x daughter board
- * duster1: 2 x 16.9344 MHz, 1 x 45.1584 MHz, 2 x PCB + 2 x daughter board
- * RollE2k: 1 x 45.1584 MHz + 1 x 49.152MHz + 2x PCB + 2x daughter board
- * shendrik: 2 x 22.5792 MHz + 2 x 24.576 MHz
- babaudio : 4 x 22.5792 MHz + 4 x 24.576 MHz
- Justubes : 1 X 25mhz
- grzegrzol : 2 x 16.9344 MHz + 2 x 22.5792 MHz + 2 x 24.574 MHz + 2 x 25 MHz 2 x 45.1584 MHz + 2 x 49.152MHz
- * jborden: 2 x 25.000 Mhz + 2 x PCB
- smanz:1 x 45.1584 MHz, + 1 x 49.152MHz, 1x TWTMC-D&D, + 2x TWTMC-C
- * jims: 1 x 6.1440 Mhz + 2 x 16.9344 MHz + 6x TWTMC-C +3x TWTMC-D&D + 6 x TWTMC-DIL
- * madkid 1 x 90.3168 MHz + 1 x 98.304MHz + 4 x PCBs
- * tbrowne 5 x 22.5792 MHz + 3 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152 MHz
- gentlevoice: 1 * 6.144 MHz
- Gunders: 5 * 24.576 MHz + 2*TWTMC-C

 

[Alexiss]

Adding 2 * 11.2896MHz AT-cut + 2 * TWTMC board

Hello Andrea.
I noticed that my orders are not included in the list. The order-from has been sent you previously.
Anyways, hook me up for 2 x 11.2896 MHz AT-cut + 2 TWTMC pcb boards.
Thanks
------------------------------------------

Interest List SC-cut Crystals from Laptech
- * andrea_mori: 3 x 11.2896 MHz + 2 x 22.5792 MHz + 2 x 24.576 MHz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- Acko: 1 x 45.1584 MHz + 1 x 49.152MHz
- iancanada: 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- Clsidxxl: 1 x 45.1584 MHz + 1 x 49.152MHz
- lindamar: 1 x 11.2896 MHz 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- * tagheuer: 1 x 45.1584 MHz + 1 x 49.152MHz
- palmito: 1 x 45.1584 MHz + 1 x 49.152MHz
- damohpi: 2x 11.2896 Mhz
- Naimster: 1 x 22.5792 MHz + 1 x 24.576 MHz
- Phi: 1 x 22.5792 MHz + 1 x 24.576 MHz (Added title )
- ambrosia168: 3 x 11.2896 MHz + 3 x 45.1584 Mhz + 2 x 49.152 MHz
- * flowerpot: 1 x 45.1584 MHz + 1 x 49.152MHz + 2 x PCB + 1 x daughter board + 2 x DIL
- jdlvfr: 3 x 11.2896 Mhz
- tods: 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152MHz
- wlowes: 1 x 45.1584 MHz + 1 x 49.152MHz
- * fralippo: 1 x 45.1584 MHz + 1 x 49.152MHz + PCBs
- * duster1: 1 x 45.1584 MHz + 1 x 49.152MHz + 1 x 90.3168 MHz + 1 x 98.304MHz + PCBs
- deanoUK: 1 x 45.1584 MHz + 1 x 49.152MHz
- * BDL: 1 x 22.5792 MHz
- * vita : 1 x 45.1584 Mhz + 1 x 49.152 Mhz
- Malvin: 1 x 45.1584 MHz + 1 x 49.152MHz
- rlim: 1 x 22.5792MHz + 1 x 24.576MHz
- Canvas: 1 x 11.2896 MHz + 1 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- * mcluxun: 1 x 45.1584 Mhz + 1 x 49.152 MHz
- grzegrzol: 2x 11.2896 2 x 22.5792 MHz + 2 x 24.576 MHz + 2 x 45.1584 MHz + 2 x 49.152 MHz
- * jims: 1 x 11.2696 Mhz + 1 x 45.1584 Mhz + 1 x 49.152 MHz
- * madkid 1 x 45.1584 Mhz + 1 x 49.152 Mhz
- * tbrowne 2 x 22.5792 MHz + 1 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152 MHz + 1 x 90.3168 MHz
- * jborden: 2 x 45.1584 Mhz + 2 x 49.152 Mhz
- * Stijn001: 1 x 45.1584 MHz, + 1 x 49.152MHz

Second batch AT-cut Crystals from Laptech
- * andrea_mori: 2 x 6.1440 MHz + 2 x 12.2880 MHz + 1 x 16.9344 MHz
- Miklos: 1 x 16.9344 MHz
- Marlowe: 3 x 16.9344Mhz + 3 x 25MHz
- carz 1x 27MHz
- 1audio : 1X6.1440MHz
- * mravinsky : 2 x 16.9344Mhz + (2 x TWCMC-C XO board +1 x daughter board) if possible
- * noizas : 2 x 16.9344 MHz
- randytsuch: 2x 25.0000 MHz
- casshan: 1 x 11.2896 MHz + 1 x 12.288 MHz + 2 x PCB + 1 x daughter board
- Fabian85: 1x 16.9344 MHz, 2x PCB + 1x daughter board
- * duster1: 2 x 16.9344 MHz, 1 x 45.1584 MHz, 2 x PCB + 2 x daughter board
- * RollE2k: 1 x 45.1584 MHz + 1 x 49.152MHz + 2x PCB + 2x daughter board
- * shendrik: 2 x 22.5792 MHz + 2 x 24.576 MHz
- babaudio : 4 x 22.5792 MHz + 4 x 24.576 MHz
- Justubes : 1 X 25mhz
- grzegrzol : 2 x 16.9344 MHz + 2 x 22.5792 MHz + 2 x 24.574 MHz + 2 x 25 MHz 2 x 45.1584 MHz + 2 x 49.152MHz
- * jborden: 2 x 25.000 Mhz + 2 x PCB
- smanz:1 x 45.1584 MHz, + 1 x 49.152MHz, 1x TWTMC-D&D, + 2x TWTMC-C
- * jims: 1 x 6.1440 Mhz + 2 x 16.9344 MHz + 6x TWTMC-C +3x TWTMC-D&D + 6 x TWTMC-DIL
- * madkid 1 x 90.3168 MHz + 1 x 98.304MHz + 4 x PCBs
- * tbrowne 5 x 22.5792 MHz + 3 x 24.576 MHz + 1 x 45.1584 MHz + 1 x 49.152 MHz
- gentlevoice: 1 * 6.144 MHz
- Gunders: 5 * 24.576 MHz + 2*TWTMC-C
- alexiss: 2 x 11.2896 MHz + 2 x TWTMC PCB boards

 

[Alexiss]

@alexiss:



Well, not that I may be the most obvious one to reply to this but for my part I can say that it is outside of my area of knowledge and so I wouldn't have much to add in this context ...

And also not that I want to imply that "because you have noted this you may also have some answers to this" but if you have I'd be positively interested in hearing them ...

Jesper

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Hej Jesper og takk for the vote of confidence.

All commercially available Master Clocks have one thing in common:
They over-promise and under-deliver, and none provide measurements (except for Tent-clock at 33 MHz)... It is VERY difiicult to find a good Master Clock for audio applications these days.

For this reason, Andreas efforts are much appreciated.

It is not my intention to be a "smart-***" and provide a list of improvements to Andreas design. That is why I asked him to revise the design (I'm not sure he's got the time though).

If you look carefully, I did mention something in the previous post.

Without digging too deep, there are many ways to lower the DL, some very simple and some a bit more complex.
But any modification (even good ones) to the present circuit will result in - you guessed it: altered phase noise figures.

Therefore it should be done by the designer and verified through measurement. So you see, it is not as simple as saying: remove this or change that....
Andrea has provided measurements for the present TWTMC design, so please stick with it and make no changes (without consulting the designer or verifying it through measurement).

I believe he is working on a simplified Driscoll design that will be interesting.

The original Driscoll has the highest short-term stability of harmonic-type designs that I've encountered, and I think it will accommodate even AT resonators (with some minor component value changes).
This design is also more forgiving about DLs and so forth, so lets wait and see...


cheers...

 

[Alexiss]

You need some good isolation even with CMOS. Any change in loading (and the internals of a DAC change a lot reflecting back to the inputs) will affect the oscillator. You will get lower phase noise with a good buffer. A 74AC buffer will give excellent results with added phase noise that will be unmeasureable in this context.

This is a valid statement, emphasizing the necessity of using the so-called "isolation amplifiers" in precision oscillators.
Specially for those who crave ultra low "jitter" (I seriously dislike that word).

cheers...

 

[1audio]

The audiophile community seems to have an outsized focus on jitter as though better will always sound better. That there is a hard floor to what lower jitter will do for the analog output seems to be a concept that is not understandable. Especially the reduced impact of close in phase noise which has a better analog to wow on a turntable. Not to mention even the worst crystal oscillator will be orders of magnitude better than the best turntable. This project started as a test of the sensitivity to close in phase noise and I don't think any real tests of that have been done yet. I would love to be proven wrong and that close in phase noise is significant since it highlights an achievable performance goal but the numbers make that seem really unlikely.

 

[bson]

One of the experiments I want to perform is exactly evaluating the audibility of clock phase noise. My plan is to do this by introducing controlled levels of white rail noise on the buffer(s) (LVC04) by mixing a function generator's white noise output on top of the 3.3V DC supply. For CMOS adding random noise to it is equivalent to adding white noise to the threshold voltage (since it's 50% of supply), which then introduces phase and amplitude noise in the output signal. Since the output noise is also white it's 50/50 amplitude and phase. So 1% white noise on the rail should result in 0.5% phase noise, at least as an analytical starting point. (Then there's the buffer's noise - FGkBT where the gain G is high since logic gates are effectively comparators though with hysteresis, and the bandwidth B is quite high for the LVC logic family.) Still looking for ways to actually measure phase noise; the spectrum analyzer method looks like the best bet (without spending crazy money), especially using one that can compensate for its own noise factor.

Actually, come to think of it a FG can act as the supply voltage for an LVC04, so maybe 10mV white noise on top of a 3.3V offset will do the job without involving a DC supply per se.