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

Status
Not open for further replies.
If you are going to design a new "low phase noise" clock here, would something like the
LMK04906 Ultra Low Noise Clock Jitter Cleaner be useful ??

http://www.ti.com/product/LMK04906

Its interesting but the datasheet is so long I would never get it running. . .

Actually its noise floor is too high fo these applications. the corner frequency is around 1 MHz above which is pretty good. We usually see that at 1 KHz and in some cases 100 Hz with the discrete crystal oscillators. And they use parts humans can solder. For its intended application its very good.
 
Hi all,

Although this is not specifically related to me (Phil does not assemble the boards for me) I notice that the assembly service appears to have been stalled for some time now.

Also, that Phil (korben69) seems to not have made any posts in diyaudio forums since 30th April. Prior to this time he posted "regularly".

(FYI: Link to statistics: diyAudio - Search Results )

I'm thinking if something might have happened to Phil? And to that end if it may be feasible to contact with other of the French forum members that live in his area to find out if they may somehow help clarify if this is the case?

Again, this is not particularly related to me - and so I will not drive this - yet the thought has popped up in my mind so was thinking that I would mention it.

Best wishes,

Jesper
 
Hi All,

I'm now at the point where I'm assembling the oscillator and while in this process a couple of questions have come up which I hope one of you may also have considered/can help with ...

- Calculating the DC power dissipation of Q1 it looks as if it could be past its rated dissipation which is 350 mW at 25 degr. C. 20 mA into 14.2 volts is 284 mWs. And the transistor is to be derated with 2.8 mW/C above 25 degr C ... My guess is that it'll be quite hot (70 - 80 degr C ??) if it is to dissipate 284 mWs from this size which, if it becomes this warm, means that the PD is down to 210 mWs. Which is beyond the specified limit ... Anyone have experience with this/tried to make a similar calculation? To this end: Could it be feasible to reduce the supply voltage or replace the J310 SST version with a TO92 version?

- If connecting the oscillator directly to an ADC/DAC is it then possible to do without the components from & including the yellow marking named (7) - possibly also omitting the 10 pf C5?

- If omitting R9, i.e. using the (7) marking components: is there a reason to use U1A & U1B if only U1D is to be used?

Cheers,

Jesper
 
Hi again,

... a brief add-on comment: I would like to keep both the oscillators (44.1 & 48 kHz base) switched on permanently and to that end need a superb switch (the "best" whatever that may be). Here I'm considering the potato semi PO3B3305A, a "High-Bandwidth, Hot-Insertion, 2-Bit, 2-Port Bus Switch with Individual Enables" ..

http://www.potatosemi.com/ (and then half-way down the webpage)

... but am thinking that maybe others here have experience with other ICs?

Cheers,

Jesper
 
I'm not sure which schematic you are using. Perhaps you can supply a link so we are talking about the same circuit.

Q1 power dissipation needs to be done in a simulator since it may not be operating in a linear region. Also the cap values need to be tweaked to manage the power in the crystal. Too much and they can break.

The output voltage of the oscillator can go negative so you do need to cap couple into the buffer gates. They don't offer a lot of isolation. Faster gates will have lower phase noise. The Potato logic stuff should be pretty good. You can either use some nand gates to switch as well as buffer or add a CMOS mux to switch between oscillators. Provide separate isolated supplies to each oscillator and the initial buffer/interface to keep the noise and any interactions down. the 74AC stuff works pretty well of the Potato stuff is too hard to get. 74AC can generate a lot of noise so be careful with layout and wiring.
 
@ 1audio: Hi - & thanks once more for very "sound" and useful advice 😉

I'm not sure which schematic you are using. Perhaps you can supply a link so we are talking about the same circuit.

For reference I've attached the document - with schematic - that I'm using for values etc. The schematic I use is on page 2 and frequencies are 22, 24, 45 & 49 MHz.

Q1 power dissipation needs to be done in a simulator since it may not be operating in a linear region.
... So I assume this means that even if the DC PD would be outside of the "safe area" of the J310 then due to non-linearities it may work anyway ... right? (Would you BTW use the J305 (as I believe zenelectro suggested) or would you stay with the J310?)

The output voltage of the oscillator can go negative so you do need to cap couple into the buffer gates.
Just out of curiosity - do you know how negative the voltages may become?

The Potato logic stuff should be pretty good. You can either use some nand gates to switch as well as buffer or add a CMOS mux to switch between oscillators. Provide separate isolated supplies to each oscillator and the initial buffer/interface to keep the noise and any interactions down.
.. I will be using the Potato semi ICs (NAND gates, thanks!) since - as I understand it - they emit less noise into the PSU rails. I may possibly buy another brand to see what the differences are.

To that end a slight digression: Some time ago you advised me regarding some way of measuring low noise levels on the PSU rails - I was making an add-on amplifier to my oscilloscope. At that point in time you suggested this component:

http://www.minicircuits.com/pdfs/RAM-8A+.pdf

which unfortunately was not easy to find for purchase in low amounts (some £80 for 20 pcs).

As it is the circuitry I made, which is based on two bipolar 1.3 GHz opamps, is quite noisy (as you assumed it could be ... ) and I am considering making a new and simpler design. However, it seems that the RAM-8A+ is still not easy to come by so if you might have an alternative suggestion (Mouser or Digikey available) I would appreciate this.

...of the Potato stuff is too hard to get.

... I am thinking that maybe you don't know of potsemi's online store - ebay based?


http://stores.ebay.com/potatosemi_W0QQsspagenameZMEQ3aFQ3aSTQQtZkm

Here it's possible to buy 1- ?? pieces of their various products. Not at prices that are production oriented but for (my) DIY purposes it may be ok ...

Thanks again for your feedback, Demian 🙂

Jesper
 

Attachments

Last edited:
Hi All,

I'm now at the point where I'm assembling the oscillator and while in this process a couple of questions have come up which I hope one of you may also have considered/can help with ...

- Calculating the DC power dissipation of Q1 it looks as if it could be past its rated dissipation which is 350 mW at 25 degr. C. 20 mA into 14.2 volts is 284 mWs. And the transistor is to be derated with 2.8 mW/C above 25 degr C ... My guess is that it'll be quite hot (70 - 80 degr C ??) if it is to dissipate 284 mWs from this size which, if it becomes this warm, means that the PD is down to 210 mWs. Which is beyond the specified limit ... Anyone have experience with this/tried to make a similar calculation? To this end: Could it be feasible to reduce the supply voltage or replace the J310 SST version with a TO92 version?

- If connecting the oscillator directly to an ADC/DAC is it then possible to do without the components from & including the yellow marking named (7) - possibly also omitting the 10 pf C5?

- If omitting R9, i.e. using the (7) marking components: is there a reason to use U1A & U1B if only U1D is to be used?

Cheers,

Jesper

Hi Jesper,

the Clapp oscillator operates in class "C", so the biasing of the jfet is not an issue. With the crystal inserted in the circuit, the quiescent current flowing in the device is almost zero. The jfet remains always cold.

As Demian said, capacitor C5 is needed to block the DC. I could estimate a negative peak of some volts.

The 74HC gates operate as the slicer, converting sine to square wave. If you omit R9 you should use components marked (7), R3-R4-R5-C6, to set the operating point of the gate to around 1/2 Vcc. If you need 50% duty cycle I suggest to use R9 only.

Potato gates are a good choice, you can easily buy 1 piece on ebay.

Andrea
 
For reference I've attached the document - with schematic - that I'm using for values etc. The schematic I use is on page 2 and frequencies are 22, 24, 45 & 49 MHz.
Thanks that helps everyone keep up.

... So I assume this means that even if the DC PD would be outside of the "safe area" of the J310 then due to non-linearities it may work anyway ... right? (Would you BTW use the J305 (as I believe zenelectro suggested) or would you stay with the J310?)
I would need to look at datasheets and prowl the internets. The goal is a low enough noise figure at the target frequency plus gain plus low noise at audio frequencies.

Just out of curiosity - do you know how negative the voltages may become?
Its very possible to get up to 10V P-P if the power is not managed. You would need to measure it. I'm hoping Andrea did his homework on the crystal current to keep everything happy.

.. I will be using the Potato semi ICs (NAND gates, thanks!) since - as I understand it - they emit less noise into the PSU rails. I may possibly buy another brand to see what the differences are.
Let us know. I have not seen any measurements of the phase noise contribution of those gates and it would be interesting. Needs a pile of expensive stuff to do it correctly.

To that end a slight digression: Some time ago you advised me regarding some way of measuring low noise levels on the PSU rails - I was making an add-on amplifier to my oscilloscope. At that point in time you suggested this component:

http://www.minicircuits.com/pdfs/RAM-8A+.pdf

which unfortunately was not easy to find for purchase in low amounts (some £80 for 20 pcs).

As it is the circuitry I made, which is based on two bipolar 1.3 GHz opamps, is quite noisy (as you assumed it could be ... ) and I am considering making a new and simpler design. However, it seems that the RAM-8A+ is still not easy to come by so if you might have an alternative suggestion (Mouser or Digikey available) I would appreciate this.

That would be one of the most difficult parts to work with. A thought I just had would be to look for a low noise preamp for TV antenna service. Not useful at low frequencies but would work at RF. I think I suggested a real spectrum analyzer would be needed to get useful info. Even really good digital scopes still have significant noise floors that a triple conversion spectrum analyzer can work around. we could get into a more involved discussion on PS noise and its impact/non-impact but that would be more useful as a separate thread.



... I am thinking that maybe you don't know of potsemi's online store - ebay based?


http://stores.ebay.com/potatosemi_W0QQsspagenameZMEQ3aFQ3aSTQQtZkm

Here it's possible to buy 1- ?? pieces of their various products. Not at prices that are production oriented but for (my) DIY purposes it may be ok ...

Thanks again for your feedback, Demian 🙂

Jesper
I am aware but have not needed to get any parts yet. Good luck and let us know how well it works.
 
@1audio:

Hi - & thanks for your extensive reply. Some comments & feedback from me:

Originally Posted by gentlevoice View Post
.. I will be using the Potato semi ICs (NAND gates, thanks!) since - as I understand it - they emit less noise into the PSU rails. I may possibly buy another brand to see what the differences are.

Let us know. I have not seen any measurements of the phase noise contribution of those gates and it would be interesting. Needs a pile of expensive stuff to do it correctly.

Unfortunately, I don't have any equipment to do phase noise measurements and so would have to find alternative ways of assessing the jitter/phase noise contribution or - if I can do without the clock when I first use/hear it - I could send it to someone here who can do these measurements (would have to be for a short while 🙄)

I think I suggested a real spectrum analyzer would be needed to get useful info.
... I just re-read your post and you did indeed say that:

The more constructive way to look at wideband noise (and find its sources) is a spectrum analyzer. I just discovered this: RF Explorer - 3G Combo [TES09102P] - $269.00 : Seeed Studio Bazaar, Boost ideas, extend the reach which I think would be a more useful tool.

As it is, however, I don't have such a spectral analyzer nor do I have access to one (would be attractive, though!) ... and money is a bit low at the moment ... What I do have is a Rigol scope and there's a built-in FFT in my AD7760 evaluation board's software that goes up to 1.25 MHz (Nyquist) ... Could it be that noise in the <1.25 MHz region may reflect audible phenomena?

we could get into a more involved discussion on PS noise and its impact/non-impact but that would be more useful as a separate thread.

Hmmm... this is something I've actually been thinking about for a long time and it could be interesting to exchange/learn more about this as the choice of decoupling caps appears to be not necessarily a simple one - apparently more philosophies to choose from. To this end I've started this thread:

http://www.diyaudio.com/forums/digi...ts-possible-design-solutions.html#post4459668

To me it could be very interesting to hear what diyaudio member's practical experiences are with digital decoupling.

@andrea mori: While at the topic of decoupling capacitors I noticed that in your BOM you suggest 1u decoupling capacitors with an X7R dielectric - the 10 pF coupling capacitor is also an X7R dielectric. I remember reading that Cyril Bateman some years ago - in his distortion measurements of various capacitor types - observed that X7R types have very high distortion levels.

So I've been wondering if there's some other reason why you've chosen X7R capacitor types (and not e.g. C0G for the 10 pF)?

Cheers to you both,

Jesper
 
I'm sorry for the situation, but it not depends on me.

I have just sent a PM to Jean-Paul asking if he has any news about Phil.
I hope he could contact Phil to clarify the situation.

Hi Andrea,

any news on that front ?. I have more than 200 euros worth of stuff oh Phil's hands ... I also understand you have your hands tied right now. Don´t you have a phone number to give him a call ? I will gladly do it

I dropped an email to Phil weeks ago to no response.

best regards
Pepe
 
Status
Not open for further replies.