Questions about Crsytal Oscillators

[fmak]

Two questions.

What is the adjustment in a TCXO for? Is it the temp aspect or the mark space ratio? I have some with very poor symmetr(about 40% only)

How important is the mark space ratio in a digital clock application.

[Cybergent]

The adjustment is for frequency fine adjustment.

The duty cycle of a digital clock normally does not affect following stages as the clock in most cases will be devided anyways, what puts the duty cycle to 50%.

But a non 50% ratio results in more aggresive harmonics which might be a problem with FCC/CE compliance.

[JohnW]

While its true that most IC’s divide the Master clock by at least 2 internally (which results in an internal 50% M:S ratio) some devices, notably modulators for digital amplifiers don’t.

In addition, if the M:S ratio is very poor, some devices could malfunction due to the narrow clock pulse width, most spec 60:40 / 40:60 as a Max.

Unless you’re building an ADC for archiving purposes, absolute frequency accuracy is not so important. However, what is more important is short term Phase Noise (Jitter).

Like for like - TCXO are internally more complicated – and as a result, normally have worst short-term phase noise then their simpler XO cousins.

I’ve seen TCXO’s that use a combination of capacitor dielectrics to achieve temperature compensation – apart from the extra (paralleled) capacitors, these circuits are identical to XO – and therefore, so should their short-term phase noise. If you’re TCXO’s are more fancy, and uses a Varicap diode and control leap – then you’re almost guaranteed worse Phase Noise – from experience, I would stick with standard XO’s.

[fmak]

Unless you’re building an ADC for archiving purposes, absolute frequency accuracy is not so important. However, what is more important is short term Phase Noise (Jitter).
-------------------------------------------------------------------------------
Trouble is, most manufacturers do not have jitter and phase noise figures for ordinary XOs. One large one doesn't even know there is such a problem.

--------------------------------------------------------------------------------


If you’re TCXO’s are more fancy, and uses a Varicap diode and control leap – then you’re almost guaranteed worse Phase Noise – from experience, I would stick with standard XO’s. [/B][/QUOTE]
--------------------------------------------------------------------------------
I am toying with the idea of trying S/C cut OCXOs such as those by Euroquartz with quoted phase noise. Have you tried these, or similar ones?

[JohnW]

I’ve been spending a large amount of time researching Clocks recently. It surprising how bad off-the-shelf units are with close-in phase noise.

At the end of the day, the fundamental requirements for low close-in noise are Large Q, Low noise transistors and low value resistors to limit their noise contribution. These are not expensive items, so I don’t understand why low noise oscillators fetch such a premium price. Its possible to reduce the noise contribution of the surrounding oscillator components to a point where the noise within the crystal becomes the dominant noise source. Only then is it worthwhile going for the lower noise SC cut crystal.

Once I have such a circuit operating – I will post it on this site.

OCXO SC cut crystals are seriously expensive, apparently they use double twisted (sounds like my Ex) crystals, these are then cut at the oven operating temperature – typically 85Deg C. all not cheap! It is possible to source low noise AT crystal – still trying to find out more.

What Frequency do you need?

[jwb]

If the manufacturer doesn't specify jitter, don't buy. Valpey-Fischer does specify jitter for all their crystal oscillators. TCXO really is weird for audio, where we really don't care about long-term stability nor absolute frequency control. If you were launching a GPS satellite, it might be important, but in your CD player it is irrelevant.

[fmak]

Only then is it worthwhile going for the lower noise SC cut crystal.

Once I have such a circuit operating – I will post it on this site.

OCXO SC cut crystals are seriously expensive, apparently they use double twisted (sounds like my Ex) crystals, these are then cut at the oven operating temperature – typically 85Deg C. all not cheap! It is possible to source low noise AT crystal – still trying to find out more.

What Frequency do you need? [/B][/QUOTE]
---------------------------------------------------------------------------------

24.576 aand 45.1584 MHz.

There is another aspect. The combination of XO and Low noise PSs generate different levels and spectra of noise in the supply line. So, what is the real perpose of fitting a low noise PS unless that is developed with the XO?

I fine different sonics with different low noise supplies, Tent , Snoeren, KKHo, LT 176x and my own, nominally 10-15 nV/sqrt Hz over a 1 MHz bandwidth driving resistors. Nothing like this when in situ.
--------------------------------------------------------------------------------


Valpey-Fischer does specify jitter for all their crystal oscillators.
--------------------------------------------------------------------------------

From memory, not for TTL XOs; I wasn't very impressed by what I can get.

[JohnW]

Are you sure about your clock frequencies?

48KHz x 512 = 24.576MHz
44.1KHz x 1024 = 45,1584MHz

Seems and odd combination to have in one box, or is it for two different designs?

Anyway, your correct in stating that the PSU must be an integral part of the clock design, there’s no point in designing a Low Phase noise Oscillator then hanging it off the end of a 7805!

One of the better (surprisingly) XO circuits for short-term stability is the good old Pierce Oscillator built with discrete transistors. Don’t attempt to build a Pierce Oscillator with CMOS / TTL logic.

As CMOS / TTL changes Logic State, a large transient current is drawn by the input stage. These non-linear, low resistance flat spots in occur at the worst possible place in the cycle – at the switching point. This increases the timing uncertainty, as when to switch from one binary state the other, and thereby degrade the phase noise performance.

For more information read Robert J. Matthys excellent book “CRYSTAL OSCILLATOR CIRCUITS” ISBN 0-89464-552-8 from Krieger Publishing Company.

See my reply under posting “49.152MHz Clock” for a typical Pierce 3rd overtone circuit.

[fmak]

[QUOTE]Originally posted by JohnW
[B]Are you sure about your clock frequencies?

48KHz x 512 = 24.576MHz
44.1KHz x 1024 = 45,1584MHz

Seems and odd combination to have in one box, or is it for two different designs?

---------------------------------------------------------------------------
Not in one bax. 24/96 and DSD units.

The PS plays an important part in the sonics department. The coupling resistor between XO and the processors also plays an important part in controlling the clock signal waveform. Optimising this seems to improve the sound as well!