Clock signal with no phase shift across long time

[ricko_uk]

Hi,
I need to generate a sinewave that I can control (preferably through a micro) between 5 and 60 Hz with zero phase shift across long time, at least a week.

I was thinking of using an IC like the AD9833 controlled by some high precision very stable crystal. What type of crystal/oscillator should I use? What parameters should I search for to ensure no phase shift across long time?


The unit always works at room temperature 15-27 degC.

SIDE NOTE... Granted that true "zero phase shift" might be impossible in practice, what is the best oscillator/solution that comes as close as possible to that within a reasonable budget (i.e. less than 50USD or so)?

Many thanks

[ricko_uk]

Any suggestions anybody?

What type of crystal oscillator should I look for? What parameters indicate the phase stability?

Thank you

[ejeffrey]

For this application I think you need a GPSDO if you can get an outdoor antenna or if you can count on AC power being available 100%, an oscillator disciplined to the 50 Hz AC.

If you want less than 1 radian phase drift at 60 Hz over 1 week that implies ~5 ppb frequency stability for a free running oscillator.  That is in OCXO / rubidium oscillator range which isn't likely to fit in your $50 budget unless you get lucky.  However, with a GPS PPS you can meet that long term stability with a relatively ordinary crystal.

[ricko_uk]

Thank you ejeffrey!

[T3sl4co1l]

Uh, phase shift with respect to what?

Tim

[ricko_uk]

Hi Tim,
phase shift in time. i.e. to itself if it was a perfect crystal. In other words if after a week a perfect atomic clock would have exactly 1 trillion oscillations (+/- 0 cycles) then this ideal oscillator I am looking for would have also the exact 1 trillion oscillations +/- o cycles and also 0 degrees shift.

Hope that makes sense?

[T3sl4co1l]

With its frequency adjustable?  How the devil would you know?  And if you did find out, just nudge it back?

Tim

[ricko_uk]

I meant that I adjust the frequency to whatever value once and that's it. And then it remains stable with no phase shift.

In practice that would be as little phase shift as possible (not zero), that's why I was asking if there are any solutions that is not some atomic clock

In practice, what are the best oscillators that achieve that (within maybe 40-50 USD) at least on a shorter time scale like a day or two and what parameter do I have to look at?

Thank you

[helius]

Oscillators work in two basic ways: they can be locked to a reference, or free running. A voltage-controlled oscillator (VCO) is free-running by default. If it is forced to lock to a reference signal, it will become a signal with a non-balanced duty factor. This causes inter-modulation distortion of the waveform, but at least the phase will stay locked to the reference.
To get balanced signals that stay phase locked, the oscillator needs to accumulate and cancel out phase errors. The phase-locked loop (PLL) does this for rational multiples like 1/2, 2/3, 4/5. To generate arbitrary frequencies, a counter and divider is needed. This is direct digital synthesis (DDS).

[T3sl4co1l]

Yeah, GPSDO or other accurate reference, feeding a DDS, is the way to go.

Tim

[BrianHG]

I meant that I adjust the frequency to whatever value once and that's it. And then it remains stable with no phase shift.

In practice that would be as little phase shift as possible (not zero), that's why I was asking if there are any solutions that is not some atomic clock

In practice, what are the best oscillators that achieve that (within maybe 40-50 USD) at least on a shorter time scale like a day or two and what parameter do I have to look at?

Thank you

It would also need to be kept stationary with the degree of precision you want.

[bdunham7]

There is short term--phase noise or jitter--and long term frequency drift.  If you have a lot of the former, it becomes difficult to know right away what the long-term frequency setting will actually be so as to evaluate drift over time.  How do you intend to set the frequency in the first place--and by that I mostly mean how precisely and also whether you are syncing with something or just setting an absolute number.  And, is absolute accuracy important?  That is if you set it for 10.0000000000000 Hz, do you expect exactly 10x3600x24 cycles every day, down to the microsecond?

If you do not need very high absolute accuracy and just stability once it is set, the very best you could possibly do on that budget would be the something like the chip you mentioned and a used OXCO like this to clock it:

https://www.ebay.com/itm/Morion-MV85-5V-10-MHz-OCXO-SC-Sine-TESTED-DATE-2008-2009-50-7dBm-2/312303291797?hash=item48b6ba0d95:g:ijgAAOSwjg1b4sn~

I used one of those recently to make something and it was pretty stable, but not 'zero' anything.  You shouldn't say 'zero' drift or shift because there is no zero.  Nothing is exact, there's always an uncertainty and any design has to have tolerances.  If you need better than a hundred parts per billion accuracy as you imply, you might need to spend more money.

And b/t/w, can we ask what this is for? 

[dmendesf]

AD9833 needs a 12.5MHz maximum clock so you can either buy an used 10MHz DOCXO or rubidium clock. This one fits the bill: https://www.ebay.com/itm/332386541547