Master Clock for your EtherREGEN

[Mops911]

8 hours ago, Clockmeister said:

if John feels sign waves work better for the eRegen

Dude, he never said that. He said square is the best IF the implementation is right which is hard and expensive. He said sine is easier to implement for a good value solution where cost is an objective.

[Mops911]

On 5/1/2021 at 3:22 PM, DarqueKnight said:

Measurement certificate for my AfterDark Giesemann Emperor Double Crown clock.

I have problems reading these plots

 

to me it looks -119.5dB at 1 and -132dB at 10....how to read/see -143?

[Mops911]

23 hours ago, DarqueKnight said:

about -143 dBc/Hz

thanks @DarqueKnight

 

My problem was actually not reading the graph but interpreting it correctly. How would I know that a line which consist of constant "spuriae" can be interpreted in a best case way ;-) Are the spuriae measurement artifacts? Or is there a more serious problem around 10Hz where the performance has an issue due to power supply, vibrations or other problems...

[Mops911]

Just pulled the plug and ordered:

- Clock module  (Best 2)

- Teddy Pardo tripple output to feed ER, ISORegen and clock

- 10cm cable 

- Low pass filter

 

Now looking into a nice housing and potentially dampening.

 

I am a bit reluctant to cover the clock in any way as it changes the temp control. 

How did you customize the clock with wrappings or covers?

 

Cheers

Tom

 

[Mops911]

Does anybody know if there is a service in the US to measure phase noise and Allen dev?

 

Ill get my clock soon and would like to test it….

 

cheers

tom

[Mops911]

My clock is warming up and my Teddy as well....more on it (maybe) later....(I made to many changes at once to have a good comparison....but will go back in time and add changes incrementally to see what matters). 

 

However, much more fun is my recent interaction with the guys on time-nuts. Fascinating folks (and very suspicious about adding a clock to the switch would do anything audible). But one guy offered to measure my clock and he just wrote me 

 

quote:

My typical process is to let the OCXO sit for a day or two to equalize with the environment, then power it up (monitoring the current to see that the oven is kicking) and letting it run for anything up to a week to settle in -- it's quite remarkable how the aging curve on many OCXOs settles down with run time, sometimes an order of magnitude per day from the first 24 hours.

There is a neat trick for phase noise testing where you can use two independent reference oscillators and with enough digital processing their noise correlates away.  That can let you measure below the noise level of either reference.  I usually do that, normally with the 10 MHz output of the maser and the 10 MHz Wenzel ULN, and should be able to pretty accurately measure your unit's phase noise at any offset.

Allan Deviation (stability over time) is harder because that correlation idea doesn't work (there's a similar concept called three-cornered-hat, but it can be really squirrely and doesn't always work).  You would ideally like the reference to be an order of magnitude better than the device under test, but that gets harder to achieve as the DUT gets better.

If your oscillator really is ~3e-13 at 1 second, that's about the same as the maser* so will be challenging to measure.  It's not impossible for your unit to be that good; I've seen other OCXOs in the wild in that range.  OCXOs that are actually *spec'd* at <5e-13 are rare and costly beasts, but surplus OCXOs can do better -- sometimes a lot better -- than their spec.

If it looks like your unit is near that floor, I have an OCXO reference that is verified to be <8e-14 at short tau, which is pretty much the best you can get with anything that's not in a bottle of liquid helium. But like any OCXO, its drift causes the stability to go down over periods of more than a few hundred seconds, so it's not good for long term measurements, or absolute frequency accuracy -- that's where the maser is so nice.

I'm telling you all this not just to brag :-) but to point out that the kind of numbers stated for your oscillator are at the edge of as-good-as-you-can-get, and measurement at that level isn't easy.  The measurement instrument can be less a problem than finding references that are good enough.  That's what turns us into time-nuts -- every time you get something better, you need something better still to compare it to!

end quote

 

I got many more emails, lots of theory, which I dont understand really....but really a very cool bunch!!

 

Cheers

Tom

[Mops911]

I though I gonna share this also. My question was about how does it take for a clock to stabilize and how much are power interruptions resets the start up time.

 

quote:

It depends on how closely you look. :-)  In general, an OCXO will reach operating temperature in less than an hour (sometimes just a few minutes) and the frequency will settle down.  But it takes days/weeks after power on for everything to reach as close to equilibrium as it will get, and for all the mechanical stresses to work out.

If you monitor the current draw, you'll see that the initial value is probably several hundred mA, and then after some time -- from a couple of minutes to a couple of hours, but usually 10 - 30 minutes -- the current will drop as the crystal and its environment reach the set point temperature and the heater reduces its output.  It may oscillate up and down for a bit as it closes in on the final temperature.

While the oven is still heating, the frequency will be very unstable over a period of even seconds to minutes.  Once it reaches the point of stable current, the drift will become less and less, often almost exponentially for the first day or two.

If you're mainly concerned about phase noise, you're probably good once the current has stabilized.  For frequency stability over periods of hours, I'd give it at least a day.  For day-to-day frequency stability to reach its final level, it may be a week or more.  This is all *very* dependent on the specific model and even specific serial number.  And again -- this all depends on how closely you look -- once a high quality oscillator has reached set point, further short term frequency change in a quality oscillator will probably be less than 1e-9 (1 Hz at 1 GHz) per hour to begin with, and around 1e-11 per day after a week of running.

For what it's worth, the specs for the HP 10811A, a very highly regarded OXCO used in a lot of HP test equipment, are stated to be valid after 24 hour warm-up.

Getting an idea of what the startup current curve looks like will give you a feel for when things have stabilized -- when the current has dropped from the starting level, and has been stable to within a mA or two for several minutes, the basic warmup is complete.

But any interruption of power will cause the thermal equilibrium to be broken, and there will be a frequency difference when the power comes back on.  While it may not take long to get back to operating temperature there is still mechanical stress from the temperature change that takes time to work out.  So normal practice for high quality oscillators is to keep them running continuously unless power consumption is a concern.  (A lot of HP counters, spectrum analyzers, etc., have the oscillator bypass the power switch so it stays powered up as long as the unit is plugged into the wall.  That lets it get into spec a lot more quickly when you turn the instrument on.)

 

end quote

 

cheers, happy weekend. 

[Mops911]

This looks interesting too

 

Alibaba Clock

 

 

[Mops911]

 

 

 

more clocks coming...;-)

 

 

[Mops911]

Just got the measurements for my Symmetricom OCXO. When I ordered it I took a leap of faith and ordered a selected "best 2" from the Alishop assuming I will get scamed likely. Today, I got the measurements from a nice HAM guy who measures my OCXO.

 

This is what I got, -113dB at 1Hz and -147dB at 10Hz...not to shabby for $400... ;-)

 

cheers

Tom

[Mops911]

Superdad was very explicit about clocks and proximity and stated several times that an external clock must be MUCH better than the internal due to the distance. However, it seems very much that this can be achieved. But yes, a lot of aftermarket external clocks will NOT improve the EtherRegens internal clock performance.

[Mops911]

2 hours ago, audiobomber said:

I assume clock cable choice is also crucial.

These were mentioned recently and I think to remember that they found Superdads "blessing"

 

https://www.ebay.com/itm/182468704902?var=485044364178

 

I got a sine wave OCXO as it was said that square wave is harder to implement aka more expensive. Sine wave plus filter seems like a cost effective solution with my 4inch cable.

 

[Mops911]

Here is the plot I got for the phase noise of two more OXCO I bought on Ali. The CTS was ~$200 and the other one was ~$50

 

Interesting that at 10HZ they are almost same...however, my Symetricomm measured much better than both here and was ~400

 

Also, the guy who measured for me looked at the impact of LPS and insists also that a good and dedicated LPS is a must for these clocks. I know, nothing new here ;-) but still nice when confirmed

 

This is probably a bargain despite the -90db@1HZ

https://www.aliexpress.us/item/3256803469184705.html?spm=a2g0o.order_list.order_list_main.10.45c11802Td4RSd&gatewayAdapt=glo2usa&_randl_shipto=US

 

Cheers

Tom

[Mops911]

I got such a "simple" board from Ali. I paid for "best 2", a preselected specimen. $420 and it measured (by a HAM guy from time-nuts) as 114 at 1Hz and 147 at 10Hz....

 

Form follows function, I would say ;-)

 

[Mops911]

Again, I got a "Top2" here https://www.aliexpress.us/item/3256802980837876.html?spm=a2g0o.order_list.order_list_main.15.64541802SszZEV&gatewayAdapt=glo2usa

 

which measured 144dB at 10HZ ... yes, it comes without a power supply and without plot....but who cares at this performance and price.....

 

Cheers

Tom

[Mops911]

5 hours ago, Superdad said:

Can I borrow $85,000 from someone for that? 

 

If you have a good business case :-D