Low cost <1 ppb 10 MHz OXCOs on ebay

[rhb]

Do you have a GPSDO?  My testing over several days with a pair of the PCB versions, the OXCO in my HP5386A and  a Bodnar 2 output GPSDO viewed on a DSO triggered on the GPSDO showed the CTIs were as good and possibly better than the OXCO in the 5386A.

I've been testing RF cables from bydpete on ebay.  I've got 10x $ value in additional cables coming.  My 2nd batch of 10 CTI units is making its way slowly in my direction.  I also ordered a bunch of multiturn dividers.  I am going to attempt to build a fixture that will let me run 5-10 units at once and switch between units automatically over a long period of time.

I have never been able in any of my attempts to set an OXCO to exactly match the GPSDO.  If I have an instrument open I usually attempt to retrim the OXCO as almost all my HPAK kit has the OXCO option installed.

That experience over the span of 5-6 years has convinced me that without an electronic means of setting the control voltage it is impossible to set exactly to 10 MHz.

If  the phase noise is Gaussian, combining 4 OXCO outputs *might*  cut the phase noise in half.  No idea yet if that is possible, much less how.  Just an interesting idea at present.  At normal OXCO prices it would be cheaper to add  a GPS receiver.  But  at $3 that changes.  If it's possible to cancel phase noise adding a GPS receiver might yield a significantly more stable reference with 3 dB lower phase noise than a single OXCO for very little additional cost.

Have Fun!
Reg

[WatchfulEye]

I've found the CTI OCXOs really impressive for the cost.

The 2 main things that separate them from more expensive OCXOs are:
unpredictable retrace (can restabilise at up to 5 ppb from where they stabilised before) as well as severe frequency drifts over the first 12-24 hours, gradually stabilising over the next few days.
High temperature coefficient (the one I've best characterised has a -500 μHz/K coefficient).

Once stabilised, the performance is very impressive, however.

Here are 2 examples of stability against my reference of another free-running OCXO, a trimble 73090 off aliexpress. This also gives excellent performance and this particular one is in a thermally isolated box and has been running for months, and shows essentially zero aging against GPS.
One is with the OSC5A2B free running, and the other is with EFC temperature compensation.

Note to the eagle eyed - the TC does not appear to be -500 μHz/K, but that is because the ambient temperature is plotted, not oscillator enclosure temperature. Because the OCXO is temperature controlled, it has a stabilising effect on the temperature in the enclosure, typically reducing temperature swing amplitude it by about 50% compared to ambient.



 

[MIS42N]

The OSC5A2B02 is more sensitive to variations of the control voltage (EFC) than more expensive OCXOs. I've characterised quite a few and they are around the specified 1ppm/V. Which is 1 part per billion per millivolt. Most run of the mill regulators are in the 1mV/C range so for better results the source of the EFC becomes important. Some design is required to avoid small voltage variations in the earthing system. The reference for the EFC source needs to have the same earth as the OCXO. I made a PCB with a layer of copper as the common earth, there was enough stray current to create small voltage differences (microvolts) between the OCXO earth and the EFC source earth. It had a noticeable effect.

[Yrrah]

Already some nice work done. Kudos! I had OCXO no 3 running for several weeks (24/7) in my test fixture and I just noticed the frequency did not change within 20 mHz. This on my HP5345A counter with a HP10811 OCXO. Always plugged, of course. No scientific accuracy claimed of course but indeed good evidence these OCXO's are really good.
Br,
Harke

[WatchfulEye]

I've had a chance to do some additional measurements, and I was a bit surprised by the results.

The measurements shown are my OSC5A2B02 based GPSDO vs a freerunning OSC5AB202 and a rather more expensive, freerunning trimble 73090 DOCXO.

In green is the OSC5A2B02, and in black is the 73090. Noise floor of my frequency counter is shown in blue. The OSC5A2B02 clearly has substantially better short term stability.

The medium (500-2000 s) tau are not reliable on this plot, because the GPSDO was locked and GNSS noise in my setup is substantial. However, at very long tau, the drift and temperature sensitivity of the OSC5A2B02 is clearly visible compared to the double oven trimble.

[rhb]

I've not been following EEVblog lately, but it confirms most of my observations.

I received  a tinyPFA from R&L and will set that up soon.  I don't have proper environmental control or monitoring and won't for some time while I set up a new lab space.  I'm building a 7 bay rack!  Bit of a TEA binge :-)

A very limited error budget analysis from mains to OXCO shows that PS noise is  a significant limiting factor.  The 20 uVrms noise of an LM399 is 0.02 ppb/s.  That figure requires *no* noise on either the LM399 or OXCO supplies.

I've moved posts of my work to the qex@groups.io list as I had mentioned these on several lists and it became too confusing.

My immediate focus is the PS as that is critical to my goals.  At 0.001 ppb/uV sensitivity getting to 0.01 ppb/yr is going to require very careful environmental and EMI control in addition to active compensation for aging of 4 OXCOs.  Ultimately I plan to heterodyne the OXCO outputs and use a zero crossing detector to get <0.001 ppb resolution.

I'm going to visit an HVAC shop today to have a desktop box made which will get insulated with Peltier thermal stabilization and shielded feedthrus, etc.

Have Fun!
Reg

Edit:  BTW before someone decides to jump up and down about the brutally difficult voltage stability problem that 0.01 ppb/yr implies I'd like to note that after following Ken Thompson's "When in doubt, use brute force." rule I plan to throw everything I know about DSP at the problem. I'll also be selecting 4 OXCOs from a batch of 20 with the intent of deliberately exploiting variations among 4 devices.  For example, if I know that each OXCO has a different response to uV level changes I intend to exploit that as a means of estimating the voltage variation and adjusting system.

In short, this is a very difficult project which I expect to span several years and may end in failure.  Were that not the case it really wouldn't be very interesting.  The only thing I know well is empirically estimating the aging rates and basic knowledge of the relevant factors.

[tszaboo]

So I got a few of these OCXOs because I clearly have a mental issue doing all the projects at the same time.
Does anyone have a PCB design for this?

[alligatorblues]

I rather fear I am among the lost as I'm about to see if the fabled <$150 rubidium standards exist.

Pray for my salvation and that I shall be delivered from the debauchery of becoming a Time-Nut. ;-)
Reg

BTW this is the board design I bought though possibly not this vendor.

https://www.ebay.com/itm/144978139139

For USA buyers, I have 2 rubidium standards. I'll put them in FS if needed. But no nation permits importation of radioactive substances through the post. I can only ship them in the US.

[807]

So I got a few of these OCXOs because I clearly have a mental issue doing all the projects at the same time.
Does anyone have a PCB design for this?

This guy designed a PCB, but didn't provide the files, or put his design in the shared section of PCBway:-

https://www.pa7elf.nl/en:projects:10_MHz_ocxo_frequency_reference

As I have a few Racal Dana 199x frequency counters, I bought a few of these PCB's from OSHPark. They are specifically for the 199x counters though:-

http://syncchannel.blogspot.com/2016/02/ocxo-upgrade-for-racal-dana-199x.html

[Kean]

So I got a few of these OCXOs because I clearly have a mental issue doing all the projects at the same time.
Does anyone have a PCB design for this?

What type of PCB design?

I've been playing with this GPSDO that uses this same OCXO, and I bought 10 extra ebay OCXO modules for some reason (well they were pretty cheap).
https://www.eevblog.com/forum/projects/budget-gpsdo-a-work-in-progress/

As I have a few Racal Dana 199x frequency counters, I bought a few of these PCB's from OSHPark. They are specifically for the 199x counters though:-

http://syncchannel.blogspot.com/2016/02/ocxo-upgrade-for-racal-dana-199x.html

Oh, nice!  I also have a pair of the 1992 frequency counters, so that is 2 of my extra OCXOs accounted for. 

[tszaboo]

So I got a few of these OCXOs because I clearly have a mental issue doing all the projects at the same time.
Does anyone have a PCB design for this?

What type of PCB design?

From what I've seen you need a reference voltage, a potentiometer or a DAC+MCU to make the control voltage and maybe a buffer for the output?
I think I can throw this together in a hurry, but I was just wondering if someone has PCB for it. Probably even breadboards would work.

[Kean]

Yeah, see that GPSDO forum topic I linked to
And https://github.com/ajcashin/budget-gpsdo

[MIS42N]

Yeah, see that GPSDO forum topic I linked to
And https://github.com/ajcashin/budget-gpsdo

You remind me, I really should clean that up. If anyone is up for it, the PCB could do with a redesign as there is some interaction between the OCXO circuit and the PIC/other bits circuit. I thought giving each its own LDO would be sufficient, but it appears having a common ground plane is a problem. And since places like JLCPCB can populate a board as well an SMD/through hole version would be nice - SMD for the common bits and through hole for the CPU and OCXO (and whatever uncommon bits not in JLCPCB's catalogue).

[vindoline]

Thanks to Reg for pointing these OCXOs out! I wanted to let people know that I've used one to make an easy GPSDO reference https://www.eevblog.com/forum/projects/yet-another-diy-gpsdo-yes-another-one/msg5064550/#msg5064550 as well as add high stability options to my Racal Dana 1992 and HP 3325B. The GPSDO was used to calibrate the 1992 and 3325B.