Metrology 100-101 Project/Inquiry for Volt and Time Nuts

[Electro Fan]

Preface

I made a post in an ADR1000 thread and then realized I needed to backup and regroup.  Here is my attempt at regrouping.

Thanks to Jacques and Dieter for getting me back closer to being on track.

I now understand that an ADR1000 EZ board is a DC voltage reference with no MHz (or any Hz) output.  I should have looked closer at the schematic and understood it better but I got mesmerized by seeing a SMA connector.

More Background

I don't have any big "need" I'm just fascinated with measuring V, I, and R and frequency with as much accuracy and resolution as possible.   However my voltnut and timenut interests are offset by the fact that my knowledge only let's me go so deep (not very) at the circuit level, so I'm more at the "system" and "system integration" level, or maybe the "Plug N Play" level.

Having said that, I've come to believe that the LTZ1000 and ADR1000 represent the top of the line or near top of the line for voltage references, so I was (and still am) intrigued by the ADR1000 "EZ" module.  As mentioned, I now understand that it's outputting a DC signal - a high quality DC signal but just a DC signal.  So presumably it could in turn become a very good PS for something else, but it's not going to generate a 10 MHz frequency (or any frequency).  No doubt, kind of a Duh moment for me 

But, hey, we all have to start somewhere....

Mission Description

What I'd like to figure out is how to build something low cost (under $30-$50?) that combines an accurate and stable 10 MHz reference with a suitably stable power supply.  The Amazon OCXO modules are as low as $13, so that leaves the power supply, and probably a case.  (Additionally, as another project, I'd also be interested in seeing how far I could drive 10 MHz accuracy, resolution, and stability but stay in the $100 to $200 range.  At about $200 or less, a BG7TBL GPSDO seems hard to beat, or maybe there is a better benchmark at an even lower price.)

I guess another way of describing my interests would be to say I'm up for experimenting in the $20 to $200 range to see what causes what through measurement, trial and error, posting results, getting feedback, and trying to learn in the process.  Kind of metrology 100-101 for rookies.

So maybe the LTZ and ADR are overkill for powering such an OCXO?  I probably got too excited thinking maybe I could jump to a top of the line voltage reference and then experiment to see what impact a super PS would have on a "good enough" OCXO. 

On the forum(s) I have made some posts about the $13 Amazon 10 MHz OCXO modules.  They have unit to unit variation but a good one seems to be stable to about .1 or maybe .01 Hz, or possibly even better. 

(I also made some posts awhile back where I learned that measuring and controlling resistance to around 1 milliohm requires 4 wire Kelvin clips and managing ambient temperature - so I'm definitely sold on the impact of temperature mgmt.  Having said that, I've found that GPSDOs seem pretty resilient to ambient temperature, so maybe an OCXO without a GPS discipline is a losing battle?)

Back to OCXOs without GPSDOs, I'm pretty sure achieving better than .01 Hz on a 10 MHz signal with just an OCXO requires controlling ambient temperature.  To try to confirm this I'm working on putting the OCXO module in a case.  Before I get to the case I've been experimenting with the OCXO module to see if a vertical vs horizontal PCB orientation has any impact on performance, but it might be hard to distinguish the physical orientation impacts without first controlling the ambient temp.  So maybe I need to first get the module installed in a case that helps deal with ambient temp.

I'm powering the OCXO module with a Korad power supply.  Before my "regroup" I was thinking the ADR would be a (much) better power supply.  It probably is but at ~$165? for the ADR board it kind of defeats the "low cost" objective.  Maybe a LM399 would be a better candidate?

Summary

The GPSDOs I've experimented with seem to be very accurate and stable for about $200 each.  Maybe it's going to be hard to do better for $200.  So maybe the question is becoming: is it possible to integrate a $13 OCXO and a ~$30 LM399 to get a ~$50 reasonably accurate and stable 10 MHz reference?  (Or maybe if I can get the $13 OCXO module in a case so it's insulated from ambient temperature changes it will improve in accuracy and/or stability enough that the Korad PS isn't a limiting factor?) 

I really don't have any big goal here other than experimenting and being able to learn in the process. Net, net:  I'd like to learn what causes what from power supply to OCXO to overall temperature management.  If you have any suggestions or guidance, or learnings from other related projects, I'd be very happy to hear whatever you think.

Thanks, EF

[ch_scr]

I've messed a bit with those OCXO modules and a DIY setup to accurately measure them. The TinyPFA is the cheapest option to get the needed resolution to actually "see stuff". To give a bit of (practical) perspective, those OCXO modules aren't perfect. As in, when you touch the case, the frequency will move because the oven regulation isn't perfect. (Sidenote: Even a second-hand double-oven unit is way out of the price range) So the question is rather "how far" will it move when I put a finger on it, not "will it move". To put further perspective on that, I use a lowly bandgap reference (OnSemi TL431BCLP in particular, still impressed with these) for the Vtune. And I can touch those just fine without seeing a frequency shift! Obviously there are long-term stability considerations as well, where a band-gap might not be the best. What I can see is the slack in the cheap 10-turn potentiometers and the slider moving in there just from hitting it with a small screw driver handle. And that is on a 10-turn pot with resistor padding to reduce the tuning range! Just my 5cent of observations so you don't bark up the wrong tree...

[Electro Fan]

ch_scr,

Thanks for all the info including the recommendation on the TinyPFA.  I have one on order and am looking forward to learning more by using it.

On the 10 turn pots I'm going to experiment with replacing those on some of the OCXO modules.  I suspect that several of the ones I have are not functioning properly; after I swap them out for new ones I'll report back.

Your experiences and comments were very much along the lines of what I'm looking for.  Hope others might jump in and share theirs too.  Thx again.

[ch_scr]

I'm not using pre-made modules but bought "raw" de-soldered OCXO's, basically a few samples of all the types I could get (for reasonable prices).
(While I'm not done testing, CTI and Isotemp have impressed so far. Though I've only looked at short-term (seconds to minutes) noise and "finger-type" thermal shock)
I've added impressions of the latest PCB, designed for the 12V "EU case style" OCXO units, using the so-much-better Helipot 22-turn trimmers.
I've made some deliberate layout decisions, notably star-gnd and power arrangements (decoupling to the gnd plane though).
The goal is to isolate the effect of the heater current draw on the rest of the circuit (influence on the Vtune notably).
I've added the top view to show the effect of the "exclusion zones" on the gnd fill, so the THT-pads don't circumvent the star-gnd.
Also the Tl431 is after the 12V reg to isolate it better from input supply variations. While this is the "minimalist" approach, I didn't omit the LC filtering.
I use an "unbuffered" type hex inverter, first as buffer with a coupling/feedback arrangement (C17/R4) that should provide close to 50/50 duty cycle,
for the usual OCXO output waveforms. Then it's also used with series resistors to drive the two 50 Ohm outputs.
There is ample space all around the OCXO to fit a 3d printed cover as well.
Hope this gives some ideas, aside from damping the LC filter resonance there's little I could think of to improve for now.
Don't read too much into the bulk smd decoupling cap values, I put whatever desoldered SMD tants I have in stock on there 
I'd be interested to hear if anything sticks out as a particularly bad design decision to anyone 

[Electro Fan]

ch_scr, Thanks for the update on your project.  Please let us know what test results you get when you have things assembled.

On my end I purchased some new trimpots (several both with the original value and some with other values).  None of the trimpots seems to enable a module that could previously get within about 1.5 Hz on 10 MHz to do any better.  As limiting as as fiddly as the standard trimpots are, I'm pretty sure it's not unit to unit variation or probably even the trimpot that is the limiting factor in the module's performance.

PS, one thing that is working well in my project is a Hakko FR-301; it makes removing the trimpots pretty easy and fast.

[ch_scr]

with the orignal trimpot as e.g. 5k, use a 500R one and put each a resistor in series with Vref and gnd; say 2.2k each.
You might have to cut the trace with a scalpel or shop knife and remove the solder resist with a fiberglass pencil,
you get a much finer trim around vref/2 then.
depending on individual ocxo, adjust the ratio to e.g. 3.3k/1k to get that finer adjust around the point where the 10MHz is.
(which is why it isn't done for mass produced units)
you could have a coarse and fine trim dual-potentiometer arrangement instead, but "padding" improves long term stability and tempco as well.

[trtr6842]

One method for better Vref/Vctrl stability would be to make a simple PWM DAC using a microcontroller and an analog switch.  In this case linearity wouldn't matter, just long term stability, so pair that with an appropriate voltage reference and I think a PWM DAC could be a great balance of cheap, easy, and well performing.  In terms of stabilizing the OCXO's themselves, I bet some basic insulation will go a long way to reducing outside effects without resorting to secondary temperature control.  But if you're already putting a cheap MCU on there to do a PWM DAC, it wouldn't be too hard to integrate a PCB heater that keeps the base of the OCXO at some relatively stable temperature by just using the MCU's ADC and a standard NTC.  Surely that would be more stable than room temperature, which would reduce the external fluctuations the OCXO has to deal with.  I'm up for the task of designing a prototype if a few others are willing to chip in for the prototypes!

[ch_scr]

The problem starts with having half-decent ocxo in the first place. If the thing itself isn't good, no amount of lipstick will make it not a pig...
With regard to oven regulation, a cap does help but can push the ocxo oven controller into thermal oscillations.
Take the Vectron C4600A1 (green trace) in the first picture. Too bad that without the cap thermal regulation (and stability) is even worse!
(note that the MV197 and OC12SC36 probably look the same because the noise floor of the measurement is limited by the ref osc used)
(note 2 the drop in the yellow trace was me adjusting the pot)
That lead me to looking at oven startup current draw behaviour, see picture 2 and 3.
Funny enough, the one where the startup current draw looks like a well-tuned control loop right out of a text book is the one thats thermally unstable.
But a startup from cold like that is a singular event in the life of an ocxo in an ideal world IMHO, so no need to optimize for that anyway.
In my mind, much like the large/small signal transient response of operational amplifiers,
we have a similar effect here where one looks good (startup), but the other (final regulation) not so much.

[dietert1]

Frequency deviations are the product of temperature deviations and TC of oscillator. While the Vectron oven appears to be well tuned, the oscillator inside isn't well tuned for low TC. The observed frequency deviations are as expected with a period of about 45 seconds. And there could be some problem with its heater circuit that causes instability near balance. Probably the deviations could be reduced using the part inside some enclosure.
That's something known from the LTZ1000 voltage reference: It has a fast, near perfect oven, yet it also has a significant TC. People use plastic covers to protect it from air draft.

Regards, Dieter

[ch_scr]

Hello Dieter,
That frequency stability graph is with the plastic cover already! Without it, just ambient air drafts make it wander much more.
In the picture attached, the other lines are the same as before, just the scale is quite different.
Before the red dotted line, you can see behavior without plastic cap. After, the excursion from putting my finger on top.
You can see why I ended the test early...

[dietert1]

Yes, Vectron sounds a bit like vintage. Probably at the time they wanted to make something good for 1E -8. Not everybody can measure to 1E-11 within some seconds..

Regards, Dieter

[ch_scr]

It took some doing. You can thank Corby for making it look easy with his DMTD board. "That doesn't look too bad, surely I can implement that"... Yeah right.
Sure did learn quite a few things though! Now with Rev 4, it seems to be getting there. If only I knew using one is as complicated as building it 
Then I had to make my own counter for 10Hz, to get the full update rate... All still WIP but it seems to work now.
I have locked the Offset Osc to the Ref one, so I can take the Freq difference from one channel directly - no need to mess with unrolling the phase.