Low phase noise 10Mhz refrence for the Lab

[Zenwizard]

I am looking to add a 10 Mhz reference for the lab I currently have a Leo Bodnar GPSDO in the lab and I love it, It has served me from some time. But my needs have continued to go and I need a sine wave output with low phase noise. I am going to be adding some spec and to the lab and need a reference that I can clock them with in the hopes to improve their performance and stability over the long run. From what I have been able to tell so far I am looking for a 10 Mhz oscillator that has both good close in phase noise and long term drift specs. But I have nothing in the lab currently that I can measure or quantify phase noise parameters nor access to a lab that would let me make some quick measurements so I cant at the moment build something and test it. Does anyone have experience with the Endrun Meridan II unit. The phase noise specs of the US-OCXO seem excellent. Even better then rubidium or is there something I am missing something?

https://endruntechnologies.com/products/time-frequency/low-phase-noise

Zen

[tkamiya]

I've been down this road.

I don't know what your budget is like, but if it's going to be a typical hobby budget these things new will be well outside of it.  Also, the phase noise graph looks too good.  It's probably idealized graph meaning reality will be a bit different.  Before you go far in thinking process, it might be better to give them a call and get a rough price quote.

Typically, GPS or anything that pushs and pulls OCXO will have much worse phase noise than same OCXO without.  Some of us use OCXO based GPSDO and disconnect the GPS side to make measurement.

[Zenwizard]

Thanks Tkamiya.

Where I am at the moment is I am at the beginning of a long path down the rabbit hole, and I am setting up the lab and trying to future proof it with in reason. I am been thinking about pulling the trigger on the SRS rubidium unit but I do not have yet a way to generate 1PPS output to lock it to the GPS in the lab. So in terms of price I know where the ball park is. To start and hope to generate the 1PPS signal I have ordered one of the Samsung modules that I am hoping to hook up to lady heather when on arrives on extended shipping time. But I might just use SRS as they are the known quality. In terms of budget rubidium is not out of reach, cesium beam yea… uh NOPE. But in learning I want my tools to be better then I am so when something is messed up I KNOW it is my fault and not the tools.

Zen

[nctnico]

Be aware that the phase noise of a Rubidium clock is horrible short term. You'll need to put a so called 'cleanup oscillator' behind it to have a signal which has a low phase noise both short and long term. I second tkamiya's suggestion to use a GPSDO in holdover mode to have a clean 10MHz.

[tkamiya]

I've actually tried that....

PRS-10 can follow a reference such as GPS by 1 pps and adjust itself.  It's handy for calibration purpose, but Adev measurement (stability) showed having PRS-10 alone is better than such combination by 10 fold.  I'd suggest you try it and see for yourself.  If nothing else, it's curious and interesting.

Rubidium's long term stability is such that training via GPS continuously make little sense.  As I understand it, making this work will take vast amount of knowledge and experience to a level someone who can design these things from scratch.  That is not me!  Instead, I have both as separate unit. 

I like PRS-10.  SRS has put 10 times the amount of rubidium in it that usual mode of failure, depleted Rb doesn't usually happen.  One thing to note that Adev has this weird hump around tau = 1 second mark. 

I know of several experts in this area.  They don't depend on one type of unit for their needs.  They do things like using OCXO alone for 100 seconds, and rest with Rb or Cs.  Take both graph and merge them. 

For my lab reference, I have both GPSDO and Rb, and they are switchable.  For measurement reference purpose, I have all of above.  After fighting with all these stuff for 3 years, I come to realize everything is a compromise.  I must accept something that is good enough.  It's far more important that my lab equipment are all based on the same clock.  For this, GPSDO is more than sufficient.  If I lose GPS, then switch over to Rb.  I calibrate Rb with GPS every 6 months.

Just get what you want and try it yourself.  If nothing else, it's fun.
 

[DaJMasta]

Phase noise is going to be a product of your OCXO, not really dependent on long term drift, so while a GPSDO can often make a decent source, it's not inherently any better.

Your current reference isn't too bad in the phase noise department, does it not output a sine wave?  You could look into a GPSDO with a better OCXO used for the absolute reference, or even a rubidium or disciplined rubidium source, but again those aren't really going to have a bearing on the phase noise inherently - it just depends what their output oscillator is derived from.  Some rubidium sources are derived directly from that physics package, so they are pretty bad for phase noise, in the case of the PRS-10 (which I agree is great but have had no trouble with disciplining), the rubidium disciplines a high quality OCXO internally, so the basic 10MHz output is very low phase noise since it's actually an OCXO disciplined by the rubidium oscillator.

This also means that if you need low phase noise only, a high quality free running OCXO should do just fine, and any piece of gear with a good OCXO reference with an output may suffice for that.


But what do you actually need a 10MHz phase noise reference for?  As phase noise is going to be a product of the PLL or disciplining loop of what you're working with, having the lowest phase noise 10MHz reference isn't actually a huge deal - unless its egregious the equipment referenced to that dirtier 10MHz signal should be able to do better if they are designed for it.  Generally you are concerned with phase noise with your LO or RF tones at whatever frequency they're required at, so unless you're going to be transmitting at 10MHz directly, it's not that critical that your 10MHz reference is super clean in terms of phase noise.  If you want a reference beyond 10MHz (or even below), I'd look at either a fixed output one like the Signal Hound PNCS-1, or a very low phase noise generator like a PSG generator or a Holzworth generator, or something comparable.  As mentioned before, these low phase noise generators don't even require an external 10MHz reference to achieve their output phase noise performance because it's a factor that depends on the internal PLL design and architecture.

[Zenwizard]

So It sounds like what I want for the lab is an ultra stable OCXO that only does a frequency check in about once per day. But then it free runs the rest of the time so as not to get the short term jitter of the GPS updating the clock. Currently all I have is the Leo GPSDO that is outputting a square wave to the frequency counter.  I would like to get a 10 Mhz distribution network going in the lab but I would like it to supply frequency counters generator and spectrum analyzers. So I need it to be better then the clock in my best spec an but this is a future purpose as my current spec an only does 40 Mhz.

[tkamiya]

Spectrum analyzers are usually not used to measure accurate frequency.  It's not that they can't do that but that's not where the instrument shines in most lbs.  In my lab, they are running on its own OCXO and it's just fine.  Typically, your bandwidth will be 1K or so, so if it moves a little, you won't even see it.  If you have a super expensive kind, it might matter but then again, they come with good enough OCXO in itself.

Frequency counter and RF generators, on the other hand, would benefit from Leo's GPSDO for accurate frequency but not much so from low phase noise and jitter.  Counters gates open 1s or so.  Any phase noise will be buried in that. 

I'd say you already have what you need.  Ultra stable and ultra low noise OCXO are useful as a measurement reference.  You'd take special care in cabling and setting it all up.  Not something you'll put through distribution amp and massive and lengthy amount of varying length of coax cables.

Also, ultra stable and low noise OCXOs are rather expensive.  New, they go for $5KUS.  Even at that price, good luck on getting one.  It's not something you can just place an order.  They are typically made to order and manufacturers don't like to make just one.  Used, you can have one starting around $100 but then problem is verifying they are indeed stable and low noise.  After 20 to 30 years in field and stored in unknown condition, many of them are no longer at their original spec.  To check this, you'll need a good/better reference and you don't have that.  So it's really not a practical solution. 

My own conclusion is GPSDO are one of the best choice for my home lab.  It's relatively inexpensive, it's self-calibrating, and if it locks well, it's pretty much guaranteed good.  I have various oscillators that I'll bring in on as-needed basis.  I'm still looking for low noise OCXO myself. 

[jpb]

Also, ultra stable and low noise OCXOs are rather expensive.  New, they go for $5KUS.  Even at that price, good luck on getting one.  It's not something you can just place an order.  They are typically made to order and manufacturers don't like to make just one.  Used, you can have one starting around $100 but then problem is verifying they are indeed stable and low noise.  After 20 to 30 years in field and stored in unknown condition, many of them are no longer at their original spec.  To check this, you'll need a good/better reference and you don't have that.  So it's really not a practical solution. 

$5k sounds quite cheap, I think the BVA ones when they made them (I don't think they are manufactured anymore) are more like $20k-$30k!

The trouble is you also need to buy 2 so you can measure one against the other!

I've recently noticed that some are made for audiophiles, though these are probably rather more expensive than buying the OCXO alone - they claim to select from a number of samples.
e.g.
https://www.mutec-net.com/product_ref10_se120.php#data
The above is about £5k I think.
I too would like to acquire one so peruse ebay from time to time but my budget is more like $100.
I have got a few HP10811s but they are not the best specced ones and I've not had a chance to measure them properly so I've no idea if they are any good or not.
here is an example on uk ebay:
https://www.ebay.co.uk/itm/224378888645
I am tempted but it would just add to the pile of unmeasured and unkown ones I already have

A bit off topic but there is a tear down of a BVA on this thread:
https://www.eevblog.com/forum/metrology/oscilloquartz-8600-3-gpsdo-internal-photos/

and some measurements:
http://www.ke5fx.com/bva.htm

[bingo600]

I'd say you already have what you need.  Ultra stable and ultra low noise OCXO are useful as a measurement reference. 
Also, ultra stable and low noise OCXOs are rather expensive.  New, they go for $5KUS.  Even at that price, good luck on getting one. 

Taka here you go ...      - My budget isn't for that one 
https://www.ebay.de/itm/174754655714

I'm 99% sure they "Butchered" it from this unit, as this one had it in on the first pics.
https://www.ebay.de/itm/224441938564

/Bingo

[tkamiya]

Problem with BVA is that what's available aren't new.  It's hard to tell how well they meet the spec.  There is no telling how they were used and stored.  I'm not going to spend $5K on that....  What I was talking about is that most manufacturers have their own top-of-the-line oscillators.  Biley included.  It's top end model is called Poseidon.  It starts around 5KUS and goes up from there.  Every improvement in phase noise, every step in stability increase, they charge dearly.  Likely, they won't make one either. 

I actually bought Biley's run-of-the-mill oscillators new from factory.  Not the best but quite decent.  but as a bonus, the tech department director had his staff measure my particular units (I bought two) and graphed phase noise for me.   Phase noise is actually excellent. 

I have a 4040A cesium standard.  Mine includes huge blue oscillator unit.  I forgot the name but it's not BVA.  You can find this 5MHz standard every now and then on eBay.

[kj7e]

I have had very good luck with this combo;

BG7TBL 10 MHz distribution block without OCXO;
https://www.ebay.com/itm/182249079841

Then add this OCXO to the distribution block above.

UCT 108663 OCXO;
https://www.ebay.com/itm/203257378007

Those are NOS, they need to age a bit, but after running for a month or two they really settle down and make great stable and low noise reference.  I have a pair of GPSDO's I check it against now and then. Almost never touch it as it just does not move.  You have to spend considerably more to get any better IMO.  One note, make sure you move the jumper (chip resistor) on the distribution block to use the OCXO's internal Vref for the Vc input, not the onboard Vref.

[Gerhard_dk4xp]

I have a Lucent KS24361, built by HP/Agilent for Lucent.
Each of the 2 Plugins has a MTI-260 5 MHz oven for
redundancy. The second plugin has no gps receiver,
which made it cheap.
Top-of-the-options MTI-260 invade into rubidium land,
but mine are very probably none of them.

I have modded the unit with the GPS receiver by adding
a doubler 5->10 MHz and some outputs: 1pps with CMOS/TTL level
and 4 outputs 10 dBm for the 10 MHz.

<    http://www.hoffmann-hochfrequenz.de/downloads/DoubDist.pdf     >

Cheers, Gerhard.

who thinks about locking 16 MTI-260 sloooowly to a common
ref and to wilkinson the 5 MHz outputs together. That should make
quite an effect on the phase noise. Yes, I have them 

[tkamiya]

Gerhard,

You mentioned SIXTEEN MTI-260.  I thought about using as many HP11801.  What would you do to make use of all of them though?  I can't think of a way to average signals.  I can injection lock but that's all I can come up with. 

[Gerhard_dk4xp]

More like a proof of concept. Poor man's replacement for BVAs.
2 Blocks of 8 would probably make more sense as references
for 3 cornered hat measurements  of different oscillators.

Injection locking is bad because it cannot be controlled. Yes,
I know that Rohde has calculated the formulas for n coupled
oscillators, but when you try to do it with real hardware, you
can't find your numbers.

I have made a board that allows locking a MTI-260, or MV89A,
or 10811 or some others to an incoming reference frequency,
or to an incoming 1pps. The 1pps has not yet been really tested.
It can also generate a 1pps from the onboard oscillator.
The digital part of the PLL/1pps is in a Xilinx Coolrunner ( $3 )
The pic with the 3 oscillators was a test if the PLL locks with
all of them. They have different tuning sense and Kvco, there
are strapping options for that.
 

[Gerhard_dk4xp]

1 more pic

[Zenwizard]

Thanks for all the information. I have a ublox LEA-6T-1-001 on the way in a Samsung case. I recently caught some deals on the spectrum analyzers for the lab so I can now answer the question as to what I am going to need to feed. I Have a frequency counter HP 53131A, an HP 3585B, slow spec an 20 Hz to 40 Mhz, An extremely well optioned HP 8596E 9Khz – 12.8 Ghz with the OCXO option and tracking generator, and a HP E4411B Option 1AX and 1DN also the tracking generator, and a HP 8657A that has the high stability time base but it is off frequency by about 50 Hz and needs a touch up.

I would like ultimately to also have some scopes with 10 Mhz ref inputs that I can hook up to the network as well. The question now is, Is the Samsung unit better than the reference in the 8596E as other then the counter that seems to be my best clock currently. Just trying to squeeze as much performance out of the lab as possible with out getting in to stupid money.

Zen

[Leo Bodnar]

My go-to 10MHz OCXO for phase noise daily use is Wenzel 500-07379.  It's one of the best among inexpensive Wenzels.

I don't know if having one expensive source for phase noise and stability actually makes practical sense. 
Typically, for lowest phase noise you want a quartz crystal with high excitation levels.  For high stability (low ADEV) you want quartz crystal with low excitation levels.

Personally, I use Wenzel OCXO for phase noise tests and Rubidiums (FRK-L or FRK-H) for stability (1s to 10,000s ADEV.) 
It's quite rare to need PN and ADEV tested at the same time.  PN tests usually take only a few minutes and ADEV tests take hours if not days so it's not too difficult to plan around this.

Leo

P.S. Here is the phase noise of two identical (one is DUT, the other - reference), so reduce the figures by 3dB to get the idea of actual PN.  Spurs are environmental - this OCXO does not have spurs.