Master Clock for your EtherREGEN

[Superdad]

3 hours ago, MartinT said:

My thinking was that the ER, with its notorious Crystek 575 oscillator, would be the better component to respond to an external clock although I have no knowledge of the clock used in the Mutec.

 

Not sure what you mean by "notorious" Crystek 575.  The CCHD-575 is one of the lowest phase-noise production XO available (about $10 each at 500 piece qty,), and our 3 original 25MHz (random production) samples came with plots showing 10Hz-offset phase-noise of -108, -110, and -112 dBc/Hz.  (And remember, its easier to get better numbers from a 10MHz clock than a 25MHz clock, though they can be equivalent in terms of jitter/phase-noise).

To do better than that you have to either spend $400-$800 on an OCXO (the OXCO alone, not in a case), one specifically meant and measured to have ultra low phase noise, or buy--and screen with an expensive analyzer--selected surplus OCXOs (as Cybershaft does for some of their models). 

 

Plus we position the CCHD-575 just 3mm from the EtherREGEN's Silicon Labs clock synthesizer and power it carefully with an ultra-low-noise LT3045 regulator.

So an external clock better be darn good with both its phase-noise and its cable if it is going to outperform our internal clock.

 

Quote

I am pleasantly in shock that £70 worth of Chinese made amateur radio clock (well, to be fair, a pretty accurate oven controlled clock made for telecoms use) could achieve anything very much, let alone what I’m hearing.

 

I am really glad you are pleased with what you are hearing, but inexpensive OCXOs, while fine for stability (lack of frequency drift which is what telecom stuff calls for) are terrible with regards to phase-noise at low offsets--which is all that matters for our applications.  So unless your BG7TBL clock is performing some miracle, I'm not at all convinced that it measures any better than our internal clock. And certainly not at the end of a modest clock cable.

Sorry, great OCXO clocks don't come cheap! 

This is the inside of your unit. No markings on the clock to attempt to look up its specs. It is inexpensive enough that I might order one and have John run a full phase-noise plot of it with his new $17K Jackson Labs Phase-Station.

 

Cybershaft does have a new entry level line, and for the price (about $1,500 for the base model MA-OP13) they are exceptional.  Note that their phase noise specs (for which one receives a individual graph for each unit) are listed at 1Hz offset. All their model numbers correspond to the dBc/Hz figure at 1Hz offset. -113dBc/Hz @1Hz is terrific! As mentioned, I think they accomplish this by having sources for--and then screening, measuring, and sorting themselves--surplus clocks.  Nothing wrong with that at all.

 

AfterDark, UpTone's successful dealer in Hong Kong, is also a big Cybershaft dealer. And he will be doing bundles of their MA-OPxx clocks with the EtherREGEN and our JS-2 later this summer.

[Superdad]

4 hours ago, MartinT said:

I found the specs for the DAPU OCXO. Would very much appreciate your thoughts on phase noise and the other specs.

 

Funny, I was just on that spec page as you posted this! 
If the O55A is really the clock module in your box then it’s a good bargain for $100.  -125dBc/Hz at 10Hz offset is indeed better than what the Crystek 575 offers.

Still you will want to use a very good—and short—clock cable and be sure the impedances match.

Enjoy! 

[Superdad]

It is starting to make more sense now.  John told me he runs into the Chinese DAPU OCXOs surplus on eBay all the time. Pulled from PCBs in older 4G towers. Typically about $15-20. Probably less in China.  
Here it is on Taobao:

https://www.taobao.com/list/item-amp/574299443769.htm?spm=a21wu.10013463.3.11

Now I dig a little further to understand that BG7TBL is the call sign of a Chinese ham radio operator who some years ago began producing a whole series of 10MHz clock units for ham amateur radio. Well covered here:

https://www.eevblog.com/forum/testgear/bg7tbl-gpsdo-master-reference/
Fascinating.  I wouldn’t mind finding one with a nice Morion clock in it. 

[Superdad]

6 hours ago, MartinT said:

I use both sinewave outputs and am not using the squarewave output.

 

Actually, the EtherREGEN prefers squarewave input.  Since you have both, why don’t you go ahead and compare them? 

[Superdad]

3 hours ago, basillus said:

The rest are 0 ohms.

 

Zero Ohms?! That’s called a PCB trace—though even those have some resistance. 
I mean yes, one can buy 0-ohm resistors, their only purpose (AFAIK) being to jumper spots where the circuit no longer needs an actual resistor.

[Superdad]

21 minutes ago, Confused said:

Out of interest, what is the technical reason for this?

 

Let's let @JohnSwenson tackle that one.  He has explained it to me several times--including recently--but today I have the memory of a goldfish as well as a pile of e-mail to return. 

[Superdad]

4 minutes ago, PYP said:

Interesting.  From the photo of the back of the MA models, it indicates 15V 1A (and 13.5 - 16v).  Isn't the JS-2 limited to 12V?  I ask because I have a JS-2 and am considering a reasonably-priced external clock.

 

Upon request we now offer to adjust and relabel one output setting of one of the JS-2's two rails to 15V. With 120/240V AC input it can deliver continuous 4.2A @ 15V (that's less than our 7.4A continuous from one output at 12V--with 120/240V mains). With 110/220V AC mains our 15V setting can deliver 2.2A max.  Plenty for a Cybershaft clock. Even loading the other rail with a modest load--say an EtherREGEN at 12V works out.

 

We have shipped a few units this way, and AfterDark in HK is starting to order all theirs with such.

By the way, I have been told that Cybershaft will, for top-of-the-line OP20A/21A or OP20A/21A-D units, be moving back to the same beautiful chassis series as JS-2, just like it had been for now discontinued Limited2 series. No need for heatsinks on top of chassis and stamped faceplate.  But that falls outside your definition of "reasonably priced clock." 

[Superdad]

16 hours ago, zerung said:

MORION 89A (Frequency range 4.096 MHz ~ 10 MHz)  Offset Noise at 10Hz-155Dbc/Hz,   Double Oven 

Sorry, but you are reading the Morion datasheet incorrectly.

That -155dBc/Hz figure is at 10,000 Hz not 10Hz.  And it is for a 5MHz clock.

For a given construction the phase-noise generally worsens by 6dB per octave of fundamental. 
The MV89 datasheet shows 10Hz typical phase-noise as -130dBc/Hz for a 5MHz clock. So that will be about -124 for a 10MHz sample.

Basically the same as the DAPU (Which from the BG7TBL guy are for certain surplus pulled from 4G towers.)

https://www.morion-us.com/catalog_pdf/mv89.pdf
 

-155dBc/Hz at 10Hz from a 10MHz clock?  You’d have to pay at least $1,000, order 10 pieces and wait 20 weeks. 

[Superdad]

4 minutes ago, elan120 said:

Might want to check with Pink Faun to see if the physical size of their Ultra Clock will fit inside the existing eR case.  Their Ultra OXCO oven case plus PCB is >21mm, and eR external case height is 30mm.

 

There is not a chance in heck that it would fit! Plus there is power supply and heat to deal with.

It is also most likely that the act of removing the Crystek 575--from its tight spot between the 'B'-side RJ45 and the BNC jack--will damage other things. The pins of our clock synthesizer are just 4mm from the 575--and tiny 0402 size capacitors on both sides of the clock.

 

The only way I could see doing something--to make use of the $1,400 Pink Faun Ultra OCXO--would be to get a 10MHz version and solder its wires directly to the two PCB pins of the EtherREGEN's BNC jack. Then set switch to External as usual.  But what really is the point of that, other than short wires, when one could just buy the the similarly spec'd Cybershaft for that amount.  Oh wait, I see that Pink Faun is not even offering a 10MHz version of their Ultra clock board. Only their less expensive ($620) OCXO board is sold with 10MHz--and for that one they do not provide any phase-noise specs (probably more typical, like what you get with the BG7TBL units--which are a deal because they use reclaimed from 4G tower clocks).

[Superdad]

3 hours ago, Clockmeister said:

Hello Superdad

Like your products very much by the way 😎

Thank you very much!

 

Quote

With regard to the Crystek 575 clock, are you looking to replace this item in any up and coming re-engineering of the Ethernet regen at all?

We have no plans to replace the Crystek 575 in the current EtherREGEN.

We have on paper an extremely elaborate future design (likely $2K-$3K), but with other projects in front of it I am sure it could be a year or more until release. Such would likely incorporate a much more expensive clock. (Remember, even at OEM 100+ pieces, the really good phase-noise OCXOs start at $350 for the part. Sorry, not going to screen surplus clocks or buy questionable Chinese OCXO production.)

 

Quote

While the specifications are quite good I have noticed a fair amount of harmonic noise does extended well into the 2Ghz territory in numerous products that use this device, which may possibly have an effect on surrounding circuitry?

I can tell from some of your posts in other threads that you are technical person with some design experience, so I will not dismiss your comment. Such should be addressed by @JohnSwenson as I am out of my depth here.  But harmonic noise in 2GHz range? Surely you are not referring to any phase-noise in that region. Is this some property of AT-cut crystals as used in XO versus SC-cut typical in OCXO? I have read a little on various crystal cuts and have not seen reference to such noise. Is this something you have measured?

 

Do remember that in EtherREGEN we utilize a precision clock synthesizer (we need 4 separate clock lines, 2 each at 25MHz and 250MHz) and run all clock lines as LVDS (differential). That's something you rarely see even in top-dollar DACs. 

[Superdad]

Well AliExpress is very easy for me (order things from AliExpress and Alibaba all the time). Much easier that the Chinese-only Taobao.com.

 

So I just went ahead and ordered one of these BG7TBL units to try.  

 

Because really, nobody can purchase new those DAPU O55A OCXOs for even 1/4th the price they are getting for the whole box (with 4G tower reclaimed clocks).

I did receive a solicitation from a fellow in China offering to sell us unlabeled low-phase noise OCXOs (typical -130dBc/Hz at 10Hz for 10MHz version) for $100 each at 250 piece quantity order. But even if we take samples and test (John now has $18K Jackson Labs Phase-Station to do that! ), there is no guarantee that production parts would be consistently as good as the samples--and plotting/sorting ever part is out of the question for any sort or reasonable price product. Takes WAY too much time--set-up, warm-up, etc.  That's what Cybershaft offers with their graded units--for which each one comes with its own phase-noise plot.

[Superdad]

31 minutes ago, LowMidHigh said:

No one has pushed the whole of Asia off the map, and I don't have to be careful with my views, unless promoting violence or hate. 

 

Enough please!  This is my forum area and I won’t permit it to devolve into geopolitical generalization.

 

It is also inappropriate for you or others to defame Cybershaft here. Opinions about their products are fine to express, but to impugn the company or its owner—who is not here to defend—is not something I will allow.

 

Thank you.

[Superdad]

2 minutes ago, Chrome said:

The Bartok has a master clock output that is 44.1 kHz. Can I use this output with the eR?

 

Hi:

Sorry, but the EtherREGEN accepts only a 10.00MHz clock as an external reference. Will not sync to audio sample-rate related clock frequencies.  So no, you will not be able to utilize the word-clock output of your Bartok with your EtherREGEN.

 

Best,

--Alex C.

[Superdad]

On 7/25/2020 at 8:37 AM, MartinT said:

It's the same AliExpress link as I posted on page 1. Send a message to the vendor Hamradioshop Store that you want a 75 ohm version. He will confirm and tell you to place the order with a reminder message. You will receive a custom 75 ohm version. Best to ask for the DAPU OCXO as it's a good one.

 

I ordered exactly that just over 1 month ago. I should have picked a more costly shipping method because the mail in and out of China are super slow and my unit has been stuck somewhere for weeks.  Hope it doe not get lost entirely. 

[Superdad]

5 minutes ago, roberto2 said:

I have found this page with data sheet of the FE-180 OCXO:

http://pmod27b6c.pic34.websiteonline.cn/upload/FE180datasheet_t3gq.pdf

It seem to be similar to the DAPU if I am not wrong...

 

Yep, -125dBc/Hz @10Hz--same as the DAPU.  Remember folks, the only reason these units are as inexpensive as they are is because they use surplus OCXOs pulled from 4G cell tower gear. The seller probably buys a lot and gets these from about $20 each. So it is not surprising that they run out of one brand and move to another. Of course there is no guarantee of phase-noise performance on these reclaimed parts. (When I finally receive mine I'll probably send it up to @JohnSwenson for him to plot it with his expensive Jackson Labs PhaseStation. ).

For the price you should be quite happy. To buy a new -125dBc/Hz @10Hz 10MHz OCXO--even as a manufacturer ordering in quantity--one would have to pay at least $275-$350.

And as mentioned previously, -125dBc/Hz @10Hz is about what it takes for a 10Mhz clock to audibly surpass the EtherREGEN's 25.0MHz Crystek CCHD-575.  25.0MHz is 1.5 octaves above 10.0MHz, so its typical -110dBc/Hz (@ 10Hz offset) can be considered equivalent to -119dBc/Hz if it was a 10MHz clock (about 6dB/octave is safe to draw comparisons between frequencies). And it seems plausible that the adjusted-for-frequency difference of 6dB--between -119 and -125--would be about the threshold for what could be audible with our switch.

This assumes use of a good short clock cable, matched impedances, and use of square wave...

 

[Superdad]

1 minute ago, Exocer said:

Yes, I am.

 

Yes, despite the BG7TBL clocks being labeled "Max 18W", I too have found that I can power mine with an UltraCap LPS-1.2 (set to 12V of course).  In fact, based on LPS-1.2 case temperature after a few hours, I am guessing the clock is drawing only about 500~650mA.  It likely draws much more in the first few minutes while warming up the oven of the OCXO, but even that is still within the 1.1A capability of every LPS-1.2.  

[Superdad]

6 minutes ago, Exocer said:

Well my wife helped me do a blind test switching cables.

....

Just a fun experiment for a Sunday.

 

Wow, your wife is good sport--a keeper for sure! 

[Superdad]

13 minutes ago, jamesg11 said:

Will they respond to requests for the DAPU version when ordering, or you get what you get?

No, you get what you get. The specs for the FE-180 are essentials the same as the DAPU. -125dBc/Hz at 10Hz offset (which, when corrected for 10MHz versus 25MHz Crystek 575 performance equals about 6dBc/Hz @10Hz better performance than the EtherREGEN's internal clock).

 

But unless someone is putting those clocks on a proper phase-noise analysis station we really don't know for sure if those surplus DAPUs or FE-180s (pulled from cell-tower boards) are meeting their specs, sample-to-sample.

[Superdad]

11 minutes ago, Exocer said:

Speaking of tests...has yours arrived yet? Still planning to test it? 

 

It has--and I have been too busy to even give it a listen. 

 

11 minutes ago, Exocer said:

It would be interesting to see measurements of at least one of these clocks.

 

I'd need to send it up to @JohnSwenson for him to graph it on his fancy Jackson Labs Phase-Station.  Such tests take time to set up properly--and he is pretty busy on our other development projects this month.

 

[Superdad]

3 hours ago, Exocer said:

For the sake of my understanding, are you saying that the use of LPS 1 or LPS 1.2 for the clock or ER vs an SMPS would not make for an audible difference? Or were you comparing the LPS 1 to LPS 1.2 when power the ER or Clock?

Just to be clear: The original UltraCap LPS-1 can not be used to power either an EtherREGEN or a BG7TBL clock. The current generation UltraCap LPS1.2 can be used with either.

This is because while both UltraCap models are capable of delivering up to 1.1A at any of their output voltage settings, the highest output voltage setting for the original LPS-1 was 7V. And:

a) The BG7TBL clock requires 12V;

b) An EtherREGEN run from 7V draws about 1.35A—a bit over the limit for out UltraCap units (whereas at 12V the EtherREGEN draws just 0.8A).

 

[And yes I saw that @R1200CL said he his is using a pair or original LPS-1 units in series (one set to 5V, the other to 7V) to obtain 12V.  But that seems to be getting lost in the shuffle of these posts today.]
 

 — Alex C.

 

[Superdad]

13 minutes ago, jacques_racine said:

Superdad,

Don’t think. Just answer. What’s this new project? Quick!

 

LOL.  Well I am hoping one of them is a time machine--to make more hours in the day to get stuff done. That or a cloning device. 

 

But seriously, this is a very fertile time for John and I.  Things are inching forward on four, possibly five projects concurrently. I really don't know which of them will enter production first--at least three are jockeying for position. Still, all are months away and it is too soon for any bean spilling. 

It is very fun. For me at least. John is doing all the really challenging brain work... 

[Superdad]

51 minutes ago, Avalfa said:

Now I am trying to buy some cable, to be precise Belden 4731R since it's had very nice specs especially at 10 MHz. But I can only find 305m reels for sale.

 

Good luck finding that big stiff 14AWG Belden for sale by the foot! 

But next best will be the new Belden 16AWG 4794R.  It is available by the foot at:

https://www.markertek.com/product/bl-4794r-ft-bk/belden-4794r-12-ghz-4k-uhd-75-ohm-16-awg-rg7-precision-video-cable-black-per-foot

And on Amazon you can find it preassembled:

https://www.amazon.com/UHD-SDI-Cable-12G-SDI-Custom-Connection/dp/B076JLSDCF

Even in lengths as short as 1 foot! 

 

[Superdad]

6 hours ago, sgr said:

My Etherregen Chinese clock system stops playing after about an hour. This causes me to hard restart the Clock, the EtherREGEN, and sometimes also my Aries G2 streamer. 

....

The EtherREGEN plays without stoppage when its internal clock is used. 

 

Try another clock cable.  We have seen this before--and with MUCH more expensive clocks.

Even the most brief interruption (3-5 µSeconds) of the clock to the EtherREGEN will cause a problem and require repowering of the switch (actually, if the interruption is not long, the 'A' side will recover but not the 'B' side). 

[Superdad]

9 hours ago, jacques_racine said:

Is there a ready made LPS you can recommend that would fit the bill?

I (and others) power the BG7TBL clock with an UltraCap LPS-1.2.  Works quite nicely. 

 

2 hours ago, GMG said:

Well, I really like Teddy Pardo power supplies with my equipment.

Teddy's units are very nice and he is a good guy!  Please remember that this thread is in the UpTone sponsored forum area. 

[Superdad]

2 hours ago, jacques_racine said:

House is full of uptone magic.

I know Jacques... 

 

2 hours ago, jacques_racine said:

My understanding was that unit performed better with 12.5v.

The spec peck above the DC jack on the BG7TBL clock quotes a range of 11.7~12.9V.  

12V works fine for me.