John Swenson tutoring paper: "Considerations regarding usage of external reference clocks" (EtherREGEN and other): Sine vs. square, impedance, cables—and a money saving surprise at the end.

[Superdad]

Many of our friends here at Audiophile Style have been waiting for us to publish John Swenson's introductory explanations concerning choosing and implementing an external reference clock—be it for connection to an EtherREGEN or some other audio device.

 

[I take the blame for the delay. John gave me the first draft of his paper just over two months ago. The concepts were pretty clear but I think he wrote it late at night so it needed a few editing/formatting passes. I finally had time this past week to prepare it for presentation.]

 

Here is a link to the PDF for offline reading: Considerations regarding 10MHz external reference clocks...

And below are the JPGs of the pages.

 

We do not expect that this paper will answer all questions about the complex realm of high-frequency reference clocking, but it is a start. Nor does this paper get deeply technical and there are some generalizations.

 

After reading it, feel free to ask questions.

 

As teased in the thread title, at the end of the paper we present a shortcut to getting stellar performance out of less than state-of-the-art clocks.

If there is interest, we may begin stocking the item mentioned--mostly for convenience of our overseas clients as the company selling the small, inexpensive item charges $120 for shipping to destinations outside the USA.

 

Enjoy!

--Alex C.

===================

 

 

[Superdad]

21 minutes ago, James Stephens said:

Are there any other clocks, besides I imagine the Mutec Ref 10 and Ref 10 SE 120 (?), for which we might expect the square wave output into an ER to surpass its sine wave output (+ of course the Mini-Circuits filter)?

 

[When you accidentally posted your question in AfterDark’s “Filter group buy” thread I answered there. But since you caught that the question is more appropriate here, I’ll just repost my long answer below...]

 

Hi James:

While I understand what you are asking, the phrasing of the latter part of your question does not really fit. You seem to imply that there are first-rate clocks that offer both square and sine waves from the same box. That’s not the case with the Mutec (square only) nor any of the better boxes from Asia (mostly sine) that I am aware of.

 

As John explained in his paper, it takes a lot of care to both produce a really good square wave AND to preserve that quality (via cabling) all the way to the target device. Not that it can’t be done.  And OXCO modules (the circuits in the cans along with the crystal) are available with square wave output. But how that is implemented and handled varies greatly.

Also, some companies (and I have no idea about what AfterDark does to offer square wave in some models) use sine>square circuits between the OCXO module and the output.  There exist a couple of chips that do this extremely well, but then there are the rest which are rather poor at it (and then the truly terrible squarer mess that is in the BG7TBL boxes).

So with inexpensive square wave output clocks you really never know what you will get. Yet with sine wave clocks you need to be very careful to filter to a very clean wave—best done at the endpoint target device end.

 

Back to your question—which likely was more meant to be something like:
“Can a really low-phase noise sine wave output clock—terminated with the Mini-Circuits filter—outperform a first-rate square wave box such as a Mutec REF10 SE20 (assuming $$$ also spent on a cable for that square wave)?

 

I certainly can’t answer that and I think John would decline to as well without making extremely careful (and rather difficult to set up) measurements.  And even the characteristics of the receiver in the endpoint target would likely have significant affect on such a comparison.

 

I write all the above primarily so that we can all begin thinking about these things from the proper angles. It’s confusing stuff! 

[Superdad]

24 minutes ago, LowMidHigh said:

Secondly, I believe you misconstrue James' question. He  was interested to know if there were other excellent square-producing clocks, apart from Mutec's, which could outperform a sine+filter.

Seems it was, but I wanted to clarify and to expound a bit on the choice and decision process. 

 

24 minutes ago, LowMidHigh said:

Put differently, could you list the brands whom he could trust to have the proper internal circuitry to produce a stellar square wave.  

We are not presently surveying the whole clock market, nor would it be possible to assess any models without looking very closely as their circuits (really good clocks are not cheap!) 

However, John did this with the BG7TBL that I sent him and found quite a lot to criticize in that piece with regards to the squarer circuit. In fact, even the sine from the OXCO itself goes through the lousy squarer circuit and then a poor filter I think. Maybe he will speak of what he found as I think it might be educational. Certainly that box was not designed with audiophiles in mind. Would have been better just to run the OXCO module in the thing straight to a BNC jack.

Still, as he pointed out, putting the Mini-Circuits low-pass filter on it at the target device end will yield cleaner and much more worthwhile sine wave.

 

19 minutes ago, LowMidHigh said:

I have a question fo my own here to John. Given we have the figures for both the Bedlen 1694A and Canare LV-77S, which of the two would he recommend over the other? Going by the numbers alone, the Bedlen is superior, but some people here claim that the Canare "sound" better. Kind of a contradiction, so I hope he could weigh in. 

Five of your six posts in this thread have been about clock cables for square wave clocks. But the thrust of the paper and the concepts we are wishing to speak about are not that. You could start a new thread on clock cables or perhaps revive one such as this which already exists:

 

 

[Superdad]

16 minutes ago, MartinT said:

1694A is a lesser cable and sounds inferior.

You too Martin. Please let's not divert this thread into clock cable chatter--at least not so early. Don't want to confuse or scare folks off! 

(Especially since one of the main points of the paper is that with sine waves and a filter the clock cable quality almost ceases to matter!)

Thanks,

--Alex C.

[Superdad]

5 hours ago, LEO SOUND said:

If square wave outputs are to be avoided, why the etherregen is "square wave"...

[I will probably state this with slightly wrong terms but conceptually its about right.]: 

 

Clocks exist in digital circuits for the chips to time everything they do from. The chips have to "trigger' off the incoming clock at some set voltage point of whatever the clock's waveform is. The easiest and most accurate (time/spacing/phase-wise) for a device to clock off of is a square-wave, because ideally its rising voltage is straight up at its frequency intervals--it's square-ish!  As John explains in the paper delivering a really good square wave is harder than delivering a good sine wave (and filtering it so that the inevitable harmonics don't disrupt the accuracy for the target device.

 

So it is not just the EtherREGEN that prefers a square wave--all digital circuits do.

[Which is not to say that a smart designer couldn't put a really good sine>square converter at the input of their device. We could have with the EtherREGEN but the really good one we like needs more power supply current that we could provide on the 'B' side with this design. Alternatively, Mini-Circuits makes a PCB-mount version of the exact same filter as the one we are recommending. But it is physically very large and there is no room for it on the EtherREGEN board.]

 

 

Quote

...and therefore you have to buy a sin wave clock with a filter to be sure to have a good signal.

It's just an easy way to assure good performance from a sine wave clock and to save you money on heavily-shielded clock cables.

 

Quote

I'm interested in a Square Wave from AfterDark that claims it's special for etherregen on their website.

@AfterDark.Can you provide details on this clock as it looks like it was produced in collaboration with Uptone Audio.

There was no collaboration on development between AfterDark and UpTone on their clocks.  

AfterDark is a successful dealer for UpTone, they sell a lot of EtherREGEN and JS-2s, they also sell Cybershaft and have gotten that firm interested in EtherREGEN and JS-2 for Japanese market. That is the extent of collaboration between us. 

 

 

Quote

But from the white booklet, we just have the impression that we need Sine Wave because almost no clock does a good Square Wave except the very expensive watches and out of budget for me.

Well it seems you got the point then.  Doing a good square wave takes care--and then the clock cable matters a great deal.  Going sine is easier and more budget-friendly.

 

 

Quote

Sotm is sine wave I think.

Talk about out of budget! If you are going to spend upwards of $4,000 on clock, you could buy a proven Mutec REF10 or top model Cybeshaft.

 

 

5 hours ago, LowMidHigh said:

It's not true that a square wave is to be shunned. Here's what the paper says:

 

"Just remember that the best possible result is still going to be using a square wave clock box with a REALLY good sine to square converter, everything being just right inside the box—AND you use a really low-attenuation-with-frequency cable."

Right. It is just that knowing and doing that all just right is neither a given, nor is it cheap. Don't know that that needs to be repeated too many times as the paper speaks for itself (or should I say for @JohnSwenson). 

 

[Superdad]

52 minutes ago, GMG said:

Would this filter work?

 

https://www.crystek.com/microwave/spec-sheets/filter/CLPFL-0010-BNC.pdf

It will. It's down a little more right at 10MHz (compared the 11MHz lowpass of the Mini-Circuits) but that should not be any problem.  It is quite a steep filter. I thought the Crystek  one was going to be a lot more expensive and a custom order item, but I see that Digikey has a few in stock for just $37 each.  

[Superdad]

5 hours ago, Exocer said:

The pin diagram of these OCXOs is fairly straight forward, however, it would be nice to find a small decent quality PCB board that would allow me to solder the OCXO from the BG7TBL (since at least the clock appears to be relatively decent for the investment on paper) for a complete sine-wave output or two via BNC outputs. Also a 2.1/5.5mm DC input on said board would be great.

....Do you know of any module kits/PCBs one could drop the FE-180/DAPU clock into?

 

For $19 you could try this board:

https://www.ebay.com/itm/OCXO-adjust-PCB-Power-supply-voltage-12V-suitable-for-51X51-51X41-package/313123416940?

 

But desoldering the OCXO from the BG7TBL board may not be terribly easy depending upon how tight those holes are since you have to get all the solder connections to let go so that you can pull the part.  Attempt at your own risk and please don't ask me anything about the above linked board. I just found it on the net and looks to be clean and simpler--a lot simpler than all the unnecessary circuitry that is on the BG7TBL boards.

[Superdad]

6 hours ago, AfterDark. said:

We are comparing with Giesemann sine wave version to Queen Square Wave SE version according to their Phase Noise and Allan Derivation as well. The phase noise table @10Hz and @1Hz is the most sensitive areas, this figure will be useful for audiophile to choose their best clock with indicated measurement.

Hi Adrian:

While we are supportive of you new fine-value clock series, we ask that you (and others) keep conversations about AfterDark products in your own sponsored forum area or in one of the public forums.

We really prefer that this particular thread stay focused on the concepts and facts put forth in John’s paper.

Thank you,

—Alex C.

[Superdad]

48 minutes ago, basillus said:

@Superdad
The 75 ohm BNC output of the  BG7TBL will fit proberly with the the Mini - Circuit BLP -10.7 50 ohm?
It will not be a little loose/flimsy, because there is a difference in the thickness of the isolation  in a 75 ohm and a 50 ohm BNC connector.

They fit just fine. Nothing to worry about.

[Superdad]

1 hour ago, LEO SOUND said:

On the sotm clock, the filter can be integrated then also?

 

“Integrated”, as in filter built into the clock box? Not applicable assuming the box is producing a clean sine wave in the first place (which for the $3,500 SOtM I expect they are). 
No, the point of the low-pass filter is to place it at the far end of the cable (in our case at the EtherREGEN) so that you then don’t have worry about using short and very stiff semi-rigid clock cabling.

 

As well, and perhaps as important, the filter (again, only for sine wave clocks—don’t put one on a square wave!) will assure a very clean, free from harmonics and noise sine, regardless of what might have been transmitted by the clock or picked up the cable.

This is very important because (seen if you read John’s paper carefully) the circuits receiving 10MHz clocks—which always involve a clock synthesizer since that 10MHz is always turned into some other frequency(s)—are quite sensitive to harmonics. They would prefer a really good square wave, but if you feed them a sine wave it would best be an extremely clean one.

 

The above is really a key point of John’s paper so I hope everyone will ponder it further.  But try not to read too much or too little into it. 

 

—Alex C.

 

P.S. Anticipating the other obvious question regarding SOtM (not that I want to talk about their products): Their large clock board, (fitted in all of the “Ultra” models as well as the $1,700 version of their switch) uses a crystal as reference to an older Silicon Labs clock synthesizer—or, when extra is paid to have a BNC jack, the synth gets programmed to use an external 10MHz reference. Like every other (as explained above), that board would prefer a great square wave but will accept a (hopefully very clean) sine wave. Their clock board input contains neither an active sine>square converter chip nor a passive low-pass filter.

So the things we wrote about sensitivity to various factors (say with EtherREGEN) apply to the SOtM units as well. (Again, I am not referring to their OCXO clock box.)

[Superdad]

13 hours ago, R1200CL said:

What version of the BG7TBL did you get ? The one claiming to be square and 75 ohm ?

Yes, the one labeled as having both sine and square wave outputs—with 75-Ohm BNC jacks.

 

13 hours ago, R1200CL said:

Have @JohnSwenson put it on some sort of measurements ? Or intend to do ?

He already looked at circuit and the waveforms. And it wasn’t pretty.

No current plans to run phase noise plots as we see no point and he is very busy with other development work.

 

6 minutes ago, R1200CL said:

But I think it’s also quite clear by now, the 50 ohm version will just do.

That’s what we have been saying from the start! (With sine wave and filter at the target end of the cable.)

[Superdad]

2 hours ago, R1200CL said:

Do you think changing some parameters in the EtherRegen SW can be beneficial when using an external 10 MHz clock ?

No. None at all.  The clock synthesizer already receives correct and optimized programming code for a 10MHz clock when the EtherREGEN is set to run from an external clock.

Other than being more precisely referenced, the 4 differential clock output lines from the synthesizer (two at 25MHz and two at 250MHz to run our special reclocking flops) are exactly the same as far as the rest of the functioning of the switch goes. Nothing to change.

[Superdad]

3 hours ago, LowMidHigh said:

"Just remember that the best possible result is still going to be using a square wave clock box with a REALLY good sine to square convertor, everything being just right inside the box—and you use a really low-attenuation-with-frequency cable."

 

42 minutes ago, MartinT said:

I can only say that sine with the filter sounds better than square with my BG7TBL clock.  Same (good) cables.

Not just a bit better - it's quite noticeable.

There is nothing inconsistent between the above since as John’s examination of the BG7TBL’s circuits revealed, it does not at all qualify as having a “really good sine-square converter.”  It is actually quite poor in that regard. On the opposite side of the solar system in comparison to a Mutec REF10 for square wave performance.

 

Plus with the filter you don’t even have to sweat the cable quality (low attenuation for square; super shielding for sine). 
 

Hence the sine output with filter is the way to go with this inexpensive clock. Heck, I bet the filter on the “squarish” wave output of BG7TBL is better than without even for those versions.

[Superdad]

2 hours ago, PYP said:

Alex, 

What kind of shielding would be considered to be optimal for the sine cable?  Thanks. 

 

Very best shielding (100%) is with the semi-rigid cables--because they literally use a metal tube!

But again, if using the filter at the target end, you don't need to worry much (at all?) about the shielding. 

[Superdad]

41 minutes ago, Mihaylov said:

Can you specify the names of those two chips?

The Linear Tech LTC6957 is state of the art for that.

 

29 minutes ago, Mihaylov said:

What criterion (indicator) do you use to assess the quality of the conversion of sine to square? On what basis do you draw this conclusion about the quality of the conversion of sine-square to BG7TBL and Mutec? Based on product schematics analysis, measurement analysis (which?) or something else? Have any measurements of BG7TBL been taken in order to infer extremely low quality of square wave at the output of the clock?

Yes, John analyzed the circuits of the BG7BTL unit I sent him, gave details of his findings publicly, and also made measurements at various places in the circuit.  Someone else here published the nasty-looking square wave output of it as well. It really would have been better if the BG7TBL guy just installed those reclaimed OCXOs on a plain board with nearly no circuitry. (That's about what AfterDark does, but with new-old-stock CTS clocks which they grade with their Symmetricom.)

 

The Mutec REF10 is in an entirely other league.

 

[Superdad]

6 hours ago, Mihaylov said:

Do I understand correctly that in order to convert the 75 ohm version of ER to 50 ohm, is it enough to replace the BNC connector with a 50 ohm connector and replace the R86 resistor with 50 ohms?

That is correct.  However, removal of the 75-Ohm BNC without damaging the circuit board is difficult. The BNC jack, in addition to it's two small pins for electrical connection, has two large diameter alignment/securing pins. Unsoldering all pins--with full release so that the jack can be wrestled off the board--takes time, patience, and good nerves. We are not responsible for torn up traces resulting in a ruined board.

I have done the conversion once--took me nearly an hour--and have sworn not to do it again. 

 

From the next run of 225 EtherREGENs (boards due mid-January), 15 units will be made with 50-Ohm BNC (and matching termination resistor). Anyone can request reservation of one of those when they order.