A novel way to massively improve the SQ of computer audio streaming

[AmusedToD]

56 minutes ago, austinpop said:

Thanks for writing this up, John. And thank you again for hosting us!

 

 

Did you mention the effect of just the reclocked switch to your chain? That was our first experiment. Since your music lives on the NAS, it benefits from the switch even for local playback. I heard what I usually do with this change, but I'll let you describe in your words!

 

 

This is consistent with my own and Eric @limniscate's experience using the Zenith SE, @mozes's recent experiments,  and several others.

 

One of the most important things I wanted to try in John's system was to power the trifecta with the SR-7. This was the burning question - was the trifecta in our previous experiments being held back by the PSUs we had on hand? And would the SR-7 catapult the trifecta above the custom server. The fact that it didn't suggests that beyond a point, the better PSU does not raise the performance of the trifecta.

 

However, I would still say the sMS-200ultra has a place. If you're handy, and willing to build a system (not technically hard, but does require attention to detail and a certain facility, especially installing OSes, drivers, etc), then an sCLK-EX modded system like @Johnseye's and @ElviaCaprice's is a very economical way to go. The SE (or other similar server) approach adds cost, but does not require assembly, while the sMS-200ultra is a lower-cost option that is comparable to the DIY server approach. Let's face it, the differences we heard between the 2 setups were quite subtle, so I wouldn't feel comfortable declaring one the winner or loser.

 

I will reinforce one common theme in all of these - the tX-USBultra at the end of the chain, right before the DAC, ideally disciplined by a 10MHz reference clock. I think this is the glue that binds all these approaches together.

 

 

This is a recurring theme, I feel. The further upstream you optimize, the more muted the improvements are. Eric and I also had a hard time distinguishing between the SE and his vanilla Roon Core when comparing the SE chain to the trifecta chain.

 

 

 

This one truly surprised me. I also heard this, but I'd like to repeat this comparison with more time. Hopefully Eric's SR-7 will be here real soon now, and we'll have that opportunity. As I've mentioned, PSU differences can vary based on the component being powered, especially based on their current demands. I have also found empirically that the tX-USBultra and the DAC are the most sensitive to power quality. I want to repeat this experiment using the SR-4 and SR-7 powering the tX-USBultra and the Brooklyn DAC+ if I still have it.

 

I also want to control for the dual rails in this experiment. I want to use just one SR-7 rail active, and compare with when both rails are active.

 

Still, the good news here is that for low current applications, given the Sr-7 isn't really available anymore, the SR-4 is a really attractive option, especially if you happen to live in UK and Europe. How often does that happen!

 

 

 

One caveat I would add is that in our tests at John's, the Zyxel switch, while being driven by the sCLK-EX, was not reference clocked. So it wasn't quite an apples-to-apples comparison. With more time, we could have swapped the Cybershaft input to the sMS-200ultra that was sCLKing the switch. This illustrates the limitation of the single-output OP-14. 

 

Still, the Linear Solutions switch seems like a winner. More importantly, it seems to be a no-brainer for those who do not have sCLK-EX components in their chain, but want to benefit from the same mechanisms.

 

Regarding the SOtM switch, don't dismiss it yet. Remember, it will have an option to purchase with a 25MHz SMB input, rather than an internal sCLK-EX board. I hope SOtM will price this configuration attractively. In this mode, it would be a replacement for the modded switch people are using. 

 

This switch space is heating up, so I think we're in for an exciting few months as SOtM, Uptone and others come out with new products for us to try.

 

 

Yes, I credit @limniscate for convincing me of the importance of level matching. We did this the first time we compared his Yggy to my Codex. We first listened without careful level matching, and formed a strong preference for one. After level matching, all that went out the window, and we were hard pressed to tell them apart!

 

 

Rajiv, for those of us without any sCLK components in the chain, the LS switch sounds like a good prospect, but I get your points.

 

Anyways, I had a revelation myself during the past several days. The magnitude of improvement was almost like adding a dCS DAC to my chain as a replacement for the Holo Spring lvl3. 

 

I am talking about a Bassocontinuo HiFi rack (completely decoupled through their DSD mechanism) which replaced my previous shelf (yes, it was just shelf, a piece of furniture and nothing more). The music simply exploded in all directions, with significant reduction of glare and astonishing depth. I am amazed, honestly I never expected an improvement of such magnitude from a rack. Only goes to show that every little thing matters in this hobby of ours.

 

 

[elan120]

19 minutes ago, Johnseye said:

I recently ordered a Holo Spring Level 3 DAC.  I'm looking forward to the many possibilities this brings.  It will be a new chapter as I'll also have a new sCLK motherboard and the ability to upsample to DSD512.  I'll be sure to compare with level matching this time around.

I am looking forward to this comparison as well.  What input will you be using to feed Holo Spring DAC (USB/I2S...)?

[Johnseye]

11 minutes ago, austinpop said:

 

It helps to think of this comparison as follows: how do I get OCXO-level clocking delivered to the switch? In the Linear Solutions design, it is to engineer an OCXO chip directly in the switch. In the sCLK-EX approach, it is to take an OCXO reference clock, connect it to the sCLK-EX master clock input, and distribute to the switch. Without the reference clock in the modded switch, we're not getting the OCXO-level quality. I realize this is a non-technical description. What does it really mean to use an OCXO reference clock to discipline an sCLK-EX clock generator? I guess in layman's terms, and based on what we are hearing, the "quality" of the OCXO reference clock seems to propagate to the outputs of the sCLK-EX.

 

If Linear Solutions offered a master clock connection it would, or could, receive further benefit from synchronized timing with the other devices.  The sCLK is very good.  How good vs OCXO without a master clock is now questionable, at least for a switch.  Is there more to Adrian's switch?  Would the sCLK benefit from OCXO?  Exciting time to be into audio  

[Johnseye]

3 minutes ago, elan120 said:

I am looking forward to this comparison as well.  What input will you be using to feed Holo Spring DAC (USB/I2S...)?

 

Great question.  Holo recently improved the USB to where it should be similar to i2s.  I'm talking with someone in the forum now about buying his Singxer for i2s.  This adds another layer of complexity however.

[elan120]

1 minute ago, Johnseye said:

 

Great question.  Holo recently improved the USB to where it should be similar to i2s.  I'm talking with someone in the forum now about buying his Singxer for i2s.  This adds another layer of complexity however.

I am looking forward to your findings as I also have Holo Spring DAC L3, but older not with the new X208 USB input.  I use Sinxger SU-1 to send signal via I2S to Holo Spring DAC, but I modded mine recently to be sCLK-EX clocked and fed by sCLK-OCX10 with great result.

[Johnseye]

1 minute ago, elan120 said:

I am looking forward to your findings as I also have Holo Spring DAC L3, but older not with the new X208 USB input.  I use Sinxger SU-1 to send signal via I2S to Holo Spring DAC, but I modded mine recently to be sCLK-EX clocked and fed by sCLK-OCX10 with great result.

 

May and Lee were talking with me about that yesterday.  That would mean another sCLK for me.  More cost.  You wouldn't happen to live in the Chicago area would you

 

[elan120]

1 minute ago, Johnseye said:

 

May and Lee were talking with me about that yesterday.  That would mean another sCLK for me.  More cost.  You wouldn't happen to live in the Chicago area would you

 

Sorry, I wish I am, especially last week when you are having so much fun with these great people.

 

I can only comment on the comparisons I did when I took different steps modding my SU-1, and the improvement from sCLK-EX / OCX10 combo was beyond all the mods I did previously.  In my case, having the other two sCLK-EX outputs to both txUSBUltra and txUSBexp sort of "even" out the overall cost a little.

[spotforscott]

Very interesting findings. I am going to have some learning of my own next month. I have ordered a couple of things from Linear Solutions Design:

 

* OCXO Switch with custom LPS

 

Modded Oppo 203 with the following Linear Solutions upgrades:

* LPS: Nichicon KA Caps/ M-Cap upgrade

* OCXO Clock 

* New custom wiring throughout with 7N OCC Copper 15Pin 
* Furutech FI-06 NCF IEC upgrade  
* EMI treatment

* Aluminum Stabilizer

 

I will be testing it against my Trifecta set up. Attraction to modded oppo player is optical output to Chord DAVE. Hope is that my Sonic Transporter i7 (bridged) connected to the Linear Solutions OCXO switch, feeding the modded Oppo player via ethernet cable with optical out to the DAVE will hold its own against the Trifecta - using USB connection to DAVE. 

 

Will be testing using a direct connection from DAVE to hi-efficiency speakers (very transparent set-up) and with recently broken in Abyss Phi headphones.

 

I will be sure to report back on findings...

 

[Ralf11]

Thanks for posting that writeup, John (and for hosting the listening test, which is helpful to others not there as well - I hope you got some free beer or wine...)

 

The Cybershaft OP-14 is Rubidium based, is that right?

[jacquesr]

On 2018-04-16 at 9:18 AM, jaspal kallar said:

I presume it can play back AIFF files (among others), however ripping only supports FLAC & WAV. Is that correct?

Yes you are correct - I checked with Innuos.

[Ben2300]

33 minutes ago, Ralf11 said:

The Cybershaft OP-14 is Rubidium based, is that right?

OP-14 is OCXO based.

[mozes]

13 hours ago, auricgoldfinger said:

2 x 12v/6A rails with dual regulator per each output (fixed voltage)

Do you know what kind of regulators they use?

[BigAlMc]

12 hours ago, AmusedToD said:

 

Thanks for the detailed analysis. You guys had some nice fun together, great to see an Internet forum bringing like minded people together like that!

 

Anyways, your collective finding that the Linear Solution OCXO switch is better than the SOtM sCLK modded switch is in line with Roy’s findings. That kind of narrows it down. The Linear Solution gets you the switch plus a linear PSU for only $649, whereas the SOtM switch (with a sCLK board) will probably be close to a grand (speculation based on their previous products pricing), and that’s without a PSU.

 

The LS switch will probably be the better buy, unless the upcoming Uptone switch kills both for less than $500.

 

I see that @Johnseye and @austinpop have addressed elements of this in subsequent posts but I think this is a very interesting area.

 

For the most part I agree with your analysis. However for me the SOtM ace card is the ability to use a good master clock on several components. So for me at least the key differentiater between the Linear Solution switch and the upcoming SOtM switch is the clock input on the SOtM. 

 

I recently added the SOtM sCLK-OCX10 to my stack and whilst it's still burning in I've heard enough to know it's a keeper. It's powering my TX-USBultra only. 

 

I also have an Aqvox SE switch (powered by an LPS-1) which has done marvels for my Tidal streaming. 

 

I therefore wonder how the Linear Solutions would compare to the the Aqvox. But on paper I assume they'd be fairly close. 

 

I also wonder how the SOtM switch would compare, especially when it was connected to a master clock. On paper here I presume there are two advantages to the SOtM route. A. The superior clock. B. Synchronisation between the switch and the TX-USBultra. 

 

And the 64,000 dollar question is whether any improvement is SQ would justify the cost of trading the Aqvox for the SOtM switch! 

 

Either way I agree with Johnseyes comment that it's a great time for our hobby. 

 

Cheers, 

Alan 

[austinpop]

12 hours ago, Johnseye said:

 

If Linear Solutions offered a master clock connection it would, or could, receive further benefit from synchronized timing with the other devices.  The sCLK is very good.  How good vs OCXO without a master clock is now questionable, at least for a switch.  Is there more to Adrian's switch?  Would the sCLK benefit from OCXO?  Exciting time to be into audio  

 

John, the master clock does not apply to the Linear Solutions approach. Consider this. A generic switch has a fixed frequency (say 25MHz) oscillator (i.e. clock) in it. When you use an sCLK-EX approach to improve the switch, you are doing this:

• sCLK-EX's on-board 10MHz reference clock ---- drives ------> frequency synthesizer ----> synthesizes 25 MHz ------ delivers over SMB to -------> switch

When you connect a 10MHz master clock to the sCLK-EX, you are using the superior OCXO clock in the master clock to "discipline" the sCLK-EX's reference clock. So the new chain, with improved SQ, looks like this:

• OCXO-based 10 MHz reference clock ---------- disciplines -----> sCLK-EX''s reference clock ---- drives ------> frequency synthesizer ----> synthesizes 25 MHz ------ over SMB -------> switch

Presumably, the reason we hear an improvement with the above is that the OCXO "goodness" of the reference clock somehow propagates all the way to the switch.

 

The Linear Solutions approach is different. It is putting the OCXO directly in the switch. It is replacing the generic 25MHz clock in the switch itself with an OCXO 25MHz clock. There is no synthesizer, so the design does not lend itself to an external reference clock. This is a different approach.

 

Does that make sense?

 

As for your point about "synchronization." I've reported on this before. Eric and I have done experiments where we constructed chains where all the components were clocked from a single sCLK-EX board (i.e. synchronized), vs. each component with its own sCLK-EX board (i.e. unsynchronized). We heard no difference in SQ between the 2, suggesting no SQ benefit from synchronization. This suggests the improvements come from lowering the phase noise, not necessarily running all components in lock-step.

 

I hope this clarifies things.

[austinpop]

2 hours ago, BigAlMc said:

 

I see that @Johnseye and @austinpop have addressed elements of this in subsequent posts but I think this is a very interesting area.

 

For the most part I agree with your analysis. However for me the SOtM ace card is the ability to use a good master clock on several components. So for me at least the key differentiater between the Linear Solution switch and the upcoming SOtM switch is the clock input on the SOtM. 

 

I recently added the SOtM sCLK-OCX10 to my stack and whilst it's still burning in I've heard enough to know it's a keeper. It's powering my TX-USBultra only. 

 

I also have an Aqvox SE switch (powered by an LPS-1) which has done marvels for my Tidal streaming. 

 

I therefore wonder how the Linear Solutions would compare to the the Aqvox. But on paper I assume they'd be fairly close. 

 

I also wonder how the SOtM switch would compare, especially when it was connected to a master clock. On paper here I presume there are two advantages to the SOtM route. A. The superior clock. B. Synchronisation between the switch and the TX-USBultra. 

 

And the 64,000 dollar question is whether any improvement is SQ would justify the cost of trading the Aqvox for the SOtM switch! 

 

Either way I agree with Johnseyes comment that it's a great time for our hobby. 

 

Cheers, 

Alan 

 

Alan,

 

If you already have an OCX10, and one unused tap on the tX-USBultra, then getting either a modded switch now, or waiting for the new SOtM switch is a no brainer.

 

I hope you noted my point earlier that SOtM will have an option to buy this switch without an internal sCLK-EX board, and just an SMB connector for external clocking. I doubt this will be as expensive as a Linear Solutions switch.

 

The Linear Solutions switch makes a lot more sense to those who don't have an sCLK-EX component, or a free clock tap on their sCLK-EX.

 

See my previous post right above this one regarding synchronization.

[BigAlMc]

23 minutes ago, austinpop said:

If you already have an OCX10, and one unused tap on the tX-USBultra, then getting either a modded switch now, or waiting for the new SOtM switch is a no brainer.

 

Thanks Rajiv,

 

But would connecting the sCLK-OX10 directly to the SoTM Switch not make more sense that going via the TX-USBUltra?

 

The OX10 comes with 4 BNC clock outputs.

 

Cheers,

Alan

[austinpop]

Just now, BigAlMc said:

 

Thanks Rajiv,

 

But would running connecting the sCLK-OX10 directly to the SoTM Switch not make more sense that going via the TX-USBUltra?

 

The OX10 comes with 4 BNC clock outputs.

 

Cheers,

Alan

 

Sure - if you want to spend more money.

 

If you have unused taps on the sCLK-EX in your tX-USBultra, then why buy another sCLK-EX in your switch?

[BigAlMc]

On 17/04/2018 at 4:52 AM, austinpop said:

The switch can be ordered either with an SMB input, to be supplied by an existing sCLK-EX board in your system, or with an sCLK-EX board in the box, with an optional reference clock input

 

Hi Rajiv - I wasn't following what you meant and had to reread your Axpona post. Think I've got it but just to clarify.

 

The sCLK-OX10 has 4 BNC clock outputs. This can be connected to a TX-USBUltra which has an sCLK-EX in it.

 

The SoTM switch without an sCLK-EX in it can accept the clock signal via "SMB Input" from an external sCLK-EX, right? But cannot accept the clock signal directly from the sCLK-OX10 - is that what you meant?

 

Assuming my understanding above is correct then that seems a bit disapointing. The sCLK-EX is essentially a clock. So if I don't need that clock because I have a superior one then why doesn't the switch accept the BNC in from the sCLK-EX. I guess there's a technical aspect I'm missing - as opposed to just more revenue for SoTM.

 

Thanks,

Alan

[spotforscott]

In discussions with Adrian from Linear Solutions, he said that they do not believe external master clocks are the best approach from a sound quality perspective because the distance between clock and the chipset is critical. That's why they put OCXO clocks inside their units. I have heard a lot of chatter on hear about cable lengths to clocks and it appears he is taking that focus to an extreme. 

 

His rationale seems to make sense but I would like to hear it with my own two ears. Roy spoke very highly of what he heard with their approach (more of "you are there") and so I took the plunge and purchased their switch and modded oppo player to hear for myself.

 

It is truly an exciting time in audio. My system has never sounded better than it does right now...
 

[rickca]

3 hours ago, BigAlMc said:

But would connecting the sCLK-OX10 directly to the SoTM Switch not make more sense that going via the TX-USBUltra?

The only SOtM component that optionally provides a master clock connector is the sCLK-EX board.  So you would need an sCLK-EX board in the SOtM switch in order to connect the sCLK-OCX10 directly to the switch.  

 

As Rajiv said, if you have unused taps on the sCLK-EX in your tX-USBultra it's more cost effective to just install a clock connect PCB in the SOtM switch and leverage the connection of the sCLK-OCX10 to your tX-USBultra.

[BigAlMc]

1 minute ago, rickca said:

The only SOtM component that optionally provides a master clock connector is the sCLK-EX board.  So you would need an sCLK-EX board in the SOtM switch in order to connect the sCLK-OCX10 directly to the switch.  

 

As Rajiv said, if you have unused taps on the sCLK-EX in your tX-USBultra it's more cost effective to just install a clock connect PCB in the SOtM switch and leverage the connection of the sCLK-OCX10 to your tX-USBultra.

 

Thanks @rickca

 

I would need to re-organise some stuff and put the switch closer to the TX-USBUltra but I guess that's not a deal breaker. All hypothetical of course as the switch is not even out and my Aqvox SE is doing a fantastic job currently.

 

Cost aside. Any thoughts on whether there is a theoretical advantage for the sCLK-OX10 clock signal to go directly to a sCLK-EX containing switch, versus via the TX-USBUltra and it's sCLK-EX?

 

Cheers,

Alan

[lmitche]

4 hours ago, austinpop said:

 

John, the master clock does not apply to the Linear Solutions approach. Consider this. A generic switch has a fixed frequency (say 25MHz) oscillator (i.e. clock) in it. When you use an sCLK-EX approach to improve the switch, you are doing this:

• sCLK-EX's on-board 10MHz reference clock ---- drives ------> frequency synthesizer ----> synthesizes 25 MHz ------ delivers over SMB to -------> switch

When you connect a 10MHz master clock to the sCLK-EX, you are using the superior OCXO clock in the master clock to "discipline" the sCLK-EX's reference clock. So the new chain, with improved SQ, looks like this:

• OCXO-based 10 MHz reference clock ---------- disciplines -----> sCLK-EX''s reference clock ---- drives ------> frequency synthesizer ----> synthesizes 25 MHz ------ over SMB -------> switch

Presumably, the reason we hear an improvement with the above is that the OCXO "goodness" of the reference clock somehow propagates all the way to the switch.

 

The Linear Solutions approach is different. It is putting the OCXO directly in the switch. It is replacing the generic 25MHz clock in the switch itself with an OCXO 25MHz clock. There is no synthesizer, so the design does not lend itself to an external reference clock. This is a different approach.

 

Does that make sense?

 

As for your point about "synchronization." I've reported on this before. Eric and I have done experiments where we constructed chains where all the components were clocked from a single sCLK-EX board (i.e. synchronized), vs. each component with its own sCLK-EX board (i.e. unsynchronized). We heard no difference in SQ between the 2, suggesting no SQ benefit from synchronization. This suggests the improvements come from lowering the phase noise, not necessarily running all components in lock-step.

 

I hope this clarifies things.

Rajiv,

 

The two of us have had this conversation several times. Money aside, if synchronized clocks make no difference in SQ than we should, at the least, be indifferent between having a great clock in each device vs. a centralized clock solution.

 

In the case of the two switches we compared on Sunday, distributed won, but not by miles.

 

If we agree, then the economics of central vs. distributed clock generation kick in. I don't know which is most cost effective, but I suspect it may be the latter approach. Less spaghetti too.

 

 

[Johnseye]

3 minutes ago, BigAlMc said:

 

Thanks @rickca

 

I would need to re-organise some stuff and put the switch closer to the TX-USBUltra but I guess that's not a deal breaker. All hypothetical of course as the switch is not even out and my Aqvox SE is doing a fantastic job currently.

 

Cost aside. Any thoughts on whether there is a theoretical advantage for the sCLK-OX10 clock signal to go directly to a sCLK-EX containing switch, versus via the TX-USBUltra and it's sCLK-EX?

 

Cheers,

Alan

 

Only the cable length.  SOtM uses longish cables in their devices anyway so it shouldn't make a difference.  I'd leverage the TX ultra if you have a tap and the time.

[BigAlMc]

4 minutes ago, Johnseye said:

 

Only the cable length.  SOtM uses longish cables in their devices anyway so it shouldn't make a difference.  I'd leverage the TX ultra if you have a tap and the time.

 

Thanks John.

 

Yeah - that does seem the way to go as I'd imagine any difference is fairly minimal.

 

And probably beats the wisdom of shelling out for another sCLK-EX and another 500 quid BNC cable!  It (the SoTM BNC cable) comes in a pizza box, it's a thing of beauty. But it's eye-wateringly expensive!

[Johnseye]

5 hours ago, austinpop said:

 

John, the master clock does not apply to the Linear Solutions approach. Consider this. A generic switch has a fixed frequency (say 25MHz) oscillator (i.e. clock) in it. When you use an sCLK-EX approach to improve the switch, you are doing this:

• sCLK-EX's on-board 10MHz reference clock ---- drives ------> frequency synthesizer ----> synthesizes 25 MHz ------ delivers over SMB to -------> switch

When you connect a 10MHz master clock to the sCLK-EX, you are using the superior OCXO clock in the master clock to "discipline" the sCLK-EX's reference clock. So the new chain, with improved SQ, looks like this:

• OCXO-based 10 MHz reference clock ---------- disciplines -----> sCLK-EX''s reference clock ---- drives ------> frequency synthesizer ----> synthesizes 25 MHz ------ over SMB -------> switch

Presumably, the reason we hear an improvement with the above is that the OCXO "goodness" of the reference clock somehow propagates all the way to the switch.

 

The Linear Solutions approach is different. It is putting the OCXO directly in the switch. It is replacing the generic 25MHz clock in the switch itself with an OCXO 25MHz clock. There is no synthesizer, so the design does not lend itself to an external reference clock. This is a different approach.

 

Does that make sense?

 

As for your point about "synchronization." I've reported on this before. Eric and I have done experiments where we constructed chains where all the components were clocked from a single sCLK-EX board (i.e. synchronized), vs. each component with its own sCLK-EX board (i.e. unsynchronized). We heard no difference in SQ between the 2, suggesting no SQ benefit from synchronization. This suggests the improvements come from lowering the phase noise, not necessarily running all components in lock-step.

 

I hope this clarifies things.

 

Somehow I missed your findings that synchronization using a master clock doesn't have any impact on SQ.  If this is the case, then neither the Linear Solutions switch, nor any other device would benefit from a master clock for synchronization purposes.  I believe what you heard, but it bypasses conventional wisdom.  I'll have to look for your test.  If this is the case then there is no reason for a master clock if your devices have an OCXO clock or potentially better like caesium or rubidium.  Depending of course whether timing is the factor.  All you are buying with a master clock is the ability to tap into an OCXO for reference, while potentially using a lesser quality clock on the device.

 

Can you expound on what you understand "discipline" to mean in this context?

 

Based on what you said, the sCLK would be better with an on board OCXO 10MHz reference, avoiding the need for a master.  Other than having another device to sell, I don't know why SOtM would not do this.