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

Forehaven · April 18, 2018

Having recovered I too wanted to add my Thanks to John for inviting us to his lovely home and for arranging all this! It was such a pleasure to meet John, Rajiv and Vivek. Add in Larry, such a powerhouse of knowledge. Also, a great pleasure to meet May and Lee at the P. Barber show! Speaking of which, Patricia was awesome!

I haven't kept up with all the SOtM terminology so I didn't fully follow all the devices. But we all heard the same as Rajiv and John as noted in their awesome posts. Thanks guys for write up guys!

For me, one big picture was the differences at times subtle despite the different approaches. We had to A/B sev. times to pick up the differences that told me we have all made great progress in our digital systems. We need to further understand what's going on with all these clocks and switches.

So Thanks again guys for a really enjoyable time and the exposure to the SOtM products and LSolutions switch!

Chris

austinpop · April 18, 2018

1 hour ago, Johnseye said:

Somehow I missed your findings that synchronization using a master clock doesn't have any impact on SQ.

Let's be very precise by what we mean by synchronization.

What I though you were asking, and what Eric and I tested empirically, was whether having multiple components clocked by the same sCLK-EX board (let's call this synchronized, although I think if we dig deep into the mechanisms, that may be a misnomer) vs. having components each clocked by their own sCLK-EX boards (let's call this unsynchronized). See this post, experiment 6.

Larry @lmitche, I think this is what you and I discussed too - back then.

1 hour ago, Johnseye said:

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.

Yup. But be careful. There used to be a mistaken view that Rubidium clocks were better than OCXO for audio, but it's turned out to be the opposite. John Swenson can explain this way better than I can:

https://www.computeraudiophile.com/forums/topic/27650-cybershaft-rubidium-clocklow-priced-option/?do=findComment&comment=580808

In essence, to quote: A rubidium standard offers NO advantage over the very good OCXO. The rubidium standard has two systems, a rubidium oscillator, which has high jitter but very good long term stability and an OCXO with very low jitter but not as good long term stability.

1 hour ago, Johnseye said:

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

I'm at the limit of my detailed understanding of how exactly clock disciplining works. At AXPONA, I asked Lee whether - when you connect the external 10MHz clock to the master clock input - the external clock overrode the internal reference clock? He said no, it was more like synchronization. I've read this mechanism being referred to as disciplining, but don't know the exact circuit topology and mechanisms beyond this.

Academics aside, the practical effect is the same. Adding the master clock improves the SQ as we have heard.

1 hour ago, Johnseye said:

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.

Yes, certainly. As for why, it's a tradeoff. OCXOs of the quality used in the OCX10 and the Ref 10 are quite expensive. You can't put them in each device in a cost effective way. The fascinating question though is - do you need to?

What Roy found was that the chain with Linear Solution (LS)'s switch:

LS switch (with internal 25MHz OCXO)

sounded better than:

Ref 10 (10MHz OCXO) > sLCK-EX (in tX-USBultra) > modded DLink switch

I asked Adrian at AXPONA what the phase noise spec for his OCXO chips was. While I don't want to divulge that number without his consent, suffice it to say it was nowhere near the equivalent phase noise number for the Ref 10 or the OCX10. Another caveat is that you can only compare phase noise at the same carrier frequency, and here we are talking about apples (10MHz OCXO) and oranges (25MHz OCXO).

I see the sCLK based chain as a leaky sieve. You start out with outstanding phase noise at the reference clock. Mutec are justifiably proud that their phase noise data is observed at the BNC outputs of the Ref 10. But then this phase noise is going to degrade as it traverses the path to the destination - the 25MHz input in the switch's internals. This path is:

50 or 75Ω cable to the Ultra box containing the sCLK-EX
BNC connector quality
cable from BNC to U.FL connector on the sCLK-EX board
quality of the terminator at the sCLK-EX
reflections due to imperfect impedance match
noise due to synthesis of the 25Mhz clock
cable from sCLK-EX tap to the output SMB connector
quality of the SMB connector
quality of the SMB cable to the switch
reflections due to minor impedance mismatches
cable from SMB input on switch to internal board

And even this is an incomplete list, I'm sure. The question is after all this chain of corruption, what is the phase noise characteristic of the signal entering the internal switch board? And could you use an OCXO of lesser quality (i.e. cheaper) within the switch that could provide equal or better phase noise?

That is the premise of the onboard OCXO, since that architecture looks like this:

25 MHz OCXO --- few mm of interconnect ---> 25 MHz clock input for switch's internal board.

Hope all this made sense - I feel like I rambled on.

Johnseye · April 18, 2018

5 minutes ago, austinpop said:

Let's be very precise by what we mean by synchronization.

What I though you were asking, and what Eric and I tested empirically, was whether having multiple components clocked by the same sCLK-EX board (let's call this synchronized, although I think if we dig deep into the mechanisms, that may be a misnomer) vs. having components each clocked by their own sCLK-EX boards (let's call this unsynchronized). See this post, experiment 6.

Larry @lmitche, I think this is what you and I discussed too - back then.

Yup. But be careful. There used to be a mistaken view that Rubidium clocks were better than OCXO for audio, but it's turned out to be the opposite. John Swenson can explain this way better than I can:

https://www.computeraudiophile.com/forums/topic/27650-cybershaft-rubidium-clocklow-priced-option/?do=findComment&comment=580808

In essence, to quote: A rubidium standard offers NO advantage over the very good OCXO. The rubidium standard has two systems, a rubidium oscillator, which has high jitter but very good long term stability and an OCXO with very low jitter but not as good long term stability.

I'm at the limit of my detailed understanding of how exactly clock disciplining works. At AXPONA, I asked Lee whether - when you connect the external 10MHz clock to the master clock input - the external clock overrode the internal reference clock? He said no, it was more like synchronization. I've read this mechanism being referred to as disciplining, but don't know the exact circuit topology and mechanisms beyond this.

Academics aside, the practical effect is the same. Adding the master clock improves the SQ as we have heard.

Yes, certainly. As for why, it's a tradeoff. OCXOs of the quality used in the OCX10 and the Ref 10 are quite expensive. You can't put them in each device in a cost effective way. The fascinating question though is - do you need to?

What Roy found was that the chain with Linear Solution (LS)'s switch:

LS switch (with internal 25MHz OCXO)

sounded better than:

Ref 10 (10MHz OCXO) > sLCK-EX (in tX-USBultra) > modded DLink switch

I asked Adrian at AXPONA what the phase noise spec for his OCXO chips was. While I don't want to divulge that number without his consent, suffice it to say it was nowhere near the equivalent phase noise number for the Ref 10 or the OCX10. Another caveat is that you can only compare phase noise at the same carrier frequency, and here we are talking about apples (10MHz OCXO) and oranges (25MHz OCXO).

I see the sCLK based chain as a leaky sieve. You start out with outstanding phase noise at the reference clock. Mutec are justifiably proud that their phase noise data is observed at the BNC outputs of the Ref 10. But then this phase noise is going to degrade as it traverses the path to the destination - the 25MHz input in the switch's internals. This path is:

50 or 75Ω cable to the Ultra box containing the sCLK-EX

BNC connector quality

cable from BNC to U.FL connector on the sCLK-EX board

quality of the terminator at the sCLK-EX

reflections due to imperfect impedance match

noise due to synthesis of the 25Mhz clock

cable from sCLK-EX tap to the output SMB connector

quality of the SMB connector

quality of the SMB cable to the switch

reflections due to minor impedance mismatches

cable from SMB input on switch to internal board

And even this is an incomplete list, I'm sure. The question is after all this chain of corruption, what is the phase noise characteristic of the signal entering the internal switch board? And could you use an OCXO of lesser quality (i.e. cheaper) within the switch that could provide equal or better phase noise?

That is the premise of the onboard OCXO, since that architecture looks like this:

25 MHz OCXO --- few mm of interconnect ---> 25 MHz clock input for switch's internal board.

Hope all this made sense - I feel like I rambled on.

All makes sense, and agreed except the synchronization. I'm not referring to single vs multiple sCLK's. I'm referring to the true job of the master clock, providing a single drum beat to time multiple devices in synchronization.

This simple explanation of a word (master) clock is what I'm referring to. Additionally, why couldn't we use these master clocks, some of which have OCXO crystals? Those are some used in pro audio and I'm sure there are more. What makes the OCX10, Ref10 and OP-14 better and do we even know if they are better?

MNG · April 18, 2018

10 minutes ago, Johnseye said:

All makes sense, and agreed except the synchronization. I'm not referring to single vs multiple sCLK's. I'm referring to the true job of the master clock, providing a single drum beat to time multiple devices in synchronization.

This simple explanation of a word (master) clock is what I'm referring to. Additionally, why couldn't we use these master clocks, some of which have OCXO crystals? Those are some used in pro audio and I'm sure there are more. What makes the OCX10, Ref10 and OP-14 better and do we even know if they are better?

I’m trying to learn and catch up regarding these clocks, but isn’t it a different thing entirely? I was getting confused myself about this, but I think that the Word Clocks do bring a benefit of synchronicity of timing but these clocks we’re talking about, OCX10, Ref10 etc, are not Word Clocks and they benefit by reducing phase noise, not improving and syncing timing. I hope that’s right anyway.

Johnseye · April 18, 2018

I think I found the answer to my questions when reading up on the REF10. It looks like it's a reference master clock and not a word master clock. I have to read up more on the differences between the two.

seeteeyou · April 18, 2018

http://thelinearsolution.com/ocxo_switch.html
http://thelinearsolution.com/ocxo_router.html
http://thelinearsolution.com/ocxo_lan.html
http://thelinearsolution.com/ocxo_usb3.html

Quote

- Military Spec OCXO
- 2oz Copper PCB +/- 10ppb
- Ultralow Jitter <1ps RMS

Same specs for all of the above, and then their USB 3 adapter was clearly showing that $55 Connor-Winfield OH4610LF-025.0M

http://www.conwin.com/datasheets/cx/cx259.pdf

http://www.conwin.com/product_locator.html?ModelFam=Oscillators&ProdType=OCXO&Recommended=Y&Pkg=14 Pin DIP
https://www.digikey.com/product-detail/en/connor-winfield/OH4610LF-025.0M/CW899-ND/5641634

The cat was already out of the bag so there were no beans to be spilled whatsoever.

Johnseye · April 18, 2018

2 minutes ago, MNG said:

I’m trying to learn and catch up regarding these clocks, but isn’t it a different thing entirely? I was getting confused myself about this, but I think that the Word Clocks do bring a benefit of synchronicity of timing but these clocks we’re talking about, OCX10, Ref10 etc, are not Word Clocks and they benefit by reducing phase noise, not improving and syncing timing. I hope that’s right anyway because it’s what I was after when buying the OCX10.

Yup, that's what I'm starting to gather. You're correct. I've missed this critical piece of info which explains the confusion on clock synchronization.

austinpop · April 18, 2018

6 minutes ago, MNG said:

I’m trying to learn and catch up regarding these clocks, but isn’t it a different thing entirely? I was getting confused myself about this, but I think that the Word Clocks do bring a benefit of synchronicity of timing but these clocks we’re talking about, OCX10, Ref10 etc, are not Word Clocks and they benefit by reducing phase noise, not improving and syncing timing. I hope that’s right anyway.

Precisely.

4 minutes ago, Johnseye said:

I think I found the answer to my questions when reading up on the REF10. It looks like it's a reference master clock and not a word master clock. I have to read up more on the differences between the two.

Word clocks operate at multiples of the sample rate frequencies of 44.1 and 48 kHz.

If someone made external OCXO system clocks at the frequencies of interest... multiples of 12 and 25MHz, that would be the equivalent. I’m not aware they exist. Not saying they don’t, just that I’m not aware.

seeteeyou · April 18, 2018

Reference clock:

10MHz

Word clock:

Multiples of 22.5792 MHz / 24.576 MHz

System clock:

Depending on what we're talking about (e.g. 25MHz for Ethernet, 12MHz / 24MHz for USB 2.0 etc.)

BTW, it's pretty much like dealing with Traveling Salesperson Problem

http://www.cs.princeton.edu/courses/archive/fall14/cos126/assignments/tsp.html

The longer route we take, the more degradation we'll have to face. Anyways, it's kinda like a "better than nothing" strategy that's less than ideal if you will. Paying a lot for an excellent 10MHz OCXO in the first place, power that with one of the best PSU (such as SR7) as well, of course throw in another expensive clock cable.

Degrade that all we want but that's still performing quite a bit better than the average system clocks, even if the same level of performance were achievable with an affordable (let's say $55) OCXO inside a switch.

Johnseye · April 18, 2018

2 minutes ago, austinpop said:

Precisely.

Word clocks operate at multiples of the sample rate frequencies of 44.1 and 48 kHz.

If someone made external OCXO system clocks at the frequencies of interest... multiples of 12 and 25MHz, that would be the equivalent. I’m not aware they exist. Not saying they don’t, just that I’m not aware.

So essentially what these reference clocks are, are a better clock than currently in the sCLK with the ability to connect with, and "discipline" other clocks, including the sCLK, at a 10MHz frequency. There is no synchronization occurring. I wrongfully assumed with the multiple BNCs that this would be taking place.

I need to learn more about what this disciplining is. I like Adrian's approach better than the spaghetti model. Larry hit the nail on the head with his earlier comment. If there is no synchronizing occurring, then why complicate things, introduce cables, potential for EMI/RF, etc.

BigAlMc · April 18, 2018

12 minutes ago, MNG said:

they benefit by reducing phase noise, not improving and syncing timing.

Hi All,

Fascinating stuff. Well, for this community. I doubt my colleagues in the office would agree!

So if its not about synchronisation and all about phase noise then does anyone have any ideas about whether there would be any benefit putting an OXCO router upstream from an Oxco switch?

Better still, anyone got any experience of this badboy or anything similar (Paul Pang or suchlike)?

http://thelinearsolution.com/ocxo_router.html

Probably heading into rapidly diminishing returns territory but from a theoretical perspective at least it'd be great to understand.

Cheers,

Alan

austinpop · April 18, 2018

2 minutes ago, Johnseye said:

So essentially what these reference clocks are, are a better clock than currently in the sCLK with the ability to connect with, and "discipline" other clocks, including the sCLK, at a 10MHz frequency. There is no synchronization occurring. I wrongfully assumed with the multiple BNCs that this would be taking place.

I need to learn more about what this disciplining is. I like Adrian's approach better than the spaghetti model. Larry hit the nail on the head with his earlier comment. If there is no synchronizing occurring, then why complicate things, introduce cables, potential for EMI/RF, etc.

I think @seeteeyou‘s post above captures it perfectly.

I also like the idea of no spaghetti, but we do not currently have a full chain available with local OCXOs in each component. Once we do, it will be a fascinating comparison, and exactly as Larry says, the economics vs. SQ will determine the result.

MNG · April 18, 2018

24 minutes ago, Johnseye said:

So essentially what these reference clocks are, are a better clock than currently in the sCLK with the ability to connect with, and "discipline" other clocks, including the sCLK, at a 10MHz frequency. There is no synchronization occurring. I wrongfully assumed with the multiple BNCs that this would be taking place.

I need to learn more about what this disciplining is. I like Adrian's approach better than the spaghetti model. Larry hit the nail on the head with his earlier comment. If there is no synchronizing occurring, then why complicate things, introduce cables, potential for EMI/RF, etc.

I agree, but I think price is the denominator here - these clocks are very expensive and it gets cost prohibitive to stick them in each separate device so stick it in one and share it seems to be the prevailing logic. Currently, I’m only clocking my tX-U and I would far rather have just the tX-U without need for a SPS-500 and CLX10 to run alongside.

Some very interesting info to be found here:

auricgoldfinger · April 18, 2018

14 hours ago, mozes said:

Do you know what kind of regulators they use?

They use LT1083, 1084, 1085 and some Toshiba, depending on the circuit design of the device. For my LPS, they are using the LT 1083.

Johnseye · April 18, 2018

9 minutes ago, MNG said:

I agree, but I think price is the denominator here - these clocks are very expensive and it gets cost prohibitive to stick them in each separate device so stick it in one and share it seems to be the prevailing logic. Currently, I’m only clocking my tX-U and I would far rather have just the tX-U without need for a SPS-500 and CLX10 to run alongside.

Some very interesting info to be found here:

$55 per clock isn't that much if it's built into the components. Especially compared to a $4k reference master. Of course when you're competing against devices without them that cost less people may not understand why the cost difference.

Ralf11 · April 18, 2018

I agree - it isn't much (even if they pay retail or unit sample prices). And adding that to your advertising might be very helpful...

MNG · April 18, 2018

16 minutes ago, Johnseye said:

$55 per clock isn't that much if it's built into the components. Especially compared to a $4k reference master. Of course when you're competing against devices without them that cost less people may not understand why the cost difference.

I must be missing something then? The clock in the tX-U is already pretty good and we’re improving it with these expensive Master Clocks. How is it only 55$? The one in the tX-U must cost that at least?

seeteeyou · April 18, 2018

For some relatively "special" frequencies such as 19.2 MHz, we could still go for stuff like Neutron Star 2

http://www.newclassd.com/index.php?page=200

In some cases sCLK-EX could very well be the only choice whenever we actually need an "oddball" frequency. For instance, 14.318 MHz is required for motherboards but only cheapest of the cheap ones could be found on Mouser.

For some reasons I looked very hard for the clock with 14.318 MHz but this particular Asus motherboard might be an exception with "abnormally" low number of system clocks

https://www.asus.com/Motherboards/PRIME-B250-PLUS/

http://www.computeraudiophile.com/forums/topic/38199-sotm-sclk-ex/?tab=comments#comment-801085

On 3/30/2018 at 6:47 PM, Chrisc10115 said:

I can resolve the issue now. Confirmed by SOTM.

For the maniboard ASUS Prime B 250 plus:

24 MHZ for the mainboard. The ehthernet comunication crystral 25 MHZ.

Voltage: 3,3 Volt.

More importantly, how do we actually determine what system clocks are present throughout an entire music server (and identify each frequency) without sending everything to SOtM? Even stuff like SSDs and optical drives etc. have their own system clocks when we crack them opened.

Johnseye · April 18, 2018

17 minutes ago, MNG said:

I must be missing something then? The clock in the tX-U is already pretty good and we’re improving it with these expensive Master Clocks. How is it only 55$? The one in the tX-U must cost that at least?

I said $55 because that is not only a number I heard elsewhere for a military spec OCXO, but @seeteeyou mentioned it as well. Here's a post from @Superdad on the topic. He's saying they can run in the hundreds for a really good OCXO. I'm sure it isn't hard to find the cost.

Considering what I heard with the Linear Solutions OCXO switch, the fact that the REF 10, Cybershaft and SOtM OCX10 use OCXO, I'm leaning that direction. I suppose it ultimately comes down to phase noise. We'll see what the EtherREGEN brings to the table. If there's a single port that filters out noise and uses a clock as good or better than the Liner Solutions switch, it could be a winner. We'll see, and it could be a while.

seeteeyou · April 18, 2018

Granted we've got two different frequencies below but it's just a matter of putting prices into perspective

$55 for ±10ppb (25 MHz)

https://www.digikey.com/product-detail/en/connor-winfield/OH4610LF-025.0M/CW899-ND/5641634

$94 for ±5ppb (25 MHz)

https://www.digikey.com/product-detail/en/connor-winfield/OH300-50503CF-025.0M/CW943CT-ND/7691417

$760 for ±0.5ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-5/535-11917-ND/3641393

$1,222 for ±0.3ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-2/535-11918-ND/3641392

$1,800 for ±0.1ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-1/535-11919-ND/3641391

If we're only looking for system clocks that are usable without "resorting" to the discipline of 10 MHz reference clock, the "only" one with ultra-high stability might be this one from Vectron that's good for anywhere between 4 MHz and 15 MHz

https://www.vectron.com/products/ocxo/dx-040.htm

Depending on the model that's ordered, they're offering either ±0.4ppb or ±0.2ppb but so far we couldn't figure out what the actual price really is without getting a quotation.

Other than that, the only useful frequencies from 4 MHz to 15 MHz should be 12 MHz for USB 2.0 and 14.318 MHz for motherboards respectively.

RickyV · April 18, 2018

12 minutes ago, seeteeyou said:

Granted we've got two different frequencies below but it's just a matter of putting prices into perspective

$55 for ±10ppb (25 MHz)

https://www.digikey.com/product-detail/en/connor-winfield/OH4610LF-025.0M/CW899-ND/5641634

$94 for ±5ppb (25 MHz)

https://www.digikey.com/product-detail/en/connor-winfield/OH300-50503CF-025.0M/CW943CT-ND/7691417

$760 for ±0.5ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-5/535-11917-ND/3641393

$1,222 for ±0.3ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-2/535-11918-ND/3641392

$1,800 for ±0.1ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-1/535-11919-ND/3641391

If we're only looking for system clocks that are usable without "resorting" to the discipline of 10 MHz reference clock, the "only" one with ultra-high stability might be this one from Vectron that's good for anywhere between 4 MHz and 15 MHz

https://www.vectron.com/products/ocxo/dx-040.htm

Depending on the model that's ordered, they're offering either ±0.4ppb or ±0.2ppb but so far we couldn't figure out what the actual price really is without getting a quotation.

Other than that, the only useful frequencies from 4 MHz to 15 MHz should be 12 MHz for USB 2.0 and 14.318 MHz for motherboards respectively.

Compare phase noise specs of the connor-winfields ocxo with the phase noise specs of a crystek cchd-957 or 575 near the same ground frequentie.

More or less the same or am i missing something here?

RickyV · April 18, 2018

http://www.crystek.com/crystal/spec-sheets/clock/CCHD-957.pdf

Superdad · April 18, 2018

25 minutes ago, seeteeyou said:

$94 for ±5ppb (25 MHz)

https://www.digikey.com/product-detail/en/connor-winfield/OH300-50503CF-025.0M/CW943CT-ND/7691417

$760 for ±0.5ppb (10 MHz)

https://www.digikey.com/product-detail/en/abracon-llc/AOCJY6-10.000MHZ-5/535-11917-ND/3641393

.....

If we're only looking for system clocks that are usable without "resorting" to the discipline of 10 MHz reference clock, the "only" one with ultra-high stability might be this one from Vectron that's good for anywhere between 4 MHz and 15 MHz

Frequency stability is not important for the application! Only phase-noise at low frequency offset is important.

Despite their still-published too-conservative plots, the Crystek CCHD-575 ($9.60 each when we order 500 at a time) are running -108, -110, -112dBc/Hz @10Hz. Those are the numbers on the paper plots they included with our first (confirmed random) samples that we got from Crystek almost 2 years ago.

The only significantly better numbers I have seen for a (semi)-production clock under $500 come from the Pulsar Clock. (Though Patrick in Texas has been hand screening NDKs for a few people for a few years; I think he said he can find about 10% of them getting to -112dBc/Hz at 10Hz).

From every clock data sheet I have look at (over 100 that's for sure) OCXOs below $500 are inferior--from a phase-noise perspective--to the cheap-and-cheerful Crystek CCHD-575.

So anytime someone brags about their use of a TCXO or OCXO (in a DAC or anything else), then you'd best ask for a link to a model and data sheet.

Lastly, I must say that I find TheLinearSolution's "spec" on its "OCXO Circuit" puzzling. They state: "Ultralow Jitter: <1ps RMS." Aside from being meaningless without a bandwidth range, 1 picosecond is laughably bad. 1 picosecond equals 1,000 femtoseconds. A $10 Crystek clock is spec'ed at 85 femtoseconds RMS jitter (12kHz~80MHz).

Ralf11 · April 18, 2018

1 hour ago, seeteeyou said:

...

More importantly, how do we actually determine what system clocks are present throughout an entire music server (and identify each frequency) without sending everything to SOtM? Even stuff like SSDs and optical drives etc. have their own system clocks when we crack them opened.

That is the crux of the bisquit...

Johnseye · April 18, 2018

5 minutes ago, Superdad said:

Frequency stability is not important for the application! Only phase-noise at low frequency offset is important.

Is phase noise really the only thing we care about in a clock? Nothing else?

5 minutes ago, Superdad said:

Lastly, I must say that I find TheLinearSolution's "spec" on its "OCXO Circuit" puzzling. They state: "Ultralow Jitter: <1ps RMS." Aside from being meaningless without a bandwidth range, 1 picosecond is laughably bad. 1 picosecond equals 1,000 femtoseconds. A $10 Crystek clock is spec'ed at 85 femtoseconds RMS jitter (12kHz~80MHz).

As puzzling as it is, and maybe they were lazy or being secretive about sharing their spec, the switch noticeably improved the sound quality.

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

Recommended Posts

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Create an account or sign in to comment

Create an account

Sign in