Uptone Audio Regen - Clock Mod

[HenSch]

Gentlemen,

 

PLEASE let us close that discussion about alternatives to the USB Audio interface at this point in this thread.

 

It already started to become painful for any fellow looking for information on the clock mod in this thread. Let us all be polite towards them and keep on topic from now on.

 

A big Thank You for all contributing here!

[Blizzard]

Gentlemen,

 

PLEASE let us close that discussion about alternatives to the USB Audio interface at this point in this thread.

 

It already started to become painful for any fellow looking for information on the clock mod in this thread. Let us all be polite towards them and keep on topic from now on.

 

A big Thank You for all contributing here!

 

Okay sorry for taking over your thread. I just wanted everyone to be well informed before dropping this kind of money on this mod.

 

I think I've made my point. I don't think anyone will be worse off learning the info I've shared.

 

Thanks.

[JonathanCatuccio]

Experiments? Ok, but subjective.

 

Archimago's Musings: MEASUREMENTS: Audiophile Sound and Operating Systems. (Windows 8.1, Windows 10, Fidelizer & JPLAY again!)

 

I'm just really questioning the reality of the USB mods proposed here...vs cost of the Bega for example.

 

I mean, upgrading the master clock, ok I can see that, but how much better do we need with the Vega?

 

"Another signature feature of the Vega is its Femto Master Clock, which yields a spectacularly low 0.082 picoseconds"

 

I think his experiments are also great. To me his results are real proof of the shortcomings of using USB for audio. If it was a great format to begin with, the tweaks wouldn't make any difference. Try making improvements to I2S transmitted via a couple inches of heavy high purity copper PCB traces to the DAC chip or DAC FPGA. You might have some difficulty.

[Blizzard]

Experiments? Ok, but subjective.

 

Archimago's Musings: MEASUREMENTS: Audiophile Sound and Operating Systems. (Windows 8.1, Windows 10, Fidelizer & JPLAY again!)

 

I'm just really questioning the reality of the USB mods proposed here...vs cost of the Bega for example.

 

I mean, upgrading the master clock, ok I can see that, but how much better do we need with the Vega?

 

"Another signature feature of the Vega is its Femto Master Clock, which yields a spectacularly low 0.082 picoseconds"

 

That 0.082 picosecond spec is the jitter at 100khz. Way above the range of human hearing. It's a useless spec used for marketing purposes. 10hz phase noise is the spec that matters for audio jitter.

 

I happen to have the same clock as the Vega master right here on my desk. It's a bit rough looking as it has some solder spattered on it. Sure it's a decent clock for $24. But if you want to spend a couple hundred you can get way better

 

IMO it's this series of clock that should be used in the REGEN upgrade. DEXA doesn't even share a phase noise plot of the neutron star.

 

Another thing is every CM of cable connected to a clock adds jitter. Even a worse clock can outperform a better one when soldered direct to the SMD contacts. Exactly why as John also mentioned in his comment, high end external clocks rarely outperform mediocre internal ones.

 

http://www.digikey.com/product-detail/en/CCHD-575-50-100.000/744-1454-ND/2742153

 

[Ben-M]

That 0.082 picosecond spec is the jitter at 100khz. Way above the range of human hearing. It's a useless spec used for marketing purposes. 10hz phase noise is the spec that matters for audio jitter.

IMO it's this series of clock that should be used in the REGEN upgrade. DEXA doesn't even share a phase noise plot of the neutron star.

 

Another thing is every CM of cable connected to a clock adds jitter. Even a worse clock can outperform a better one when soldered direct to the SMD contacts. Exactly why as John also mentioned in his comment, high end external clocks rarely outperform mediocre internal ones.

 

CCHD-575-50-100.000 Crystek Corporation | 744-1454-ND | DigiKey

 

Finally, back on topic. You guys lost it there for a while. I really hope this topic doesn't go sideways again like that. It's only 5 pages long and 3 pages were taken up by Ethernet DAC vs. USB vs. money comments. Be great if we had mods that could swing in here and wipe those comments off the books, or at least hide them in "discussion not on topic, OP tried to rein in, but dissenters didn't listen" kind of expandable section that readers could choose to open if they wanted something else to read. This is a mod thread, let's leave it at that.

 

I also agree about that terribly long clock interconnect. Whenever I have seen the Neutron Star clock used in mods, I think, I have seen it directly inside the same enclosure and using a very short run. I know that's to due with space requirements, but given that, wouldn't putting this just above or beside the Regen be best? Then only having to use a small ~5cm run of wire to the Regen. In imagining this I see it and the Regen inside the same, larger enclosure.

 

I've been wondering since I first saw the mod whether or not the supplied cable can be shortened and reterminated?

[HenSch]

 

I also agree about that terribly long clock interconnect. Whenever I have seen the Neutron Star clock used in mods, I think, I have seen it directly inside the same enclosure and using a very short run. I know that's to due with space requirements, but given that, wouldn't putting this just above or beside the Regen be best? Then only having to use a small ~5cm run of wire to the Regen. In imagining this I see it and the Regen inside the same, larger enclosure.

 

I've been wondering since I first saw the mod whether or not the supplied cable can be shortened and reterminated?

 

You are very right with the lengthy connector being probably suboptimal. I initially left it as long as it came with the Clock, but Dexa clearly recommends cutting it as short as possible. So yes, the cable can be shortened.

 

After I have found a decent case for the clock I'll probably cut the cable to its minimum length that suits my installation.

 

However, right now I have no motivation to even put the clock into a case. Do you know these situations with audio when suddenly everything sound just right and involving you completely? That's where I'm in right now. A bit anxious to destroy this brittle beauty by any further change.

 

So it might sound even better with a shorter cable. My head says yes, my heart says put your hands off it.

[Blizzard]

Finally, back on topic. You guys lost it there for a while. I really hope this topic doesn't go sideways again like that. It's only 5 pages long and 3 pages were taken up by Ethernet DAC vs. USB vs. money comments. Be great if we had mods that could swing in here and wipe those comments off the books, or at least hide them in "discussion not on topic, OP tried to rein in, but dissenters didn't listen" kind of expandable section that readers could choose to open if they wanted something else to read. This is a mod thread, let's leave it at that.

 

I also agree about that terribly long clock interconnect. Whenever I have seen the Neutron Star clock used in mods, I think, I have seen it directly inside the same enclosure and using a very short run. I know that's to due with space requirements, but given that, wouldn't putting this just above or beside the Regen be best? Then only having to use a small ~5cm run of wire to the Regen. In imagining this I see it and the Regen inside the same, larger enclosure.

 

I've been wondering since I first saw the mod whether or not the supplied cable can be shortened and reterminated?

 

I would like to see someone just try this mod using the Crystek 575 or 957. The 575 looks to me like the exact same form factor. It can be swapped and still appear like nothing was touched. Who knows may even outperform. For $24 not much to lose.

 

Does it have to be exactly 24MHZ? Off the shelf they come in 24.576, but Crystek can supply exactly 24 if requested.

[Superdad]

That 0.082 picosecond spec is the jitter at 100khz. Way above the range of human hearing. It's a useless spec used for marketing purposes. 10hz phase noise is the spec that matters for audio jitter.

 

Oh lord, now you are conflating time domain with frequency domain measurements--AND trying to tell us that phase noise frequency offsets are related to the bandwidth of our hearing. Please Bliz, I'm no engineer, but at least I'm not tossing all little bits I know into a blender and then pouring them out and trying to pass them off as authoritative advice.

[Blizzard]

Oh lord, now you are conflating time domain with frequency domain measurements--AND trying to tell us that phase noise frequency offsets are related to the bandwidth of our hearing. Please Bliz, I'm no engineer, but at least I'm not tossing all little bits I know into a blender and then pouring them out and trying to pass them off as authoritative advice.

 

Maybe you should sit down and have a talk with John. He seems to be much better informed in this area. Obviously the reason he designs your products.

[Ben-M]

You are very right with the lengthy connector being probably suboptimal. I initially left it as long as it came with the Clock, but Dexa clearly recommends cutting it as short as possible. So yes, the cable can be shortened.

 

After I have found a decent case for the clock I'll probably cut the cable to its minimum length that suits my installation.

That's awesome news, so it can be shortened! It's definitely a PRICEY add-on, but learning more about it is always interesting. It'd be cool if someone could compare some measurements of this Neutron Star clock vs. other ones where measurements are available. In the end I think that "there is no rule" rule applies, though. It's always YMMV, so it's good just to see what other guys are doing and see how you could implement something in the same direction, if it tickles your fancy(I don't think I've ever once used that saying before...)

 

However, right now I have no motivation to even put the clock into a case. Do you know these situations with audio when suddenly everything sound just right and involving you completely? That's where I'm in right now. A bit anxious to destroy this brittle beauty by any further change.

 

So it might sound even better with a shorter cable. My head says yes, my heart says put your hands off it.

I know what you mean. As the other saying goes, "if it's not broken, don't fix it". But like the other thread that was recently started on here, as an Audiophile, how long do you think you can hold out and be content, knowing that there is another arrangement that you could test that could (and should) be better sounding yet again...? It's in our blood to yearn for better sound, even when what we have is probably better than what most others will ever have, outside of the audiophile subset anyway.

[Blizzard]

That's awesome news, so it can be shortened! It's definitely a PRICEY add-on, but learning more about it is always interesting. It'd be cool if someone could compare some measurements of this Neutron Star clock vs. other ones where measurements are available. In the end I think that "there is no rule" rule applies, though. It's always YMMV, so it's good just to see what other guys are doing and see how you could implement something in the same direction, if it tickles your fancy(I don't think I've ever once used that saying before...)

 

 

I know what you mean. As the other saying goes, "if it's not broken, don't fix it". But like the other thread that was recently started on here, as an Audiophile, how long do you think you can hold out and be content, knowing that there is another arrangement that you could test that could (and should) be better sounding yet again...? It's in our blood to yearn for better sound, even when what we have is probably better than what most others will ever have, outside of the audiophile subset anyway.

 

Well here's the phase noise plot of the Crystek 575-50-24.576. It would be cool to see the plot for the DEXA. Perhaps Alex can share the plot of the stock one.

 

[Blizzard]

Oh lord, now you are conflating time domain with frequency domain measurements--AND trying to tell us that phase noise frequency offsets are related to the bandwidth of our hearing. Please Bliz, I'm no engineer, but at least I'm not tossing all little bits I know into a blender and then pouring them out and trying to pass them off as authoritative advice.

 

Copied this directly off the Neutron star clock page:

 

 

A note on so called Femtosecond Clocks, and clocks with Femtosecond Jitter performance.

 

 

 

A new breed of audio clocks are emerging on the market claiming jitter performance as low as two digit femtoseconds, like 50 or 70 femtoseconds. (We claim <1 pS = 1000 femtoseconds). The 50 femtosecond performance does not have to do with a spectacular new technology for building clocks, but rather the way jitter is measured. If you wiev jitter graphs forany clock you will find that the jitter is lower at higher frequencies. The jitter in 'femtosecond' clocks are usually measured from 12 kHz to 20 MHz, so mostly out of the audio band. This gives a very nice low jitter figure on paper, however it does not necessarily measure sound quality. We always specify our jitter performance from 10 Hz to 50 kHz.

[Superdad]

Does it have to be exactly 24MHZ? Off the shelf they come in 24.576, but Crystek can supply exactly 24 if requested.

 

Yes, it absolutely must be 24MHz since USB high-speed data rate is 480Mbps/sec. (24MHz*20=480Mbps).

 

Crystek will make custom clocks in 24 weeks, 200 piece min. (unless you really want to pay through the nose).

 

But as John said, super low phase noise clocks should not be needed for the USB clock. I recall him suspecting that the benefit being heard may be from separately powering the clock. Really needs more reasearch.

 

In general, super-duper low jitter "femto" clocks are lost on most DACs since the rest of the DAC generates at least an order of magnitude more jitter than the clock is capable of. That's why there are DACs with average clocks and lower overall jitter than others with fancy clocks.

[Blizzard]

Yes, it absolutely must be 24MHz since USB high-speed data rate is 480Mbps/sec. (24MHz*20=480Mbps).

 

Crystek will make custom clocks in 24 weeks, 200 piece min. (unless you really want to pay through the nose).

 

But as John said, super low phase noise clocks should not be needed for the USB clock. I recall him suspecting that the benefit being heard may be from separately powering the clock. Really needs more reasearch.

 

In general, super-duper low jitter "femto" clocks are lost on most DACs since the rest of the DAC generates at least an order of magnitude more jitter than the clock is capable of. That's why there are DACs with average clocks and lower overall jitter than others with fancy clocks.

 

Crystek will make custom clocks with a minimum order of 10 with a 4-6 week lead time. But chances are they already have some 24's kicking around.

 

Anyone interested contact luiz oricchio. He's the man to talk to.

 

[email protected]

 

Here's the last quote I got from them:

 

[HenSch]

 

But as John said, super low phase noise clocks should not be needed for the USB clock. I recall him suspecting that the benefit being heard may be from separately powering the clock. Really needs more reasearch.

 

In general, super-duper low jitter "femto" clocks are lost on most DACs since the rest of the DAC generates at least an order of magnitude more jitter than the clock is capable of. That's why there are DACs with average clocks and lower overall jitter than others with fancy clocks.

 

That was actually the main feature what drew me towards the NeutronStar. The decoupling of the clock power supply and ground from the Regen. Altough every decoupling transformer adds jitter again. But still, having the clock to work independently with its own low-noise power supply intrigues me very much.

 

Thus from my personal believes I doubt that a 1:1 swap of the Regen clock with a better spected one would give such a good result the NeutronStar gives.

 

It would be interesting though to compare the NeutronStar with the lot more affordable small brother Neutrino. In September I'll receive my new Regen (I restocked an original Regen after having killed one, see earlier post on first page) maybe I give it a try with the Neutrino then.

[Superdad]

Copied this directly off the Neutron star clock page:

 

 

A note on so called Femtosecond Clocks, and clocks with Femtosecond Jitter performance.

 

 

 

A new breed of audio clocks are emerging on the market claiming jitter performance as low as two digit femtoseconds, like 50 or 70 femtoseconds. (We claim <1 pS = 1000 femtoseconds). The 50 femtosecond performance does not have to do with a spectacular new technology for building clocks, but rather the way jitter is measured. If you wiev jitter graphs forany clock you will find that the jitter is lower at higher frequencies. The jitter in 'femtosecond' clocks are usually measured from 12 kHz to 20 MHz, so mostly out of the audio band. This gives a very nice low jitter figure on paper, however it does not necessarily measure sound quality. We always specify our jitter performance from 10 Hz to 50 kHz.

 

Yes, well-known that phase noise is lower at higher fs. But they really should be stating it as phase noise. Yes there are ways of convering frequency domain PN measures to time domain jitter measures. But the fact remains that jitter at 100KHz--offset from 24.xxx clock frequency--has zero to do with aural frequencies!

 

Maybe read some more: http://www.sitime.com/support2/documents/AN10007-Jitter-and-measurement.pdf

[Blizzard]

That was actually the main feature what drew me towards the NeutronStar. The decoupling of the clock power supply and ground from the Regen. Altough every decoupling transformer adds jitter again. But still, having the clock to work independently with its own low-noise power supply intrigues me very much.

 

Thus from my personal believes I doubt that a 1:1 swap of the Regen clock with a better spected one would give such a good result the NeutronStar gives.

 

It would be interesting though to compare the NeutronStar with the lot more affordable small brother Neutrino. In September I'll receive my new Regen (I restocked an original Regen after having killed one, see earlier post on first page) maybe I give it a try with the Neutrino then.

 

You could be right. But we are talking $24 vs $808 plus case etc. It may be worth a shot for someone on a bit more of a budget. Besides you can always power the Crystek clocks independently as well.

[Superdad]

Crystek will make custom clocks with a minimum order of 10 with a 4-6 week lead time.

 

Well, the 2,000 pcs. of 24MHz C3391 I have on order with them are taking a LOT longer than 4-6 weeks. They must have stopped so they could make your 10 pieces first.

[HenSch]

You could be right. But we are talking $24 vs $808 plus case etc. It may be worth a shot for someone on a bit more of a budget. Besides you can always power the Crystek clocks independently as well.

 

Yes, indeed. But it still will have to share the same ground plane with the Regen.

[Blizzard]

Yes, well-known that phase noise is lower at higher fs. But they really should be stating it as phase noise. Yes there are ways of convering frequency domain PN measures to time domain jitter measures. But the fact remains that jitter at 100KHz--offset from 24.xxx clock frequency--has zero to do with aural frequencies!

 

Maybe read some more: http://www.sitime.com/support2/documents/AN10007-Jitter-and-measurement.pdf

 

The 82 femto second jitter of the 575-50-100 master clock in the Vega is the phase jitter spec at 100mhz.

 

100mhz is out of the audio band. Spec that matter for audio are phase jitter specs within the audio band.

 

Here's what they say on their datasheet about it. It's the 10hz phase noise spec that's the one to pay attention to.

 

Note on the 100mhz model, 10 HZ phase noise is -90DBC/HZ. Which is good but the 24mhz model is much better at -100DBC/HZ.

 

There's 2 different frequencies here. The clock speed frequency, and the signal frequency it's clocking.

 

Phase noise usually improves at lower clock speeds. This is not to be confused with signal frequency which is the opposite.

 

[Blizzard]

Yes, indeed. But it still will have to share the same ground plane with the Regen.

 

You can just connect the same 2 leads your connecting now to the REGEN, and solder wires on the other 2 leads for 3.3v and separate ground. No different than the neutron, just much smaller.

 

Cheaper and smaller to source a 3.3v supply as well.

 

[HenSch]

You can just connect the same 2 leads your connecting now to the REGEN, and solder wires on the other 2 leads for 3.3v and separate ground. No different than the neutron, just much smaller.

 

Cheaper and smaller to source a 3.3v supply as well.

 

 

Soldering a separate ground wire to a direct coupled oscillator doesn't make an insulated ground plane. Every clock without coupling transformer is not directly comparable to approaches with such a device, being NeutronStar or others.

 

Just have a close look at Paul Pang's USB cards for another example.

[Superdad]

The 82 femto second jitter of the 575-50-100 master clock in the Vega is the phase jitter spec at 100mhz.

 

100mhz is out of the audio band. Spec that matter for audio are phase jitter specs within the audio band.

 

Phase noise at different offsets from the clock frequency have NOTHING to do with the audio band! Someone else is going to have to come in and explain this to you.

[Blizzard]

Well, the 2,000 pcs. of 24MHz C3391 I have on order with them are taking a LOT longer than 4-6 weeks. They must have stopped so they could make your 10 pieces first.

 

As you can see on my attached picture, the bigger the quantity, the longer the lead time.

[Blizzard]

Soldering a separate ground wire to a direct coupled oscillator doesn't make an insulated ground plane. Every clock without coupling transformer is not directly comparable to approaches with such a device, being NeutronStar or others.

 

Just have a close look at Paul Pang's USB cards for another example.

 

Well you can't expect a $24 clock to preform as good as a $500 one. But I'd say worth a try experimenting anyways. There's quite a few components under the case in the Crystek's as well. Not just the oscillator. They just have smaller hands over at Crystek .

 

I took one apart if anyone wants to see.