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

[LTG2010]

2 minutes ago, mansr said:

I'm not telling anyone what to do.

My apologies.

[lmitche]

40 minutes ago, LTG2010 said:

My apologies.

Thanks to both of you for your concern about my audiophile health!

[mansr]

I'm just offering some facts about what these things actually do. What you do with them is up to you.

[jabbr]

3 hours ago, LTG2010 said:

The variations in the process clock frequency modulation can cause measurable jitter over 20ns according to some chipset manufacturers. The Jittered signal might explain the audible difference, unless you are of the opinion that jitter is dealt succesfully by the DACS buffers, then its not an explanation.

 

See here's the thing: adding spread spectrum by definition adds waaaaayyyy more than 20ns jitter, so it totally swamps variations in logic process (very real) as well as clock oscillator.

[jabbr]

3 hours ago, vortecjr said:

John S. has a hypothesis that oscillators can negatively  impact each other. Unfortunately, I don't think that his research is far enough along to comment on the exact mechanism. My contention has been since this is so hard to measure its impact most be very small. Understand that he is having to build very low noise hardware to even start the investigation. Having said that....it's probably not a good idea to add more noise:) 

 

Einstein described this long before John S. was born, nor hypothesized. It remains an active area of research ?http://www.sciencemag.org/news/2018/04/einstein-s-spooky-action-distance-spotted-objects-almost-big-enough-see

[mansr]

20 minutes ago, jabbr said:

See here's the thing: adding spread spectrum by definition adds waaaaayyyy more than 20ns jitter, so it totally swamps variations in logic process (very real) as well as clock oscillator.

Uh, could you clarify that a bit? For PCIe and USB3, we're talking about modulating the unit interval by 1 ps or less. Where are you getting 20 ns?

[jabbr]

13 minutes ago, mansr said:

Uh, could you clarify that a bit? For PCIe and USB3, we're talking about modulating the unit interval by 1 ps or less. Where are you getting 20 ns?

 

I was quoting to quickly. The 20ns number isn't right because PCIe allows no more than 12ns skew. Its also roughly the width of a DAC clock cyle (if roughly 50 Mhz). Process phase error variation is often in the ps to 10s to 100s range. Misread ns as ps...

 

The point is that going with a "femotosecond" clock which is the topic of this read doesn't make sense when spread spectrum is adding 0.5% jitter -- the actual number. The amount in ps depends on the particular clock.

 

https://www.microsemi.com/document-portal/doc_view/135439-white-paper-spread-spectrum-clocking

 

 

 

[mansr]

40 minutes ago, jabbr said:

The point is that going with a "femotosecond" clock which is the topic of this read doesn't make sense when spread spectrum is adding 0.5% jitter -- the actual number. The amount in ps depends on the particular clock.

Nobody is applying spread spectrum to the DAC clock. That would be insane.

[jabbr]

11 minutes ago, mansr said:

Nobody is applying spread spectrum to the DAC clock. That would be insane.

Yes but what is being talked about is reclocking the system clock. 

[mansr]

4 minutes ago, jabbr said:

Yes but what is being talked about is reclocking the system clock. 

A lot of things are being talked about.

[jabbr]

1 minute ago, mansr said:

A lot of things are being talked about.

Let me summarize the actual situation: 

 

Modern miniaturized logic uses sharper and sharper rise times, which leads to increased ringing at logic interfaces as well as increased EMI in certain situations, as well as increased switching noise. These noises can be averaged out by adding phase noise to the clock and de syncing the logic. 

 

For audiophile purposes the true solution is to redesign  the logic itself and properly reduce ground plane noise as well as EMI. 

That ain’t likely to happen in a mass market PC motherboard. Good luck trying to fix that with good power supplies and Femto clocks.

[lmitche]

Ok, so I just disabled spread spectrum for PCIE and had a listen.  Not good, everything sounds dead. I turned it back on and quality was restored. Phew!

 

I'll try disabling SS for the VRM next. Stay tuned.

 

FYI - this test was done with the BLCK SS turned off.

[mansr]

1 hour ago, lmitche said:

Ok, so I just disabled spread spectrum for PCIE and had a listen.  Not good, everything sounds dead. I turned it back on and quality was restored. Phew!

 

I'll try disabling SS for the VRM next. Stay tuned.

 

FYI - this test was done with the BLCK SS turned off.

So SSC isn't all bad then.

[lmitche]

14 hours ago, mansr said:

So SSC isn't all bad then.

No, of course not. We just need to know where to use it and where to disable it. It seems to be a simple trade-off of a more stable clock vs. an increase in EMI, although I suspect the underlying mechanisms are much more complex.

 

No matter, I have excellent measurement tools so it's easy to find the best combination of the three motherboard settings.

[mansr]

29 minutes ago, lmitche said:

No matter, I have excellent measurement tools so it's easy to find the best combination of the three motherboard settings.

What would you be measuring?

[lmitche]

1 hour ago, mansr said:

What would you be measuring?

SQ with these ?

[davide256]

Hmm,  I always wondered about quality management tech needs for AC power grids .

https://www.vox.com/energy-and-environment/2018/6/5/17373314/electricity-technology-efficiency-software-waste-3dfs

 

[mansr]

13 minutes ago, lmitche said:

SQ with these ?

Sure... Nothing wrong with listening, but measurement it ain't.

[lmitche]

1 hour ago, mansr said:

Sure... Nothing wrong with listening, but measurement it ain't.

?

[mansr]

Anyhow, it'll be interesting to learn what you find. What settings do you have available?

[lmitche]

4 minutes ago, mansr said:

Anyhow, it'll be interesting to learn what you find. What settings do you have available?

There are three switches for ssc, enable/disable for bclk, vrm and pcie. Best SQ so far is bclk disabled, pcie enabled, both with vrm enabled. Will test vrm today.

[LTG2010]

22 hours ago, mansr said:

So SSC isn't all bad then.

 

8 hours ago, lmitche said:

No, of course not. We just need to know where to use it and where to disable it.

Some of it must be friendly jitter

SSC does mess with power management so if you disable PCIE SSC and its drawing more current whilst VRM SSC is limitting current, might not play - well worth trying all disabled in your tests.

I switched SSC - ON with my motherboard ( there was only one general setting ) Maybe a slight loss of focus, but nothing major on my setup.

Now as for disabling HPET much more interesting, cleaner more 3D sound and an annoying micro stutter sometimes when using remote desktop disappeared.

[mansr]

4 minutes ago, LTG2010 said:

Now as for disabling HPET much more interesting, cleaner more 3D sound and an annoying micro stutter sometimes when using remote desktop disappeared.

If you're getting stutter, you should look at increasing audio buffer sizes wherever possible.

[Johnseye]

19 minutes ago, LTG2010 said:

 

Some of it must be friendly jitter

SSC does mess with power management so if you disable PCIE SSC and its drawing more current whilst VRM SSC is limitting current, might not play - well worth trying all disabled in your tests.

I switched SSC - ON with my motherboard ( there was only one general setting ) Maybe a slight loss of focus, but nothing major on my setup.

Now as for disabling HPET much more interesting, cleaner more 3D sound and an annoying micro stutter sometimes when using remote desktop disappeared.

 

I thought enabling HPET was a good thing.  It's also a setting in Audiophile Optimizer.  Will have to try that myself.

[LTG2010]

33 minutes ago, mansr said:

If you're getting stutter, you should look at increasing audio buffer sizes wherever possible.

Yes have put all buffers back to default.

 

23 minutes ago, Johnseye said:

I thought enabling HPET was a good thing.  It's also a setting in Audiophile Optimizer.

Yep so did I, I think it was the stuttering - when I noticed the setting in BIOS I recalled that turning HPET off improves stuttering in Gaming, so gave it a go. Perhaps it doesn't play well with sCLK or it's just my server, but its well worth a test. I also think AO's HPET works independently of the BIOS so you could have BIOS off and AO on.