Finally I can sell my stillpoints!

[Mivera]

 

[Mivera]

Now that video is a great example of a great way to clearly demonstrate objective performance in the modern age of the internet. Video is the only convincing method this day and age.

 

[amirm]

Only if those measurements are the output of the DAC, not the oscillator.

 

[Mivera]

Only if those measurements are the output of the DAC, not the oscillator.

Yes but the engineer must choose an oscillator as well. And the performance of the oscillator in a real world application will have an impact on the end result. I don't think anyone out there is disputing that jitter isn't detrimental to digital audio performance. you will never improve upon the jitter performance of an oscillator, downstream of it. All you can do is layout the board properly to ensure jitter isn't introduced between the oscillator and what it's clocking.

 

[fas42]

Out of interest, is anyone aware of any experiments done where an audio circuit using a high performance oscillator was doctored, so that one could arbitrarily adjust the level of jitter of the clock, from performing at its best, to absolutely terrible - and have listening tests so see when the jitter was "detectable", using music recordings as the signal?

 

[Mivera]

Out of interest, is anyone aware of any experiments done where an audio circuit using a high performance oscillator was doctored, so that one could arbitrarily adjust the level of jitter of the clock, from performing at its best, to absolutely terrible - and have listening tests so see when the jitter was "detectable", using music recordings as the signal?

It's very easy to reduce jitter performance of a clock. All you need to do is increase noise on the power supply. Like clearly demonstrated in this proper scientific test.

 

[Mivera]

Notice that's there's well over $250000 worth of test gear in that setup. Unfortunately that's what's required to take digital measurements at that level.

 

[fas42]

It's very easy to reduce jitter performance of a clock. All you need to do is increase noise on the power supply. Like clearly demonstrated in this proper scientific test.

But, if the noise of the supply is increased, how can one be certain that this increased noise is not affecting the overall performance via another, interference, route? To do the experiment properly, to my mind, would be for everything to be fixed, except for some circuit addition which 'artificially' varies the jitter - and listen ..

 

[amirm]

Out of interest, is anyone aware of any experiments done where an audio circuit using a high performance oscillator was doctored, so that one could arbitrarily adjust the level of jitter of the clock, from performing at its best, to absolutely terrible - and have listening tests so see when the jitter was "detectable", using music recordings as the signal?

There are a couple. This Dolby paper is one of them: https://secure.aes.org/forum/pubs/conventions/?elib=8354

"Theoretical and Audible Effects of Jitter on Digital Audio Quality"

Here are a few graphs I have saved up from it:

The audible thresholds were considerably above what we typically see in DACs.

 

[Mivera]

There are a couple. This Dolby paper is one of them: https://secure.aes.org/forum/pubs/conventions/?elib=8354

"Theoretical and Audible Effects of Jitter on Digital Audio Quality"

Here are a few graphs I have saved up from it:

View attachment 6625

View attachment 6626

The audible thresholds were considerably above what we typically see in DACs.

Why does Mark Levinson use premium clocks then? What about clean power supplies? Just a marketing gimmick?

That paper is way outdated. They have learned lots about jitter and audio over the last 19 years.

 

[amirm]

Why does Mark Levinson use premium clocks then? What about clean power supplies? Just a marketing gimmick?

It is called engineering excellence. A requirement in my book for high-end audio.

That paper is way outdated. They have learned lots about jitter and audio over the last 19 years.

Listening tests don't have expiration dates on them. There can be criticism against the paper and study as with anything. But its overall message unfortunately is not friendly to creating ultra-low jitter products.

 

[Mivera]

It is called engineering excellence. A requirement in my book for high-end audio.


Listening tests don't have expiration dates on them. There can be criticism against the paper and study as with anything. But its overall message unfortunately is not friendly to creating ultra-low jitter products.

If there's no audible effect by using more expensive low phase noise clocks, how is using them engineering excellence? Sounds like a great way to increase the BOM, with no benefit to the end user.

 

[fas42]

There are a couple. This Dolby paper is one of them: https://secure.aes.org/forum/pubs/conventions/?elib=8354
The audible thresholds were considerably above what we typically see in DACs.

Thanks for that, Amir. Unfortunately, music was not used for the test - I was specifically curious if there was any testing using such.

 

[Mivera]

Thanks for that, Amir. Unfortunately, music was not used for the test - I was specifically curious if there was any testing using such.

1000's of tests have been done with music by audio gear manufacturers. This is why they use low jitter clocks. It's not a practice limited to the snake oil brands of the industry.

 

[fas42]

1000's of tests have been done with music by audio gear manufacturers. This is why they use low jitter clocks. It's not a practice limited to the snake oil brands of the industry.

Informally, yes. But it would be nice to have a "proper paper" to point to, that ticked all the boxes for doing the test in a manner that audiophiles could relate to.

 

[Mivera]

Informally, yes. But it would be nice to have a "proper paper" to point to, that ticked all the boxes for doing the test in a manner that audiophiles could relate to.

I suppose if folks don't have faith that the engineers who designed the gear they enjoy listening to know what they're doing, there's always the option to buy something lower end. Perhaps even some manufacturers could custom build with lower cost, higher jitter clocks if requested.

 

[amirm]

Thanks for that, Amir. Unfortunately, music was not used for the test - I was specifically curious if there was any testing using such.

It was. It just wasn't in that graph that I showed. Here are the threshold results in nanoseconds for music:

Due to masking effects, the thresholds are much higher (i.e. we tolerate a lot more jitter distortion) with music than single tones.

 

[amirm]

If there's no audible effect by using more expensive low phase noise clocks, how is using them engineering excellence? Sounds like a great way to increase the BOM, with no benefit to the end user.

My car runs no faster or better due to better stitching of the leather on the dash. It is there because it is expected that when cost factor is taken out, better quality goes in.

Can it make a difference? In extensive testing I did of much lower quality AVRs with high jitter HDMI input, I could not show any of them having jitter distortions above masking thresholds. With high-performance DACs with far lower jitter, I am confident even trained listeners would fail such tests.

 

[Mivera]

My car runs no faster or better due to better stitching of the leather on the dash. It is there because it is expected that when cost factor is taken out, better quality goes in.

Can it make a difference? In extensive testing I did of much lower quality AVRs with high jitter HDMI input, I could not show any of them having jitter distortions above masking thresholds. With high-performance DACs with far lower jitter, I am confident even trained listeners would fail such tests.

The higher end the DAC, the more noticeable it is. Because all the jitter isn't buried in all of the other hash. I have several DAC boards that are identical besides the clocks soldered on them. And my clients who compared them blindly, could clearly pick out the superior clocks. However if a client wants to custom order a $1 Fox xpresso clock from Digikey for their DAC, I have no problem soldering it on the board for them. In fact I'd prefer to save the good clocks for folks that have good ears. Giving a good clock to a man with poor hearing, is like giving Stevie Wonder a set of sunglasses with Maui Jim's special "Polarized2" tint. I highly doubt he'd benefit from it.

 

[fas42]

It was. It just wasn't in that graph that I showed. Here are the threshold results in nanoseconds for music:

Thanks. There is some range in the results, showing variability of sensitivity to the resultant distortion - a good start to better understanding.