Mutibit vs Delta Sigma (old vs new)

[Arg]

Bill, when I built and installed a high performance Guido Tent clock onto my CDP in the 90's, including a DIY over-the-top PSU that I was inappropriately proud of designing, and did the same for a friend for his (different brand) CDP, we both reported hearing a significantly different sound quality from our respective players.

However, the nature of human perception is such that I cannot possibly pretend that such casual listening tests comprise evidence that they ACTUALLY sounded different.

C'est la vie.

[Drizt]

However, the nature of human perception is such that I cannot possibly pretend that such casual listening tests comprise evidence that they ACTUALLY sounded different.

C'est la vie.

Exactly right.

[zenelectro]

Bill, when I built and installed a high performance Guido Tent clock onto my CDP in the 90's, including a DIY over-the-top PSU that I was inappropriately proud of designing, and did the same for a friend for his (different brand) CDP, we both reported hearing a significantly different sound quality from our respective players.

However, the nature of human perception is such that I cannot possibly pretend that such casual listening tests comprise evidence that they ACTUALLY sounded different.

C'est la vie.

There is a high possibility that you did indeed end up wiith different SQ from the respective players.

Here are a few reasons:

- Even clocks themselves vary, the biggest factor is the crystal itself. The only way to guarantee against this is to use higher qulaity crystals that are individually tested. Not at the price point of $30 oscillators, however good they are.

The clocks I use are made with super high Q factor crystals and they are individually tested for phase noise. Unfortunately they cost a lot more than 30 bucks.

- Crystals produce vibrationally induced phase noise (jitter). This is a known phenomenom and is measurable. You can get certain oscillators that are made to be specifically low G sensitive, also certain cuts of crystal are less

sensitive to 'G' induced phase noise. Again, you won't find it in a $30 XO.

- Probably the biggest factor is the difference in clocking circuitry (maybe even DACs) in the two CD players. Also, even though you made same PS for them both, the raw DC would have an effect.

As Steve N has alluded to, clocking is a very complex area of dig audio and we do indeed have audible differences between many different design approaches to clocking.

[bhobba]

Bill, when I built and installed a high performance Guido Tent clock onto my CDP in the 90's, including a DIY over-the-top PSU that I was inappropriately proud of designing, and did the same for a friend for his (different brand) CDP, we both reported hearing a significantly different sound quality from our respective players. However, the nature of human perception is such that I cannot possibly pretend that such casual listening tests comprise evidence that they ACTUALLY sounded different.

Again IN YOUR OPINON. I have heard more than enough gear to know audible differences exist outside of controlled tests. If you don't agree - fine - but I do not agree that controlled tests are the be all and end all.

I mentioned what I had heard and its fine to point out its was not a controlled test but to want to rehash this issue is IMHO quite pointless.

Thanks

Bill

Edited March 27, 2012 by bhobba

[Drizt]

De Nile is not just a river in Egypt

[bhobba]

De Nile is not just a river in Egypt

Correct. There is no 'De Nile' my position is different to yours.

Thanks

Bill

[empirical]

Steve,

What equipment do you use to measure jitter?

Have you confirmed that there is no other change other than the jitter spectrum/level when moving between clocks?

I use a 7GHz scope with jitter software and a spectrum analyzer for direct measurements.

The only difference is the clock that I change.

Steve N.

[jkeny]

Instead of using the term "properly controlled" as a get-out-clause - can you define "exactly" what set-up & equipment was used for these tests? It's simply not enough to say they were DBT & therefore "properly controlled". It's very like the term often used in a similar type arguments when trying to "prove" one device sounds the same as any other - the term "competently designed". This is the get-out clause for all such arguments because it's a tautology - it it sounds different then it's not "competently designed" but nobody will ever give you an exact definition & the details of what "competently designed" means. Funny, isn't it

[jkeny]

I use a 7GHz scope with jitter software and a spectrum analyzer for direct measurements.

The only difference is the clock that I change.

Steve N.

Steve, can you give some more details on the scope & jitter software? Not trying to trip you up, just qualify the results.

In a clock, you know that these are not just different crystals but also different internal components or circuitry inside the clock can! This can well effect the RFI, PS noise & other issues besides just jitter! The ones that that you mentioned, Audiocom, Superclock, Ultraclock, Turboclock, have they not got different external circuitry surrounding the clock cans?

[zenelectro]

Steve, can you give some more details on the scope & jitter software? Not trying to trip you up, just qualify the results.

In a clock, you know that these are not just different crystals but also different internal components or circuitry inside the clock can! This can well effect the RFI, PS noise & other issues besides just jitter! The ones that that you mentioned, Audiocom, Superclock, Ultraclock, Turboclock, have they not got different external circuitry surrounding the clock cans?

Hi John,

The most important factor is the quality of the quartz crystal itself.

After that obviously the oscillator circuitry and then the power supply. Different oscillator circuits have various degrees of power supply rejection and it is not

a given across the board.

As an example the clocks I am getting made now have a phase noise spec of -120dBc at 10Hz offset which is very low jitter, however they do this with

an internal reg that has hundreds of nV/rt Hz noise at 10~100Hz. I am currently implementing discrete 2nV/rt Hz regs to replace them but this may

not yield much better measured phase noise (jitter). I am pretty sure they will sound better. We will see.

Generally the design will require some amount of trade off WRT phase noise distribution. If you want super low phase noise close to the carrier (low freq), the

noise floor at higher freq can suffer and opposite also is true. Some XO's with super low phase noise floor (>170dB) will have poorer LF (10Hz~100Hz) phase noise.

Further to this there are optimum base frequencies which are 5 to 10MHz. Once you go above around 10MHz low freq phase noise becomes less optimal - I'm not sure

why this is but some manufacturers are using multipliers to get better phase noise (at low frequencies) by multiplying, say a 10MHz x 5 to get a 50MHz clock with

much better low freq phase noise than a 50MHz unit. This will trade off some HF noise floor.

[jkeny]

Zen, if we are talking purely about jitter in isolation to everything else then I would rephrase your statement to read "the LIMITING factor is the quality of the quartz crystal itself". I wouldn't say that it's the most important factor when you consider all the sources of jitter. In fact, as signal correlated jitter is the most audibly noticeable, I would say that PS disturbances, SPDIF receivers, cable distortions, ground bounce, & probably a few others are more important sonically.

If looking at the digital audio chain more holistically, I would probably cite ground & CM noise as more important sonically than jitter & that's why I asked Steve about being sure that it was only a change in jitter which is the cause of the sonic improvement he reported when changing clocks

[zenelectro]

Zen, if we are talking purely about jitter in isolation to everything else then I would rephrase your statement to read "the LIMITING factor is the quality of the quartz crystal itself". I wouldn't say that it's the most important factor when you conside r all the sources of jitter. In fact, as signal correlated jitter is the most audibly noticeable, I would say that PS disturbances, SPDIF receivers, cable distortions, ground bounce, & probably a few others are more important sonically.

Yeah, I was talking in context of the oscillator - which is where it all starts.

Obviously you are dead right and all these things make a big difference. Having said that I am amazed how important the oscillator is despite often how far back or removed in chain it is

If looking at the digital audio chain more holistically, I would probably cite ground & CM noise as more important sonically than jitter & that's why I asked Steve about being sure that it was only a change in jitter which is the cause of the sonic improvement he reported when changing clocks

Agreed, this is all really important. Ultimately the clock(s) should be in the DAC. Sadly it so rare that we have this ideal situation (without ASRC etc).

Wouldn't it be good to go back and redesign digital audio. Separate clock line, no pro / consumer sample rates etc.

[empirical]

The ones that that you mentioned, Audiocom, Superclock, Ultraclock, Turboclock, have they not got different external circuitry surrounding the clock cans?

The Audiocom devices have exactly the same external buffer and impedance termination. The Turboclock uses faster devices and Hynes regs.

It's differences in jitter spectra that I'm hearing. I also modded these clocks and these mods only affect jitter. I can plainly hear the improvements. Improving the slew-rate of the buffer improves jitter by reducing the switching threshold variability at the receiver.

Steve N.

[empirical]

Wouldn't it be good to go back and redesign digital audio. Separate clock line, no pro / consumer sample rates etc.

That's exactly what I2S is. Separate clock lines, no encoded data.

Steve N.

[jkeny]

The Audiocom devices have exactly the same external buffer and impedance termination. The Turboclock uses faster devices and Hynes regs.

It's differences in jitter spectra that I'm hearing. I also modded these clocks and these mods only affect jitter. I can plainly hear the improvements. Improving the slew-rate of the buffer improves jitter by reducing the switching threshold variability at the receiver.

Steve N.

So I take it from your answer that you didn't measure for any other change?

[georgehifi]

Back to the Multibit vs Delta Sigma debate,

CarlosFM: Quote> Phillips called Bitstream to their 1-bit dac technology.

Crystal calls theirs Delta-Sigma.

Technics/Panasonic (Matsushita) calls it Mash.

All fancy names for basically the same thing.<Quote

Here is another very interesting link to read on the subject, with a good expaination by Thorsten Loscech (alias Kuei Yang Wang).

http://www.diyaudio.com/forums/digital-source/15439-how-does-delta-sigma-dac-work.html

Cheers George

[DoggieHowser]

I guess the EMM Labs and Playback Designs custom DSD DACs are variants of the delta sigma concept but they don't sound anything like them.

Maybe it helped that both founders were involved in DSD engineering tech and working on recording equipment prior to this.

[zenelectro]

That's exactly what I2S is. Separate clock lines, no encoded data.

Steve N.

Of course, but there are no standards for optimum inter device transmission. IOW a) separate dedicated lines b ) proper RF connectors (bnc etc) c) stated impedance ie; 50/75/100 ohm

[Drizt]

The ultimate would be the DAC has the high quality clock and data is transferred to the DAC via ethernet as pure data None of this crap of having to rely on the timing being dependant on the transport, its just plain silly

[Telecine]

The ultimate would be the DAC has the high quality clock and data is transferred to the DAC via ethernet as pure data None of this crap of having to rely on the timing being dependant on the transport, its just plain silly

Ethernet is a poor choice, the delivery of packets of data isn't guaranteed.

Edited March 28, 2012 by Telecine

[Drizt]

Ethernet is a poor choice, the delivery of packets of data isn't guaranteed.

Depends on the software and protocols used

[Telecine]

Depends on the software and protocols used

No it doesn't.

There are some suitable fibre channel technologies.

[Drizt]

No it doesn't.

There are some suitable fibre channel technologies.

Yes it does For this scenario it is a none issue. Send data to a buffer, then let the DAC clock out the data at its leisure. There would be no problems at all with it. Why not use a ubiquitous connection / transport mechanism ?

Transferring files across a local network at gigabit speeds is a trivial task. Makes it a non issue really when we are talking about audio files.

Using TCP works fine for the squeezebox

It has been done, obviously it works

Edited March 28, 2012 by Drizt

[New Sensations]

Yes it does For this scenario it is a none issue. Send data to a buffer, then let the DAC clock out the data at its leisure. There would be no problems at all with it. Why not use a ubiquitous connection / transport mechanism ?

Isn't this what the PSAudio Network Bridge does? Receives data via ethernet, reclocks it via the Digital Lens (buffer) and then parses it to the DAC chip via I2S.

[Drizt]

Isn't this what the PSAudio Network Bridge does? Receives data via ethernet, reclocks it via the Digital Lens (buffer) and then parses it to the DAC chip via I2S.

I'm not familiar with that device but I will look it up. The process I describe is exactly what the SB units already do .... data across Ethernet (TCP) to a buffer and then clocked out of the buffer using its own clock.