Design Criteria in DACs

[Ralf11]

from a post by @austinpop on his review of the Denafrips Terminator; as he points out, it is not a complete list:

• the clock architecture (and phase noise, close-in)

• the PSU design

• the analog stage

• noise isolation
  • at inputs
  • RF/EMI
  • between stages
• quality and selection of discrete components and modules, like capacitors, resistors, regulators, etc
• mechanical isolation
• quality of board layout, especially connectivity between key subcomponents

 

 

 

- several of these factors are things a consumer can look for (in ads or in white papers)

   ... and might be better things to focus on than what DAC chip is used

 

There have also been a number of threads on the various criteria (e.g. the RAM / RFI thread) and numerous scattered comments by @barrows, @jabbr, @et al.

[marce]

Its not PSU its the whole power delivery system that is critical in a DAC, be it SMPS or LPS (LPS's are not the panacea for all evils), slapping a LPS on the front end does not guarantee better quality in fact often it is worse because most noise is in the audio frequency range... But it has become a Audiophile belief/mantra use a LPS!

First EMC/signal integrity, these are two sides of the same coin, EMC is the effect of noise from and to the external world to the PCB/unit, signal integrity is the local effect of noise etc. on the signals running around the board (as well as transmitted signals). Plenty of information out there on mixed signal design, here is a nice started for 10:

http://www.x2y.com/filters/TechDay09kr_hpa_Track2_1_Precision_Analog_Designs_Demand_GoodPCBLayouts _JohnWu.pdf

As to determining all this in a commercial audio DAC, how are you going to do it, most Audio based commercial DAC's are not exactly brimming with relevant information, mostly marketing blurb and without some serious measuring gear you cant determine much from looking at a PCB, especially as most people don't know what they are looking at or what to look for. What do you look for for noise isolation...

Component quality, how do you determine that, what factors do you use... Would small size X7R capacitors near device power pins put people off,  Same with crystals and clocks, how without measuring the actual devices in situ do you determine how good the clock is, and how good does a clock have to be, as I often see unattainable figure regarding jitter etc and claims of femto second jitter (!!!!!), in fact it would seem that an audiophile DAC has to have better clocks than some microwave DAC/ADC sections I have seen and worked on.

How much mechanical vibration is required for a component in a domestic environment...

There are so many factors, do we try and reverse engineer a DAC to determine how good it is... believe the marketing (LOL).

This is one situation where I think how it sounds would be a better way of determining whether you like a DAC or don't... I know how to lay out low noise precision DACs (and could write a book on the subject, but as so many others have why bother),  but without some serious reverse engineering its harder to judge ultimate quality from looking a a product, though naff designs and low quality should be pretty obvious.

And each of the sections you mention would need their own thread, especially once we start adding the Audiophile factors to the discussion...?

 

 

So firstly a study of EMC, start with Henry Ott, Ralph Morrison, Keith Armstrong, then you'll get an idea what to look for for good EMC practices.

Signal Integrity, Eric Bogatin, Dr Howard Johnson just two names to start with.

Power integrity again books and articles regarding it.

Mixed signal design, just google it, tons of info, some conflicting such as multiple or single grounds.

The only way I belkieve is to discuss one topic at a time otherwise it will meander all over the place as a discussion.

[jabbr]

New "king of the hill" in ultra low noise linear regulators, yes, no point in having a good clock if you don't have a great PSU.

 

Got to love Analog Devices (manufacturer of the well known LT3045) who not only develop the latest and greatest RF linear regulator but provide phase noise measurements of their DAC chips as affected by the new regulator:

 

http://www.analog.com/en/analog-dialogue/articles/improved-dac-phase-noise-measurements-enable-ultra-low-phase-noise-dds-applications.html

http://www.analog.com/en/products/adm7155.html#product-overview

 

and all you ever wanted to know:

 

http://www.analog.com/en/analog-dialogue/articles/analyzing-and-managing-the-impact-of-supply-noise-and-clock-jitter-on-high-speed-dac-phase-noise.html

 

[esldude]

All of you are going about this backwards. 

 

Define design criteria for the analog result.  Turn people loose to achieve this any way they desire.  There is no single perfect best way for most parts of a DAC.  There are good and bad examples of nearly every part using every design approach.  The end result is all that should matter. 

 

One of the German fellows making a revered $40K plus DAC, also makes an equally revered phono stage.  It uses an SMPS.  An extraordinary SMPS.  Why?  Because he can achieve noise levels and stability that are simply not possible with an LPS approach. 

 

The end result is where you put your criteria for a design.  

[Miska]

On 9/13/2018 at 2:57 PM, marce said:

Same with crystals and clocks, how without measuring the actual devices in situ do you determine how good the clock is, and how good does a clock have to be, as I often see unattainable figure regarding jitter etc and claims of femto second jitter (!!!!!), in fact it would seem that an audiophile DAC has to have better clocks than some microwave DAC/ADC sections I have seen and worked on.

 

Usually they just quote specs of the oscillator component. But when you measure actual conversion jitter on the analog output, things certainly tend to look quite different... Usually the clock has been poisoned one way or the other right after leaving the oscillator (signal integrity), or even before if the oscillator power feed is dirty.

 

Of course just jitter figure in pico/femto-seconds doesn't tell anything. One needs to look at phase noise plot of the oscillator (low frequency phase noise matters the most there). Or more preferably J-test output of the entire DAC device.