ESS chips

But but, my Ayre has the 9018.

Still wipes the Oppo's ass thou.

 

Really interesting.

I'm not sure about Amir's reviews. He wrote a critical review of the Exa e32, and then he compared it to the Yggdrasil and showed how much lower its distortion was. But I've heard the Yggdrasil. It's one of the best DACs I've ever heard, and better than the two Exasounds I've had here. As always, I find measurements interesting, but interpreting them is little better than reading tea leaves, a lot of what they're measuring is inaudible and I suspect that a lot of what's audible they aren't measuring, like the overshoot in the ESS chips.

Whatever the reason, the DACs I've used sound far more different than the measurements would suggest. Why, for example, does my Lynx e22 have such a palpable image? I can't think of a measurement that would correlate with that.

I haven't read anything about the JA controversy, but I was puzzled by his reaction to the Yggdrasil. He seems to have rejected it on the basis of measurements rather than listening. But it's one of the best DACs I've ever heard -- the descriptions here correspond very much to what I've heard, most recently a few weeks ago at AXPONA. I'd likely own one myself if I didn't need a multichannel DAC for triamping. Anyway, I just thought the quote was interesting because it seemed to match my own initial impression of the e38 Mk II.

I wasn't ithat mpressed by the DAC3B at AXPONA -- it was playing through the AHB2, which I have and know, and their line buffers. It sounded extremely clean, but too bright and lacking in substance. I hesitate to judge something on the basis of show sound, though, could have been the speakers, and the rooms at AXPONA had serious midrange glare. And I heard a lot worse. The Soekris sounded awful -- really grainy to the point at which I wonder whether there wasn't something wrong with their unit -- and the big Magico system with the Dave (!) was really disappointing -- not the Dave's fault, of course, the system was playing in one of the worst spaces I've ever seen.

 

Heh, I wasn't actually referring to hearing the sound quality of the DAC, but rather to the transient characteristics. At one extreme, stats tend to have etched highs, and those would sound even more etched with an underdamped filter. At the other extreme, something like a soft dome tweeter could benefit from it, having more liveliness and sense of air. Tweeter polar pattern would also have an effect. So in a sense, a case of component matching.

 

"Paper tearing highs" says it all. Wonder what Ayre did to tame them?

Sadly, agree about what's happened to the audio mags.

 

Of course, TAS says everything they review is the best thing they've ever heard.

 

A custom filter

 

Just being tongue in cheek. I certainly agree with him here. I find that you can usually read between the lines to get a sense of what they really like.

 

I had to laugh at the review of the Mytek.

I think it's interesting that you said of the Matrix "There is a slight glare ror hardness to how the last octave is rendered" and in the Invicta, referred to a issue in the upper mids and lower treble. Different, but I also had treble issues with the e28 Mk II, another 9018 design. And yet the symptoms you describe were different.

I've been listening more to the e38 Mk II, and it seems to me I'm also hearing issues in the highs, but, again, different from the issues in the e28. I wonder if they have something to do with the infamous bump.

I don't think I've ever heard similar phenomena with non-ESS chips. They do seem hard to tame.

 

Not being an idiot of a company? I called the Codex the last Charlie hurrah since it came out before his passing.

He was like cray cray on crack but he never made a digital sounding shit product.

 

Ouch, are the previous gens that terrible? The 9038Pro the first good one (when well implemented)?

 

As mentioned before, 9018 in the Geek family was well done.

 

It may depend on how you define "competitiveness".

Put it differently, passable yes. Contestable (against dac frontiers) no: there are passable 9018 and 9028's (some meaningfully better than benchmark). But I can't really think of any single product that can contest against Gungnir Multibit & Yggdrasil -- or convert-2 & solaris. So in this context, 9038 pro was the first to prove the benefit of the doubt.

Of course I didn't hear all the implementations, so be aware of type-2 error!

Bottom line is that I can comfortably recommend benchmark dac 3 (9028) to anyone I hate any day.

Again, I'd like to emphasize that 9038 can be fucked up. Kingwa proved it ("must avoid even in a pinch"). Topping (dx7s) was kinda mixed bag (=more tolerable) but meh at best.

 

DAC design isn't easy! Charlie was a master and I miss him, despite having been the target of one of his infamous attacks.

 

ESS/Sabre history from a key designer, the human auditory system, noise, and what they are trying to do with hyperstream (their attempt at fixing ΔΣ) . Good talk on jitter, pll’s, vol control. Interesting this: Non-periodic steady state noise— explaining how the noise surfaces during transitions, and why measurements aren’t picking this stuff up… noise vs. dc offset graph of ΔΣ is an interesting way to try and see it.

He addresses directly as to why there’s no documentation out there!

I’m pretty neutral on Sabre, only heard a few in the (way) past— always thought it added a “tinsel” (fake detail) sound to everything, but honestly did not spend enough time with those. Their approach seems very complicated (math intensive) but interesting with the research into human hearing, it could probably go very wrong, meh, or very right! Anyway, tough beast to tame, ΔΣ!

Maybe we should bug Martin Mallison for a follow up talk, see where they are now and what they learned getting to hyperstream2.

Aside: Martin says he worked for a company in Boston before ESS. I think that's possibly Analog Devices, they are headquartered there.

 

The DAC chips account for less than 10% of the general sound impressions. The matter of the fact is, within a well-engineering D/A converter of modern topology (sigma/delta modulators), price of a single DAC chip is negligible. A properly designed analogue stage that follows it costs a lot more, not to mention a quality power supply circuit. They cost more because they mean more.

You have to realize that chips like the offerings from ESS come with myryad facilities like integrated time-domain correction and algorhythms that predict which bit is going to be mistimed next so the error is corrected before it happens. So if I'm a manufacturer of D/A converters, I buy the chip from ESS because all the advanced data manipulation is already contained within that chip so I don't have to worry about providing a low-phase-noise oscillator as much. It would cost me a lot more to do it myself so I simply buy a chip that mostly takes care of it for me.

The thing is, if I want to build a high-performance D/A converter, then I run into all sorts of problems. I then realize that the ESS chip I have chosen works best with discrete voltage regulators and that, in order to take advange of it's high dynamic range, I must now use low-profile Holco resistors that cost 2€ a piece and I need oh, about fotty of them. So I decide against the ESS in the end and use a more affordable voltage-output DAC chip from AKM or Cirrus Logic because it will give me similar results with the standard roll of SMD resistors.

What you need to understand is that industry is profit-driven. If somebody turns up with a more advanced part, it will immediately be used in a way that will devalue it's advanced properties because the rest of the part lot will now be cheaper. ESS might have superior technology but ultimately does not care where it's being implemented as long as it's being licensed to the IC manufacturers.

Cheers!
LB