Real audio science - exploring beyond the obvious and bringing back true science and exploration

Yup. What the title says.

Most of you guys know that I've been pretty bored with gear for several years now. My setups remain stable. I continue to build speakers for fun, only to deconstruct them, purely for the sake of seeing how different designs sound. In the personal audio space universe, I don't think being bored is a bad thing. The fact is, personal audio gear (10-15 years after when I first started) is so much broader with actually better higher-value gear. We don't have to build our own EHHAs, AMB B22s, or CMOY variants anymore. Headphones that were already great 10-15 years ago continue to be great. The high-end headphones that were good continue to be good. And it seems that most people here on SBAF don't give a damn about the latest and greatest overpriced exotic headphone because ultimately, they are just different, not necessarily better.

Twelve years ago (and I can't believe it's been that long), a lot of headphones sounded like shit. To be precise, many had horrible treble peaks or irregularities that hurt my ears (and the ears of @Hands, @rhythmdevils, etc.). My reference back then were speakers that I had built that met a neutral target (neutral on-axis / B&K AES 1970 target at listening position), so many headphones left me scratching my head. Why the f**k is the treble so rough and peak I was thinking! I was avid Head-Fi poster back then trying to convince to fellow Head-Fiers what I heard was real. My approach ended up as this: https://www.head-fi.org/threads/headphone-csd-waterfall-plots.566929/



This scientific approach seemed to piss some people off, including Jude, who I had heard secondhand that I was trying to destroy Head-Fi's business of being a marketing machine. (Yes, I was trying to put a stop to overpriced undeserved hype, e.g. Sennheiser HD700, Beyer T1). Some HF members thought I was trying to do measurements for the sake of measurements. The latter was never the case because these CSDs, once one learned how to understand them with trained their ears, could actually correlate the measurements with subjectively perceivable phenomena (namely different types of treble nasties).

A few years later Changstar was born. I added frequency response and distortion vs. frequency to CSDs to measuring headphones. After that was SBAF where I added distortion vs. frequency at varying SPL and few experimental ideas such as attack and decay envelopes at five different frequencies and burst decay "wormsign". Most recently I explored transient response to identify the sound of different tube brands. Early on we dabbled in measuring amps and DACs, but never moved forward because none of these measurements had much correlation to what readers were hearing - well with the exception of stuff that was broken***.

What hasn't been said is how many dead-ends, where measurements had very little or no correlation fo sound, I encountered. I think that Elrog 300B tube thread demonstrated this. That's really just the tip of it. The fact is, I've never written about the dead-ends and fails; and there've been a ton of them. With measurements taken using steady state signals, we've probably exhausted them to the point where some sites are doing measurements for the sake of measurements, usually with a single number "SINAD" without giving consumers any information of how a piece of gear may sound.

This is not what I want to do. Such an approach is ridiculous. The "SINAD" (or precisely SINAD at V level into Z load at B bandwidth with F frequency stimulus) is but a single steady state measurement. This is the equivalent of trying to sort out cars by top speed or spirits by alcohol percentage. Even if we have a plot SINAD vs output level, this is still limited because there are infinite permutations, what about frequency, or load. We would need a five dimensional matrix have a full view! I digress however. The point I wanted to make is that we must avoid measurements for the sake of measurements to the point where they are not telling us anything. I doubt with blind testing anyone can distinguish SINAD > 50db. Heck, some of the people doing measurements for the sake of measurements have been shown to have demonstrably unable to hear any differences through blind testing in situations where I (and I'm sure many others on SBAF could**.

Anyway, I wanted to say simply is than readers can maybe expect me to review less gear, but when I do review gear, to present different ideas for measurements (including the dead-ends), particularly looking into transient response, or even different presentations and analysis of steady steady measurements. I wish I had the knowledge and measurement gear (the APx55 does make things easier) when I was younger a decade ago when I had more time. I expect many of my ideas will be half-baked and hoping that younger intelligent and hungry individuals pick it up*. I'm willing to admit I don't know everything and that there is so much we don't understand. Those who cannot admit this are not scientists.

*Sadly, it seems that most people, even the seemingly smart ones, are too lazy for independent thought, and want to swallow shit like Sean Olive target curve (the three or four versions / moving targets of it) hook line and sinker.

**LOL, it's just pointless trying to argue about measurements and correlation to sound to people who just cannot hear a difference. To be honest, I think part of this is obstinance / confirmation bias (thinking there will be no difference in sound purely from SINAD measurements).

***Steady state measurements aren't useless, however they do tell us if something is broken. This is different from measurements for the sake of measurements, e.g. 128.2db "SINAD" vs 126.4 on the Excel bar graph.

 

I know that you are a manufacturer, but I think you should participate. I bet 99% of readers on SBAF (1% always gonna hate) would be super interested.

You've never come off as contentious*, I don't remember ever hearing you crap** on other manufacturers' products. The way you've come off is "this is what we do, take it or leave it, there are many other options out there".

If anything, if you reveal your knowledge, all that possibly can do is leave you at competitive disadvantage.

The thing is this: Audio designers keep a lot of knowledge to themselves. Many of these guys, at least the good ones, are dead or will be soon. You are the last generation, the youngest among them. When you go, that's it. 50 years later after the oceans rise a few inches (and they will regardless of any green movement), someone will dig up schematics of your best sounding amps and go "oh shit". Any tidbits you can share would benefit this small community and maybe the world because no one else is going to do it.

--

*If other manufacturers see your involvement as contentious want to bitch about it, then they can do the same and offer their own insights. Many MoTs are incredibly dumb / cheap / bitches. For example, they can bitch all they want about SBAF coverage in favor of X, Y, or Z MoTs, but they don't bother sending us product to review. And if they are afraid that we won't give an expensive hi-end product glowing praises, then that's their problem (especially since I know many SBAF members other than myself have been willing to participate in prototype evals, etc.)

If readers find your involvement as contentious, then they can also stop reading your Head-Fi thread.

**Well, other than Mike's T-shirt that DS DACs suck. Although I don't think people got the joke that most DACs you sell are DS DACs.

 

I am baking in the MIL-B. Initial impressions are that the MIL-B is finally worthy to be among the other Yggdrasil flavors, the LIM and OG. Let me put in some more listening time. I want to establish subjective impressions first, write them down, compare them to MIL version 1. Then after noting subjective differences take a look at the higher frequency THD performance using appropriate bandwidth.

 

If you like the OG and it works with your setup, then stick with it. Again, the MIL, LIM, and OG are flavors. The original MIL sucked subjectively speaking and I wasn't afraid of pointing that out back then and now. (The irony is that Schiit best "AmirNAD" measuring DAC sounded the worst). The current MIL DAC with the TI "B" chip is still a proto and I have no idea whether @schiit plans to release this or not. The point of this discussion is not about Schiit or their product. I know Jason felt unease talking about this as do I. The only reason we are is because we want to further the science of sound - or try to correlate measurements to sound. The point of this discussion is about how simplistic measurements such as "AmirNAD" (SINAD 1kHz at fullscale output at 20kHz bandwidth) is insufficient in capturing qualitative aspects of sound. In this case, the intent is to examine the measured performance past 20kHz. Signals past 20kHz despite technically not being audible by humans (heck make that 12kHz for old farts like me)... all the old school amplifier designers like Jason, Craig (EC, JBL, UREI), Andy Hefley (of GAS fame) that I have spoken too care about this region. Sure they did not have APx555's like Amir, Jude, Golden or I, but I know in my discussions with them, they most certainly looked in their 'scopes to see what was happening. They all had their secrets with respect to understanding measurements.

That being said, I would have to state that the "MIL-B", unlike the original MIL, is worthy of being included as a different flavor to the LIM and OG Yggdrasils. Let me repeat myself with respect to the original MIL's suckyness:

N A NUTSHELL:​

The Schiit Yggdrasil More is Less performs spectacularly in AmirSINAD, or more precisely in THD+N when using a 1kHz steady state sine wave at full scale into a load of 200-ohms (balanced). If you believe this measurement to be the primary determinant of state-of-the-artness, performance, sound quality, then by all means, the Schiit Yggdrasil MiL must be be considered and would be highly recommended.​

From a subjective point of view, the Yggdrasil MiL is in my opinion, the worst of the Yggdrasils in that it lacks the emotion, lacks the resolution, and imparts an odd spittyness to the highs (which attempts to component match with laid-back amps and headphones could not resolved). It does have the most neutral tonal balance however. Many have asked me if the MiL should be considered if one wanted a neutral sounding R2R DAC (instead of LIM which is slightly dark or OG/A2 which is slightly V-shaped). Perhaps, but I feel the MiL is a poor value for its subjective technical performance.
​

The MIL-B no longer lacks the emotion, lacks the resolution, or nor has that spittyness to the highs. In fact, I'd day the MIL is easily as resolving as the OG, with both the OG and MIL-B outresolving the LIM. Among the LIM and OG, it still retains the most neutral sounding signature. It doesn't have the harmonic richness of the OG or the timbral naturalness of the LIM. With respect to the "sturm and drang", the OG has always led over the LIM (the LIM is dark, maybe staid sounding in comparison), the MIL-B fits in between them. The MIL-B has an exciting sound. The OG and LIM seem to have a bit more dynamic heft in the lows, but the MIL isn't weaksauce or thin at all.

To put things into context, the LIM is my daily driver (if I had infinite money I'd have the Rocka Wavedream) where I use it as my primary DAC next my computer for listening to music and for gaming. The LIM is nice for music through headphones. The dark tonality of the LIM works with the JAR600 and RS2X (F pads) or various other headphones (which generally have less than smooth highs).

I do most gaming with speakers actually and this is where the MIL-B shines. The LIM is slightly too dark. The MIL-B is a bit more incisive, sparky, less staid. This helps cut through so I can hear things better. I did test with music too, with my suite of typical pop tracks that could be problematic (Dua Lipa, MSFL Pixies, Nirvana) and found no "treble f'ckry" issues in the highs. However my preference was still for the LIM for music.

Anyway, what I wanted to do here was to establish that I did hear a sonic difference, not insignificant. The next step will be measurements (outside of 20kHz bandwidth).

*And BTW, this is what good audio designers do - they listen to shit they make. They look at measurements that they believe matter, often at beyond the audio band. Good audio designers do not rely solely on AP-555 1kHz SINAD. Anyone who suggests that they do is an armchair quarterback full of himself at best or a shill for deaf uncreative audio companies who are supply (as opposed to demand) driven.

 

I will have to get to this tomorrow. Here is what I have slated:

• AmirNAD* (SINAD at 1kHz, 200k-ohm load, 20kHz bandwidth) - this is intended as as the simplistic apples to apples comparison between the MIL and MIL-B
• Comparison of what I have that I took for original MIL: 8khz, 17kHz, 19kHz
• THD vs Frequency at various output levels

Curious to see how the new B chip from TI performs.

* The reason I keep stating "AmirNAD" because I keep people hearing "SINAD". SINAD is f'ing MEANINGLESS unless it's qualified with stimulus signal, output level, frequency, load, and bandwidth. Heck, I'm pretty sure the APx555 manual states this somewhere, that SINAD must always be qualified with this information.

Teaser: AmirNAD is more or less the same (0.2db - 0.6db difference) between MIL-B and MIL despite the DACs sounding different.

MIL-B proto AmirNAD


Compare with original MIL


To be continued.

 

OK, so now let's take a look at SINAD not purely based on 1kHz. Let's try 8kHz - take a look at the full spectrum here. And also let's not limit ourselves to 20kHz bandwidth, but consider the signal out to 90kHz.

original MIL
8kHz, 200-ohm load, 0dbFS signal, balanced outputs from the DAC
44.1k (looks worse, less "grass", than 48k but wanted to do apples and apples)


MIL-B proto
8kHz, 200-ohm load, 0dbFS signal, balanced outputs from the DAC
44.1k (looks worse than 48k but wanted to do apples and apples)


Summary: MIL-B proto has much lower 2nd and 3rd harmonics than the original MIL at 8kHz,

 

Let's now compare 14kHz and 19kHz:

Original MIL
14kHz, 200-ohm load, 0dbFS signal, balanced outputs from the DAC
44.1k (looks worse than 48k, more "grass" but wanted to do apples and apples)


MIL-B proto
14kHz, 200-ohm load, 0dbFS signal, balanced outputs from the DAC
44.1k (looks worse than 48k but wanted to do apples and apples)
View attachment 36356


--

Original MIL
19kHz, 200-ohm load, 0dbFS signal, balanced outputs from the DAC
44.1k (looks worse than 48k, more "grass" but wanted to do apples and apples)
Note: Averaging turned off for this one


MIL-B proto
14kHz, 200-ohm load, 0dbFS signal, balanced outputs from the DAC
44.1k (looks worse than 48k but wanted to do apples and apples)
Note: Averaging turned off for this one


It's possible I took the original MIL measurement improperly. I dunno. There does seem to be a pattern of THD getting worse as we go up in frequency with the original MIL though, so it's likely I didn't bork the measurements. Let's take a look at other presentations / measurements in the following posts.

Before we go on, I would like to tackle one interesting question. Most of this stuff, it's so far up in our hearing range at such low levels that none of us can hear this. Maybe the 2nd harmonic of the 8kHz signal could be heard by a 15 year old child, but anyone older, it would be unlikely (read here for experiment). In theory, none of these measurements should matter, but somehow maybe they do. I mentioned in a prior post that many hi-fi audio gear designers do consider the ultrasonic range (and no it's doesn't require an APx555). Maybe steady-state measurements in the ultrasonic range could be a loose indicator of what happens in signal transients (essentially what real music is). The fact is, we can guess, but at this point without more data (subjective and objective), we just don't know. My hunch is that there is something here.

 

Here is THD+N vs Frequency, apples to apples, at three different output levels

Original MIL
THD+N vs Frequency


MIL-B
THD+N vs Frequency

 

Here are my GONAD panels as opposed to AmirNAD (note AmirNAD is displayed on the GONAD panel as well). Not totally apples to apples because I am using the coaxial output with the MIL-B proto (which could be a slight disadvantage because the Unison USB is always cleaner).

Yggdrasil MIL original


Yggdrasil MIL-B proto


Anyway, just some ideas thrown about here. Hopefully someone else will run with them because I don't have all the answers. Please be smart and don't regurgitate from self-appointed / anointed experts (don't get me started on those awful Sean Olive targets - take studies for what they are and by all means do not put them on pedestals). There is a ton with what we know about measurements. There's a ton with what we don't know about measurements.