Holo Audio - Spring DAC – Level 3 – “Kitsune Tuned Edition”
[size=20.006999969482422px]Initial comments …[/size]First those of you that just want the highlights and don’t want to read a novel, there’s a TL;DR; summary of my thoughts here.
Then let’s hit the disclaimers …
The DAC I auditioned is the personal unit of the proprietor of Kitsune HiFi/Holo Audio USA. It is the final “Level 3” tuned version of the basic Spring DAC.
I am not receiving any consideration or perks for reviewing this unit (i.e. no discounts/no special pricing, no early-shipping or queue-jumping or anything else of that nature etc.).
I approached Tim for a demo unit as part my on-going DAC-search. Although they were all out with various far-more-prominent reviewers, he was gracious enough to loan me his personal DAC for a long weekend, along with a smattering of cables and other relevant goodies.
History/Context:I’m not going to bore you with a page of largely-unrelatable-nonsense about my audio journey; if you want to know my overall gear preferences it’s all in my profile, and you can use the search function here and/or refer to my (“so far”) on-going “Life after Yggdrasil?” thread for perspective on my DAC-auditioning experiences and preferences.
My principal point of comparison here is going to be to Schiit’s “Yggdrasil” as this remains my primary DAC, despite more than half a year spent looking for something that I might enjoy listening to more (and that meets some other, non-specifically-audio, criteria). That is largely because it’s something I was able to do back-to-back but, also, due to it being a usefully-common reference for a broad array of potential Spring DAC customers/existing Yggdrasil owners.
Any commentary that compares to other DACs, unless otherwise noted, is based on memory and my listening notes from those sessions – some of which are months-old now. As such, that means such comparisons are of limited worth and should generally be regarded as high-level impressions.
Finally, I will save my technical discussion of the DAC and its implementation for another post, as that is going to be of interest more in the context of a formal “all-up” review. However, I’ll say a few words about the principal (from my perspective) innovation in this unit.
Claim to Fame: Discrete R2R DAC with “Linear Compensation”:If you’re read my “Life after Yggdrasil” thread, you’ll have observed that my principal interest has been R2R DACs and/or DACs that do more than just couple a “standard audio” DAC chip (most of which are Delta/Sigma designs) with high-spec clocks, fancy power-supplies and excessively “unique” casework for what are, commonly, relatively silly prices.
The Spring DAC falls into the broader discrete R2R “resistor ladder” category of DACs. Discrete resistor-ladder R2R implementations use individual resistors (or groups of resistors), combined with high-speed switches, to implement the different output levels for each bit-value of a given sample. Leaving aside, for a moment, the realities and limits imposed by thermal and quantum noise in components, a 16-bit DAC needs differentials of 1:65536 (2^16) in the resistances in the “ladder” and a 24-bit DAC requires a ratio of 1:16,777,216 (2^24).
I’ll talk about this in more detail in a “technical” follow-up, but for now all you need to know is that such implementations are at the mercy of both a) the tolerances of the resistors in the “ladder”, as well as other impedances etc. from the switches and traces that are also part of the circuit etc. and b) these tolerances/resistor values can change with the operating temperature of a unit.
TotalDAC attempts to get the best accuracy by using naked, trimmed, high-tolerance resistors (currently I believe they’re using resistors toleranced to 0.01%). Metrum uses multiple ladders to span a smaller range of required values (e.g. 2 x 12-bit modules per channel), with additional processing/electronics to combine these values to yield a 24-bit result.
The Spring DAC uses two separate sets of R2R ladders, one to derive the primary output level and the second to linearly compensate for errors/drift in the primary level. A FPGA provides the necessary platform for the logic and control necessary to make this scheme work.
Beyond the linear compensation, which is the headline, and (as far as I know) unique, feature for this unit, the Spring DAC also has switchable oversampling/non-oversampling operation and separate decoding path for DSD.
What is the “Level 3 – Kitsune Tuned Edition” of the Holo Audio Spring DAC?
This DAC is offered in three versions, and the changes over the standard “Level 1” version of the Spring DAC are unique to Kitsune HiFi (hence the “Kitsune Tuned Edition”) and are installed at the factory at the time the unit is built.The changes are as follows:
- 99.99% silver, custom (hand-wound), 100 VAC O-type transformer
- All copper wiring is replaced with 1.5mm OCC silver
- Input film capacitor is replaced with a Mundorf silver/oil capacitor
- PCB plugs/connectors are removed and direct-wire connections are used
- Standard IEC inlet replaced with Oyaide pure silver/rhodium plated connector
- Standard fuse replaced with Audio Horizon Platinum Reference fuse
- Upgrade to use Jensen capacitors
I cannot speak to any audible differences that may be attributable to these changes, since I have not heard the “Level 1” or “Level 2” versions of the Spring DAC.
One neat thing I saw on the DAC I auditioned, was an image of the internals of the "Level 3" version applied to the top, outer, casing:
Sources and Interfaces:
I tested with several different sources. My normal setup runs either a Focusrite RedNet 3 or an Auralic Aries (w/ femto clocks and LPS) via AES/EBU XLR into Yggdrasil fed from Roon. I also tested with the USB output from the Aries, a Sonore microRendu and the I2S input via a Singxer SU-1 DDC.
I did some experimentation with a JCAT USB Isolator (based on the Intona opto-isolators) for the USB feeds as well.
The vast majority of my testing was done with Red Book audio, with some high-resolution PCM material and maybe a dozen or so “proper” DSD-format albums.
Amps, Headphones and Speakers:
I listened through both my loaded/upgraded Woo Audio WA-5LE Mk2 and Schiit Ragnarok, via JPS Labs “Abyss”, Focal “Utopia”, Audeze “LCD-4” and Sennheiser “HD800S”. Balanced connections were used throughout (the WA5-LE is a single-ended design but has “convenience” balanced inputs/outputs).
The speaker configuration was a Linn Akurate DS/1 into Akurate 4200 bi-amping Linn Akudorik speakers and a JL Audio Dominion D-108 sub.
The XLR outputs of the Spring DAC run at 5Vpp RMS compared to Yggdrasil’s 4Vpp RMS. This was immediately noticeable during simple back-to-back listening and you’ll want to match levels to avoid the typical “louder is perceived as better” perception.
For my testing, I matched levels to within 0.1 dB using a special box I built for performing comparisons and facilitating solo blind-testing. I’ve talked about that a little before if you’re interested.
I spent my first day, day and a half, prior to getting into any “formal” listening, experimenting with different settings on the Spring DAC, letting it warm up, and just generally relaxing with some music through it. It didn’t take very long to come to the conclusion that I liked the Spring DAC best with PCM sources in “NOS” mode (Non-Over-Sampling).That’s a point you should note.
Almost ALL my serious listening was done playing PCM material with the DAC in NOS mode!
Outside of DSD-listening, which I’ll come back to, NOS replay is the only mode I would personally use this DAC in. It’s good in over-sampling mode, but other DACs simply do that better/more enjoyably for me, and even driving it with HQ Player I found no combination of settings that I enjoyed as much as I did the Spring’s native NOS replay mode.
But in NOS mode, well, this is one of the most enjoyable DACs I’ve ever had the pleasure to listen to.
It’s worth noting, here, that oversampling on this unit is handled by an AK4137 sample-rate-converter (SRC). This is a good quality SRC, but is no match for implementations like that in Yggdrasil that use custom software, DSP processing and Mike Moffat’s “megacomboburrito” time and phase optimized filter.
Bear in mind that when NOS mode is selected the AK4137 is completely out of the audio path.
Thus, unless stated otherwise, all comparisons are using NOS mode:
I found the Spring DAC to slightly out-resolve Yggdrasil. This was more noticeable using RCA outputs, but is still apparent using the balanced outputs. Unlike the Vega, which has a tendency to exaggerate detail (which superficially is easy to misinterpret as better actual resolution), I did not get that sense from the Holo Audio unit.
Another thing I consider when I consider resolution is whether I’m really just hearing brightness, but that isn’t the case here. I hear micro-detail and texturing with the Spring DAC that is modestly better delineated than Yggdrasil’s delivery. I tend to hear this most commonly with brushwork over cymbals, and sometimes drum-skins.
With Yggdrasil I’m used to having the sense of every individual wire on the brush being more or less discernible as they impact/traverse the surface of the cymbal or drum. On the Spring DAC this was augmented by a perception that I was also hearing wires skipping up and breaking contact with the instrument and then remaking contact.
Now, if you put the Spring DAC into oversampling mode that perception goes away and I find Yggdrasil moves to the front-runner position on resolution/detail. In fact, I would say that, the advantage Yggdrasil exhibits against the Spring in OS mode is bigger than the Spring has over the Yggdrasil with the Spring in NOS mode.
But, nevertheless, in NOS mode the Spring was just out-resolving Yggdrasil.
Attack, Decay, Transients & Dynamics:
Backgrounds are unearthly quiet with both units. Notes emerge from an inky-black void, and disappear into nothingness with equal aplomb.
The leading edge of notes seem a tad cleaner with the Spring. I’ll emphasize a “tad” there (and for you non-English types, a “tad” is best thought of as “really not much at all”).
The plucking/picking, of strings in “Along this Road: Kono Michi (Yv)” (Ottmar Leibert, “One Guitar”) provides an excellent demonstration of the impressive attack, and instant emergence from the, entirely-black background, of individual notes, that the Spring DAC is capable of.
Percussion is similarly portrayed, with attack and transient response of the impact of a stick on a snare seeming more visceral than via Yggdrasil. This tends to emphasize one’s perception of the raw dynamics of the unit as well.
Decay was natural, and while it took listening to pieces with predominantly electronic instruments (that can exhibit literally instant decay), the end of notes in that case also were slightly cleaner than with Yggdrasil.
Dynamics are excellent and, with speakers, highly visceral. Shock-you-out-of-your chair level stuff with big dynamic bursts out of silent passages. The super-fast attack of the Spring DAC does, as I mentioned, emphasize that a tiny bit, and I find that a bit addictive.
Once again, and this will be a bit of a theme here, turn on oversampling in the Spring DAC and the Yggdrasil switches places and delivers the better performance. As it happens I did not find a single instance in which I liked the Spring DAC better than Yggdrasil if the Spring DAC was using oversampling … which is a long-winded, and repetitive, way of saying “you buy the Spring DAC for its performance as a NOS DAC first – anything else is a bonus/convenience”).
While resolution was the first thing to strike me about Holo Audio’s product, the second, was the soundstage it projected and how it imaged.
Even with headphones, the soundstage projected from the Spring is noticeably wider and deeper (with Abyss and HD800S, since depth is less apparent with any of these DACs on the LCD-4 and, to a certain extent, the Utopia) than I’m used to with Yggdrasil. The image within is very stable. It’s not at all hard to pick out individual instruments spatially.
As with Yggdrasil, the Spring DAC imaged well enough that I can easily hear where, spatially (left to right) where a note is struck on a piano and it’s not at all hard to follow scales and progressions as they run up and down the keyboard. Chords are clearly left-biased and melody right.
Now, I never leave my soundstage/imaging comparisons to just headphones as they’re simply not as capable of exploring that the way a properly set up speaker system is …
And here it becomes apparent to me that Yggdrasil is projecting a more realistic soundstage. I felt the Spring DAC was presenting things as being deeper and wider than they actually are. I’ve sat at the keyboard of my piano enough to know how wide it is supposed to sound and there, and in recordings of that (whether from my playing or visiting concert-level pianists), and Yggdrasil just completely nails that. In contrast the Spring throws the image wider than I find entirely natural or realistic – which is fun, and maybe even desirable with some headphones, but I would go for Yggdrasil’s rendering here every time.
Yggdrasil also manages to convey height more convincingly. This is really hard to realize properly even for a well setup system with appropriate source material and a cooperative room. There is a sense of height to the Spring’s portrayal, but it’s behind Yggdrasil with the Spring’s image being both much less convincing and less stable.
Both of these units will let you hear who moved their chair, which violin flubbed a note, or even where on a Timpani the mallet actually (from spatial cues, if the recording has them, as well as the change in tone), just they’ll be spread out more on the Spring.
In oversampling mode I found Spring’s soundstage to compress a fair bit. This reins in the exaggerated (for me) width and height (winds up narrower and shallower than the Schiit DAC, but it also hampers the precision and stability of the projected image. That’s another way of “just run this thing in NOS mode”!
Tonality & Timbre:
I hear Yggdrasil as being essentially neutral and realistic. I don’t hear coloration here. Whether this is down to me being so familiar with it, or whether it’s just really that neutral is hard to say. Though as one of my principal references, the piano, sounds exactly how I expect it to via Yggdrasil, I’m inclined to believe it’s simply uncolored.
While I wouldn’t go as far as saying that I hear any specific coloration in the rendering from the Spring DAC, it has a definite “sweetness” to its sound. I’ve heard this with a lot of NOS DACs, and while I’ve described it differently (I used the term “glow” in regards to the Metrum Pavane, for example),
I find that “sweetness” with the Spring to be quite delightful. It’s not quite a euphonic thing, but it’s very engaging and enjoyable, seductive even. It doesn’t intrude on balance or detail, it doesn’t even feel like it’s straying from neutral, but still it sounds, well, “sweet”.
I don’t know what causes this; BUT, whatever it is, I find I like it a great deal.
It’s worth noting that using OS mode on the Spring essentially kills this effect for me.
That said, Yggdrasil may just be tonally purer, and still provides the best overall rendition of piano that I’ve heard in the digital world. In fact, Yggdrasil is, I think, the best reproduction of recorded piano I’ve ever heard period, as with sustained notes I can usually discern pitch variation even from the best of the turntable systems I’ve had the pleasure to run into.
Getting back on point, though …
Instruments sound natural and are just as easy to identify with the Spring DAC as they are with Yggdrasil. Even when there’s a lot going on, you’re not going to confuse violins with viola, or a glockenspiel with a xylophone (might be a stretch that you’d confused those last two anyway, but I’ve certainly seen that happen).
Discordant piano or brass keeps its edge and bite on Yggdrasil without feeling strident or uncomfortable, but in NOS mode on the Spring sometimes felt a little too edgy. Not much. Not problematically so; but still a bit raw. In OS mode, the stridency went away but so did a little of the bite. Not a big deal, but Yggdrasil handles this better in either case.
So … this is more of a signature preference and mood, music (and maybe my blood-alcohol level) have as much to do with which signature I prefer at any given time. Both are wonderful, and while I give the nod to Yggdrasil on tonal accuracy overall, that’s not to say I necessarily prefer it.
Presentation is very even handed, even with the apparent “sweetness” of the Spring DAC and there is no apparent emphasis or recession anywhere I could hear.
Bass is solid and properly extended with good slam, articulation and weight and provides a firm foundation for the rest of the music. Yggdrasil may have a wee bit more presence in the lower registers, but I’m by no means sure that’s not expectation bias, since I didn’t hear it that way consistently.
One might say the Spring DAC is a bit moister than Yggdrasil, but I wouldn’t describe it as wet.
Again, sweetness excepted, I’d put the mid-range performance at essentially the same level, excepting when there is a LOT going on (busy, full-blown orchestral pieces), in which case there seems to be an occasional hint of … I can’t find a suitable word … “something” the puts a very feint mist over things in the lower mid-range.
I suspect that is down to some IMD product or other NOS artifact, as it is one of those things that goes away in OS mode. However, it’s a very small effect when it is present, and if you’re not listening in “audition” mode, I’d be very surprised if it was apparent at all.
One thing that was more apparent, but that is also fixed by running in OS mode, is a very slight tendency towards sibilance with some recordings. The Spring DAC is not adding/causing it (I don’t think, or see how, really), but it’s more apparent with the Spring than it is with Schiit’s unit. Turn on OS mode and this goes a bit too far the other way.
Interfaces and Inputs:
For those that worry about such things I found that there were small but audible differences, between the Spring DAC’s various inputs. I would rank these as follows:
I2S > AES >= COAX >= TOSLINK > USB
The I2S input was fed via a Singxer SU-1. I did not find any benefit in putting the JCAT Isolator ahead of the SU-1.
The AES input was driven by my Aries and RedNet 3. I could not reliably tell these apart via the Spring DAC, even with a fully-synchronized blind test. The Aries is less hassle, the RedNet 3 is cheaper, both have their own quirks, and I’ll talk more specifically about those in a dedicated thread on AOIP/NAI devices.
The USB input on the Spring DAC did benefit from the JCAT Isolator, even when fed out of the microRendu, which surprised me. Coming off the Aries’ USB output I found no difference with, or without, the JCAT Isolator in the chain.
Note: If you want to run native-DSD (i.e. not via DoP), or run multi-rate DSD, then you’ll need to utilize either the I2S or USB interfaces. And in that case I’d go with the I2S interface, even though it requires something like the SU-1 (at $400).
I’m on record as not having much interest in DSD. The catalog is tiny. The files are huge. I’ve never heard a convincing demonstration of its alleged sonic superiority. Any time DSD has come out ahead in my listening it has been down to the masters (the most common case) or some artifact of the way the DAC in question processes, e.g. the PS Audio Direct Stream units, which convert everything to DSD before it reaches their D to A stage.
The Spring DAC renders simply the best/most enjoyable DSD playback I’ve heard so far.
If you are a DSD proponent, have a lot (relatively speaking) of DSD content, or want a DAC that has great performance with DSD as a format/content-hedge, then you should spend the time and energy to listen to this thing.
And it’s worth nothing that, via I2S or USB, Holo Audio’s offering doesn’t have the dropout issues that I ran into, and was unable to resolve, with Chord’s DACs (more of an issue with macOS, but they still occurred under Windows).
Caveats about audio-memory and note-based comparisons apply here (albeit that is very recent memory, and lots of time, with the Vega and Direct Stream Junior), but this is the first time any DAC’s performance with DSD has prompted me to go buy a bunch of DSD content due to how much I was enjoying it.
And I still wouldn’t put DSD ahead of PCM in terms of actual/general realized quality, but for the first time I’ve found a DAC that plays DSD content at a level where I wouldn’t hesitate to buy either DSD-only/native content, or to more readily explore DSD versions of content that exists in both DSD and PCM formats (and, in fact, I did just that, roughly tripling the size of my DSD collection so far).
Other Thoughts:Going beyond how the thing sounds and works …
Build/Fit & Finish:
Build is very solid and the Spring DAC is a lot weightier than I would have figured. The chassis is nicely finished, the copper components look quite classy against the black. The large display is dimmable and defeatable.
It looks and feels the part, without devolving into making one wonder if the casework consumed the bulk of the parts budget.
Spring DAC “Levels”:
Kitsune HiFi/Holo Audio offer the Spring DAC in three variations. I was listening to the “top of the line” model, the “Kitsune Tuned” Level 3. This adds several tweaks, mentioned earlier in these comments.
I have no idea how this unit compares to the Level 1 and 2 units, as I’ve not heard them. This would be the next most interesting thing I would like to explore, since for many if they cannot tell the difference, then there’s a $700 price difference between the Level 1 and the Level 3.
However, for me personally that price differential isn’t significant enough for me to expend the time to actually look at the less expensive models and, I’d simply opt for the Level 3.
In Summary:The bottom line here is that I enjoyed the Spring DAC enough to want one in my system. That means I’m buying one (ordered 11/07/16, expected delivery late November). It’s the most musical and enjoyable NOS DAC that I’ve heard so far, and doesn’t give up anything on the technical side.
While I would need to run it back to back to be 100% sure, right now I have to say that I’d personally rather have the Spring DAC than the Metrum Pavane (which, up to this point, had been my favorite NOS R2R DAC). That this is half the price and/or that I could have both the Holo Audio unit and an Yggdrasil for the same money, makes that a relatively easy decision for me.
With the same caveat, it easily bests my interest in the DAVE, Direct Stream Junior and pretty much every other DAC I’ve auditioned this year.
I would not sell my Yggdrasil to get the Spring DAC. But I wouldn’t do the opposite either.
If I didn’t have an Yggdrasil already it would be a very hard decision as to which way I would go. There is no question that Holo Audio are offering are more versatile unit with its ability to operate in both NOS and OS modes, as well as delivering first-rate DSD replay. Its oversampled replay doesn’t quite reach the level that Yggdrasil does for me, but then that’s all Yggdrasil does.
For me, the perfect solution right now is to keep Yggdrasil and add a Spring DAC. This gives me the best of all-worlds.
If I didn’t have Yggdrasil already, and was only in the market for one DAC then the Spring DAC would be a worthy alternative and is something I consider to be properly competitive with Yggdrasil.
And if I wanted a fully competitive DAC with DSD support, then this would be an easy choice.
Very impressive, entirely enjoyable and well worth auditioning.
And since much of what might drive me one way or the other, if pushed to only have a single DAC, is based on signature preference, you’ll probably need to audition these two units to make an appropriate choice.
And, at this point, it probably ends my “Life after Yggdrasil?” search for now: Schiit Yggdrasil + Holo Audio Spring covers me quite nicely.
It would also be nice to know what sort of files gave you the best performance in NOS mode. I'd suspect HiDef was awesome as the post filtering is easy but did those differences hold up for 16/44 where the Schitt should have a theoretical advantage with it's 'symmetrical' upsampling. Something Naim started years ago in a more sophisticated way. You should try their top dac with a PS at some point. It's a bit costly that way but it's a clear upgrade even with their cheapest outboard supply. They a use true 24 bit ladder DAC unlike Schiit's 20 bit ones along with a sophisticated up/oversampling algorithm and a bunch of clocks.
I would think that with 24 bit material, the Holo in NOS mode would have a clear advantage and with 16 material, the Schiit might, simply due to applicable filtering. Doesn't always work out that way but I'm curious what you found in these more specific instances.
Thanks for the review.