Audioquest Dragonfly Cobalt

I received the Cobalt as part of the TTVJ tour and kept it for a week to listen and then forwarded it on to the next tour participant. I have no financial interest in Audioquest or TTVJ and received no other incentives to compose this review. I'd like to thank Todd for providing me with the opportunity to listen to gear I might not have been able to otherwise. Todd the Vinyl Junkie carries a ton of cool stuff, so if you find yourself out shopping for things audio, have a look at his site.



Packaging:

The box the Cobalt came in had a large Sticker over most of the front but my understanding from others I have seen is the only thing covered by the sticker is a large dragonfly graphic. All the pertinent information is still visible in the photos below. Inside the box is a plastic tray with the manual on top and the case, dac, and USB-A to USB-C adapter. It is not what one would call upscale packaging or a large kit, but the case and cable are of good quality and give the impression they should last well, and the DAC itself (while tiny) appears well constructed.



Build/Internals:

Those who have had a chance to play with any model Dragonfly, will recognize the Cobalt immediately. It is about the size of a pack of Wrigley's Spearmint with a large dragonfly that changes colors to indicate bit-rate of the file being played. The Cobalt is slightly narrower than either of its predecessors, but otherwise the same size and roughly the same weight. At one end is the USB-A port, while at the other a single 3.5mm TRS connector is fitted. A provided cable allows for connection to usb-C but is not an OTG cable so did not allow for direct connection to my Moto Z3.

The big draw for the cobalt is the improved DAC although power supply filtering improvements do contribute to improving the output from the amplifier section as well. The DAC used in the Cobalt is the ESS ES9038Q2M, a 2 channel mobile version of the 9038. For sake of reference, this is the same chip used in the Burson Swing and the Khadas Tone Board. This is not to be confused with the 9038pro as that chip is 8 channel or the ES9038K2M which includes a 2 Volt amp on the same silicone to minimize footprint for applications like tablets or phones. The ES9038Q2M offers several filters but the Cobalt is set to the minimum-phase slow roll-off by default which Audioquest claims gives the Cobalt a bit more natural sound than the previous generations that both utilized a fast roll-off filter instead.

By using the version that does not have internal amplification, Audioquest has the option of utilizing its own in-house design. That is exactly what Audioquest chose to do, and as such the amplifier section of the Cobalt is largely unchanged from the previous Dragonfly Red. Both utilize the ESS 9061 amplifier although as mentioned previously, the cobalt cleans up input power a bit better than the Red, and also uses a lower powered main processor (Pic32MX274 (Cobalt) vs Pcic32MX (Red)) that frees up more of the input power to be used to run the DAC and Amp sections of the Cobalt.

One detail worth noting is that some play exists between the USB and jack and the shell. I can physically rock the jack by placing an finger on it and watch the usb move so the board is all that supports the jack and USB. I am not sure if this is by design or the result of a tour unit that has taken more abuse than it should before it arrived (it does not look abused) so production models may have tighter tolerances.




Operation:

Operation is straight forward as the DAC was detected by Windows without having to install drivers or fuss with it. Once Setup as the output in Foobar, I was able to throw a variety of tracks at the Cobalt and watch the little dragonfly on top change colors as I went. MQA is lavender, other than that, the dragonfly identifies bit rate rather than file type with 44.1kHz being green, 48kHz blue, 88.2kHz yellow, and 96kHz a lighter blue color. The Cobalt has a 64 position volume control that gives it good range but for me it lacks a bit of volume on some full sized cans. I found it was best reserved for use as the DAC to an external amplifier in those situations.

When paired to an android phone, the dragonfly worked equally well with USB Audio Player Pro providing the source material. With the Camera connection kit, I was also able to get the Cobalt going on an Iphone 8 without any extra hassle. I did see a bit of battery drain associated with using the cobalt paired to mobile devices, but overall it was less than several other devices I have tried. (more on that in comparisons).




Sound:

I have never been a fan of the dragonfly models, and my complaint has always been that they lacked control and sounded a bit loose. I am happy to say that the Cobalt has gone a long way to correct this as bass is impactful without becoming boomy and treble is definitely tighter and more articulate than the Red. It is moves in the correct direction. Overall, I would describe the Cobalt as slightly laidback and musical but with reasonably good control and detail. Dynamics are better than the Red as well. Lower treble is slightly forward and may seem overly bright if your source material already leans that way or you are particularly sensitive. If there is a knock in the sound department, it is as the output demand goes up it begins to lose a bit of low end as it just cant provide the needed current to really drive hard to drive headphones. I would avoid things like 600Ω Beyers or the big inefficient planars like the He-560 as it isnt particularly successful with either of them. Best pairings are low impedance / High sensitivity earphones or headphones. At times I struggled to get the amount of volume I would have liked with some cans (Cascade/HD700).



Limits:

Sound of the cobalt is dependent on what is providing the source material. I found connecting to an android tablet would only show green (48kHz) regardless of what it was fed. Turns out this is a limit of android and when paired with UAPP it works correctly. The problem is in android and not the Cobalt. With the UAPP drivers, it handles input appropriately and the color changes when different bit rates are presented. I also found noise was slightly lower when I used the same content through the UAPP driver vs the built in android systems.

The 24/96 limit of the cobalt has nothing to do with the DAC, instead it has to do with universal support. The trade off comes from the fact that Audio type 1 (required for driverless operation) only supports up to 24/96. Had the Cobalt been configured as a USB-type 2 device it could have gone up to 32/384 (limit of the dac) but with the added hassle of specialty drivers being required to do so. Audioquest clearly chose ease of use over absolute fidelity.



Comparisons:

I tried to pick a few other daps that have something in common with the Cobalt so 3 of these are competitors in the portable world and the 4th is a desktop board that uses the same DAC chip albeit lacking the amplifier stage of the Cobalt.

Khadas Tone Board - The Tone board has been quite popular as it combines the ES9038Q2M DAC with a USB input and RCA outputs at a roughly $100 US Price point. In use, I find the Tone board to have good detail and a broad range of supported formats including DSD up to 256k and MQA as well as the standard Flac, Alac, mp3 etc. The tone board makes no attempt to amplify the signal for use as a headphone amp so cannot be used as a stand alone like the Cobalt can, but supports higher resolution files 24/384 vs 24/96 and offers better DSD decoding. Ultimately, I think the tone board is a bit cleaner and tighter than the Cobalt but this may well be impacted by the amp attached to the tone board as well. If you have to have portable use, the Cobalt is a better option, if for desktop use feeding another amp, I'd opt for the KTB.

Zorloo ZuperDAC - in comparing the Zuperdac vs the Cobalt, we have two devices that both share the same form factor and same amplifier chip but two different generations of DAC. The Zuper uses the 9018 series chip for its DAC. Perhaps oddly the Cobalt is limited to 24/96 while the older device handles 24/192. The Cobalt fires back with MQA and DSD support which are absent on the Zuper. Sound is better controlled and a bit tighter on the Cobalt while the Zuper seems to lose a bit of control at both ends. This would be an easy win for the Cobalt except for that nagging lack of support for 24/192.

Apogee Groove - Here we have a dogfight. The Groove uses a full 8 channel version of the ESS series DAC and sums 4 channels to create its output. This lowers noise and THD and allows the Groove to post numbers not unlike the Cobalt even though it uses a much older chip. Apogee doesn't disclose other than to say it uses and ESS with 4 channels summed. Best guess is the chip is either a 9008 or 9016s as those are the most cost effective possibilities. The groove supports 24/192 but not DSD or MQA so format support isnt going to separate the two unless you have to have DSD or MQA hardware unfolding. Upsides for the Groove are its micro-usb female connector that allows a broad range of cabling options for connections to PC or tablet and its built in volume control keys. Upsides for the Cobalt are its lower power draw and smaller form factor. Those using android or apple phones will prefer the Cobalt as while its output is not as potent as that of the Groove, it also doesn't have the demand on the battery that makes the groove all but unusable from a phone.

Centrance DACPort-HD -This is another tough fight for the Cobalt. The Dacport is larger and uses a female port like the Groove, but has a form factor closer to that of the Cobalt. The Dacport-HD utilizes the AK4490 chip at its core and while it does not support MQA, it does support 32/384 flac and DSD64 and 128. The DACport-HD feels more solid than the Cobalt but is nearly double its size and weight so again those using it to pair with a portable device will find the Cobalt more convenient. Centrance does not market the DACport-HD for use with android or IOS devices and while I was able to get android to recognize it using USB Audio player pro, it drained the battery quite quickly. Power is no contest as the DACport can easily drive my 600Ω Beyerdynamics while the Cobalt begins to struggle considerably before reaching anywhere near 600Ω.



Conclusions:
The Cobalt is certainly the best of the line and an improvement over previous generations in both sound and power, but still suffers from some of the same limits as its predecessors. Trade-offs made for ease of use and portability prevent the high-end chip used from accomplishing its full mission. For those where portability reigns king, the Cobalt will find an easy niche, for the rest of us, those same trade-offs may make it look less than attractive.

DAC Chip: ESS Sabre ES9038Q2M & 9601 Amp
PCM 24bit – 96khz MQA, SNR: Undisclosed, Power: 2.1 Vrms, USB Type-A Male, 3.5mm SE (Microphone: Unspecified), Aluminum Chassis, DragonTail USB-C Adapter

PROS​

• Organic neutral timbre, natural analogue tonality
• Vibrant and engaging dynamics
• Crisp Treble with great extensions
• Mids with natural placement, warm tonality
• Impactful Mid-Bass response with natural texture, slow decays
• Audible Sub-Bass presence, slow decays
• Realistic (warm) guitar and stringed instruments tone
• Commendable balance between technical and musical ability
• Spacious soundstage with good imaging
• Excellent details retrieval, great transparency
• Remained cool even after used for long hours
• Does not drain host battery too fast
• Seamless MQA integration
• High resistance to phone RF signals
• Excellent synergy with fast DD and BA

CONS​

• Slow bass performance, slow attack, performs better with already fast IEMs
• Dismal speed handling when driving demanding IEMs/Headphones
• Does not work well with demanding Magnetic Planars, very unstable timbre coherence
• Grainy edged Treble that gets worse when volume cranked higher
• Audible Bass distortion when subjected to heavy load (driving over 60/100)
• Sporadically warm nasal tonality for Mids on some songs when paired with warm IEMs/HPs
• Native USB-A instead of C makes it larger and heavier than most Thumb drives

VERDICT​

AudioQuest DragonFly Cobalt. I paid dearly for my Cobalt. $350 in total including shipping from Amazon. Expectations were high and decidedly I was very critical with this one. Scrutinizing everything to ascertain if the price demanded justify the quality and experience offered.

Tuning focus for Cobalt is “Natural Sound” with the implementation of Minimum-phase Slow Roll-off Filter on the ES9038Q2M DAC. This Cobalt being my 6th DAC/Amp with ES9038Q2M so I have some good idea of what to expect. The actual tuning separates them from one to another. Cobalt clearly is the most warm sounding among them all with Hidizs S9 Pro being the most neutral and Topping D10s the most dry. E1DA 9038D is very well balanced and then followed by Shanling UA2. Lastly the dual ES9038Q2M of Cayin N6ii (E02) being the most versatile with very succinct filters option.

Honestly it is hard for me to be excited even as I write this. My personal experience with Cobalt was underwhelming to say the least from the moment of first listen until now. What I am hearing is a performance I have heard between a $20 JCally JM20 and JM04Pro – Yes it is vibrant and engaging. It has the warmth in contrast to JM20 brightness, the analogue tonality of JM04Pro. But hell, I paid $350 to get this level of sonic indulgence?

Now let’s get to the details why I was underwhelmed, bear with me on this, but it must be understood how I came to this POV:

AUDIOQUEST DF COBALT VS SHANLING UA2​

• Headphone: FOSTEX T40RP MK3 Magnetic Planar 91db 50Ω
• Filters: Cobalt (Stock), UA2 (Apodizing Fast Roll Off)
• EQ: Always OFF
• Both on ES9038Q2M with dedicated Amp stage

SOURCE: WINDOWS 10, NATIVE WINDOWS USB
Hoff Ensemble “Polarity” FLAC 32/44.1

• 8 passes for each Dongles
• Volume: Cobalt 50/100 (SE), UA2 30/100 (BAL)
• Slightly forward intimate for Cobalt vs UA2
• Slower attack on Cobalt, UA2 slightly faster
• Nasal warm edged piano tone on Cobalt, crisper neutral on UA2
• Percussions crisper on UA2, smooth decays, Cobalt slightly pale, soft decays
• Macro and Micro details equal on both

As the crescendo peaked up between 1.18-1.26 and 4.05-4.28, Cobalt exhibited sibilance & micro distortion while UA2 remained smoothly controlled.

SOURCE: iPhone 7 USB 2.0 (DDHIFI ADAPTER)
Diana Krall “California Dreaming” Deezer FLAC

• 6 passes each Dongles
• Volume: Cobalt 80/100(SE), UA2 70/100(BAL)
• Slightly cleaner crisper, punchier bass responses on UA2, fuzzy on Cobalt
• More neutral vocal staging for Cobalt, slightly intimate for UA2
• Mildly recessed backing vocals on Cobalt, better nuanced on UA2
• Soft edged piano tone for Cobalt, crisper on UA2
• Crisper micro details on UA2, softer on Cobalt
• Coarse edged vocals decays on Cobalt, smoother on UA2
• Overall warmer presentation on Cobalt, natural on UA2

Hans Zimmer “Mombasa” Deezer FLAC

• 2 passes each Dongles
• Cobalt 80/100(SE), UA2 68/100(BAL)
• Very well controlled dynamics on UA2, jittery on Cobalt
• Bass distortion on Cobalt, tight, deep and well controlled on UA2
• Transients and speed compressed between 4.12-4.38 for Cobalt, UA2 survived cleaner

Alison Krauss “When You Say Nothing at All” Deezer FLAC

• 1 pass each Dongles
• Volume: Cobalt 80/100(SE), UA2 68/100(BAL) Matched Loudness
• Nasal jittery Mids for Vocals and Guitars on Cobalt, UA2 clean and truthfully neutral

And I stopped there, can’t bear to go on and place myself into more sonic grief.

As the tests revealed, DF Cobalt was outright dismal when subjected to driving demanding headphones. For the pedigree and price it asked, I expected stable performance across the range from the easiest to drive to the more demanding ones. And I consoled myself perhaps it will perform better with much lower load. So I plugged in my Moondrop Aria which has 122db of sensitivity versus the insane 91db of T40RP. Unfortunately the sound is average at best. Half of my Deezer HiFi songs exhibited tell tale signs of coarse edges and slow loose Bass responses. Simply put, my Aria does not like Cobalt.

Next I plugged in my VE Monk GO. Hey hey, Diana Krall voice actually sounds sexy neutral on this one. No hint of being nasal. I must admit this is the most flat neutral output I have heard from Cobalt as of yet. Impressive soundstage with spacious layering. Despite being very resolving, the Monk GO wasn’t sounding grainy edged at all. Treble was smoother than observed earlier with the other IEMs/HPs. However, there’s still some hint of peaky vocal sibilance especially on Toto’s “Africa”. Ah well can’t have it all can’t I?

Then I switched on the my Heart Mirror. Finally some decent output that’s actually enjoyable. The ever beautiful Heart Mirror always forgiving with her kind persona. The sound is now less grainy and less prone to Bass micro distortions. The issues with nasal tonality for Mids all gone. Speed and transients audibly better and less prone to compression and congestion on pacey complex passages.

And finally, I plugged in my beloved Etymotic ER4SR. With 45Ω and 98db of sensitivity, the ER4SR offered middle ground between high sensitivity to hard drivability. I am actually very pleased with what I am hearing. Honestly surprised that my normally choosy ER4SR synergize well with Cobalt. The natively matured, polished characteristics of ER4SR BA eliminated any hint of roughness Cobalt has. I believe this due to the exceedingly speedy behavior of ER4SR that blended greatly with Cobalt slow filters. Bass was tightly controlled and with enjoyable seismic responses. I swear if I were to use just my ER4SR I would even say this is a match made in heaven. There’s that welcoming analogue tonality to make ER4SR highly musical and still technical at the same time. The power of Cobalt was more than enough to make my ER4SR sing with vibrant dynamics.

At this point I decided not to test Cobalt with my Shure KSE1500. What I have heard so far discouraged me immensely. The ultra-resolving, ultra-transparent nature of KSE1500 will be as merciless as the FOSTEX T40RP MK3. To serve as a great AUX feed for Electrostatics, the basic requirement is a clean, grainy free feed. Cobalt does not qualify for that.

All in all, the fidelity level offered by Cobalt reminded me a lot of the sound I get from the Plextone Type-C. It is vibrant and engaging, it has power, it is musical and detailed at the same time. The saving grace for Cobalt was the pairing and synergy with my HZSOUND Heart Mirror and Etymotic ER4SR (both of which are emotionally special to me). But it is hard for me to be merciful to Cobalt because I literally expected it to perform great on a wide variety of options to melt my brains – that didn’t happen.

Cobalt is the latest flagship from from the Audioquest Dragonfly series of portable dongle style digital audio converter with built in amplification products. Price new is currently at $299, the unit I've gotten in for review is a temporary loan from Todd of TTVJ and I was not compensated for my thoughts.

Just to set the tone, I don't feel at this time the Cobalt is competitive given the iFi xDSD is on the market and even at the xDSD's full $399 asking price I feel it's overall just a better product given all factors.

So that said here's what I like about Cobalt

• Size
• Aesthetics
• Included Accessories

An what I wasn't a fan of was it's performance. Starting with my all stock HE 4XX I experienced the following technical drawbacks

• Slow to start - literally took about 10 seconds before my V20 and G8 recognized the device and were listening ready
• Volume often reset to MAX - Neutron was the worse but uDAPP had it happen to. After a pause and resume the volume would sometimes be at MAX output

Music wise I listened to all of the 1995 Chesky Demonstration Disc before moving into one of my usual track sets consisting of

• System of a Down - Mind Vinyl RIP from the 2018 Reissue
• The Eagles - Hotel California Hell Freeze's Over Simply Vinyl 180g RIP
• Precious Fathers - Brad Quin, Medicine Man

I'll also add I don't necessarily like bright sounding systems and headphones but have a higher tolerance for it or I'm less offended/distracted or fatigued by them. Bright it usually not as offensive to my ears as others, so for me when something is TO bright or leaves me fatigued it's indicative of what I'd considered not really "worthwhile" performance.

And yes I volume match with pink noise and these days my average listening is right around about 79 dbs, peaks of 87 an dips as low as 71.

Overall sound wise I felt the unit was

• A bit exaggerated sounding and often congested
• Kinda bright with evident top end emphasis
• Presentation and envelope were a touch aggressive
• Emphasized macro detail or "noise"
  • So breathing, fret work, foot steps the occasional air conditioner and any other incidental noise captured during the recording process

Dare I say noisy?

• Drum kits often had obvious overlap to my ears and with a lot of instruments it presentation makes it not as easy to really discern a clear position and sense of space between everything.
• Really a lot of top end spectra was just overly aggressive and it was hard to discern as much as I would have liked.

As a heads up I had some technical issues with it namely my subjective impression may have been a result of poor amplification among other things. I also had issues with my V20 maxing out it's volume into the Cobalt and again when I switched between Neutron, Tidal and Amazon HD Music and the steps were not as precise as I would have liked. I did volume match as close as possible but it could be that the device simply did not behave as it should have with my V20. It could be possible those sudden jumps to max volume left my ears more sensitive and my brain with a stronger impression of how cacophonous it sounds at MAX Output.

Where as the xDSD was both more natural on it's "listening" filter and presented less overall congestion. And for what it's worth I also feel that xDSD can be a touch bright

Still starting with the envelope I felt the xDSD presented;

• A more vivid release or reverb trail
• Tactile without excess emphasis on the attack

Dynamics were also more vivid as well so the difference between quiet and soft sounds were more discernible an there was more evident micro detail with xDSD as well, an to be clear I also feel the xDSD is kinda bright an sometimes aggressive but it's not so overdone!!!!

I will say it Cobalt was more detailed than my Shanling M2S and both had issues with congestion though the M2S presents more of an overlap of “reverb” or the body of different instruments. So it sounds kinda smoothed over, VS Cobalt which presents congestion more so as an overlap between the initial attack of sounds.

Frankly I’d rather experience congestion as a smoothing over rather than a cacophony of noise. An despite being technically more detailed I again wasn’t a fan of Cobalt at all. I also didn’t really feel compelled to listen to Cobalt with any of my other headphones as I’d have to pair it with something dark like my E-MU Purpleheart to get a nice “pairing” though again you can enjoy the Purpleheart more with other ultimately more resolving devices with a"bright" presentation.



Now I did take a moment to switch over to a different playback solution this time I used USB Android Player pro with exclusively MQA featuring the following tracks

• Beck - Guess I’m Doing Fine
• Schubert Piano Quintet - A Major D667 “The Trout”

Headphones used here where the HD 600 and a WabiSabi MapleSleeved BlackLimba Magvnum V7 Build which sounds like a more refined/detailed Grado RS2E.

As mentioned subjectively the Cobalt has a rich organic tone due in part to it’s heavy top end roll off. Overall with these dynamics I found Cobalt to have;

• Soft but full bass response
  • More quantity but with a loss of texture and impact
• Tonally rich presentation of Vocals
  • This helped mask a lot of coarseness or excessively forward breathing
  • An honestly I felt Cobalt’s tuning really reproduced a more natural presentation to vocals
• Heavy Unevenness in the mid to upper mid transition
  • Stringed instruments sounded off just simply wrong
  • Lacked detail and was overly “romantic”
  • Literally slowed some faster more complex string passages
• Fairly closed staging with what was often congested imaging
  • Lacked air and precision when presenting passages with multiple instruments and musicians within a shared space

An to make matters worse I often had the volume output maxed with my HD600 to achieve a meager 81 dB peak with more dynamic/quite classical pieces.

While it wasn’t as “noisy” it still lacked detail and presented a congested image, overall I still feel it’s not a worthwhile purchase. xDSD sounds more tonally even and presents better detail, interestingly enough Bluetooth with xDSD was tonally flatter and about as detailed as cable’d listening with Cobalt. An from time to time I often use xDSD in such a manner, tucked away in my pocket connected wireless to my cell.

In conclusion I just don’t find any value in Cobalt unless you listen with easier to drive headphones and enjoy mostly vocal heavy music. Big thanks of course to Todd for sending the unit my way!