AudioQuest Dragonfly Cobalt USB DAC/AMP - Official Thread

AudioQuest just released a new member of their “Dragonfly” family of thumb-drive sized USB DAC/amps, the “Cobalt”:

Upgrades/changes over the “Red” version include:

  • ESS 9038Q2M DAC chip
  • Minimum phase, slow roll-off, filter
  • 33% faster microprocessor that consumes less power
  • Improved power supply filtering
  • Includes a DragonTail cable (USB-A male to USB-C)

Looks like it has the same power output as the Red.

Further details can be found here, and this thread is the place to discuss the new little guy.


Oooh this is cute. I have been looking for something with a little headroom but still super portable for IEM’s with my Macbook Pro. MP and gentle roll of in a well designed filter is my general preference so I’m going to have to give this a listen.


After a recent road trip to Colorado, my lack of a great super portable solution showed. In a perfect world, that would probably mean a good DAP, but I’m not spending that kind of money. I have a GO2A Infinity, but that’s too much for the situation (high power draw and hot make it less than ideal from a phone).

The Dragonfly Cobalt might do the trick. It’s a bit pricy at $300, but if the performance is superior to its predecessors it might very well be worth it.

If I have one gripe before getting my hands on it, it’s that they’re touting the slow rolloff filter but don’t let you change filters to your preference. It’s likely that it’s the filter I would prefer given my experience with the LH Labs unit, but I like options.


The default, unchangeable (not that they could easily make it changeable here) choice of filter was what stopped me ordering one “just because”. I’ll see how it is when I get my ears on it, but that was enough to stop my normal automatic “buy it just to try it”.

1 Like

Really interested in this. Dragonfly Red has been pretty well regarded, and with its upgrades, this should sound better.

I would be interested to know if this little guy has the oomph to drive my Mr Speakers Aeon Flow Closed headphones. At 2.1V out, I imagine it would be perfectly able to do so. If I’m correct, this would make a great companion at the office.

It’s going to depend on what you listen to and how loud you listen. The AFC were more power hungry than I’d expected. The bass was woolly and they were a rather flat sounding out of the DFR, which has the same power output capability as the DFC. Switching to something rather more powerful made a significant difference.

Remember, impedance and voltage aren’t the only things you care about in determining whether something has the grunt to drive a given headphone well. You also need to know the sensitivity of the headphone and the total power output capability of the amp.

It’s also worth noting that sensitivity is often measured at 1 kHz, which is much less demanding, power-wise, than bass and sub-bass frequencies, so that needs to be accounted for.


Thank you for sharing your experience with the AFC + DFR pairing. I do have the Gilmore Lite MK2 amp paired with my AFC’s currently. Did you try the DFR as a stand-alone DAC? If so, how did it perform in that capacity?

I did.

It sounds fundamentally the same as it does in DAC/amp form. But you need to run it at, essentially, it’s maximum volume when using it to drive another amp, and that’s not a configuration I am particularly fond of - since you’re amping an amp, that’s operating in at its limits … noise and all. There’s a specific discussion of the issues regarding amping-amps starting here.

Ages ago I did a comparison, and extensive review, of a bunch of portable DAC/amps, both as DAC/amps and also just as DACs. You should be able to find what you need there, but bear in mind its spread over several posts, so don’t just stop on the first one.


I don’t quite follow. For simplicity, let’s take a headphone with a flat impedance like a planar (e.g. the Aeon Flow). Let’s say that sensitive is given in Volts and that 1V at 1KHz produces 100dB SPL. Assuming that the frequency response in the bass is not too whacky, we would expect the SPL to be relatively close to the SPL at 1 KHz for the same signal, which in this case is 1V. Since the impedance is flat, the power requirement at 100 Hz should be basically the same as at 1KHz, right?

Those assumptions only hold if sensitivity is frequency invariant.

It isn’t.

It varies … and more than you think.

Try it … with measurements. Voltage may remain consistent on a flat impedance driver, but you’ll watch the current draw rise as the frequency drops for the same SPL.

How do you get the same SPL out of a smaller driver at low frequencies (think small sub-woofers vs. large ones)? Increase the driver excursion!

How do you make a driver move further? Feed it more power.

1 Like

I love experiments! I just tried this.

  1. Feed my D10 / Atom a digital sine wave at 1000Hz
  2. Set the volume to a level where I can comfortably hear it
  3. Insert an RMS digital multimeter in series to the left channel signal
  4. Measure 1.15 mA AC current
  5. Change the sine wave to 60 Hz without adjusting digital volume nor touching knob on Atom (harder to hear now since our hearing isn’t very sensitive down there)
  6. Measure 1.13 mA AC current

Two very different frequencies, basically the same current. The slightly lower draw at 60 Hz makes sense since the LCD2C rolls off slightly at low frequencies.

This result is what I would have expected since the signal amplitude is just a voltage, that voltage is the same at both frequencies, and the driver’s impedance is flat, so with the same voltage and impedance, per Ohm’s law I would expect the current to be the same as well.

How do you make a driver move further? Feed it more power.

While the driver has to move further, it has to do so less frequently, so the power requirement is consistent.


Key point here … for the same SPL … not just the same volume setting on the dial.

Also, excursion isn’t necessarily linear relative to power.

Higher the SPL and lower the frequency the more you’ll see a difference.

1 Like

Frequency response of LCD2C is near flat in the bass, so at a given volume setting SPL at 60 Hz and 1000Hz are nearly the same.


Near flat isn’t the same as flat.

Try lower frequencies, the effect is more pronounced. There should be a shift in distortion product too

Enclosure also has an effect (easier to isolate with a driver operating in free space).

That planar drivers exhibit smaller deltas in frequency related impedance doesn’t mean they have none. Nor are all drivers the same.

The AFC is still lame out of the DFR … most notably in that the bass goes to hell, and the only reason I can think of that it should be so is insufficient power.


Signal processing technology aside, what would be a good pairing with the DFC?

The Cobalt looks very interesting. It might be what I’ve been looking for to take a step up in my listening experience.

I’m using my DFB to drive my HD58x. If I upgrade to the DFC, and want sound isolation, where should I be looking?

1 Like

Curiosity got the better of me, so I picked up a DragonFly Cobalt today (pictures etc. tomorrow). I did a few minutes listening with it, just to make sure it was working properly, and then ran it with the two headphones I’ve been fiddling with this last week (I’ll do more later).

  • MySphere 3.2 needed almost everything it had to get to proper volume levels, and the top-end is noticeably rolled off when you do get there (far more so than the FR plot would suggest, and to the point of having little life, sparkle or air), and the bass was started to come apart on bass-centric tracks (it was fine on the couple of rock and pop tracks I tried). The math says it shouldn’t be clipping, but the bottom end sure sounds pretty ugly.

  • Rosson Audio RAD-0 had ample volume headroom and the bass remained full, tight and clean up as loud as I was willing to push it (about -13 dB, or two thirds of the available adjustment). Some high-end roll-off was apparent but here it was more inline with what I would have expected a) from an ESS 9038 in minimum-phase/slow-roll off mode and b) how it ultimately measured.

Then I fired up the analyzer to see what the frequency response plot actually looks like for the Cobalt itself, and in comparison to the Red and Black models:

Colors match those of the respective DragonFlies. Test signal was 24-bit/44.1 kHz. The Cobalt is -12.5 dB down by the time it crosses 22,050 Hz.

Initial, subjective, listening impressions tomorrow …

1 Like

Interesting. I have one on order also. I expect to compare to my DFB. Your comments on inadequate power of the DFB to drive the Sennheiser HD6xx and listening to those phones with the Lyr3 made me look for a very portable solution.

Although now I listen frequently to the Grado RS1e’s, perhaps the DFC will pair well with them. Why do you think AudioQuest rolls the extreme highs? Or is that a chip characteristic?


The treble roll-off in the Cobalt is a conscious choice made by AudioQuest as it is a function of the filter option they selected to use with the ESS 9038.

The ESS 9038, like most modern DAC chips (including the 9010 and 9016 in the Black and Red models respectively), has several selectable (or configurable) internal filter implementations. Some DACs let you switch between them yourself, some don’t. Absent an app, or a button, there’s no interface to allow this on the Cobalt, so you’re stuck with what the designer selected.

The profile for the Black and Red are more typical, being fast roll-off filters and differ mostly due to them being different ICs with different output stages, but they’re properly extended for Redbook purposes.

Now why did AudioQuest choose the slow roll-off filter?

That’d be speculation on my part.

It could be they wanted it audibly different from the Red and the Black. This certainly achieves that.

Perhaps they have data that shows the DragonFly units are most commonly paired (by a disproportionate factor) with headphones that have treble issues or are otherwise considered bright, and they wanted to take the edge off there.

Maybe it’s an attempt to appease the masses … whereby younger listeners seem to tend towards wanting a less treble-happy presentation and older listeners wouldn’t necessarily be able to hear the difference anyway.

The upper-treble on the Black is a bit uncouth and uneven at times, and can sound strained and grainy and is frequently described as bright (not uncommon when there are treble issues). The Red is smoother and cleaner (which may be down to the different output stage as much, or more, than the different DAC IC used vs. the Black). Though it, too, tends to get described as being a tad bright more than anything else.

So my personal thought is that AudioQuest wanted to get away from that. And in that regard I think it’ll be the general perception that they’ve been successful.


Yes, last night I was reading the Audioquest site carefully, and I did see that it was their deliberate choice. Your point about brighter headphones is well taken. The phones that I just stick in my travel bag without any thought are the Grado SR-60e, that are bright (better with the L-cushion tape mod). And if not that, then one of several pairs of IEMs that tend to be a bit bright. And at 65, my hearing while still good, does not let me locate objects in the dark.

While on the AQ site, I sent them an email inquiring about the “line-out” mode. I saw several references on the usual suspect boards (SBAF, Head-fi) about the DFR line out mode, most of which showed a profound lack of understanding. One person said that AQ said it is not true line-out, which might be supported by another comment that it sounded better backed off a bit (AQ says to put volume to MAX on both the OS (if applicable) and the volume for line-out). While the OS to USB would just be digital, this comment would be consistent with just running the internal amp to the rails and getting some clipping.

I will post any reply I get from AQ on this here.

1 Like