AudioQuest DragonFly Copper: Review and Measurements. A classic dongle DAC returns. (And Rubber Soul - 2026 remix.)

Audiophiles following the news cycle know by now that after seven years, AudioQuest has released a new DragonFly USB DAC. The DragonFly Copper (US$250, CAD$345) was spoken of at High End Vienna 2026 and started shipping in September, replacing the Cobalt at the flagship. As usual with a new Dragonfly, the forums and review sites lit up within days of the announcement (see sites like this, or this). Recently, I even got E-mails from blog readers asking when I was going to review this device, so here it is by request. 😉

Over the years, I've measured these little USB DACs going back to the DragonFly v1.2 in 2014. More recently, I wrote the two-part series on the DragonFly Black in 2017 with a good chunk of that examining MQA "rendering", and then the three-way comparison of the Cobalt, Red, and v1.2 in 2019. In that last article, the Cobalt did not exactly manage glorious performance on the test bench, clipping at 100% volume with 0dBFS material, THD merely on par with the ancient v1.2, and higher idle current draw than the Red despite all the "low power" marketing. The Red, as I wrote back then, was the winner for lowest distortion among the Dragonflies.

So in my mind, the Copper arrives with something to prove at least to this "more objective" audiophile if it truly is to be a better device. In the company literature, AudioQuest claims to have made progress in 3 main areas: "twice the output power" as a headphone amp, "25% lower current draw", and the "high-conductivity copper-plated case" allows "for improved RF-noise drainage". We'll have a look.

I bought this DAC through consumer channels and this post is in no way company-sponsored.

With this new device, they're using the integrated ESS ES9218 DAC/headphone-amp which was released in 2016 and had been implemented in other mobile devices like the well-reviewed-for-sound audiophile-targeted LG V20 smartphone; LG has since exited the smartphone market by 2021.

The main feature change is that MQA rendering is now gone. Given MQA's 2023 trip into bankruptcy/administration, Lenbrook's involvement, and Tidal's subsequent move to hi-res FLAC as its preferred format, that's hardly a surprise, and frankly the better path.

The DragonFly designer Gordon Rankin has held the line at 24/96 as the top PCM data rate. They've maintained UAC1 for compatibility like the other Dragonfly DACs before which also means no native DSD support (software will have to do DSD-to-PCM conversion).

As a new 2026-year DAC, it's very much a design based on "keep it simple, stupid" principles. Here's a table comparing all the models:

At a glance: the DragonFly family

Model	Year	US launch price	DAC chip	Max resolution	Output	Notes
DragonFly v1.0	2012	$249	ESS ES9023	24-bit/96 kHz	Analog volume, 64-step	Gordon Rankin design; Streamlength async USB; dual master clocks
DragonFly v1.2	2013	$149	ESS ES9023	24-bit/96 kHz	1.9Vrms measured, Analog volume, 64-step	Refined analog stage and power supply
DragonFly Black	2016	$99	ESS ES9010 (32-bit)	24-bit/96 kHz	1.2Vrms, analog volume	Smartphone-ready via DragonTail
DragonFly Red	2016	$199	ESS ES9016   (32-bit) + ES9601 amp	24-bit/96 kHz	2.1Vrms, 64-bit bit-perfect digital volume	MQA rendering via firmware update
DragonFly Cobalt	2019	$299	ESS ES9038Q2M + ES9601 amp	24-bit/96 kHz	2.1Vrms, 64-bit bit-perfect digital volume	MQA renderer; slow roll-off minimum-phase filter
DragonFly Copper	2026	$249.95	ESS ES9218 SoC (DAC + amp)	24-bit/96 kHz	"2× power" of others, 64-bit bit-perfect digital vol	No MQA; copper-plated chassis.

Notice that compared to other companies these days (like FiiO discussed last week), AudioQuest doesn't publish specific numbers like the output voltage or power capabilities for the Copper. So let's make sure to run measurements and put some numbers out!

Here's the box. Quite typical of the AudioQuest look with product picture and some graphics on a black background.

The packaging is quite minimal. It's just a cardboard skeleton inside with the DAC+DragonTail (female USB-A to male USB-C) connected with a twist-tie holding it to the cardboard.

Other than the DAC and converter cable, there's a 1-page folded "Quick-Start Guide":

The DAC dongle itself looks good. The shiny, copper-plated zinc shell is quite attractive and eye-catching. When shaken, it does not rattle so the fit seems to be a step above what I experienced with the previous early-release Cobalt (hope this got better). Size is approximately: 2.25" long, 0.75" wide, 0.5" tall, weighing 20g. The DragonTail cable (available separately for US$30) looks and feels well-made. It's about 5.75" long and weighs 11g. The metal shell barely gets warm after hours of normal use. 

Other than the USB-A connector, the DAC only has a single-ended 3.5mm stereo headphone jack. As per previous models, there's a translucent dragonfly icon on the top that changes color based on sample rate. Unlike previous DragonFlies, this one did not come with a USB cap nor was there a little pouch/sleeve which was included in the earlier models.

The flat bottom shape looks similar to the Cobalt (I believe they're identical although I don't have a Cobalt to confirm).

This is not a complex device so the 1-page manual is fine. Perhaps the most useful bit in there for reference is the correlation of LED color and sample rate.

There's no indication on the body of the DAC where this was made. On the box it speaks to being "Assembled in the USA from Imported Parts". Probably fair to say that the guts, maybe even the shell originated in China and assembled in the USA.

So then, how does this device perform compared to previous models? Let's run it through the test bench.

I. Oscilloscope and Digital Filter

That's a great looking 1kHz waveform on the oscilloscope. We'll have a better look later at relative output levels, but already we see no visible clipping and excellent channel overlay which bodes well for channel balance. As a 0dBFS tone into a high impedance load, we see that the maximum output level is just shy of 2.1Vrms - basically equivalent to the Cobalt and Red devices.

Therefore, when the company says this has "twice the output power" of other Dragonflies, they presumably are implying that for a low-impedance headphone load, there must be more current available, hence higher potential wattage. We'll measure that later.

Impulse response, right channel intentionally inverted and overlaid.

For those who have looked at impulse responses over the years, it's obvious that the DF Copper has a minimum phase (no pre-ringing), slow roll-off filter (relatively short length). Normal polarity. Let's have a look at this in the frequency domain:

It's about the same as the DragonFly Cobalt with some imaging artifacts seeping past 22.05kHz Nyquist, well suppressed by 28kHz. In the ES9218 datasheet, this is the standard ESS "Minimum phase slow roll-off filter".

[If you want to compare what a steeper filter looks like, just have a peek at the OS curve for the FiiO K11 R2R last week.]

Because it's a minimum phase filter, there will be some phase shift:

A reminder that even though there is no pre-ringing in the impulse response, this does not mean there are no time-domain anomalies! In fact, minimum phase filters create phase shifts in the higher frequencies as seen above which we would not find with linear filters.

II. RightMark comparisons

Unfortunately the RightMark Audio Analyzer software is very old these days and hasn't been updated since 2018 but it is still useful as a multi-test battery that allows me to compare and easily overlay data (just be mindful of the occasional glitch and double check). I had hoped to be able to compare the Copper with Red and Cobalt, but unfortunately I had lost much of the measurement data from 2019 due to a back-up issue here at home; we can make due with the data published in 2019. 😕

I do have the data from the AudioQuest Dragonfly Black from 2017 though so we can see the relative improvement. And let's compare with the inexpensive Topping D10s, and last week's FiiO K11 R2R starting with the standard resolution 16/44.1 test.

Clearly, the FiiO K11 R2R and old AudioQuest Dragonfly Black have more distortion. The crosstalk is also a bit higher with the old DF Black. Otherwise, not much else to differentiate the DACs when it comes to CD-resolution measurements.

The Dragonfly Copper performs quite well as a line-level DAC; similar to the Topping D10s. While not really audible, we can see the earlier roll-off of the Copper's frequency response due to the less steep digital filter.



Now let's look at the hi-res performance of the Copper with this 24/96 RightMark summary:

Numerically the DragonFly Copper performs well as a hi-res DAC. Notice the clear improvements compared to the old Black and significantly better compared to the FiiO K11's R2R conversion. It's not unexpected to see resolution improvement using a modern sigma-delta chip DAC compared to R-2R.

The graphs again highlight the superiority of the Copper, D10s, and DX9 (XLR output for comparison) over the DragonFly Black and K11 R2R.

While I can't overlay the Dragonfly Red and Cobalt test results (due to loss of the original measurement data), here are their 24/96 numerical values along with the old Dragonfly 1.2 and a few other comparison DACs I've used over the years:

(Note the glitchy DF Cobalt multitone frequency result data from the old RightMark software. Ignore that, it's fine...)

While noise level has always been pretty good with the DragonFlies, notice the relatively high distortion with the Cobalt hence my impression that the Red was the superior, higher-fidelity device at a lower cost with similar output power and overall lower energy consumption.

The Copper has now improved upon the distortion levels of the Red. See... The company claims that this new model "sounds better", isn't it nice that measurements actually confirm that indeed, it's a higher-fidelity device? 😉

III. Distortion: THD+N, Stepped Sine, Triple-Tone

Now that we've seen some relative comparisons, let's dig deeper into the Copper's performance starting with the 1kHz 0dBFS THD+N; both channels driven into the 100kΩ E1DA Cosmos Scaler and ADC measurement "stack".

Can't complain about a THD+N of -95dB (SINAD 95dB) for a small, low-power, dongle-DAC with 3.5mm headphone jack. No 60Hz hum (copper/zinc shield working well perhaps). Note the excellent interchannel balance of only 0.02dB difference in levels.

Let's have a look now at the stepped sine result with THD vs. Output Level to make sure there are no anomalies at different volumes:

BTW: Notice there's none of the R2R ladder DNL "bumps" every 6dB like we saw last week with the FiiO K11 R2R DAC.

Overall, looks good and as we've seen with other DACs, there can be little humps like the one around -40dBFS but these are low-level. I've put the cursor at the lowest distortion and noise point which is around -10dBFS where the DAC can achieve down to THD+N of -102dB; again, great for a small single-ended dongle DAC.

How's the output linearity looking?

Excellent! It stays below 0.5dB deviation all the way down to -120dBFS.

How about the stepped sine across audible frequencies? Let's examine at 0dBFS, -8dBFS, and -20dBFS output levels:

Predictably clean and boring as it should be with ESS sigma-delta DACs. 😊

I was curious whether this ESS DAC had the undulating pattern we saw in other HyperStream II devices published previously - let's have a look at a higher resolution THD vs. Output stepped sine every 0.1dB from -25dBFS:

Nope, there's no anomaly. That's interesting. Not that I expected the undulating pattern to be audible, but it's nice to see that the ES9218 SoC as implemented here is unaffected.

Finally, let's round off distortion testing with the Triple-Tone TD+N I've used over the years:

We can see some intermodulation distortion in the FFT which is expected. Highest distortion peak around -97dB, and the average TD+N across the two channels is a very respectable -94dB.

IV. Jitter

Looks great. We can see a few symmetrical jitter sidebands in the 24-bit test result but this is insignificant (way down at -130dB). (Jitter IMO is a non-issue these days unless horrendous.)

Typical pairing of Dragonfly with an IEM. Here's a KZ planar IEM which sounds pretty good paired with the DF Copper. Decent bass and treble not harsh, even with '80s pop like Reel Life.

V. Headphone Output and Power

First let's consider the output impedance of the headphone jack. Here are the frequency responses into various dummy loads and into the Polk Ultrafit 2000 headphone with impedance dip down to 15Ω (impedance graph discussed here):

Excellent very low output impedance. Notice the Y-axis is only 1dB here. Measured output impedance only 0.2Ω. Frequency response fluctuation with the Ultrafit 2000 headphone load was only 0.05dB (as compared to 0.3dB with the FiiO K11 R2R).

Well, since AudioQuest doesn't publish power values for the headphone out, let's have a look at what it can do.

 

Already, we know the maximum voltage is 2.1Vrms, a typical modern line-level, like the Red and Cobalt devices. Let's see it under load:

Maximum non-clip output into 20Ω is 1.33Vrms. Notice that even with a moderate 75Ω load, the Dragonfly still clips above -1dBFS. No problem by 560Ω:

These results translate to a headphone amp capable of up to 88mW into 20Ω, and 47mW into 75Ω. It means that the Copper likely is capable of higher power output than the previous Cobalt and Red variants which are both based on the ESS 9601 headphone amp (I see the Cobalt was measured at 53mW into 25Ω, ES9601 spec of 40mW into 32Ω).

How about power consumption off the USB port? Well, let's use the metric I published previously - how much current did the DAC consume when producing a 0.5Vrms 1kHz signal into 30Ω?

                            Idle      0.5Vrms@1kHz     

Dragonfly v1.2              0.05A       0.13A

Dragonfly Red               0.04A       0.08A

Dragonfly Cobalt            0.06A       0.10A

Dragonfly Copper            0.04A       0.05A

SMSL iDEA                   0.11A       0.15A

That's great. Indeed, the Copper is able to produce the same amount of output at significantly lower current draw from the USB port. This device should be able to increase the duration of music playback dangling off your phone. The highest current draw I saw when pushing the DAC/amp into clipping territory with the low-impedance 20Ω load was 0.19A.

While the DragonFly Copper is by no means a high-power headphone amp, it looks like they have achieved better energy efficiency and overall a bit higher power output as well.

VI. Subjective Listening

As usual, I made sure to listen for a few hours to get a subjective impression of the DAC/amp prior to running measurements. Given the nature of this device and the intended use case - it's only got headphone output - I just listened using headphones including my Sennheiser HD800, AKG Q701, and the KZ planar IEM shown above.

Honestly guys and gals, I really don't think any flowery words are needed for what I heard. Pop music like West End Girls' "I'm Not Scared" was appropriately bass-heavy, detailed with tons of propulsive pacing. Tonality on Szabolcs Szilágyi & László Borbély's performance of Prokofiev's "Sonata for flute & piano in D major" (2022, for best version check out the DSD256) was excellent without any sense of artificiality or piercing quality especially on the flute which can be difficult to record well.

While live albums are not typically the best recordings, I enjoyed the energy on Elvis' Recorded Live On Stage in Memphis (1974, also available streaming multichannel/Atmos) performance, probably the best sounding of his live concert albums. The country-styled "Let Me Be There" and then swingin' "My Baby Left Me" are fun.

Soundstage is difficult to evaluate with headphones and IEMs (no HRTF or pinna filtering among other differences) unless I put on a binaural track like Ottmar Liebert + Luna Negra's Up Close (2008). Excellent rendering of the 3D space with the single-driver KZ 13.2mm planar IEM! While I don't follow the headphone/IEM market closely, inexpensive IEMs seem to have really come a long way with impressive value. Close your eyes and be immersed in the realistic-sounding hand claps and percussion on "La Luna".

To me, this is simply clean-sounding, modern, ESS sigma-delta chip DAC performance; highly resolving, delivers the music's dynamics, and full frequency response. I would not say there's anything particularly "warm" or special otherwise. Having said this, it's pretty clear when I A/B listen compared to my Drop+THX AAA 789 headphone amp that the DragonFly Copper's bass "slam", the rush of an orchestral crescendo at higher volume isn't quite as impressive when paired with more demanding headphones like the AKG Q701. To be expected in light of the headphone output measurements above.

Have a listen to the DAC quality for yourself, and as usual, make sure to compare with other DACs; in particular, consider ABX'ing with the AMPT of the Fiio K11 R2R from last week in OS or NOS mode with hi-res gear and good headphones/IEMs. These DACs are objectively clearly different and utilize different conversion technologies. At the very least, I think the rational audiophile will gain some perspective on the audibility of DACs if they've never tried something like this!

AudioQuest DragonFly Copper AMPT (153MB)

24/96 capture through RME ADI-2 Pro ADC

VII. Summary:

The AudioQuest DragonFly DACs have been fun and are easy to measure as they're about as straight forward a DAC as one can imagine these days with single USB input, single 3.5mm output, and UAC1 device compatibility with maximum 24/96 PCM data rate.

Following from the listening impressions and objective data, the AudioQuest DragonFly Copper (US$250) is a positive evolutionary step. IMO, it's a better device that the previous Cobalt in that it genuinely improves on the resolution of the Red, while reducing power consumption (promised but not achieved by the Cobalt that I tested). While the Copper I believe is able to provide more power for low-impedance headphones, I'm not sure about the "twice the output power" claim; maybe they were comparing with an older model like the DragonFly Black and alas I don't have a Red or Cobalt to double check. So long as maximum 2.1Vrms output into ~100Ω load, and ~88mW into a 20Ω load is enough power for your reasonably efficient headphones, it'll sound great.

[While we can compare the FiiO K11 R2R DAC and note the lower resolution from the FiiO's R-2R design, on the headphone amp side, the DragonFly Copper is a lot weaker, clipping into a 20Ω load 9.5dB below the FiiO's single-ended output with hi-gain.]

The DAC's digital filter is a standard ESS minimum phase slow roll-off setting which is fine even though like the Cobalt, there's some imaging distortion past Nyquist. Better than the useless MQA "rendering" with very poor filters at least. 🙂

I have no problem recommending the DragonFly Copper if this form factor with USB-A + USB-C adaptor is what you want, compatibility is more important than playing with higher sample rates or DSD, the headphone output power is adequate, and you view the asking price as reasonable.

These days, there are all kinds of USB dongle-type portable DACs available at different price points, features, and power capabilities - for example, the Moondrop MoonRiver 3, Fosi DS1, Hidizs S9 Pro, Shanling UA4, iFi GO Link 2 Max, Questyle M15i Portable, Schiit Vestri Portable, and Antelope Audio ZEO Portable to name a few. Have fun shopping if you're in the market for such devices!

--------------------

Notice in the very top image I'm playing the latest 2026 remaster of The Beatles' Rubber Soul (1965) on my phone which also is available as multichannel/Atmos on Blu-ray! Finally.

For this remaster, Giles Martin & Sam Okell also used "de-mixing" machine learning to extract the instruments and vocals from the original master tape. It has been said that with Rubber Soul, The Beatles started using the studio production itself as a creative instrument, not just to capture sounds.

Unfortunately, complex multi-layer productions were limited by the 4-track tape of the time so they had to overlay ("bounce") multiple layers on a single track on the master tape itself. In order then to remix the album, the AI/machine learning algorithm was used to pull the layers apart. We've seen/heard the technique used already in 2021 with the Get Back documentary, the 2022 release of Revolver (1966), and the single "Now And Then" in 2023 where they pulled John Lennon's vocals off a low-quality cassette recording baked in with piano and other noises.

While 8-track tape recording wasn't used by the band until 1968's White Album, the 2017 remix of Sgt. Pepper's (1967) didn't need the de-mixing technology as pre-bounced 4-track first-generation tapes were still available in the archives.

I found this new Rubber Soul mix tastefully done, and much more natural with vocals centered than the early-stereo extreme left/right panning on the original 1987 CD release. Things sounded a bit better in the 2009 stereo remaster (not remix) with clean-up for clarity and narrowing of the extreme panning so vocals at least sounded more centrally placed on tracks like "You Won't See Me". The multichannel version uses the surround channels for ambience and stereo expansion, there's good clarity, certainly nothing extreme or surprising; more of a 3-channel stereo than quad-like presentation and appropriate for well-loved music of this vintage.

Here's "Michelle", original US stereo mix, lead vocals panned right:

And here's the 2026 Mix:

A little blast from the past, but updated IMO for the better. A bit like the return of the DragonFly for another iteration, I guess. 😉

Hope you're enjoying some great music, audiophiles.

Happy Canadian Thanksgiving, eh? 🍁 🦃