AudioQuest Dragonfly Cobalt Review (Portable Headphone Adapter)

[Deleted member 31750]

nice one, thanks

 

[THW]

I'm quite sure the distortion characteristics of this dongle is intended. They seem to follow the same erroneous path like PS Audio.

feature, not a bug.

in all seriousness, this desire for distortion just seems very strange to me. it kind of reminds me of the myth floating around on the Internet saying how the Bren light machine gun is “too accurate” (which is weapons-grade BS, since if you want a bigger beaten zone you can just hold the gun more loosely and let recoil do its thing)

my bemusement at this is similar, why would you want an inherently less accurate device? with a more inherently accurate device, you can, by your own volition, make the output less accurate if you want to. you can’t really do it the other way around.

 

[michman66]

Hi, I have tried several dongle dacs (Dragonfly Red and Cobalt, Tempotec Sonata HD Pro, Meizu Hifi Pro, Zorloo Ztella, Lotoo PAW S1, E1DA 9038S G3, and I found the Lotoo PAW S1 and the G3 to be the most musical between them.
I suggest you to try the E1DA 9038S G3. I think it is the most mature 9038Q2M implementation by far. With their Tweak9038 app you can change the filters, control 2nd and 3rd harmonic distorsion (built in features of the chip),etc... so you can tailor the sound signiture to your needs or even to the music you listen to, on the fly. I have seen this feature until now only on desktop ESS devices like SU-9. There are some firmwares with unique setting, that you can use as presets. The linear phase slow roll-off_SE firmware is quasi 'tube emulation'. I use the dac with this firmware, and it is very musical with totally black background. If I wish i have the opportunity to change for a more technical sound. The G3 combines great measurements/professional engineering with musical sound and perfect usability. The achilles heels are: higher power consumption, sensitivity to RF noise and only 2,5 balanced out. The sensitivity can be reduced mostly with the usage of DDHifi TC05L USB-C cable, and since there is the E1DA 9038D with the 3,5mm output available. In summary the G3 is a very innovative product for 100USD. From the listed dacs, I kept the G3 and the Meizu Hifi Pro (Lotoo I just sold because I started to build a desktop rig . I think, Audioquest should have come out with a product like this based on the 9038Q2M.

I'm in dongle hell too right now. I like your comments and see you have listened to a fair share of the top ones here on ASR. I may have missed it, have you heard the Hidez S9 vs G3. I was also considering the Topping DX3Pro. The Linsoil is 119 on Amazon like your comments reviews are very high. Amir did praise the S9 very high too. Thx

 

[Bimbleton]

Here’s a cross post from HF:
Based on reviews I ordered a Hidizs S9 Pro. By all measurements, it should be vastly superior to my Cobalt!

I’m shocked to say that the Cobalt sounds better.

I have absolutely no explanation why, but I have A/B tested it with my A12T’s, volume matched, and every time the Cobalt sounds better. A better sense of space, slightly sharper (in a good way) and better sense of positioning. The Hidizs, in contrast, almost sounds a little wooly. I can’t believe it, and I have no idea why the better measuring item (Hidizs) sounds noticeably worse than the Cobalt.

Is this a situation where the distortion or the flaws in the sound are actually enhancing it (similar to tube amps)? I just don’t understand…

 

[Raindog123]

I have absolutely no explanation why

I do… A hypothesis rather, but anyway...

In many systems involving sensors there is a known ‘wobbling effect‘ that is used to focus, get on target quicker, get the signal from under the noise, etc… By the same token, presence of certain-spectrum distortions, coloration can be (is?) perceived positively by our brain. Just like the infamous ‘tube‘ sound. And putting aside purists‘ ‘transparent reproduction‘ and its pros, the known (eg measured) coloration of Cobalt is acknowledged by many as [subjectively] very positive… When I had a Cobalt I rather enjoyed its sound too, but sold it anyway mainly due to the price/competition.

 

[Bimbleton]

I do… A hypothesis rather, but anyway...

In many systems involving sensors there is a known ‘wobbling effect‘ that is used to focus, get on target quicker, get the signal from under the noise, etc… By the same token, presence of certain-spectrum distortions, coloration can be (is?) perceived positively by our brain. Just like the infamous ‘tube‘ sound. And putting aside purists‘ ‘transparent reproduction‘ and its pros, the known (eg measured) coloration of Cobalt is acknowledged by many as [subjectively] very positive… When I had a Cobalt I rather enjoyed its sound too, but sold it anyway mainly due to the price/competition.

This is all a learning experience for me, but I think, going forward, I’ll take raw measurements with a grain of salt.

Objectively, devices may measure perfectly, but I am not perfect and (it turns out) neither are my sonic preferences.

 

[A-lexx]

I don't get all that complains about "bad" THD and oh so high distortion on DFC. Everyone seems to forget to convert THD numbers from dB into percentage. THD of 78dB means 0.01%. Everything below 1% is not audible for humans. 0.01% is 100 times lower than 1%. There is no point in discussing this, it doesn't matter if the THD numbers could potentially be better or not, as it's already much better than what a human could hear.
On the other hand, the SNR numbers and dynamic range are good, better than many other devices tested here, which is actually easily audible - just take a sensitive IEM like SE846 that hisses on almost anything. It's pretty silent on DFC, wasn't silent on DFR. A big improvement here. Something you can really hear, not only measure.
Another point - the filter they implemented. Ok, it's a minimum phase slow roll-off. They are clearly saying that in the 'specs'. So, why doing so surprised when measuring the frequency response? Of course it measures like advertised, it's a 'feature' of this filter. Who cares if it's -4 dB at 20 kHz, no adult would be able to hear this frequency anyway. And there is an easy way to fix it - you don't like the filter, do oversampling on your source. Already oversampling by x2 would move the cut-off frequency of DFC's filter way higher than any human could hear, so you are basically 'bypassing' the effect caused by the filter, that is only relevant for 44.1k content

 

[xaviescacs]

If you want a brand name, get the THX Onyx which washes the floor with it.

The guys at What Hifi have an opposite opinion on that:

https://www.whathifi.com/reviews/thx-onyx

I quote: "The five-star Audioquest DragonFly Red (£169, $200, AU$280) – the class-leading portable DAC at this price – provides a much wider window into a song, bringing musical details and instrumental textures to the surface that the THX overlooks."

And the Cobalt is supposed to be much better than the Red: https://www.whathifi.com/reviews/audioquest-dragonfly-cobalt

Many reviewers seem to be capable of hearing distortion and interpret it as detail, like with the PS Audio DACs. The often use the word "richness". Well, if I understand it correctly, there are indeed some exaggerated artifacts there as distortion gets higher, so in that sense it is "richer", as there is "more".

 

[A-lexx]

The guys at What Hifi have an opposite opinion on that:

https://www.whathifi.com/reviews/thx-onyx

I quote: "The five-star Audioquest DragonFly Red (£169, $200, AU$280) – the class-leading portable DAC at this price – provides a much wider window into a song, bringing musical details and instrumental textures to the surface that the THX overlooks."

And the Cobalt is supposed to be much better than the Red: https://www.whathifi.com/reviews/audioquest-dragonfly-cobalt

Many reviewers seem to be capable of hearing distortion and interpret it as detail, like with the PS Audio DACs. The often use the word "richness". Well, if I understand it correctly, there are indeed some exaggerated artifacts there as distortion gets higher, so in that sense it is "richer", as there is "more".

Nope that richness doesn‘t have anything to do with harmonic distortion. DFR has a pretty good THD of around 0.003%. That is not audible by no means. The threshold for audible distortion, for a single sine tone, is around 0.2%. For music, it‘s above 1%. So, THD of 0.01% is already absolutely beyond audible threshold. 0.003% of DFR even further. That would never add any “richness“.

 

[xaviescacs]

Nope that richness doesn‘t have anything to do with harmonic distortion. DFR has a pretty good THD of around 0.003%. That is not audible by no means. The threshold for audible distortion, for a single sine tone, is around 0.2%. For music, it‘s above 1%. So, THD of 0.01% is already absolutely beyond audible threshold. 0.003% of DFR even further. That would never add any “richness“.

Thanks for the clarification. I wonder then what this people are hearing to describe such a big difference between the Audioquest and the THX.

 

[amirm]

The guys at What Hifi have an opposite opinion on that:

What hifi is an advertising masquerading as reviews. If you believe them over proper objective analysis and formal listening tests I do here, then there is no hope for you. I suggest forgetting stuff like that. There is no there.

 

[amirm]

Many reviewers seem to be capable of hearing distortion and interpret it as detail, like with the PS Audio DACs. The often use the word "richness". Well, if I understand it correctly, there are indeed some exaggerated artifacts there as distortion gets higher, so in that sense it is "richer", as there is "more".

This is all incorrect. Have these people show just one of these factors to be true when they don't know what is playing and I will buy you one of these dongles! For your own sake, forget all of that. We are here to show you conclusions that are strongly supported by evidence. Don't listen to nonsense over that.

 

[xaviescacs]

What hifi is an advertising masquerading as reviews. If you believe them over proper objective analysis and formal listening tests I do here, then there is no hope for you. I suggest forgetting stuff like that. There is no there.

I've never trusted them very much, but I did gave them some credit, as commentators, so to speak. In this case where you have measured both devices, I was trying to find an explanation of what they are possibly hearing, since the tests doesn't show anything but higher distortion in the Audioquest. I was just hoping to find an explanation other than yours

What hifi is an advertising masquerading as reviews

which is too disappointing to me. If this is really all about brands and marketing, and these people don't hold a bit of honesty, it is just too sad. Good to know though.

Thanks.

 

[xaviescacs]

If you believe them over proper objective analysis and formal listening tests I do here,

I don't think it's a matter of believe or trust. It is a matter of reproducibility. Some technical measurements or papers are reproducible and therefore deserve credit and attention. Some others are not, although they look very "technical". I don't like to think that people should trust or believe in what is published here, they have to trust or believe in the method of being transparent enough to enable any person to perform the same exact test. This is what they call science.

Of course I'm not discovering anything, just wanted to dig into a bit.

 

[A-lexx]

It‘s a difficult topic. Some people are strong believers in objective or subjective tests. I consider myself agnostic. My background is EE MSc with decades working in the industry.
There are many objective measurements that you cannot neglect.
I also do not hear any differences in cables unless the cable design is really screwed introducing high impedance values.

However I also cannot explain why an Apple dongle, which measures perfectly fine, same level as DFR or even better in THD than DFC, sounds not that good to me (and also to my wife, who doesn‘t know anything or doesn‘t care about hi-fi). Apple dongle has the same or even better frequency response, the same low output impedance, even the filter on an apple dongle and DFR is the same, minimum phase fast roll-off. Also the load - an easy to drive over ear or IEM, doesn‘t put much stress on current driving capabilities. And still, Apple dongle sounds cold and clinical, with diffuse instrument placement, depth perception is suffering. DFR on the other hand sounds richer, with warmer and more powerful low end, also better controlled, and with holographic instrument placement.
Many audiophile folks claim, perception of position and depth is related to a filter used, and minimum phase fast roll-off filter with lots of post-ringing negatively affects the perception. But apple dongle and DFR both use the same filter that also measures the same, but still they sound nothing alike.

I would be happy to ditch the DFR in my chain. It‘s not very practical, is draining the battery 10 times faster… Tried to ditch it multiple times, but the apple dongle just doesn‘t sound to my liking…

 

[Raindog123]

Apple dongle, which measures perfectly fine, same level as DFR or even better in THD than DFC, sounds not that good to me

Most probably, still due to its very low output power that is not sufficient for your headphones. Are all your IEMs sound equally 'not that good'?

 

[A-lexx]

Most probably, still due to its very low output power that is not sufficient for your headphones. Are all your IEMs sound equally 'not that good'?

yes, all of them.
As an example - SE846 or T5p are really easy to drive, Apple dongle could blow a driver on T5p if you max out the volume, I‘m not talking about anything like HD600.
It stays the same, if I use it as DAC only, adding a really powerful Fiio A5 into the chain - I can still easily hear distinct sonic characteristics of an Apple dongle vs. DFR or DFC (don‘t hear much difference between DFR and DFC though).
Actually, my mobile setup for driving HD600 is an iPhone + DFR + A5. Countless times did I try to replace the DFR with an Apple dongle, sometimes for days, one time even for a couple of weeks, but it just doesn‘t sound as good as it does with DFR, so I finally gave up on this, even though DFR drains the iPhone battery flat in 7-8 hours, which is not very convenient.

 

[Bimbleton]

I have the deepest respect for Amir and ASR. The purpose of this forum seems to be to provide a specific set of objective, measurable variables… and by those standards, boy the DFC seems pretty poor!

But perhaps there’s factors that are just not measurable, or change person-to-person or moment-to-moment. Maybe sound is more like art, where precision doesn’t always produce the most enjoyment.

I posted on Head-Fi as well with the same remarks — that I have no idea why the DFC sounds good, and clearly more enjoyable than a brand new Hidizs S9Pro. I like the Dragonfly Cobalt! It just sounds good to me.

 

[Veri]

But perhaps there’s factors that are just not measurable, or change person-to-person or moment-to-moment. Maybe sound is more like art, where precision doesn’t always produce the most enjoyment.

It's also the perfect excuse to sell quickly made, cut corners, shit measuring gear. "But it's like art!"

 

[A-lexx]

Well, in case of a DFC, I wouldn't mind having better THD numbers, but actually that's not really important, as THD of 0.01% is not audible, to no one. So, it's a pure academic discussion, whether having a THD of 0.0001% would give any improvement in terms of sound quality.

The design target for DFC was clearly achieved, I'd guess. Ok, using a minimum phase slow roll-off filter was clearly a marketing decision, just to differentiate from other versions of Dragonfly. Also, some users of DFR complained, that it's treble is too harsh, so using this filter to reduce the treble by 1-2 dB at 18kHz wasn't such a bad decision neither.

What is a clear differentiator for a DFC, and an actual reason I've bought it - it has been designed with power consumption in mind. Instead of using an off the shelf USB Driver IC, the designer went for a customized solution, using a general purpose MCU in order to keep the power consumption low (remember, the newest cut of ESS chips combining USB driver+DAC+HPAmp in one IC was not available at that time).
Target achieved, even in 2021 DFC has a lowest measured power consumption among the non-single-chip dongles that can deliver around 2Vrms. And this is really important if you use it with a smaller iPhone or even an iPod Touch - the battery in those devices is not exactly huge.

The second aspect, where DFC delivers, is a very good SNR. Even though the HP-Driver is blasting at fixed gain all the time, with volume control done digitally in the DAC, and considering the max. 2 Vrms it's capable of, the noise level is exceptionally low. It's one of few dongles that does not hiss with very sensitive IEMs like an SE846.

One other aspect often forgotten - many other dongles or mobile DAC/Amps, that are actually designed to work with a mobile phone, are very susceptible to EMI coming from that phones. So, what is a perfectly measuring mobile DAC good for, if in order to use it you need to put your phone into a plane mode, so you can't even stream your music content? DFC however is dead silent, even with most sensitive IEMs.

So, basically, I consider the discussion where a DFC is trashed for some of the performance numbers that only matter on academical level, not really fair.
Other aspects, like low power consumption to save battery of a mobile device and a low noise level, that is easily audible, are more important than THD numbers already far beyond audible level that could be better but actually are not affecting the audible quality of sound reproduction.