AudioQuest Dragonfly Cobalt Review (Portable Headphone Adapter)

[toobs]

No, I commented on how misleading this whole thread is...and your reply is not at all helpful. Anything else is better for high sensitivity phones? really? That's certainly not true. The measurements shown at the beginning of the thread indicate the Cobalt is better for the UE than e.g. the Hidizs S9. But you read this thread and you would think something completely different. Show me I'm wrong about that! I'm confident that you can't. Would you say a single ended 300B amp is shit just because it can't drive a ribbon speaker? That's just nonsense.

 

[Rottmannash]

No, I commented on how misleading this whole thread is...and your reply is not at all helpful. Anything else is better for high sensitivity phones? really? That's certainly not true. The measurements shown at the beginning of the thread indicate the Cobalt is better for the UE than e.g. the Hidizs S9. But you read this thread and you would think something completely different. Show me I'm wrong about that! I'm confident that you can't. Would you say a single ended 300B amp is shit just because it can't drive a ribbon speaker? That's just nonsense.

9038D for one. And killer specs to boot.

 

[toobs]

9038D for one. And killer specs to boot.

That's certainly a good candidate, but just going off of the measurements the Cobalt has a better signal to noise/distortion ratio at the low power output levels needed for the UEs. That's going off the 300ohm chart. I didn't see a 30ohm chart.

The chips (DAC and amp) that the Cobalt uses are very good. They clearly underspecified the current delivery of the power supply, but it is very low noise and from the charts the current delivery isn't an issue until you get to higher power levels that are irrelevant with the UEs. Don't get me wrong: I wouldn't recommend the Cobalt for anything but highly sensitive headphones, but the Cobalt measures very well at low power levels and that point is clearly missing in this thread.

 

[mansr]

The chips (DAC and amp) that the Cobalt uses are very good. They clearly underspecified the current delivery of the power supply,

The weakest part of the power supply is the charge pump providing the negative rail, and that's part of the output driver.

 

[toobs]

The weakest part of the power supply is the charge pump providing the negative rail, and that's part of the output driver.

Interesting. ESS rates the 9601K at -100dB THD+noise at 30mW into 32ohms versus less than -70dB for the Cobalt as measured in this thread. I would have guessed that's an external power supply issue, not a chip issue given ESS's significant engineering resources, but have no insights beyond that.

That 9038D looks pretty interesting for my new lower sensitivity IEMs (my UE 10s broke after more than 10+ years of service).

 

[Raindog123]

but the Cobalt measures very well at low power levels…

…and still costs $300

(While E1DA goes for under $100 and EarMen Sparrow for $200)

 

[toobs]

I was a student until I was 30 so I get that, but that was a long time ago. I was just reading about all the useability issues (radio interference, drivers, popping) with the 9038D, not to mention it's sold out and I have to import it from China versus going to the local audio store, trying it out, getting a warranty and local support if something goes wrong, the tidal app sends bit perfect data without uapp .... it's easily worth 200 to me now.

Anyway, I made my point, choosing a dongle based upon thd+noise at 2v ptp isn't necessarily going to get you a good result. The Cobalt has a use.

 

[mansr]

Interesting. ESS rates the 9601K at -100dB THD+noise at 30mW into 32ohms versus less than -70dB for the Cobalt as measured in this thread. I would have guessed that's an external power supply issue, not a chip issue given ESS's significant engineering resources, but have no insights beyond that.

If you look at the output with a scope while decreasing the load impedance, it is plainly visible that the negative part of the waveform starts distorting well before the positive.

 

[toobs]

If you look at the output with a scope while decreasing the load impedance, it is plainly visible that the negative part of the waveform starts distorting well before the positive.

Could be due to undersized or high esr external caps in the charge pump circuit no?

That's something that could potentially be modded if that's the problem. hmmmm

 

[mansr]

Could be due to undersized or high esr external caps in the charge pump circuit no?

It's a ceramic type, don't know the value though I could check.

 

[toobs]

looking at the datasheet for the 9601K, there are two capacitors for the charge pump: flying capacitor (min 2.2uF/max 4.7uf, max 100mOhm ESR) and hold capacitor (min 22uF, "low ESR"). Would be interesting to know the values and types. Larger capacitors improve regulation. For the fly capacitor they say that the larger value reduced output resistance to "an extent" and add that transistor resistance dominates already with a value of 2.2uF.

It's also looks pretty easy to bypass the charge pump entirely and provide an external -3.3V supply, but then you'd be 3D printing a new enclosure to fit the supply. That said, the performance potential is pretty interesting for higher sensitivity phones.

 

[toobs]

just found this: https://audiophilestyle.com/ca/reviews/objective-audioquest-dragonfly-cobalt-review-r832/

But I guess you know about that already

 

[Raindog123]

The Cobalt has a use.

Some would/did sorta agree. “For easy loads”... While many simply lost their patience and got turned off by that shameless AudioQuest’s marketing of Cobalt compared to Red: Including their (unsubstantiated) claims of “improved” consumed power and [alleged] increase of the output level — all to solely justify the the price hike without actual merit.

 

[toobs]

So providing a proper regulated external -3.3V supply looks fairly easy

charge pump for -5V supply: https://www.pololu.com/product/2839
3.3V stepdown regulator: https://www.pololu.com/product/3791

it looks like there is a trace between a through hole and one end of R8 and the trace continues on to PIN 12 (charge pump output), which must be connected to PIN 11 (analog negative supply). It would be easy to solder the -3.3V supply to this through hole.

then cut PIN 12 to disconnect the charge pump supply voltage.

You could ground PIN 20 to turn off the charge pump, but I suppose it doesn't hurt much to keep it running.

 

[Raindog123]

@toobs May I suggest you starting a new - "how to revive/improve Cobalt" - thread? The "product review" threads we're in are traditionally kept with the info about "the consumer goods (and bads?) you get out of the box".

I think many of us would be interested to see if this idea of fixing the negative rail of the supplied power would improve the measured performance... but again, in its proper place.

[And, welcome to ASR!]

 

[mansr]

"Fixing" an AQ product is just throwing good money after bad. Buy a cheaper, better device begin with.

 

[toobs]

"Fixing" an AQ product is just throwing good money after bad. Buy a cheaper, better device begin with

I'll drop this, but the whole point is that you can't buy a dongle that matches the noise floor of the Cobalt and if providing a better supply fixes the clipping issue then you've got a product that is better than anything you can buy for sensitive headphones at any price.

 

[Rottmannash]

I'll drop this, but the whole point is that you can't buy a dongle that matches the noise floor of the Cobalt and if providing a better supply fixes the clipping issue then you've got a product that is better than anything you can buy for sensitive headphones at any price.

There's not another dongle with a lower noise floor??

 

[Jimbob54]

There's not another dongle with a lower noise floor??

Or put another way, how low does the noise floor have to be to be inaudible on the most sensitive iem?

 

[mansr]

Or put another way, how low does the noise floor have to be to be inaudible on the most sensitive iem?

Below the noise level of the music is enough for sure. Depending on playback volume, even that might not be necessary. Let's say 100 dB below peak music level to be on the safe side.