Qudelix-5K Bluetooth DAC & Headphone Amp

[jkim]

In case anyone is interested, I measured the 5K's PEQ filter response as well as its dynamic range in Bluetooth (LDAC) mode and posted here (to compare the FiiO BTR15 with it).

The 5K is exceptional!
--------------------
EDIT. Actually, I think the test deserves a spot in this thread, too.

PEQ responses in Bluetooth (LDAC) mode were tested between Qudelix 5K pitted against FiiO BTR15:

Note. The Bluetooth (LDAC) playback of test signals was from the USB Audio Player Pro (UAPP) installed on Google Pixel 8.

There is certainly a problem with the BTR15's LDAC performance, not achieving the sample rate of 96 kHz.

Measured the dynamic range of both devices' LDAC modes:

It seems that the BTR15's Bluetooth LDAC mode has some kind of clocking/jitter issue, making its FFT unstable---the skirt around the test tone is fluctuating over rolling averages.

The Qudelix 5K's Bluetooth performance (and its PEQ accuracy) is exemplary, explaining why it has been regarded as the best of its kind for quite a while.

 

[Florin Andrei]

Distortion compensation is explained on page 18 here: https://www.scribd.com/document/869292167/ES9219-Datasheet-v1-2

 

[jkim]

There must be quite a few happy Q5K users out there. Here's a very interesting (and good) thing I found recently.

It turns out that the 5K is actually one of the very few DAC/HP amp combo devices that support analog volume control. The ES9219C chip in Q5K includes digitally controlled analog volume control in its headphone amp stage. It supports a 24 dB range of analog attenuation. Q5K is designed to combine it with digital attenuation for its on-device volume control.

When the noise performance of a combo device with analog volume control is measured, a test signal should not be attenuated digitally but by the device's volume control, just as when a headphone amp is tested. But Amir's test in this review used a digitally attenuated signal.

Here's a measurement of Q5K's SNR at 50 mV output attenuated by on-device volume control:

Note. USB input, Fs = 96kHz, BW = 20Hz - 20kHz, 3.5mm output into no load. Left channel shown; right channel has similar performance.

Q5K unweighted SNR@50mV (unbalanced output) = 94.3 dB.

Amir's AP SNR measurement is CCIR-2K weighted, which is usually a little worse. So, I also measured its SNR using the CCIR-2K weighting:

Note. USB input, Fs = 96kHz, BW = 20Hz - 20kHz, 3.5mm output into no load. Left channel shown; right channel has similar performance.

Q5K CCIR-2K-weighted SNR@50mV (unbalanced output) = 92.9 dB.

Yes, this is state-of-the-art noise performance.

EDIT. Essentially no difference under realistic 32-ohm load:

This level of noise performance sets Qudelix 5K apart from other brands' portable USB or Bluetooth DAC/HP amp products with onboard PEQ. Its carefully designed functionalities, their correct/accurate operation, and user interface all make it an unbelievable value. I am starting to wonder why one would ever need another portable combo device. Only wish it provided volume-adaptive loudness compensation. Currently no portable device does that...

 

[Somafunk]

I wish someone would print off a number of these 5k button operation stickers, I love my 5k but use it infrequently so when I do pick it up I often spend 5mins getting the hang of it

https://www.reddit.com/r/headphones/comments/v627gl/qudelix_5k_sticker_final_design_with_file

 

[jkim]

Here is another aspect of thoughtful, clever engineering in Q5K.

RME ADI-2 series users do not have to worry about the so-called 'global preamp cut/gain' when setting up PEQ filters. It is simply because the device's digital volume control naturally integrates a preamp function. A user's volume attenuation will take care of overloading. This also helps a user interpret the volume reading on its display intuitively by just having PEQ effects in mind. RME also supports what they call the "Auto Ref Level" function that automatically switches analog gain from High to Low mode to achieve lower noise at low signal levels.

Q5K does essentially the same thing as RME by combining its digital volume control with analog gain adjustment. In fact, it does it even better because the analog gain can be adjusted in a wide 24 dB range in small 1 dB steps.

Q5K applies digital volume control b/w 0 to -18 dB (in High Gain mode). In fact, on the Q5K app's scale this range corresponds to +6 to -12 dB, but here I am showing values more natural on my mind, i.e., maximum = 0 dB. Below -18 dB, analog attenuation kicks in to exploit its low-noise benefit. A nice aspect of this scheme is that digital attenuation is designed to serve as a preamp cut being added to any user PEQ. So, as long as a PEQ transfer function's peak is 18 dB or lower, a user does not have to worry about a preamp cut. In fact, 18 dB is huge; so practically it eliminates the need for a manually entered global preamp cut.

To show this functionality in action, I set a PEQ filter:

PEAK: Freq = 500 Hz, Gain = 18.0 dB, Q = 0.5, Preamp Gain = 0 dB

A 500 Hz test tone > -18 dBFS would make the DAC overloaded unless digitally attenuated before DAC. First I set Q5K's volume to -24 dB and measured a response to a 500 Hz tone at 0 dBFS:

No clipping as expected, because the digital portion (-18 dB) of the volume setting works as a preamp cut. The measured voltage (0.993 Vrms) is also just as expected 6 dB below the max 2 Vrms from its unbalanced output.

Now, I increased Q5K volume to -12 dB. The DAC will be overloaded by 6 dB. Here's the response:

The signal is certainly clipped---it is not supposed to reach the theoretical +6 dB level (4 Vrms), anyway. But what is interesting is that Q5K applies some kind of soft clipping, which must be done in the digital domain, preventing screeching distortion behavior. This is a very nice touch.

For example, below is a response from another DAC product with onboard PEQ support in precisely the same 6-dB-overloaded condition (i.e., same peak filter with a preamp cut short of 6 dB):

Yikes. Nasty, screeching distortion. You may wonder what DAC product it is. I won't tell you for now

Anyway, I just wanted to show this is a thoughtful design aspect of Q5K. In fact, every time I discover a new thing about Q5K, I am amazed at how carefully and meticulously it was designed and engineered. I think it is a prime example of a well-engineered consumer audio product with great attention paid to detail.

 

[jkim]

Another interesting fact about Q5K.

Amir's SINAD measurement of Q5K is often misinterpreted as representing its mediocre performance:

As for the Q5K, the ES9219 chip on it produces better SINAD numbers when its output level is somewhat lower and load impedance is lower. Yes, this is counter-intuitive. But that is the property of this chip. In fact, these scenarios cover most of actual use cases. Even if you connect some power-hungry headphones, how often do you think you would hear the peaks of music tracks? Typical recordings have crest factors of 12 to 18 dB range (ratio of peak to average RMS levels).

Suppose you connect a 32-ohm headphone to Q5K's balanced output and listen at Q5K's maximum unclipping level defined with an input signal at 0 dBFS. It is about 3 Vrms at 0.1% THD. Assume a typical crest factor of 15 dB, meaning an input signal would be -15 dBFS on average. An average output voltage from Q5K would be about 500 mVrms. In this condition, Q5K produces the following SINAD:

Yes, SINAD is 99.4 dB. Absolutely no problem at all.

Now suppose you connect a 300-ohm headphone to Q5K's balanced output and listen at Q5K's maximum unclipping level defined with an input signal at 0 dBFS. It is about 4.0 Vrms. Assume a typical crest factor of 15 dB, meaning an input signal would be -15 dBFS on average. An average output voltage from Q5K would be about 720 mVrms. In this condition, Q5K produces the following SINAD:

SINAD is 96 dB. Again this is fine performance.

In fact, even Amir's measured SINAD has no problem, although it is worse than actual performance I get in the same condition for some reason. I have an idea of why this discrepancy can happen. It must be due to another unique feature of Q5K, which I suspect is closely related to the 'soft clipping' behavior I showed earlier. Whenever I measure Q5K, it often requires some settling time. No problem when listening to audio content, but FFT measurement is a different story, which can be extremely sensitive to a slight signal change. I don't think Amir has time to deal with such settling time.

In some testing conditions Q5K's SINAD is relatively low because harmonics (THD) are dominant in THD+N. It is just this ESS chip's characteristic, not necessarily indicating its amp stage is being pushed to its stress limit. Anyway, harmonics from headphone transducers will completely dominate Q5K's THD.

 

[jkim]

Yet another fun fact about Q5K. Yes, it keeps coming up, often unexpectedly!

In response to a question from @CedarX, I tested Q5K's interesting feature "Active Ground." One potential benefit of this feature is higher 'current' drive capability under a stressed load. So, using the lowest resistance, 10.4 ohm, on my load board, I measured 1kHz tone THD vs. level and compared Q5K's balanced, active ground, and unbalanced connections:

Under this load (10.4 ohm), active ground (blue curve) works! It provides higher output than the unbalanced connection (orange curve). But if balanced connection (green curve) is possible, we definitely want to utilize it to obtain higher output power in a similar condition.

Eventually I think I could write an unofficial review of Q5K titled something like "(Unofficial) In-depth Measurements and Review of Qudelix 5K Portable DAC/Headphone Amplifier." With all these data, I have enough material.

 

[Power Pop 23]

Eventually I think I could write an unofficial review of Q5K titled something like "(Unofficial) In-depth Measurements and Review of Qudelix 5K Portable DAC/Headphone Amplifier." With all these data, I have enough material.

Please do.

The performance you have measured in the Qudelix 5K merits the attention of a standalone listing in the Master Review Index.

Thank you

 

[patoulol]

I agree, this work deserves another topic for better visibility

 

[jkim]

Please do.

The performance you have measured in the Qudelix 5K merits the attention of a standalone listing in the Master Review Index.

I agree, this work deserves another topic for better visibility

Writing a dedicated review is always taking a lot of time---I actually have more data to show .
It would motivate me if some of you guys could help a review like that to be promoted to one of ASR's homepage reviews by sending a message to Amir.

Can you?

 

[Power Pop 23]

Writing a dedicated review is always taking a lot of time---I actually have more data to show .
It would motivate me if some of you guys could help a review like that to be promoted to one of ASR's homepage reviews by sending a message to Amir.

Can you?

Yes

 

[DrSpan]

I had a Qudelix 5K for a bout a year and yesterday it got stolen along with my backpack …… Bummer.

Today while looking for a replacement i landed here to get an update of the market and see it there is some competitor i should
also consider.
Bluetooth ability is very welcomed but not a must

Also Line in ability would be nice for when Bluetooth or Usb is not wanted

Do you think the 5K ist still the best for portability, Peq ability and Mac compatibility?
What about the Fiio KA15?
Also saw the Fiio Btr13 which looks like a cheaper Qudelix 5K alternative?

Am open for suggestions and would greatly appreciate some

 

[staticV3]

Do you think the 5K ist still the best for portability, Peq ability and Mac compatibility?

Yes.

What about the Fiio KA15?

Worse App.

 

[DrSpan]

Yes.


Worse App.

Ok you have convincing arguments i admit.
Man that stings. I hate thieves. Just had it for a year and now its gone and need to buy again

 

[Limopard]

Yes.


Worse App.

The Fiio app is a pain in the backside, that much is true. Other than that I'm satisfied mith my BTR17.

 

[Phorize]

I had a Qudelix 5K for a bout a year and yesterday it got stolen along with my backpack …… Bummer.

Today while looking for a replacement i landed here to get an update of the market and see it there is some competitor i should
also consider.
Bluetooth ability is very welcomed but not a must

Also Line in ability would be nice for when Bluetooth or Usb is not wanted

Do you think the 5K ist still the best for portability, Peq ability and Mac compatibility?
What about the Fiio KA15?
Also saw the Fiio Btr13 which looks like a cheaper Qudelix 5K alternative?

Am open for suggestions and would greatly appreciate some

I don't see anything else that is remotely as well engineered for it's purpose at any price point. It's brilliant.

 

[Curupira]

Do you think the 5K ist still the best for portability, Peq ability and Mac compatibility?
What about the Fiio KA15?
Also saw the Fiio Btr13 which looks like a cheaper Qudelix 5K alternative?

There are only two detractors to the Qudelix 5K as of 2025, relative to the competition: it's less performant than the CD Redbook format (on higher loads) (it should clear 96db SINAD for that) (EDIT: not an actual thing, when not playing fullscale the Q5K is quite transparent); its power output is not as high as the recent FiiOs, which use beefier OPAmps (a third here would be that its balance output is 2.5mm, a more or less abandoned format).

For me both of these faults are irrelevant in day to day use, as 1) can you really tell the difference from >90 dB devices? Most people can't; 2) do you really need more than 100/200mW for portable devices? When not falling for the "moar power = moar quality" fallacy, there's not much actual demand for very high power devices. The Qudelix can pretty much drive 99.9% of IEM devices, even with heavy EQ, and also most full sized headphones.

 

[solawind]

i hope the developer releases a new version someday, q5k is a fantastic device, but buttons could be more convenient and not so easily pressed in a pocket , and a better mic with noise reduction would be a plus. Also the clip is not so practical and makes the device bulkier, i would prefer a removable clip or something like that.
As for the dac and app part, they are perfect, no reall changes are needed there. Case and controls could be better. I have been using it for almost 3 years

 

[staticV3]

it's less performant than the CD Redbook format (it should clear 96db SINAD for that)

See @jkim's posts on page 53.

 

[Curupira]

See @jkim's posts on page 53.

Hmm, I've never seen this before. Still, under more common loads it should default to Amirm's findings?