======================
Important Update (8/1/25)
FiiO is the first audio company reacting to the issue of DRE-induced distortion associated with the CS431xx chips! A
new firmware release for the KA15 from FiiO supports an option to enable or disable DRE. With DRE disabled, it does not exhibit any distortion caused by DRE while retaining reconstruction filter options (NOS mode is no longer necessary as a workaround). Further, the 'DRE Enable' mode is different from original DRE operation: it minimizes DRE artifacts by lowering the DRE activation threshold. An analysis based on measurements was added to the bottom of this review. Previously posted measurements of NOS mode (an alternative way to disable DRE) were removed since there's no reason to use NOS with this new firmware.
Note to KA15 owners: Please actively discuss your impressions and experiences of any aspects/features of the KA15 (not just about this new firmware). I believe FiiO deserves user recognition.
Important Update (7/23/25)
A workaround for eliminating the Cirrus hump distortion has been found.
Measurements confirming the absence of the distortion behavior, when the KA15 is set to NOS mode, were added to the bottom of this review.(see the update as of 8/1/25). For more detail on how this hack works, refer to the
review of TRN Black Pearl which also applies to the KA15.
======================
View attachment 451164View attachment 451166View attachment 451167
The Fiio KA15 is a portable USB DAC/headphone amplifier recently released (Fall 2024) by Fiio. Key features are:
- High output power: 560 mW per channel into 32 Ohm (in balanced desktop mode)
- Dual CS43198 (DAC chips) + SGM8262 (op-amps)
- 10-band on-device parametric EQ
- Not a dongle-type yet compact form factor
If all these are true and it measures well, then it will be a very nice device worth its asking price, currently $126 (as of May 2025).
Measurement Setup
- AD converter: E1DA Cosmos ADCiso Grade A in Mono Mode, and Cosmos Scaler as a buffer.
- Analog preamp: E1DA Cosmos APU for 1 kHz SINAD and low-level tests.
- Software: Room EQ Wizard (v5.4 beta 80) for most of the tests, and Paul Kane's Multitone Loopback Analyzer (v1.2.9).
- Some test results were compared to the performance of the JCALLY JM20 and JM20 MAX whose measurements were made in exactly the same setup.
- Unless otherwise noted, the device's balanced outputs were measured in its Desktop and Class AB mode.
Results
Starting with a 1 kHz sinusoidal test of its balanced outputs under no load (20 kOhms):
View attachment 451184
This is on par with a desktop device's performance.
Another 1 kHz sinusoidal test of its
unbalanced outputs under no load (10 kOhms):
View attachment 451187
Excellent performance. Only a slight decrease in SINAD due to weaker fundamental tones.
Low-level, 50 mV output performance:
View attachment 451205
As expected, for the
same 50 mV level, its unbalanced output has a lower noise floor.
It provides choices of LP filters:
View attachment 451214
Frequency response (w/ Fast-PC filter) is as expected:
View attachment 451215View attachment 451216
Jitter test:
View attachment 451217
Jitter suppression is not perfect, but these small errors should not be audible.
Multitone test:
View attachment 451220
SMPTE IMD measurements versus output levels:
View attachment 451247
The IMD results of the three devices shown---Fiio KA15, JCally JM20 and JM20 MAX---are essentially the same as each other, except small differences near their maximum output levels. In particular, the JM20's IMD rises near the max output under this 300 Ohm load because its headphone driver integrated into CS43131 has lower current capability than the other two devices' separate op-amps.
THD+N versus frequency:
View attachment 451256
Nothing to be concerned about.
Output power of its unbalanced output:
View attachment 451259
Its
unbalanced outputs are as powerful as many similar devices' balanced outputs!
Let's see its balanced output power:
View attachment 451260
Output power from this small device is incredible. This is the record-shattering, highest-output portable USB DAC/headphone amp among the devices tested at ASR so far.
In case you want to see the same results on the output voltage scale:
View attachment 451262View attachment 451261
(Output impedance measurements of both balanced and unbalanced outputs will be posted when I get a chance to measure.)
All the measurements reported above indicate excellent, if not state-of-the art, performance of the KA15. In particular, its measured output power is unbelievable. However, the story does not end here.
There is something very interesting that is not found by the standard package of measurements. We saw the results of a 32-tone test earlier, which was performed at the maximum unclipped signal level. Normally, if we feed the same signal at a lower level, the ratio of the signal to distortion/noise deteriorates only slightly because of the weaker signal. For some CS43131- and CS43198-based devices, this is not the case---see
here for more information.
Unfortunately, the FIIO KA15 turns out to be one of the devices producing elevated noise and distortion when fed with multitone signals at lower levels. See below for
normal FFT results of 32-tone signals at high levels:
View attachment 451274
These are great results.
But as the signal gets weaker, we get the following problematic responses:
View attachment 451275
Yes, for a very wide range of signal strengths, we get quite nasty responses, losing several bits of resolution!
The response becomes normal when the signal reaches a much lower level:
View attachment 451277
To see the entire picture, a sweep of 32-tone tests across a range of signal strengths was performed:
View attachment 452225
Notice a huge "hump" which indicates the rise of total distortion + noise in a range of signal strengths. This may be dubbed a "Cirrus hump." In comparison, the JCALLY JM20 and JM20 MAX do not show this behavior in the same measurements.
One may think this measurement condition, made of 32 equal-amplitude tones, is unrealistic which may never occur in real audio contents. In fact, these signal processing defects can be found with any multitone signals, even with dual tones, as long as tone components are in similar amplitude. See below:
View attachment 451284
This dual-tone signal, called "TDFD Bass" in Room EQ Wizard, is simply composed of 41 Hz and 89 Hz sine tones. As with multitone signals, the KA15 distorts in a wide range of signal strengths. Again, to see the big picture, a sweep of these dual-tone (TDFD Bass) IMD tests across a range of signal strengths was performed:
View attachment 452226
Conclusion
So, what can we say? It is still difficult to tell
how clearly audible these measured distortions would be in reality. Of course, it should depend on audio content. But given the fact that this problem is observed even in a simple dual-tone test reported above, it is no wonder Roman at
RAA was able to easily spot a movie soundtrack to demonstrate the distortion. Obviously this is an engineering flaw that should not occur in a DA converter targeted at Hi-Fi markets and consumers.
The FIIO KA15 would have been an ideal recipe for a portable USB DAC/HP amp for those looking for high output power and hardware parametric EQ in a small form factor. Only if it had been implemented properly. In its current implementation, I cannot recommend it. In fact, I am in the process of testing quite a few CS431xx-based devices, some of which are mine and others were sent by an ASR member. Eleven of them in total! Unfortunately, except for a few dongles (including the JCALLY JM20 and JM20 MAX), most of them exhibit this problematic behavior. Based on these observations, like Roman at RAA I also believe it can be resolved by firmware design. But the problem is that there is no clear register in the chips' datasheets that must be related to this behavior. No one knows for certain, perhaps except for Cirrus Logic engineers (or even worse, former Wolfson engineers). I am planning to write another special review on this issue, covering these CS431xx-based devices..
Other Remarks:
- The distortion behavior is not affected by the Class H or Class AB mode in the firmware setting.
- The distortion occurs under any impedance load.
New Firmware (V1.0.5) with DRE Toggle Options
FiiO is the first audio company reacting to the issue of DRE-induced distortion associated with the CS431xx chips! A
new firmware release for the KA15 from FiiO supports an option to enable or disable DRE. As shown below, with DRE disabled it does not exhibit any distortion caused by DRE while retaining reconstruction filter options (NOS mode is no longer necessary as a workaround). Further, the 'DRE Enable' mode is different from original DRE operation: it minimizes DRE artifacts by lowering the DRE activation threshold. Previously posted measurements of NOS mode (an alternative way to disable DRE) have been removed since there's no reason to use NOS with this new firmware.
Let us first see the noise levels of the two modes (DRE Enable & Disable) of the new firmware:
View attachment 467002
The CS431xx chips, by default, activates DRE when the signal is below -12 dBFS (see
TRN Black Pearl measurements). In the new firmware FiiO lowered the threshold to -44 dBFS (red solid line). That is, DRE is turned on when the signal level is below -44 dBFS. So, FiiO takes a very conservative approach, minimizing DRE artifacts even in the 'DRE Enable' mode. This mode will still increase its dynamic range, by about 4 dB, when measured with a -60 dBFS tone. Also, it will lower the noise floor when near-silent tones (e.g., < -50 dBFS) are played. It seems to be an excellent trade-off.
Below are measurements comparing the effects of three different DRE modes on 32-tone total distortion plus noise (TD+N):
View attachment 467001
In the new firmware's 'DRE Enable' mode there is still a slight Cirrus hump (red solid line) when the DRE activation threshold is lowered to -44 dBFS. But the signal levels in that hump range are very, very low, leaving the distortion nearly indistinguishable from the measurable noise floor. Put another way, this distortion should be inaudible, or indistinguishable from noise (noise is unnoticeably low as well). With DRE disabled (green solid line), even this hump is gone but the noise reduction for very low-level signals is also gone.
In the same recording condition as used in my
TRN Black Pearl review, the spectrogram of recording of the Dune soundtrack clip
in either DRE Enable or Disable mode shows no distortion:
View attachment 467016
Recording of the C Major test clip
in either DRE Enable or Disable mode:
View attachment 467017
Clean as well. A faint sign of distortion in the C Major recording at the start is suspected to be from some other mechanism of the KA15. This is not from DRE since it's there whether DRE enabled or disabled.
Again, note that distortion does not show up even in 'DRE Enable' mode. This is not surprising as the DRE threshold is so low at -44 dB. Essentially, its DRE Enable mode is not different from DRE Disable mode for practically all use cases. The only difference will be in testing situations using near-silent tones < -50 dBFS.
SNR at the output of 50 mV (measured in the same condition as the 50 mV SNR reported in the
TRN Black Pearl review):
Device | SNR @ 50 mVRMS |
|---|
FiiO KA15 w/ previous firmware (3.5mm out) | 88.9 dB |
FiiO KA15 w/ new firmware V1.0.5 (3.5mm out) | 84.8 dB |
TRN Black Pearl w/ DRE on (3.5mm out) | 96.5 dB |
TRN Black Pearl w/ DRE off (NOS) (3.5mm out) | 85.2 dB |
Slightly worse noise performance compared to the TRN Black Pearl (w/ DRE off) is just as expected from the fact that the KA15 includes an additional headphone driver with +1.5 dB gain based on op-amps (SGM8262-2). Still, its SNR with DRE disabled (84.8 dB @ 50 mV) is decent enough for any practical use case.
In terms of dynamic range (DR), the new firmware gives:
KA15 Output | DRE Enabled | DRE Disabled |
|---|
Unbalanced 3.5mm out referenced to 2.4 Vrms | 122.1 dB | 118.2 dB |
Balanced 4.4mm out referenced to 4.0 Vrms | 123.5 dB | 119.6 dB |
Balanced 4.4mm out referenced to 4.8 Vrms | 125.1 dB | 121.2 dB |
Lastly, the following multitone test confirms that the standard reconstruction filter (Fast-PC here) is in place:
View attachment 467013
We finally have a CS431xx-based device with no DRE artifacts. The FiiO KA15, as shown in the review, provides incredible power output (550 mW into 32 Ohm) and onboard PEQ in a small portable form factor. Kudos to FiiO's prompt reaction.
EDIT. In fact, this firmware update was made possible by an ASR member's tenacious communication with FiiO. FiiO was also responsive to his request and my measurement-based feedback along the way to this firmware revision.
. I even own a pair of Neumann KH420 with subs, dsp system, whole room treatment with 16 or so gik acoustics bass traps but I'm being miserably stingy with this dsp DAC purchase. It doesn't make sense to me either hahah. Humans are beautiful.
