FiiO E9 Headphone Amp
YAFR (Yet Another FiiO Review): I’ve been accused of favoring FiiO and even working for them. The fact is they make some of the more popular reasonably priced headphone amps and that’s why I’ve been testing them. Some were surprised I rather liked the $19 FiiO E5 and and $99 FiiO E7. I’ve had lots of requests to review the FiiO E9 and, by popular demand, here it is.
WEDNESDAY ARTICLES: I’m going to try to publish a blog article every Wednesday starting with this one. That should make things a bit easier for those who are constantly checking to see if anything is new here.
SPLIT PERSONALITY: The FiiO E9 is a “desktop” headphone amp and can be used with the usual sources. It also serves as a dock for the FiiO E7 turning the E9 into a USB headphone DAC. The E7 is a portable USB DAC but doesn’t quite have enough output for some large power hungry high impedance cans. The E9 greatly boosts the E7’s output. It also neatly handles the PC USB connection and charges the E7 while its docked.
BUILD QUALITY: The build quality was typical FiiO which is to say most would guess it costs more than it really does. The fit, finish, and details are hard to fault.
UNUSUAL EXTRAS: The E9 sells for a reasonable $129. Check the E9’s Specifications for all the details. There are some unusual features at this price including a protection relay that enables the output a few seconds after power on, and hopefully will disconnect the headphones preventing damage if something goes wrong (DC can damage headphones). There’s also a very useful gain switch on the back that lets you optimize the E9 for low or high efficiency headphones. I explain why this is a good thing in my Headphone Amp article. It’s also nice to have both 3.5 mm and 1/4 inch headphone jacks although I don’t recommend using the mini jack—more on that later.
MINOR NITS: I was a bit disappointed the E9 doesn’t have RCA input jacks for the line input. There’s only a 3.5 mm connector. And none of the connectors on the E9 are gold plated.
IN USE: The big volume knob is nice with an OK feel—it’s a little bit “rubbery”. The power switch is built into the volume control. The small blue power LED is tastefully discrete rather than blinding like some. With high efficiency headphones there’s a little bit of “rustling noise” when changing the volume but it’s masked by music. It runs fairly cool with the metal enclosure getting only slightly warm.
HISS & NOISE: In the low gain mode, and connected to a muted source, there is some slight audible hiss with my efficient Ultimate Ears IEM headphones. The hiss is worst at half volume for reasons I explain in the Tech Section. In the high gain mode, which nobody with efficient headphones should be using, there’s a bit more hiss. But with typical full size cans, it’s dead silent in either gain mode.
MEASUREMENT SUMMARY: The measurements are summarized below. Power output was excellent and distortion was generally very low into higher impedances. What distortion there is was generally benign and very likely inaudible. The E9 didn’t suffer any of the AMB Mini3’s virtual ground woes. The main problem is an output impedance of 10 ohms on the 1/4” jack and 43 ohms on the 3.5 mm mini jack. Using the small jack will almost certainly affect frequency response and/or bass performance with many headphones—especially balanced armature IEMs (Shure, Etymotic Ultimate Ears, etc.). While using the bigger jack should work reasonably well for full size higher impedance headphones—80 ohms or higher. There was also more noise and channel balance error than I was hoping for. Here’s a comparison with the Mini3, E7 and E5:
|Measurement||FiiO E9||AMB Mini3||Fiio E7||FiiO E5|
|Frequency Response||+/- 0.1 dB Excellent||+/- 0.1 dB Excellent||+/- 0.1 dB Excellent||+/- 0.1 dB Excellent|
|THD 1 Khz 150 Ohms||0.005% Excellent||0.002% Excellent||0.003% Excellent||0.005% Excellent|
|THD 1 Khz 15 Ohms||0.037% Good||0.017% Good||0.03% Good||0.012% Good|
|THD 20 hz 15 Ohms||0.05% Good||0.01% Very Good||0.09% Good||0.6% Poor|
|THD 20 Khz 15 Ohms||0.003% Excellent||0.45% Poor||0.06% Excellent||0.05% Excellent|
|IMD CCIF||0.05% Good||0.043% Fair||0.03% Good||Not Measured|
|IMD SMPTE||0.002% Excellent||0.009% Very Good||0.008% Excellent||0.006% Excellent|
|Noise (ref 400 mV)||-88 dB Fair||-94 dB Excellent||-91 dB Good||-86 dB Fair|
|Max Output 15 Ohms||1067 mW Excellent||104 mW Excellent||113 mW Excellent||108 mW Excellent|
|Max Output 150 Ohms||317 mW Excellent||38 mW Good||23 mW Good||22 mW Good|
|Output Impedance||10 Ohms Fair||0.9 Ohms Very Good||0.13 Ohms Excellent||0.7 Ohms Very Good|
|Crosstalk 15 Ohms||63 dB Very Good||40 dB Poor||63 dB Very Good||46 dB Fair|
|Channel Balance Error||1.8 dB Fair||1.14 dB Fair||0.2 dB Excellent||Sample Problem|
- Plenty of power to drive most any headphone
- Very low distortion into higher impedances
- Dual gain modes to better match different headphones
- Protection relay
- Nice build quality
- Versatile companion & dock for FiiO E7
- Output impedance is high enough to seriously effect IEM headphones
- 3.5mm mini jack output handicapped by 43 ohm output impedance
- Some volume tracking channel balance error
- Mediocre noise levels even in low gain mode
- Mediocre low frequency distortion into low impedance loads
BOTTOM LINE: I had high hopes for the FiiO E9 and while it measured mostly very well I was a bit disappointed. The 10 ohm output impedance is less than ideal while the distortion and noise measurements were not as good as I was expecting. As explained in my article on impedance the 10 ohm output will degrade the sound with some headphones. But if you have high impedance headphones (80 ohms or higher) the E9 is worth considering—especially if you already have a FiiO E7. Besides the convenience of the dock, you can always use the E7 to drive any low impedance headphones you might have (something it does better than the E9). The E9 is otherwise well engineered and the price is attractive for what you get.
DESIGN COMMENTS: The Micca Store claims the E9 uses TI OPA2134 op amps and the TI TPA6120 for the output buffer. The OPA2134 is optimized for audio use and has impressive specs. The TPA6120 is marketed for headphone applications and offers high power, high current, extremely low distortion and wide bandwidth. It’s a stereo IC with both channels on a single surface mount IC with a large heat sink pad on the bottom. It’s used in a number of high-end headphone products. The output is relay switched and hopefully there’s a DC protection circuit for the headphones. The FiiO description says it merely “protects amplifier output” so it’s possible the relay only eliminates turn on/off noises and protects the TPA6120 rather than the headphones. It appears to be AC coupled so I couldn’t simulate a failure by running DC into it.
NO VIRTUAL GROUND: The E9 runs from a single voltage 15 volt DC power supply. It draws about 160 mA with no load and 300+ mA when driving a sine wave at high levels into 15 ohms. Both the input and output grounds measured zero volts DC with respect to the power 15 volt power supply. FiiO says it uses an “internal power converter module” and that appears to be true. As can be seen in the Slew Rate test, it’s capable of about 20 volts peak-to-peak output which would be impossible using a virtual ground and a 15 volt supply. This is a much better design choice than the AMB method of using a third channel or virtual ground. And it’s impressive in such an inexpensive product (bipolar DC-DC converters are relatively expensive).
DC OFFSET: The DC offset measured 1.5 mV in one channel and 3.0 mV in the other—both plenty low.
FREQUENCY RESPONSE: The frequency response was near ideal being 1 dB down at 7 hz and 48 Khz. It was essentially the same with 15, 150 and 600 ohms and you can also see the channels are closely matched to 0.08 dB with the volume control all the way up. Note the entire graph scale is only +/- 3 dB.
POWER OUTPUT: The E9 has enough power to drive most any pair of dynamic headphones. FiiO claims 80 mW into 600 ohms and the E9 measured 84 mW at 1 % THD (clipping) meeting the spec. They also claim 1 watt into 16 ohms, and I measured 1.067 watts. The rise at low levels is mostly noise as it’s a THD+N measurement and the noise remains constant and dominates the measurements at low levels. Note these measurements are using the large 1/4 jack. From the 3.5mm mini jack, maximum output into 15 ohms was only about 250 mW:
OUTPUT IMPEDANCE SMALL JACK: As Maverickronin pointed out in the comments, the output impedance depends on which jack you use. It measures 10 ohms at the large 1/4” jack and 43 ohms at the 3.5mm mini jack. Either is too high for use with most balanced armature IEM headphones. To understand why, see my article on impedance. Here’s the frequency response, from the 3.5mm jack (43 ohm output) of the E9 into my Ultimate Ears IEMs referenced to 1 Khz. Note there is about 12 dB of total variation:
OUTPUT IMPEDANCE LARGE JACK: Here’s the same plot as above but via the large jack with a 10 ohm (vs 43 ohm) output impedance. Note there’s still 6 dB of variation which is plainly audible. I discuss this issue more at the end of the article below. With normal full size cans like the Sennhesier HD600, there would only be a very slight rise (less than 1 dB) around 100 hz:
THD+N 150 OHMS: Here’s the wideband spectrum measurement showing both channels at 400 mV into 150 ohms. The distortion is impressively low and the channels are well matched indicating a reasonably symmetrical PCB layout:
THD+N 15 OHMS: Into 15 ohms the distortion rises considerably. Some rise is to be expected, and the good news is the higher harmonics are all well below –80 dB which is my “rule of thumb” for potential audibility. Only the generally benign 2nd harmonic is above the threshold. This is a respectable result but I expected better from the TPA6120. I have to wonder if there are not some internal power supply issues or poor grounding causing such a rise into 15 ohms:
THD RESIDUAL 150 OHMS: Into an easy load the THD residual is just the 2nd and 3rd harmonics and noise:
THD RESIDUAL 15 OHMS: Into the more challenging 15 ohm load, you can see the strong 2nd harmonic and little else. There’s almost no sign of crossover distortion here which means TI did their homework with the TPA6120:
THD 20 hz 150 OHMS: Low frequency distortion with an easy load was around –100 dB which is excellent:
THD 20 hz 15 OHMS: This is an interesting result. At 20 Khz (where there’s generally less feedback available normally resulting in higher distortion) the E9 does much better into 15 ohms than it does as 20 hz. This implies a power supply problem. This test only draws 38 mA of peak current but the current is drawn for a much longer duration and apparently that’s causing problems with the switching DC-DC power supply. The good news is the distortion, as at 1 Khz, is mostly the benign 2nd harmonic. All the other harmonics are below the –80 dB threshold. Plus human hearing is much less sensitive to low frequency distortion. So it’s safe to say this is inaudible but the TPA6120 is capable of much better:
THD vs FREQUENCY: Below is the THD vs frequency at 400 mV with a measurement bandwidth of 22 Khz (hence the drop above 10 Khz). The yellow plot into 150 ohms shows impressively low distortion across the band. The blue trace, however, shows rising distortion as seen earlier in the 20 hz test above. This is likely a power supply problem:
IMD CCIF: This is interesting as the number by itself doesn’t tell the whole story. The 0.055% number is respectable but slightly worse than the AMB Mini3’s 0.044%. But note the odd-order sidebands near the 19 Khz and 20 Khz signals are at –88 dB which is excellent. With the Mini3 these much more audible odd-order distortion products were about 30 dB higher reaching almost –50 dB. The benign 2nd harmonic shows up as the 1 Khz difference signal and dominates with the E9. Even though the percentage is higher, this is much better overall performance than the Mini3 shown in the 2nd graph below. I’m only showing the 15 ohm result here as the 150 ohm result was excellent.
SMTPE IMD: The SMPTE IMD into even 15 ohms was very low and the only obvious issue is the same benign 2nd harmonic seen earlier:
INTERCHANNEL IMD: I ran a new test for the first time on the AMB Mini3 driving each channel to 1 volt but at different frequencies. The Mini3 performed rather poorly creating a dense “forest” of complex distortion products. I decided to repeat the same test on the E9 which uses a proper split power supply instead of a virtual ground. There’s the expected crosstalk at 300 hz while the rest of the spectrum is relatively clean (everything is well below –80 dB) compared to the Mini3:
CHANNEL SEPARATION (CROSSTALK): Into 150 ohms the crosstalk was an excellent 80 dB across most of the range and it degraded to about 63 dB into a 15 ohm load which is still very respectable performance. Both are a huge 20 dB better than the Mini3:
NOISE: The noise performance was a bit disappointing even in the low gain mode. As noted in the first part of the review, the noise is higher at half volume than full volume. This is likely due to Johnson Noise from the volume potentiometer which is worst at half volume (assuming a low impedance source). So first I measured at half volume referenced to my usual 400 mV:
Then I measured at full volume and noted about a 3 dB improvement in the hiss portion of the noise but the hum is about 10 dB worse. It’s interesting that 3 dB was the difference with my SuperFi IEM’s between plainly audible hiss and barely audible hiss. The hum is stray pickup and moving the E9 around didn’t seem to help much. But at –100 dB, it’s safely inaudible.
CHANNEL BALANCE & VOLUME TRACKING: Here the top (white) trace is max volume followed by 3 O’clock, 2 O’clock, 1 O’clock, and 12 O’clock top to bottom. You can see at 12 O’clock there’s about 0.5 dB channel balance error and the total range covered by the top half of the volume control is a bit over 8 dB:
Here’s the lower range of the volume control from 12 O’clock down to 9 O’clock covering another 30+ dB of range:
HIGH GAIN: The 10 ohm output impedance makes the gain somewhat load dependent . At max volume in the high gain mode into 150 ohms it’s close to the FiiO specified 18 dB:
LOW GAIN: At the low gain setting into 150 ohms it’s about 10 dB (into 150 ohms the gain switch boosts/cuts the gain by about 7 or 8 dB):
PHASE: The phase performance, as expected from the frequency response, isn’t perfect but 7 degrees error at 10 Khz is respectable. It’s really 180 degrees out of phase as the E9 inverts absolute polarity:
INPUT CLIPPING: The FiiO E7 suffered from “premature” input clipping at 1.25 V RMS due to its battery powered design. The E9 has a more typical clipping voltage of about 2.1 volts as seen below. Here the volume is turned down for the standard 400 mV of output and the input voltage is shown on the right. This allows typical 2 volt consumer line level sources to be used with the E9 without any problems:
SQUARE WAVE PERFORMANCE: The slightly rolled off high frequencies caused the expected rounding of a 10 Khz square wave but the good news is even driving a reactive load (Sennheiser CX300 headphones in this case) there’s no sign of any ringing, overshoot, or other instability. The 10 ohm resistor in series with the output isolates the reactive load and this is a very stable result. Note it’s inverting absolute polarity as mentioned above under Phase:
SLEW RATE: The ultimate slew rate was plenty fast at about 7 volts per microsecond. The E9 only needs about 1.4 V/uS to not be slew limited. FiiO probably chose to limit the slew rate (the TPA6120 is spec’d at 1300 V/uS!) and, for stability reasons, I would do the same. You can also see the E9 is producing almost 20 volts peak-to-peak as further proof it’s not using a rail splitter, third channel or virtual ground (which would only allow about 14 volts peak-to-peak from the 15 volt supply):
OUTPUT IMPEDANCE COMMENTS: The extremely wide bandwidth of the TPA6120 makes it inherently less stable. Unfortunately, TI made it much faster than it needs to be for audio use and it also is not a pure buffer—it has gain. For these reasons, it’s easy to understand why TI recommends a 10 ohm series output resistor to isolate reactive loads. The National LME49600 and TI/Burr-Brown BUF634 are competing parts that do not require an output resistor for stability. The TPA6120, however, is significantly cheaper on a per-channel basis so I can understand why FiiO chose it for a $129 amp. But the 10 ohm output resistor is a bit of a “band-aid”.
TECH SECTION SUMMARY: In many ways the E9 measured very well. The distortion is impressively low—especially into higher impedance loads. But I do have some concerns:
- The biggest disappointment was the 10 ohm output impedance. As I’ve shown above, even 10 ohms causes 6 dB of frequency response variation with balanced armature IEM headphones. Using the 3.5mm jack is even worse with its 43 ohm output impedance. 10 ohms should not cause any serious problems with normal dynamic headphones rated 80 ohms and higher. See my article on impedance.
- I’m guessing FiiO added the extra 33 ohms to the mini jack to help avoid accidental headphone damage with low impedance headphones. But it seriously compromises the performance of the amp. And, ironically, nearly all balanced armature IEMs have a 3.5mm plug but you very much want to use the E9’s big jack with these headphones (or even better, the FiiO E5, E7 or another amp with near zero output impedance). A better solution would have been for FiiO to further lower the E9’s gain for the low setting to limit the output to a reasonable level for low impedance headphones.
- If you have an E7 you can always use it to drive low impedance headphones and use the E9 only for high impedance cans.
- The noise performance, even in the low gain mode, was disappointing—especially at half volume. The measurements were consistent with what I heard with my sensitive IEMs.
- There was some channel balance error around 0.5 - 1 dB at several settings and a max of around 3 dB at –50 dB. This isn’t awful, but it’s marginal if you plan to use really sensitive headphones and listen at very low volumes.
- The low frequency distortion into 15 ohms was much higher than the the TPA6120 is capable of. It’s not alarmingly high, but it’s a sign a corner or two was cut or someone got sloppy with the design. This is a classic case of getting nowhere near the datasheet performance from the OPA2134 and TPA6120 due to implementation issues.