FiiO E7 USB DAC & Amp

RELATIVE BARGAIN? Since measuring the ill-fated NuForce uDAC-2 many have asked me to recommend a similarly priced DAC with fewer flaws. I suspected the FiiO E7 might be a good choice but a bench test was needed.

RAISING THE BAR: This review is the first with several new measurements. I’ve added some tests and improved others to be more consistent with other reviews like John Atkinson’s well respected work for Stereophile. I’ve gone even further beyond the usual difficult to compare RMAA results.

BY ANY OTHER NAME… If a serious headphone-loving audiophile encountered the E7 with no brand name they might easily suspect a stealthy black Apple product. It’s similarly sexy, solid, and hewn largely from black anodized aluminum with impressive tolerances and finishes—much like an iPhone 4. Most wouldn’t blink if you told them it was $300. But, the thing is, its only $99—or even less on eBay. Build quality, however, isn’t everything. The NuForce also made a decent first impression before that train derailed on the test bench.

JUST THE FACTS: The E7 is a 16 bit 48 Khz USB DAC with a line input for analog headphone amp duty. It’s powered either by the USB port or an internal Li-Ion battery. The sundry details have been well hashed in countless other reviews but nearly all omit meaningful measurements. So I’ll just hit the highlights and focus on the measured performance. More info: E7 features and specs.

GETTING PHYSICAL: The footprint is smaller than Apple’s iconic phone but the E7 is more Motorola Droid-like in thickness. The buttons are high quality and the display is easy to read when on and disappears into the stealthy black face when off. It also comes with a nifty silicone protective shell, all the cables, and a soft carrying case—more than you’d expect for $99, or for that matter, than you get with an iPod.

PLUG-AND-PLAY... REALLY: Like the uDAC-2, the FiiO is a Class 1 USB Audio device meaning no drivers for Windows, OS X or (in theory) the right Linux distros are required. And it worked first try with Win 7 and XP.

VOLUME CONTROL: The upper two buttons on the left control the volume in 60 gradual steps with the E7 never leaving you wanting a step in-between. And the balance between the channels remains anchored dead center over the entire range.

Nothing’s ever perfect, however. The first time you press a volume button nothing happens. You have to press it again to change the volume similar to many portable players where the first press “wakes up” the display but doesn’t do anything until you press again. The weird thing is the display is already on. FiiO could easily fix this with a firmware tweak. And, just so you know, the master volume in Windows is forced to maximum when using the E7 as your primary audio device. You can still use the volume control in your player (i.e. Windows Media Player, Foobar, Media Monkey, etc.) but FiiO apparently wants the PC delivering all 16 bits of resolution for maximum audio quality (reducing the volume in windows can also reduce the bit resolution). I didn’t try it on OS X or Linux but I’m sure you can find those details elsewhere.

MENUS: The lower two buttons on the left access the setup menus. You can enable the sleep timer, turn USB battery charging on or off, set a preferred maximum volume setting, enable a button “lock” (which also “sleeps” the display), and enable one of three bass EQ options. It’s all intuitive and the max volume is an especially nice feature if you have really sensitive headphones and/or want to save your hearing.

BATTERY & POWER: I don’t know how accurate the battery meter is, but in all my testing and listening over many hours while running from battery (no USB connected), it only dropped one bar. It’s rated for 80 hours and I can believe it. The battery charges via USB using a computer, inexpensive USB wall charger, or other power source. In fact, using a charger turns the E7 into a desktop amp/DAC you never have to charge or worry about turning off. FiiO gets extra points for using a regular mini-USB connector for both the PC connection and charging. If you lose the supplied cable you probably already have a spare. I didn’t try, but the E7 appears to stay on forever unless you set the sleep timer. If it’s running from battery, it might run itself dead. FiiO could easily fix this in the firmware.

DISPLAY: How many small DAC’s even have a display? Isn’t that more like a $2000 DAC feature? The organic LED (OLED) display looks much like the one in the Sansa Clip right down to the yellow section at the top. Without using the key lock feature, it’s on all the time. As others, like dfkt, have pointed out in their E7 reviews, OLED displays can be prone to eventual “burn in” where the most used pixels lose brightness. This could be a minor problem if you leave the E7 display on for months on end. The key lock option is a somewhat awkward solution if you’re worried about it. This would be a another simple firmware fix if anyone at FiiO is listening?

SUM OF PARTS = ?: On paper the E7 looks impressive for $99. it sports an audiophile-grade Wolfson DAC, Analog Devices op amp, Li-Ion battery, display, adjustable bass EQ, line input, precision volume adjustment, and Apple-like industrial design. In short, there’s a a lot more hardware here than in the uDAC-2. Yet, somehow, the E7 sells for about 30% less.

FiiO E9 UPGRADE: If you have you have some power hungry high impedance cans, FiiO has an unusual solution: The $129 E9 desktop amp. The E7 cleverly “docks” to the E9 with the E7 handling DAC duties while the E9 has the cojones to drive just about any headphone. It also powers and charges the E7. So you get a portable DAC/Amp, and a high output desktop/home amp/DAC, all for less $$ than many portable amps alone. I haven’t yet tested the E9 but keep reading for the E7’s end of the bargain.

SUBJECTIVE SOUND QUALITY: Before making any measurements I spent some time listening to the E7 as both a headphone amp (driven by my Benchmark DAC1) and as a USB headphone DAC connected to a PC. There were no obvious flaws or red flags like obvious hiss, channel balance problems, rolled off frequency response or obvious distortion.

VOLUME, POWER, & GAIN (updated 5/22): I did max the volume playing some tracks with my Beyerdynamic DT 770 Pro 80’s (which are 80 ohms and fairly inefficient) and 250 ohm Sennheisers. So the E7 does have its limits. The overall maximum gain is well chosen for typical headphones in the USB mode but it’s a unusually low for high impedance phones when using the analog input. This means if your source doesn’t have a lot of output the E7 may not produce full output from the line input. The electronic volume control is quiet when it’s stepping up or down and the step size works well. The maximum output is similar to the NuForce but the max gain in USB mode is lower making more of the volume range usable.

HISS AND NOISE: Even with my uber-sensitive SuperFi 5 Pro headphones the FiiO had only slight audible hiss at any realistic volume setting. It’s notably quieter than the UCA202 and with more normal headphones it’s close to silent at any volume setting.

BASS EQ: The Bass EQ was more useful than some might imagine. When using headphones that are bass shy, like my Eymotic IEMs, I found the Bass EQ “1” setting added some useful punch without any real downside—i.e. it didn’t thicken the lower midrange, etc. I wasn’t as crazy about the “2” and “3” settings which are more heavy handed.

MEASUREMENT SUMMARY (revised 5/22): Overall the E7 measured well as an under $200 DAC and even better via the line input as a headphone amp. The analog amp was clean with low zero distortion into 150 ohms, very flat extended response, an unusually low output impedance, and enough output for most headphones likely to be used with a portable amp. After hours of torment on the bench, I could elicit only some minor bad behavior: The E7 has limited output into high impedance loads which could be an issue for certain power hungry cans. And this might be made worse as the line input only has 4 dB of gain (USB is fine). There’s a possibility of the line input clipping with some fixed level home sources. And there was some noise well above the audible range that, despite being completely inaudible, some might object to. Here’s a summary (for all the details and graphs, see the Tech Section):

Measurement	Fiio E7	Nuforce  uDAC-2
Frequency Response 20hz - 20 Khz	+/- 0.1 dB Excellent	+/- 0.1 dB Excellent
THD 1 Khz 0 dBFS USB	0.05% Good	0.67% Poor
THD 1 Khz 150 Ohms 400 mV Line In	0.003% Excellent	Not Applicable
THD 1 Khz 150 Ohms 400 mV USB -3 dBFS	0.03% Good	Not Measured
THD 1 Khz 15 Ohms 400 mV USB -3 dBFS	0.03% Good	0.05% Good
THD 20 hz 15 Ohms 400 mV	0.09% Good	0.06% Good
THD 20 Khz 15 Ohms 400 mV	0.06% Excellent	1.0% Poor
IMD CCIF USB	0.03% Good	Not Measured
IMD SMPTE	0.008% Excellent	0.08% Fair
Noise (relative to 400 mV A wtd)	-91 dB Good	-94 dB Excellent
Max Output 15 Ohms	113 mW Excellent	47 mW Very Good
Max Output 150 Ohms	23 mW Fair	32 mW Good
Output Impedance	0.13 Ohms Excellent	6.0 Ohms Fair
Crosstalk	63 dB Very Good	47 dB Fair
Channel Balance Error Typical Vol	0.1 dB Excellent	1.4 dB Poor
Channel Balance Error Low Vol	0.2 dB Excellent	10.3 db Poor
Jitter USB 16/44 Jtest	< -110 dB Very Good	< -105 dB Fair
Square Wave	Very Good	Very Good

FiiO vs NuFORCE: This really isn’t a fair fight. The E7 has much less distortion and channel balance error along with a much lower output impedance, lower jitter, more output power with typical headphones, optional bass EQ, a line input, portable battery operation, more accessories, and costs less. There are not many well recognized USB headphone DACs under $150. I mention some of the other options in the NuForce review, but so far I’ve only tested these two at the E7 is the clear winner overall.

BOTTOM LINE: The FiiO E7 seems like it should cost more from the moment you lay your eyes and hands on it. And that impression remains when evaluating the performance. The E7 is hard to beat for the price unless you have power hungry full size high impedance cans. It’s also makes a respectable battery powered amp. It’s well worth $99.

FIRST CLASS:

• Improves on the sound of many PCs, laptops and portables with weak headphone outputs
• Respectable measurements and overall performance for the price
• Fine volume steps with excellent channel balance
• Apple-like build quality
• Line input and optional desktop amp/dock (E9) for added flexibility
• Useful accessories included
• Very long battery life

ECONOMY:

• Might not have enough output for some high impedance headphones
• Line input has only 4 dB gain
• Display on all the time
• No auto shut off to save battery
• Disables master volume control in Windows
• Line input can clip with some fixed output home gear
• Headphone amplifier IC is the weak link in an otherwise solid signal path
  

---

TECH SECTION:

NEW AND IMPROVED: Because I’ve made a lot of changes, and the E7 is really an amp and a DAC, this Tech Section is longer than usual. I’ve added new measurements, improved a few others, and made the entire blog slightly wider to allow displaying the dScope screen shots in their native resolution—no more click-to-zoom disruptions. While I’ve tried to explain some of the background behind the changes, stay tuned for an upcoming article that discusses all the following measurements in more detail and documents the baseline capabilities of my instruments. I’ve also been making revisions to the article on how I test:

• Testing Methods

HARDWARE DESIGN: At the heart of the E7 is the well respected Wolfson WM8740 DAC. That’s a relatively high-end part to find in a $99 portable product. The WM8740 lacks USB support so another relatively expensive chip is required for the USB interface—the ubiquitous TI PCM2706 which, in this application, only converts USB to I2S and its analog DAC output is unused. The E7’s analog section is reportedly from another boutique chip company--the low noise 0.0006% THD Analog Devices AD8692. Output duties are handled by the TI TPA6130 which also handles the electronic volume function. When you factor in the expensive Apple-like industrial design, high build quality, graphical display, battery, accessories, etc, it’s hard to imagine selling it for much of a profit at $99.

WINDOWS VOLUME MYSTERY: Does anyone know why the PCM2706 as used here defeats the Windows master volume control but doesn’t when used in other products? I haven’t looked very deep into this but it does seem rather odd. The PCM2706 isn’t exactly a programmable microcontroller. Is the microcontroller in the E7 also talking to the Windows driver? That would require splitting the endpoints between 2 devices which would be a bit odd in this application. Or perhaps the E7 doesn’t really use the PCM2706?

ANALOG FREQUENCY RESPONSE: Via the Line Input the E7 has excellent frequency response and very precise channel balance. It’s within +/- 0.1 dB from 20 hz – 20 Khz. Even at 80 Khz it’s only down 1.5 dB and about 1 dB at 5 hz. The channel balance is also excellent (the lines for each channel, for a given load, are difficult to tell apart) and it changes very little with loading. There’s no meaningful change dropping from 150 ohms (pink/red) to 15 ohms (blue/yellow). The output is obviously direct coupled with solid response down to an ear drum rattling 5 hz:

USB FREQUENCY RESPONSE: Via USB the channel balance error rises from 0.08 dB to 0.22 dB likely due to a slight gain error between the DAC channels. Anything less than 1 dB is generally considered inaudible so this is still very good performance. The graph covers a more narrow frequency range due to the limitations of 44 Khz CD audio. The response has a few more ripples but is still very flat at +/- 0.2 dB from 20 hz to 20 Khz. Unlike many DACs, it doesn’t fall off a cliff as it approaches 20 Khz—showing superior filtering in the Wolfson DAC chip:

ANALOG THD: The distortion via the line input at 1 Khz with a 150 ohm load and my usual headphone reference level of 400 mV (which is 0 dBr for these tests) was nil. The THD+N is very low around 0.003% These are audiophile-class numbers and, unlike in previous reviews, both channels are shown (in blue and yellow). Note the distortion is mostly 2nd harmonic which is considered the least objectionable sort of distortion but when the numbers are this low, it doesn’t matter:

This is a new measurement showing the residual distortion (in blue) from the spectrum above with the same 150 ohm load in the time domain rather than the frequency domain. The yellow trace is the 1 Khz output scaled in volts: 400 mV RMS is +/- 560 mV peak. The blue trace is the output with the 1 Khz signal removed by the dScope in real time so you’re left with only what the E7 is adding to the signal. Notice the blue scale on the right is in microvolts. The peak distortion is only 0.000004 volts from a 0.4 volt signal and it’s mostly noise with no obvious crossover distortion. This is a test run by Stereophile and yet another example of something you can’t do with RMAA or any soundcard-based test setup I know of:

Here’s the E7 driving a much more challenging 15 ohm load showing a more “closeup” view of the spectrum to only 20 Khz. The distortion rises to a still respectable 0.012% and notice the output level barely drops. But the TPA6130 is having more trouble here than with 150 ohms:

Here’s the residual distortion into 15 ohms which now shows signal related components including some apparent crossover distortion:

USB THD+N: Here’s the same test as above except using a – 3 dBFS digital signal via USB. There’s more 2nd and 3rd harmonic distortion now and you can also see artifacts of the 44 Khz sampling frequency above the audio band. The Wolfson DAC and FiiO low pass filter are doing an impressive job controlling out-of-band noise as it’s mostly below –120 dB. The THD+N, however, increases by about a factor of ten to 0.032%. This is still plenty respectable, and very likely completely inaudible, but the Wolfson DAC is capable of better. The good news is it’s mostly 2nd harmonic which is, by far, the least audible:

Here’s the residual USB distortion from above in blue. You can see it’s mainly the strong 2nd and 3rd harmonic seen in the spectrum above with no other obvious problems:

FULL SCALE USB THD+N: Some DACs show increased distortion reproducing a full scale 0 dBFS signal. This test indicates how well the DAC can handle clipped signals. A lot of pop recordings are intentionally clipped and frequently hit 0 dBFS (presumably to make them sound “louder”). The NuForce uDAC-2 had dramatically higher distortion on this test as, due to a NuForce design flaw, the DAC itself was driven into analog clipping. With the E7 the distortion rises only slightly from about 0.032% at –3 dBFS (see above) to 0.053% in the worst (blue) channel. The 0.2 dB DAC gain error is likely responsible for the THD difference between the channels. The output level was referenced to approximately the same ~400 mV to keep the analog distortion the same. This is a very good performance:

ANALOG THD 20 Khz 15 OHMS: This is a punishing test  (also sometimes called “THD-20”) as the negative feedback in most amplifiers starts dropping well before 20 Khz and 15 ohms is a very challenging load. With less negative feedback available to correct errors, the distortion rises—sometimes dramatically. This test also often reveals poor high frequency stability, feedback loop design errors and PCB layout mistakes creating undesirable parasitic coupling or ground problems. The NuForce uDAC-2 was an epic FAIL on this test with about 1.0% THD+N at 20 Khz. The E7 is twenty times cleaner at only 0.057%. The measurement bandwidth for this test runs out to 80 Khz to include the 2nd and 3rd harmonics of the 20 Khz signal. Also note the frequency response is only down 0.1 dB vs the 1 Khz reference which is also excellent performance:

USB THD 20 Khz 150 OHMS: Same test as above but via the USB digital input. The distortion is lower because it’s into 150 ohms instead of 15 ohms but the idea is to show the DAC itself doesn’t contribute any significant high frequency distortion components with the measurement extending out to 80 Khz:

ANALOG THD 20 hz 15 OHMS: Measuring THD at 20 hz into a difficult load stresses an amplifier in different ways. It requires sustained amounts of current from the power supply and stresses any electrolytic coupling (DC blocking) capacitors in the signal path. It can even reveal thermal distortion in the output devices. The distortion of the E7 rises from 0.012% at 1 khz into 15 ohms to about 0.09% at 20 hz. But the good news is it’s almost all 2nd harmonic which is the least audible. The third harmonic is down over 90 dB. The ear is also less sensitive to low frequency distortion. This is likely related to the negative rail power supply and/or power supply filtering in the E7. The FiiO E5 also suffers from rising low frequency distortion but it’s about ten times higher. Overall this is respectable performance for a battery powered device generating its own power supply and nearly anyone would consider this inaudible:

USB THD 20 hz 150 OHMS: Again, the distortion is much lower into the easier load and the DAC isn’t contributing any unexpected distortion so this is a good result. You can see the 44 Khz sampling artifacts:

ANALOG THD+N vs FREQUENCY: Here’s the THD vs frequency for both channels into both loads. The lower set of traces are the left and right channels into 150 ohms and the upper set into 15 ohms (the caption has them backwards). You can see the rising low frequency distortion mentioned above. The measurement bandwidth here is out to 22 Khz which explains why the 15 ohm distortion falls above 10 Khz. But it’s still extremely impressive the distortion stays so low at high frequencies (see THD 20 Khz above):

USB THD+N vs FREQUENCY: This is the same as above but feeding a digital signal from the dScope to the DAC and only at 150 ohms. Again the result is impressively flat with frequency and the low frequency rise above up to 0.02% at 20 hz is masked by the 0.03% THD of the DAC. The dip before the rise is due to the analyzer cutting off harmonics that are above the audible range. This is very good performance all the way around:

ANALOG IMD CCIF: This is another new test that’s both challenging and revealing of particularly audible kinds of distortion. Stereophile performs this measurement with an Audio Precision analyzer and it cannot be done with RMAA. Two high level high frequency tones at 19 Khz and 20 Khz interact with any non-linearities creating distortion products at multiples of the difference between the two frequencies. Interestingly, the more friendly even-order distortion products show up at 1 Khz, 2 Khz, etc, while the generally more objectionable odd-order distortion products show up as “side bands” to the high frequency signals (i.e. at 18 Khz, 21 Khz, etc.). Here the E7 does well at 0.003% overall but it’s worth noting the two sidebands at about –74 dB which are some slight cause for concern and it turns out they’re related to the punishing 15 ohm load. Again the TPA6130 is showing its limitations:

USB IMD CCIF: This is a repeat of above except into 150 ohms via the digital USB input. Note the distortion product at 1000 hz is 20 dB higher than above. This is from the higher 2nd harmonic distortion of the DAC (and the main reason for the increase from 0.003% to almost 0.03%). But also note the sidebands marked above have dropped from –74 dB to around –85 dB. This demonstrates these are caused by the more punishing 15 ohm load used above. The DAC itself is mostly free of problems here besides the relatively benign 2nd order spike at 1 Khz:

SMPTE IMD: This is the SMTPE version of IMD using 60 hz and 7 Khz tones at different levels. Here the analog and USB performance were very similar so I’m only showing the analog result. The only area of note is the 2nd harmonic of the 60 hz signal at –72 dB. This is a THD component, not an IMD distortion product and is due to the rising low frequency distortion into 15 ohms noted earlier. This is not included in the 0.0076% number because it’s not at a frequency related to the interaction of the 60 hz and 7 Khz tones. The calculation of true SMPTE-DIN IMD is relatively complex and something RMAA appears to get consistently wrong. The IMD products of interest are all clustered around the 7 Khz signal and impressively below 90 dB. This is very good performance:

OUTPUT IMPEDANCE: This is a critical measurement for anything designed to drive headphones. As discussed in my article on impedance the lower the better for most applications. As show below, the E7 produced 402.0 mV into essentially no load (100,000 ohms). Into 15 ohms (see above) it managed 398.6 mV which calculates out to a very impressive 0.13 ohms:

MAXIMUM LINE INPUT LEVEL: The E7’s input clips around 1.25 volts RMS. That should work fine with the line output on an iPod, PC, or anything with an adjustable output. But some home audio equipment, like home CD players, can have fixed outputs producing around 2 volts RMS at 0 dBFS. Such gear, if you can’t lower the output, would cause some peaks to be clipped by the E7. 1.25 volts is 3.5 v p-p. With the E7 running on a 3.7 volt battery it’s easy to see why it clips at 3.5 volts. Fixing this correctly would require a true split rail supply for all the analog circuitry—not just the headphone power amp in the TPA6130.

MAXIMUM OUTPUT & THD vs OUTPUT: I’ve improved this test to provide much more information. And, again, this is something you just can’t do with RMAA or any soundcard-based measurement system I know of. The dScope plots the THD+N along the left over the entire range of output power along the “X” axis. Because it’s a THD+Noise measurement, noise will dominate at very low levels on the left and clipping distortion dominants on the right hand side. This is directly comparable to the Audio Precision THD+N sweep Stereophile uses.

When I first wrote this review I didn’t realize the E7 is using the TI TPA6130 as the output amplifier but someone was kind enough to point it out in the comments below. I ran this test running from the internal 3.7 volt battery using the line input and the TI datasheet shows the TPA6130 produces about the same output from 3.7 or 5 volts with a 20 ohm or higher impedance load (see graph to right). The graph allows estimating the power into various load impedances. The TPA6130 uses a charge pump to generate a negative power supply that “mirrors” the positive supply. This doubles the total power supply voltage and eliminates the need for distortion-inducing output capacitors or the added complications of a rail splitter.

Even on battery power, the E7 manages about 113 mW into 15 ohms which handily beats the uDAC-2 even on USB power. Connected to USB, according to the graph, 140 mW could be expected. And into 150 ohms it still manages a respectable 23 mW. The 15 ohm performance is due to the TPA6130’s limited output current capability . In summary, the E7 should do fine with all but power hungry high impedance headphones:

GAIN (added 5/22): With a full scale USB digital source the E7 has about 12 dB of added gain beyond 400 mV. It has no problem reaching clipping via USB. With an analog input, however, it only has about 4 dB of gain at the maximum volume setting (Volume 50 is 0 dB or unity gain and Volume 60 is the maximum). 4 dB is about 1.6 times. So to reach 1.3 volts out it needs about 0.8 volt input and that’s more than a lot of portable players can manage. With a more typical 0.45 volts maximum input, you’ll get 0.72 volts out. That’s 32 mW into 16 ohms but only 6 mW into 80 ohms and a downright weak 2 mW into 250 ohms. I suspect at least part of the problem here is the volume control is in the headphone chip amp, and any extra gain before the chip amp would just make the input clipping problem mentioned above worse. So perhaps this is the best compromise FiiO could come up with? Still, just to be clear, it’s not a problem if you intend to use the E7 as a USB DAC.

ANALOG INPUT NOISE: I deviate a bit from traditional noise measurements. Signal to noise ratio is often measured against maximum output (sometimes shown as “dBc”) to obtain a more impressive number. It also often changes with the volume control setting. Sometimes the only way to reproduce a manufacture’s lofty spec is to crank the volume up and push the device into clipping to obtain the reference level, then they turn the volume all the way down with no input and take the noise measurement. That yields a completely unrealistic number. Even leaving the volume alone and using clipping as the reference makes it hard to compare between different devices. Instead I use a typical fixed volume control setting and a more typical reference level. For headphone outputs I use 400 mV RMS. The result is usually at least a few dB less favorable than if I measured the unrealistic way. The benefit is you can compare my noise measurements across products and have a much better idea of how much noise there will be in typical real-world use.

The E7’s noise measurements require some explanation. The unweighted measurement is oddly marred by noise at 90 hz and 180 hz. I’m guessing these are created by the TPA6130’s charge pump (generating the negative supply rail). The good news is the ear isn’t very sensitive at low frequencies and, try as I might, I couldn’t hear the 90/180 hz noise components even with my most sensitive headphones. This is reflected in the respectable A-Weighted number of nearly –91 dB. The weighting is an industry standard adjusting the measurement to the ear’s sensitivity at various frequencies. It’s typically what you see in manufacture’s S/N specs (even if they don’t disclose the weighting). By comparison the FiiO E5, which is also respectably quiet, measured 3 db worse at –88 dB A-Weighted. It’s worth noting the actual “hiss” portion of the noise spectrum is relatively low. With the volume set to 80% the spectrum above 600 hz is at about –125 dB which is slightly better than the NuForce uDAC-2 and much better than the E5 or UCA202. And, as shown in red, it gets even lower if you turn down the volume to a more typical setting. This is a long way of saying the subjective noise performance of the E7 is better than the following might lead you to believe:

USB NOISE: Here’s the digital version of the above referenced to the same 400 mV using a digital 1 Khz signal at –115 dBFS to keep the DAC from going into mute—that’s the spike you see at 1 Khz. Here, with the analog input disabled (which may remove some gain and hence noise) and the volume set to 35 (determined by 0 dBFS level), the E7 does significantly better than above but the weird components at 90 & 180 hz are still there, just lower in level:

STEREO CROSSTALK: This shows the channel separation with a worst case 15 ohm load. Crosstalk tends to be worse into low impedance loads due to greater power supply interaction, higher modulation effects, and increased crosstalk between PC traces due to higher currents and electromagnetic coupling. The E7 does very well here with > 60 dB across the entire spectrum even up to 80 Khz:

VOLUME TRACKING: The graph below shows both the fine steps of the volume control (about 0.5 dB each) and the tight tracking between the channels (something the uDAC-2 did very poorly with):

Even at the lowest volume settings the E7 tracks very well (slightly better even!). This is where the NuForce went from bad to horrible. Here are the 5 quietest settings and you can see the pairs of traces for each are still very closely matched between the channels:

BASS EQ MEASUREMENTS: Here’s the effect of the three bass EQ settings (four if you include “off” shown in green). The “1” setting, shown in red, is probably the most useful unless you like more boom than thump in your bass:

JITTER: The E7 uses the TI (formerly Burr Brown) PCM2706 to convert USB to I2S. The PCM2706 does a reasonable job on its own with jitter. To quote the spin-meisters at TI: “analog PLL’s with SpAct enable playback with low clock jitter.” The E7 adds the Wolfson DAC’s jitter rejection but it’s all very implementation dependent. Anything that skews, distorts, or adds noise to the internal digital signals and clocks can easily degrade jitter performance. The devil is very much in the details. Please see my jitter article for more.

Periodic jitter is revealed as symmetrical pairs of “sidebands” around the main signal. While “spread” at the base of the main 11,205 hz signal indicates random low frequency jitter. With the E7 the side bands are about 1000 hz apart and impressively below –110 dB. The spread is also relatively minimal and stays below –110 dB. This is significantly better than the NuForce and UCA202. The absolute frequency (clock) accuracy is right on (the dScope adds approximately 0.2 hz to the frequency reading):

DAC LINEARITY: The E7 had virtually perfect linearity at –90 dB as can be seen by the center number below. The THD figure is real and is a good indication of how much distortion is present when the DAC is reconstructing a signal from very few bits and the dither noise. 1% is relatively respectable:

PHASE: Here’s another new measurement. The phase response of the E7 via the line input is within +/- 10 degrees over the audio range and off by only 6 degrees at 10 Khz. This is acceptable if not stellar performance:

USB SQUARE WAVE PERFORMANCE: Here’s the 1 Khz square wave performance with a USB digital input. The yellow trace is the dScope’s conversion of the digital signal, and the blue trace is the E7’s output into Sennheiser CX300 16 ohm headphones. A realistic load, like the CX300’s can reveal instability, ringing, and other problems. There are no surprises here although the “ripple” is somewhat different than I’m used to seeing:

ANALOG SQUARE WAVE PERFORMANCE: The analog square wave response requires some explanation for the noise and rise time—see the RF Noise and Slew Rate sections below. The red trace is the lab-grade 10 Khz square wave input and the upper blue trace is the roughly 1 volt p-p output of the E7 driving Sennheiser CX300 16 ohm headphones as viewed on a 100 Mhz wideband oscilloscope:

RF NOISE EXPLAINED: If I was making the usual soundcard-based, or even dScope/Audio Precision analyzer measurements, you would never even see the noise shown on the square wave above. It’s not something you’ll find in say a Stereophile review but I’m all for telling it like it is. The RF noise has an average value of only about 35 mV and it decreases with lower signal levels and easier loads. The spectral energy is entirely above 400 Khz and hence well above the audio spectrum.

E7 FORENSICS: Because the noise is worse on the negative half of the square wave it’s clearly from the 400 Khz charge pump power supply built into the TPA6130 output amplifier. If it’s any consolation this sort of noise is very common with Class-D and similar amplifiers including some that cost many thousands of dollars and receive very favorable audiophile reviews. The FiiO E5 has a similar noise RF noise signature and likely uses the same TI output amp.

DESIGN CHOICES: I didn’t run this test on the NuForce uDAC-2 but it likely has similar noise as well. There are only four choices open to a designer for a USB (or single cell battery) powered DAC:

• Single 5 Volt Supply – Using only the 5 volt USB supply saves money but limits performance. It limits the output voltage swing to about 1.3 volts RMS, or even worse into more difficult loads. This makes it a poor choice higher impedance headphones. It also requires the headphone output be capacitor coupled with a large and relatively expensive output capacitor to avoid low frequency roll off and increased distortion into low impedance headphones. So you end up with more low frequency roll off and potentially audible distortion. The Behringer UCA202 headphone output suffers these problems but it also sells for only $29.
  
• Rail Splitter – This is employed in some “Cmoy” style op-amp designs. It uses an internal reference at half the power supply voltage that becomes a sort of “fake ground” for the headphones and allows the amplifier to be direct coupled. The benefit is you get rid of the output capacitor. But in its place, you have any flaws in the rail splitter compromising the performance and the better rail splitters tend to be power hungry making them a poor choice for battery powered devices. And, even worse in this case, the total voltage swing is still just as limited as with the single 5 volt supply. I don’t like fake milkshakes and I don’t usually like fake grounds either.
  
• Fully Isolated DC-DC Bipolar Converter – This is the most expensive solution. It also takes up more space and uses more power. The 5 volt USB source (and/or battery) powers a DC-DC power supply that generates two isolated +/- power supply “rails”. The DC-DC converter will generate some amount of noise above the audible range much as you see above from the E7. How well it’s kept out of the audio depends on the circuit design, type of converter, how much physical isolation can be obtained (difficult in a pocket device like the E7) and other factors. The benefits are wider voltage swing, true +/- supply rails with no “fake ground” compromises, true isolation, and no output caps required. The HRT DACs use this solution.
  
• Inverting Charge Pump – This method also yields a true bipolar power supply but costs less, takes up less space, and increases efficiency (important for battery operation). Filtered USB/battery power is used for the positive rail and only the negative rail is generated. A charge pump cleverly charges a capacitor, disconnects it from the positive supply, flips it over, and connects it to the load creating a mirrored negative supply. And while that cap is powering the load, a 2nd cap is charged. It switches back and forth between the 2 caps always charging one while the load runs from the other. It does this very rapidly—400,000 times per second in the E7. As you might imagine this switching creates some electrical noise in the negative supply from all the switching and positive side from the spikes of current drawn to charge the capacitors.
  

DESIGN ERROR? Did FiiO make the best choice? If I were designing a pocket portable DAC to sell for $99 I’d probably also use an inverting charge pump to generate the negative rail. It’s the best compromise in this application. The TPA6130 offers a cost effective, battery-friendly solution that performs relatively well as the measurements have shown. But it is one of the weaker links in the E7’s entire signal chain. Running the entire design on a split supply using a dedicated charge pump and using a better headphone amp would offer some improvements in several areas. But it also would have added cost, decreased battery life, and possibly even made the E7 larger, with little audible benefit to most users.

SLEW RATE: Despite what you may have heard or read you only need 0.2 V/uS slew rate per volt of RMS output to perfectly reproduce any signal you’ll ever find on a CD. That number has been verified by several well respected audio engineers like Douglas Self. And it’s conservative--It assumes a worst case full output at 20 Khz which pretty much never happens anywhere but on a test bench. The E7 has about 1.8 volts maximum output, so it needs 0.36 V/uS for the absolute worst case.

The measurement table at the top of the graph below shows a dV (delta or change in voltage) between points 1 and 2 on the graph of 0.48 volts and a dt (delta time) of 1.3 uS. This works out to a measured slew rate of 0.37 volts/uS which is coincidentally so close to the theoretical value one might suspect FiiO and/or TI (with the TPA6130) designed it that way—and they probably did. Designing an amplifier to be faster than needed often involves other design compromises (faster op amps are typically more noisy and power hungry than slower ones for example). I verified the E7 can produce a 20 Khz sine wave at 1.8 volts without a significant rise in distortion so I’m entirely satisfied it’s as fast as it needs to be for any real world signal it will encounter. A faster design would certainly perform no better reproducing 44 Khz sampled audio and may well perform worse in other ways.

In this case, TI apparently worked hard to keep the power consumption of the TPA6130 down. Lower bias levels, and low power amplifiers, tend to be slower. But, in this case, some would consider this careful engineering. They made the TPA6130 (and FiiO made the entire E7) fast enough so the slew rate cannot limit its performance:

DAC IMPULSE RESPONSE: This is another new test showing an 8 sample digital pulse. The response shows the E7 has accurate polarity (it doesn’t invert via USB or the line input) and exhibits nearly symmetrical pre and post ringing indicating the Wolfson DAC is using a classical steep slope output filter. There’s nothing to worry about here unless you were hoping for the much less common linear phase or other more esoteric filtering:

THE FINE PRINT:

• FiiO E7 Firmware: FW01JLY31JA
  
• Audio Analyzer: Prism dScope Series III with Version 1.4 software
  
• Operating System: Windows 7 Ultimate 64 bit & XP SP3
  
• Volume Settings: Unless otherwise noted, line input analog tests utilized unity gain (Volume set to 50) and USB tests set for a 400 mV RMS output with 0 dBFS input (Volume set to 35).
  
• Line Input measurements made while running from battery. USB tests conducted with USB Charging disabled. This may have reduced maximum output possible with a power source connected.

GEEK’s VERDICT: The biggest compliment I can pay the E7 is I can’t find much I would change for a portable product in this price category. But, in a perfect world, I’d add probably consider:

• A better headphone output amp with lower noise and more output swing for high impedance headphones.
• A true split rail supply (versus the one built into the TPA6130).
• Use the split rail supply to raise the line input clipping point above 2 volts RMS.
• Reduce the Wolfson DAC distortion to be closer to the analog performance.
• Tweak the firmware to improve usability as I suggested in the first part of the review

The biggest real-world issues are the maximum output into high impedance loads and the input clipping level. If you have relatively inefficient cans and like to listen loud, or to wide dynamic range content, the E7 might clip the peaks. And if you plan to use a home CD player, or similar, that only has a fixed output level make sure it’s under 1.25 volts RMS max output.