Pages

November 30, 2011

Objective Desktop Amp (ODA) & DAC

INTRO: Ok, right up front you need to know this article is not one of my usual in-depth write ups. But here’s the deal. The battery powered O2 headphone amp has been more popular than I imagined with 2000+ PC boards already in circulation and more on the way. After thousands of O2 messages and comments I’ve learned many are interested in purely a desktop amp and many O2’s have gone into service sans batteries. Those considering an O2 for desktop use might want to read this “preview” article.

THE ODA: “ODA” simply stands for “Objective Desktop Amp”. The idea is to take the same objective principles behind the O2 and apply them to an amp optimized for desktop-only use (no battery power). Those principles include:

  • Genuinely Transparent Performance – I wrote an article for InnerFidelity that describes what’s required of a headphone amp for transparent performance. I was even more stringent in defining the requirements for the O2. The ODA will meet or exceed the same requirements so you can listen to everything your music has to offer without your being forced to listen to your amplifier’s unwelcome contributions. This means no audible hiss, distortion well below audible levels, and a suitably low output impedance to maintain flat frequency response and proper damping with all headphones.
  • Wide Headphone Compatibility – Music lovers shouldn’t have to worry about trying to figure out what headphones work well with what amp. The O2 has demonstrated it’s entirely possible to have one modestly priced amp that can accurately drive 98% of current headphones on the market.
  • Focus On Design Rather Than Designer Components – The O2 demonstrates design and implementation is often far more important than using trendy expensive parts. The ODA places a similar emphasis on proper implementation to get the most from reasonably priced high quality components. As with the O2, I expect the ODA will outperform many other headphone amps using more exotic components and topologies.
  • DIY Friendly Design – Several first time DIYers have already built an O2 and reported it worked the first time they powered it up. By avoiding surface mount parts, point-to-point wiring, and chassis mounted components, the ODA should be equally simple to build. Because DACs require surface mount components, and for other reasons, the optional DAC board will be available pre-assembled and tested (see below).
  • Real Ground – The ODA will not have third channel masquerading as a virtual ground to mess up the performance. It uses a proper bipolar power supply referenced to true ground and direct coupled outputs.
  • Maximum Value – Like the O2, the ODA is being carefully “value engineered” which takes into account everything from the size and complexity of the PC board to minimizing the number of different components on the parts list. Upgraded components are used where there’s a meaningful benefit or little price penalty. The ODA will use a low cost off-the-shelf enclosure.

ODA VS O2: So what makes the the ODA different and more desirable for desktop use? The main O2 thread on diyAudio is flirting with 1000 posts. And, collectively, there are another 1000+ posts and messages elsewhere. Some clear consensus stood out in what many wanted in a desktop headphone amp:

  • Optional Internal 24/96 High Resolution USB DAC – The number one requested “upgrade” to the O2 is a DAC that’s designed and measured to similar standards of high performance. Just such a DAC is well under development for the ODA (see below).
  • Higher Quality Inputs & Outputs – The O2, for size reasons, has 3.5mm input and output jacks. The ODA adds a 1/4 inch (6.5mm) Neutrik headphone jack for use with high-end full size headphones and RCA input jacks for higher quality connections and even lower crosstalk.
  • Rear Panel Connections – The ODA has the power and input connections in back to keep the cables out of the way for a cleaner look and better ergonomics (the downside being another relatively expensive panel to buy).
  • Wider Source Compatibility – The ODA should work with anything from an iPod Line Out Dock (LOD) to high output home DACs with a wide range of headphones. It has more flexible gain options compared to the O2.
  • Higher Quality Power Supply – The half-wave power supply in the O2 is something of a compromise for size, cost, and battery charging reasons. The ODA power supply improves on the O2’s in several ways.
  • Headphone Protection Relay – The O2, like many amps, produces a “click” in the headphones when you turn it on and a soft “thump” when you turn it off. The ODA uses a headphone relay to eliminate these noises and provide added protection for expensive headphones.
  • Other Possible Upgrades – Depending on how the details sort out, the ODA may have some other upgrades as well such as a possible preamp/line output.

ODA + ODAC: What’s an “ODAC”? It’s a USB DAC designed the objective way—just like the O2 and ODA. It fits inside the ODA turning it into a desktop headphone DAC. Here’s some preliminary information:

  • High Resolution USB – Relatively few reasonably priced commercial DACs, and almost no DIY DACs, support anything beyond 16 bits at 44 or 48 Khz over USB (despite most having 24/192 DAC chips). The problem is, until recently, there were no suitable options for high-resolution audio over USB without needing special drivers or spending lots of money. The ODAC will support up to 24/96 over USB.
  • 111+ dB Documented Dynamic Range – The most dynamic range you can get from a DAC operating in 16 bit mode is around 96 dB with most falling several dB short of that. While that can be enough if you control the volume in the analog domain somewhere after the DAC, it’s often not enough if you want to control the volume at your PC in the digital domain. That’s where those extra high resolution bits show their stuff. Even with the PC volume turned down, you can still get 16+ bits worth of resolution and dynamic range. That can mean the difference between hearing noise or pure silence in your headphones. And, unlike nearly every other DAC out there, the Dynamic Range of the ODA will be fully documented. Many DACs just quote the spec for the chip off the datasheet which is usually nowhere near the entire DAC’s actual performance. See the Tech Section for a preview.
  • USB Audio Class 1 Compliant – Unlike most pro-audio interfaces that support 24/96, the ODAC requires no problematic proprietary drivers for XP, Vista, Windows 7, OS X or Linux. It’s true Plug-And-Play. It also does not require UAC2 drivers, like some DACs, which are not provided in any current version of Windows.
  • ODAC vs DAC1 – I’ve done some preliminary blind testing against my $1600 Benchmark DAC1 Pre with a variety of music and my best headphones, and so far, a least two different people cannot tell them apart. The DAC1 Pre has won a lot of professional accolades for being one of the better DACs money can buy at any price. I can see a formal ODAC NwAvGuy blind listening challenge coming up.
  • Redbook Compliant Output – For many reasons, it’s good to have a DAC that complies with the Redbook standard for digital audio output levels. Many USB DACs, like the AMB gamma, can’t produce the required 2 volts RMS of output and suffer lower dynamic range. And some go way over the specification which risks overloading the inputs of a lot of equipment. The ODAC is Redbook compliant. Among other benefits, this makes it much easier to do level matched A/B comparisons with other DACs and digital sources that are also Redbook compliant.
  • Pre-Assembled Board – Because USB and DAC chips are typically fine pitch surface mount devices they’re not very DIY friendly. To make matters worse, if you want 24 bits over USB with native Windows drivers, I’m not aware of any suitable chips that are available to DIYers. All the current solutions require licensing and/or some sort of contractual agreement. They’re not stocked by DigiKey, Mouser, Farnell, etc. And to further complicate things, they also require custom programming before they will operate correctly. That’s 3 strikes against DIY for high resolution USB DACs. The obvious solution was finding someone to co-develop the board with and handle all the contractual, assembly, programming and financial details so they could offer a pre-assembled board.
  • Line Input Retained – The ODAC won’t defeat the line input on the ODA. You can use both sources or even plug another Redbook compliant DAC in for direct A/B comparisons (even better if a friend does the switching out of the listener’s sight!).
  • Standalone Capable – The ODAC is designed so it’s also usable by itself for other applications while operating entirely from USB power. No ODA required.

NO MONEY FOR ME – There’s been some speculation that an assembled DAC board is a chance for me to finally make some money. But that’s not the case. There’s no money in the ODAC for me. As with the O2, and ODA, I’m leaving that part for others to sort out. This blog, and hence my reviews and commentary, remain entirely non-commercial.

WHAT THEY’RE NOT: A few things the ODA and ODAC are not:

  • No S/PDIF – There’s little need for an S/PDIF output as it seems unlikely anyone would want to use the ODAC as a USB-S/PDIF device for another DAC. And while there are a few applications for an S/PDIF input it’s not something most will ever use. For everyone else, an S/PDIF input would add complexity, size, and substantial cost with zero benefit and may degrade jitter performance. That’s not keeping with the “best performance for the lowest cost” O2/ODA approach.
  • No Balanced Outputs – Balanced audio is great for pro use with long cables and electrically hostile environments. But it’s generally more of a liability than an asset in home audio gear. All else being equal, balanced stages usually have higher overall noise and distortion than their unbalanced counterparts. Headphone drivers don’t magically somehow work better when driven by a balanced source. Most of the stuff you hear and read about balanced audio for home headphone gear is myth, hype and even snake oil.
  • No Arc Welding – If you’re looking for an ultra high output amp for those vintage AKG K1000s, or to weld that broken light fixture, the ODA isn’t it. Just like with S/PDIF, it doesn’t make sense everyone should pay for power they can’t use (and that could also more easily damage their headphones) just to support a tiny minority of potential users with unusual requirements.
  • No Battery or USB Only Power – The ODA isn’t intended to be portable. That’s what the O2 is for. The ODA needs a proper power supply for the best performance and should outperform any USB powered headphone DAC. The ODAC board by itself, however, can be powered from USB for use in other applications.
  • No Recording – The ODAC is playback only. If you want a microphone input for Skype calls, check out the $12 Syba CM-119 I reviewed. If you want to record a string quartet, there are lots of great reasonably priced professional USB audio interfaces optimized for recording rather than playback.

DAC INTERPLAY: I’ve held back on some ODA design details waiting to see if the ODAC was going to be a reality. So in some ways, the ODAC has come first. The ODA PC board layout and other details are subject to change greatly depending on what form a DAC option might take. Because I want to remain non-commercial and not profit from any of this, I had to find someone interested in offering the ODAC for sale in pre-assembled form. It might be a bit premature, but so far the ODAC is looking viable and I’m very pleased with the performance and testing so far. It’s been a joint effort.

HOW MUCH? Many have asked about the cost. I’m guessing the ODA itself will be around $30 more than a complete O2. That would put the total DIY price somewhere around $130 without the DAC option. Just the board would be under $70 complete and fully functional if you build it yourself. And I’m told the optional assembled, programmed and tested ODAC board should be under $100. These are all just estimates and subject to change. The more popular the ODA and ODAC become, the lower the price will probably go.

WHAT’S NEXT & TIMING: My current plan is to publish another ODA/ODAC article sometime in December with more details, more test results, etc. Depending on how things go, the ODA documentation package (including PCB artwork), and even the ODAC board, could be available as early as late January. Being realistic, February is more likely and it might slip into March. It all depends on how many board revision cycles are required, what changes/problems come up, etc. The O2, for example, had a last minute re-design and delay because the main distributors ran out of volume controls and a few other critical parts. For now I just wanted to let everyone know there’s progress, and if you’re thinking about an O2 for desktop use, or perhaps another desktop headphone amp or headphone DAC, you might want to wait for the ODA.

 


TECH SECTION


DYNAMIC RANGE BACKGROUND : Probably the biggest spec thrown around by DAC chip makers for bragging rights is the dynamic range (DNR) performance of their chips. Some sleazy audio DAC marketing types just publish the number from the datasheet as the spec for their completed DAC. But that’s usually cheating in a big way and a bit like saying the tires on your car are rated for 155 MPH so that must be how fast the car can go. Uh, no. But then again some of those designing and selling USB DACs probably have no way to measure the actual DNR so perhaps that’s why they cheat. Getting even close to the chip’s ultimate performance requires great care with the power supply, PC board layout, grounding, and more. For 112 dB of dynamic range with the Redbook standard output of 2 V rms at 0 dBFS, there can be only 5 microvolts of total noise. That means all the noise in the audio band added together has to be less than 5 microvolts! That’s hard to do even if the DAC chip itself were somehow noiseless and that’s hardly the case with lots of noisy digital signals just a few millimeters from the analog pins and noisy USB signals not much farther away.

ODAC DYNAMIC RANGE: Shown below is the performance of an early prototype of the ODAC. The final design could be worse, but hopefully might be even better. Regardless, the results show nearly 112 dB of genuine dynamic range (A-weighted or “dBA” as is the industry standard). Most USB DACs are limited to 16 bits and around 90 dB plus or minus a few dB of dynamic range. I used a –60 dBFS signal instead of my usual –90 dBFS as the former seems to be more of an industry standard for this measurement:

Odac 1 Khz -60 dBFS Noise & Dynamic Range 10K Ohms BW=22Khz (ref 2 Vrms)

 

WHY 24 BITS CAN MATTER: Many find it convenient to leave their headphone amp turned up and use the volume control in your PC’s operating system or player software. For one thing it allows locating the headphone DAC out of reach and/or out of sight. But when you do that with a typical DAC capable of only 16 bits over USB you get less than 16 bit audio delivered to the DAC at anything less than maximum volume on the PC. Turn it down only 6 dB and you have 15 bit audio. Another 6 dB and you’re getting only 14 bits. And so on. Here’s a graph of the ODAC running in 24 bit mode in blue. It’s playing a 6 Khz test track just below maximum at –1 dBFS but the volume control in Foobar was turned down about –45 dB. So the end result is a signal that’s about –46 dBFS being sent to the DAC. In 24 bit mode the THD+N was impressively low at only 0.01% relative to the –46 dB signal. Shown in yellow is the much higher noise floor in 16 bit mode without changing anything else. The dScope can’t show FFT readings from previous sweeps easily, but the 16 bit THD+N was nearly 20 dB worse at about 0.08%—or eight times as much noise and distortion in 16 mode (As has been pointed out in the comments, the readings in this graph may not be correct but the spectrum is correct. I plan to re-run the test):

Odac 6 Khz -1 dBFS Vol = -45 dB THD & THD N 24-44 (blue) 16-44 (yellow)

 

MORE TO COME: Check back in December for more on the ODA and ODAC!

November 23, 2011

FiiO E6 Amp

fiio e6 package dscopeINTRO: I reviewed the amazingly small and inexpensive FiiO E5 back in this blog’s infancy. I was fairly impressed with the build quality and even found the audio performance acceptable considering the $20 price. But it did have some significant flaws. FiiO saw fit to release a supposedly improved successor to the E5 and named it, appropriately, the E6. It’s also interesting to see how it compares to my recent review of the $65 FiiO E11.

THE FiiO E6: The street price has gone from $20 for the E5 up to $25 – $30 for the FiiO E6. FiiO unfortunately ditched the dedicated EQ switch and has combined that function with the power switch. The good news is the controls are now easier to use without looking at the amplifier with the multi-function switch and volume control on opposite sides. It comes with a USB charging cable and a short 3.5mm male to 3.5mm male audio cable to connect to a portable player, etc. Like the E5, the E6 uses a very common mini-USB connector for charging so if you lose the cable you probably have another one lying around. Just to be clear, the mini USB port is only for charging--the E6 is not a DAC. Unlike the E11, you can use the E6 while it’s charging.

GOING BACKWARDS: For the extra 25% – 50% FiiO removes from your bank account (compared to the E5) you get an easily scratched shiny black plastic enclosure instead of the sexy satin finished metal one on the E5. The solid spring loaded metal clip on the back of the E5 has been replaced with a removable clear plastic clip seen flanking the amp in the photo above (click for a larger view). Both of these appear to be cost saving measures and a step backwards in build quality. It’s sad to see a company substantially raise the price and give you less. But the cheaper case and clip may not matter much to some—especially those who don’t want a clip.

SIX IN ONE SWITCH: The sliding spring loaded power switch (which also has a “lock” position) provides six functions in the E6: On, Off, EQ OFF, EQ1, EQ2, and 2 volt mode, There are two levels of bass EQ versus the single EQ option on the E5. There’s also a new mode to allow 2 Vrms of input without overloading the amp. These three options are indicated by a tiny LED on the back that glows red, blue, or purple depending on how many times you momentarily move the power switch to “ON” to cycle through the options. It remembers the last setting. Holding the switch for 3 seconds turns the E6 on or off. Because the switch does different things with long and short activations, and you have to remember the order and/or LED colors, it can be a bit confusing. In my opinion, a power switch should just be a power switch and a dedicated EQ switch, as on the E5 and E11, would have been nice.

fiio e6 e5 ipodNO AUTO SHUTOFF & DESKTOP USE: One of my biggest complaints about the original E5 is the lack of an automatic shutoff of any kind. If you accidentally leave it on, it runs the battery dead. That’s hard on Li-Ion batteries and can easily leave you without a working amp while you’re away from home. Even worse, if you want to use it as a desktop amp plugged into a USB port on your PC, it will still run itself dead when you shut your PC off unless you remember to turn the amp off with the PC every time (or your PC has “stay alive” USB ports designed for device charging). Sadly, the E6 suffers the same problem. I left mine untouched for hours and it was still running and there’s no mention in the very brief operating instructions of an automatic shutoff.

SUBJECTIVE LISTENING: I listened to the E6 with several headphones. With my Ultimate Ears IEMs it has plainly audible hiss but it’s not seriously objectionable with music playing. With my Mee M11+ 16 ohm dynamic IEMs the E6 was reasonably quiet and had plenty of volume but the M11’s already edgy high frequency response sounded even more edgy than normal. I thought I also noticed some harshness with my Etymotic ER4s. But it’s hard to say for sure without a blind test. My inexpensive Sennhesier HD201s sounded about the same as usual but they tend to gloss over flaws. Trying my HD650s resulted in rather nasty clipping on peaks. Unless you listen to highly compressed music, and don’t like it very loud, the E6 is a poor match with headphones like the HD650, HD600, etc.

BASS EQ: I didn’t much like either EQ setting with any of my headphones or music. EQ1 is relatively muddy and bloated. EQ2 is a bit more restrained but still boosts way more than just the deep bass. Both settings seriously “thicken” male vocals. Some, however, might like the EQ—especially if they have really bass shy headphones—like say those you get for free on airplanes. The E6’s EQ came closest to being useful with my sterile sounding Etymotics but even then, with a lot of music, I couldn’t get past the muddy lower midrange and upper bass.

GAIN & OVERLOAD: The E6 has the same 2.5X (~ 8dB) gain as the E11 and O2 on low gain. This should work well for most users. But if you have higher impedance headphones that need a lot of voltage, beware you need a source capable of 1 Vrms to get the most out of the E6. An iPod LOD, for example, is only good for about 0.5 volts. The input overload in the default and EQ modes was around 1.1 Vrms. In the special 2 volt mode, it was about 2.1 Vrms. For more on gain in general see: All About Gain

BATTERY LIFE: I didn’t test the battery life. FiiO rates it at 10 hours. As mentioned above, however, there doesn’t seem to be an automatic shut off. So expect to find the battery dead when you forget to turn the E6 off. Unlike the E11, the battery is sealed inside (AFAIK).

fiio e6 dscope ipodMEASUREMENT SUMMARY: The E6 results are mixed. FiiO improved the low frequency response, low frequency distortion and noise a bit over the E5. But they went backwards on the CCIF IMD high frequency distortion, clipping behavior, and ultrasonic noise. Rumor has it the E6 is based on the same TI chip amp as the E5 so I’m not sure where some of the differences come from (although I speculate in the Tech Section). Overall the two are relatively similar in their performance. The E6 offers a lower output impedance and a bit more output than most portables can manage on their own. But it also may add some distortion to higher quality portables like the iPod Touch. Here’s a summary and comparison table:

.
Measurement FiiO E6 FiiO E5 FiiO E11 O2 AMB Mini3
Frequency Response +/- 0.5 dB A +0/-3 dB C +/- 0.1 dB A +/- 0.1 dB A +/- 0.1 dB A
Phase Error 100-10K 4 deg N/A 4 deg < 1 deg < 1 deg
THD 1 Khz 150 Ohms 0.008% A N/A 0.004% A 0.0016% A 0.002% A
THD 1 Khz 15 Ohms 0.03% C 0.012% B 0.011% B 0.0023% A 0.017% B
THD 20 hz 15 Ohms 0.1% C 0.7% D 0.017% B 0.0023% A 0.01% B
THD 10 Khz 15 Ohms 0.04% C 0.05% C 0.011% B 0.010% A 0.45% F
IMD CCIF 15 Ohms 0.006% C* N/A 0.002% A 0.001% A 0.043% D
IMD SMPTE 0.019% C 0.006% A 0.0079% A 0.002% A 0.009% B
Noise A-Wtd -95.2 dBu -93.8 dBu -101.3 dBu B -114 dBu A -103.2 dBu B
Max Output 15 Ohms 122 mW A 108 mW A 63 mW B 337 mW A 104 mW A
Max Output 33 Ohms 59 mW C N/A 101 mW C 613 mW A 98 mW C
Max Output 150 Ohms 24 mW D N/A 52 mW C 355 mW A 38 mW C
Output Impedance 0.25 0.7 0.5 Ohms A 0.54 Ohms A 0.9 Ohms A
Crosstalk 15 Ohms 48 dB B 46 dB B 48 dB B 65 dB A 40 dB C
Channel Balance < 0.1 dB N/A 1.1 dB B 0.6 dB A 1.14 dB B
Battery Life 10 hr rated ? ~10 hr rated ~8 hrs / ~30 hrs ~5 hours

* Note: the IMD distortion letter grades take into account the spectral graph and not just the raw number.

FIRST CLASS:

  • Seriously small and light
  • Very low output impedance
  • Some improvements over E5 
  • USB battery charging

ECONOMY:

  • Falls 25% short of FiiO specs for output power into 16 and 150 ohms
  • Lower build quality, poor quality clip, and more plastic compared to cheaper E5
  • Single control has 5 functions which is not intuitive and may be confusing
  • Alarming levels of ultrasonic/RF leakage of DC-DC charge pump into audio output 
  • 50% higher price than E5
  • Very marginal clipping performance
  • Excessively broad bass EQ results in muddy/bloated sound
  • Moderate phase error

HEADPHONE COMPATIBILITY: For most headphones with high sensitivity the E5 should be OK. In short, the E6 is best used with headphones rated at a minimum of 108 dB/Volt or 97 dB/mW. If you don’t like to listen very loud, you can subtract up to 5 dB from those numbers. If you listen to wide dynamic range classical or jazz you may want to add 5 dB. For InnerFidelity reviews, the headphones should be rated at less than 0.13 volts for 90 dB SPL. See my More Power article for more details on matching headphones to amps. 

BOTTOM LINE: The E6 is not much of a step up from the E5 and in some ways is a step backwards. Overall, they’re fairly similar in most ways. Both are best suited for driving portable low impedance and balanced armature IEM headphones on a budget or when something ultra portable is desired. Both offer a lower output impedance than many portable players and cell phones which is a significant help with balanced armature IEMs. For a tiny $30 headphone amp the E6’s performance is reasonable enough and improves on the E5 in a few areas. But if you’re serious about sound quality, and have headphones around 50 ohms or higher, the FiiO E11 might be worth the extra $35 for mainly portable (but not desktop) use. The E11 can better drive higher impedance less sensitive headphones, has much better clipping behavior, lower distortion on many tests into 150 ohms, much less ultrasonic crud in the output, and much lower noise. For 16-32 ohm headphones, however, both the E6 and E11 have some problems if you’re genuinely concerned about sound quality. For an even cleaner, more powerful, and quieter portable amp that can drive low and high impedance headphones well, you might want to check out the O2 but it’s significantly less portable and more expensive in assembled form than even the E11.

 


TECH SECTION:


HARDWARE: Rumor has it the E6 uses the same Texas Instruments TPA6130 “chip amp” as the E5 and E7 and despite FiiO’s claim of a “new power amplifier circuit” my measurements seem to confirm it is the same chip. The TPA6130 is rated at 127 mW into 16 ohms when running from a nominal Li-Ion battery as used in the E6. It’s max rating, with 5 volts of USB power, is 138 mW. This makes FiiO’s claim of 150 mW into 16 ohms more than a bit suspicious. As documented below, it only managed 114 mW into 16 ohms on a nearly fully charged battery.

FREQUENCY RESPONSE: The FiiO E5 rolls off the bass to an audible –3 dB at 20 hz. Fortunately, FiiO improved the E6 to where it’s only down an inaudible 0.75 dB at 20 hz. It also has slightly flatter response at 20 Khz for slightly less high frequency phase shift. The channel matching is nearly perfect. There are 4 lots below, 2 for each channel at each load, but they’re all right on top of each other. Somewhat ironically even some high-end gear can’t match the channels anywhere near as well. That’s one of the advantages of using IC’s instead of discrete components:

FiiO E6 Frequency Response Both Ch 16 Ohms (blue) 150 Ohms (yellow) Vol=Unity 400 mV

 

PHASE ERROR: The goal is less than 1 degree of error from 100 hz to 10 Khz where the ear is most sensitive. The E6 fails to meet the goal at both ends of the audio spectrum. The phase is off by about 4 degrees at both 100 hz and 10 Khz. This isn’t awful, but it’s not ideal either:

FiiO E6 Phase Response 150 Ohms Vol=Unity 400 mV

 

1 Khz THD+N vs OUTPUT 15 OHMS: The E6 falls short of its claimed specs. The graph below looks a bit jumbled but shows the E6 vs the $65 E11 and the $150+ AMB Mini3. Interestingly, from about 0.75 volts to 1.25 volts the E6 has significantly lower distortion than either of the more expensive amps. That’s likely because the 3 channel design of the E11 and Mini3 harm their performance into low impedance loads. The max output was 1.35 Vrms into 15 ohms or 122 mW. Into 16 ohms that’s 114 mW which is well short of FiiO’s 150 mW claim and a bit shy of TI’s claim of 127 mW for the chip. The E6, however, outperforms the Mini3 both in overall distortion and maximum output. I tried this test with the E6 running on USB power fully charged and got a nearly identical result—I suspect because the charge pump in the TP6120 is the limiting factor (see: Clipping Performance). Into 33 ohms (not shown), the E6 managed 1.4 Vrms at 1% THD+N for 59 mW:

FiiO E6 1 Khz THD N vs Output 15 Ohms vs Mini3 (yellow) & E11 (orange)

 

1 Khz THD+N vs OUTPUT 150 OHMS: Here’s the same test as above except with 150 ohms running on USB power. The results were essentially identical on battery power. Most notable is the E11 has much lower distortion across the board and much higher output into this load. Into the easier load the E6 produced 1.9 Vrms which is 24 mW into 150 ohms and 12 mW into 300 ohms. FiiO claims 16 mW into 300 ohms which is 2.2 Vrms. TI specifies the TPA6130 at 5.3 Vp-p at either 3.6 V or 5.0 V supply voltage. That’s 1.87 Vrms which is very close to the 1.9 Vrms I measured. Given this, and the result above, FiiO seems to be applying some very creative marketing to their specs:

FiiO E6 1 Khz THD N vs Output 150 Ohms vs Mini3 (yellow) & E11 (orange)

 

CLAIMED OUTPUT vs ACTUAL OUTPUT: FiiO Claims 150 mW into 16 ohms but, on USB or battery power, the E6 can only manage 114 mW. Even TI only claims 127 – 132 mW for the chip used in the E6. Into 300 ohms, the E6 only managed 12 mW against the 16 mW claimed. In both cases the real output is about 25% lower then FiiO’s spec. Given there’s nothing in the Texas Instruments TPA6130 datasheet to support FiiO’s claims, and my rather black and white measurements, it’s likely another case of creative, and misleading, marketing. This sort of thing is unfortunate as I believe it makes FiiO’s specifications for all their products more difficult to trust.

 

100hz OUTPUT IMPEDANCE & MAX POWER: At 100 hz the E6’s distortion starts to rise rapidly above 1.23 Vrms into 15 ohms. Removing the load at that level causes the voltage to rise to 1.25V. This works out to a very low output impedance of 0.25 ohms. It’s almost as if the E6 has current feedback as that’s a surprisingly low number. Current feedback senses the output current and uses it to provide an error signal to the amplifier to correct for voltage drop. I measured a more normal 0.7 ohms with the E5. In any event, both are well below the desired 2 ohms which is great:FiiO E6 100 hz Max Output & Impedance 15 & 100K Ohms

 

THD+N vs FREQUENCY: The older E5 had some fairly serious low frequency distortion. Into 15 ohms the E5 exceeded the 0.05% threshold at 200 hz, was over 0.4% at 30 hz, and off the top of the graph at 20 hz. At the same 400 mV RMS, the E6 does better and doesn’t hit 0.05% until 30 hz—that’s eight times less low frequency distortion—and much less likely to be audible. I suspect FiiO beefed up the power supply capacitance, especially on the charge pump, which is likely what hurts the E5. The E6’s distortion into 15 ohms is still a bit bothersome as it’s about 0.03% across most of the spectrum which is above the ideal 0.01% but below the worst case threshold of 0.05%. So it’s overall acceptable but mediocre. Into 150 ohms it’s below the magic 0.01% from about 100 – 2500 hz and rises from there to a modest 0.025% at 9 Khz (it would keep rising if the audio analyzer didn’t cut off at 22 Khz). Into 32 ohms (not shown) it’s around 0.02% through the midrange. This isn’t great performance but it’s better than the E5 at least:

FiiO E6 400 mV THD N vs Frequency 15 Ohms (blue) 150 Ohms (yellow)

 

SMPTE IMD 15 OHMS: At the same 400 mV I tested the E5 at, the E6 is surprisingly worse for SMPTE IMD. The “mountain” of E6 IMD products at the base of the 7 Khz signal is much higher reaching up above –80 dB. The THD (harmonics) from the 60 hz signal were similar to the E5. My best guess is the higher IMD is due to the PC board routing/layout, or perhaps the EQ/input level circuitry which is different in the E6 vs the E5. It could also be due to power supply changes. Regardless, this is marginal IMD performance but not unacceptable for a $30 amp:

FiiO E6 SMPTE IMD ~400 mV 15 Ohms (ref ~ 400 mV)

 

CCIF IMD 15 OHMS: The E6, not surprisingly, is also marginal on the CCIF test. The first pair of “sidebands” flanking the 19 Khz and 20 Khz signals are well over –80 dB at about –72 dB. The difference signal at 1 Khz is just below –80 dB. Like the SMPTE performance, this isn’t awful, but it’s not great either. As is often the case with IMD, the reading doesn’t tell the whole story. It’s at least better than the AMB Mini3 (I didn’t test the E5 for CCIF IMD):FiiO E6 CCIF IMD ~400 mV 15 Ohms (ref ~ 400 mV)

 

NOISE: The E6 is a significant 6 dB more noisy than the FiiO E11 and a huge 19 dB more noisy than the O2. But it’s a few dB quieter than the E5. The E5 was 93.8 dBu while the E6 is 95.2 dBu A-Weighted. The absolute noise was about double the E11 at 27 uV. This is acceptable performance but there will be some audible hiss with highly sensitive headphones (including most balanced armature IEMs):

FiiO E6 Noise Vol=Max Ref 775 mV (dBu)

 

CROSSTALK: By avoiding a three channel design the E6 has respectable crosstalk of –66 dB at 150 ohms:

FiiO E6 Crosstalk 150 Ohms (ref ~400 mV)

 

CHANNEL BALANCE: Unlike my particular E5, which has some sort of channel balance flaw from the factory, the electronic volume control in the E6 (built into the TPA6130 amp chip) had flawless channel balance at any level. It’s hard to even tell, but each of the plots below are really two lines—one for each channel. Even at –45 dB below 400 mV the balance was still within 0.025 dB! There is, however, a bit of odd behavior. The plots below –25 dB were made starting at –45 dB and increasing the volume a single “click” at a time to judge the size of the steps. Most of the steps are around 2 dB but every now and then it jumps by around 3 dB (the wider gaps shown between some lines) up to –27 dB (there are a couple clicks to the –20 dB line). This isn’t a huge problem but sometimes you might find a 3 dB step a bit too much leaving you to choose between a bit too soft and a bit too loud:

FiiO E6 Volume Steps & Tracking (ref 400 mV)

 

GAIN: The E6 has 7.9 dB of gain or about 2.5X—the same as the E11 and O2 set to low gain. This is a good choice for this amp. To obtain the maximum 2.9 volt output it needs 1.16 volts in. But to hit the maximum 1.3 volts into lower impedance headphones, it needs only about 0.5 volts which is the LOD output of an iPod. Beware, however, if you’re trying to drive power hungry higher impedance headphones you’ll need a source capable of at least 1 Vrms output: For more see: All About Gain

FiiO E6 Gain Vol=Max 150 Ohms (ref 0225 mV)

 

INPUT OVERLOAD DEFAULT SETTING: Related to the above is the issue of input overload. Both the FiiO E5 and E7 overload, regardless of the volume setting, with any input much over 1 Vrms. That’s also true of the E6 in its default mode. The graph below shows it overloading with 1.18 Vrms input:

FiiO E6 Input Overload 150 Ohms (ref ~400 mV)

 

INPUT OVERLOAD 2 VOLT MODE: FiiO provides a “LO 2v high level input” option. To enable it you have to activate the power switch 3 more times after the E6 is already on until the EQ LED turns purple. This feature is not available together with EQ nor is it intuitive. It increases the 1% THD point to about 2.2 Vrms with a fully charged battery and around 2.0 Vrms with a typical battery. This allows using the E6 with normal home sources that follow the Redbook standard of 2 Vrms maximum output. But, as is discussed in the O2 articles, many home sources and even some portable ones exceed 2 Vrms. These are not compatible with the E6 unless they have their own volume control:

FiiO E6 Input Overload LO 2V Mode 150 Ohms (ref ~400 mV)

 

EQ: I recently criticized the E11 for having bass boost that reached significantly above 100 hz which tends to result in more “boom” along with the bass boost lower down. The E6’s EQ is significantly worse with a much more broad boost that will still be audible even above 1000 hz. The first EQ1 option (one extra “power on”) results in a red EQ LED (the red plot below) and provides the most dramatic EQ. It drops the upper frequencies by over 3 dB and boosts the bass by nearly 5 dB. The total boost is about 8 dB broadly centered around 75 hz. The EQ2 mode (blue LED and plot) is about a 4.75 dB boost centered at about 50 hz but still over +1 dB at 1 Khz without the high frequency volume drop. The subjective result with most headphones is more boom than deep thump. The E5 had only a single EQ option and it was a more subtle 3 dB boost centered at around 80 hz. For comparison the E5 and E11’s EQ is show in the second and third graphs:

FiiO E6 Freq Response EQ OFF (yellow) EQ1 (LED red) EQ2 (LED blue) Vol=Unity Ref 400mV

FiiO E5 ~400 mV 150 Ohm Load Frequency Response With Bass EQ (blue)

FiiO E11 Frequency Response EQ=2 EQ=1 EQ=0 Top to Bottom

 

SQUARE WAVE PERFORMANCE: The high speed scope shot below shows the E6 driving Sennheiser CX300 headphones with 0.01 uF of added capacitance. The “thick fuzz” is massive leakage of the DC-DC charge pump in the TPA6130 chip amp. For some reason it’s even worse in the E6 than it was in the E5. It’s much worse on the negative half of the waveform as that’s the half supplied by the charge pump. The sloping sides of the square wave are a clue this is a relatively slow amplifier (more on that below). The only good news here is there’s only a hint of overshoot on the top of the waveform meaning the stability is relatively good (click for full size view of all Agilent scope graphs):

FiiO E6 10 Khz Square Wave CX300   10 nF 2Vp-p

 

DC-DC CHARGE PUMP NOISE: Here’s a close-up of the “fuzz” from the bottom of the waveform above. It’s a very high 400 mV peak-to-peak which is 140 mV RMS. To put that in perspective, with sensitive headphones that’s greater than the audio signal much of the time when listening at moderate levels! The scope also shows it’s around 770 Khz which is a bit higher than the 400 Khz TI specifies for the TPA6130. So I have to wonder if FiiO has altered the charge pump capacitors which may have changed the operating frequency. They may have made one or more of the caps larger to improve the low frequency distortion performance mentioned earlier. Whatever they did it made the leakage worse. Ultimately, nobody can hear 770 Khz, but one does wonder what it might do to the sound quality when it can be even greater in level than the audio signal. At low listening levels, you’re essentially listening to a modulated 770 Khz RF carrier that’s being “demodulated” by your headphones. As an aside, the E6 not surprisingly wrecks havoc with an AM radio as it’s essentially a low power radio transmitter operating right in the 510 – 1600 Khz AM (USA) band. I suspect there’s no way it can properly pass FCC or CE certification with headphones plugged in acting as an antenna:

FiiO E6 DC-DC Converter Noise 400 mVp-p at ~ 770 Khz

 

SLEW RATE: TI rates the TPA6130 at 0.3 V/uS slew rate. As shown below, the E6 required 9.2 uS to slew from –2 V to + 2V yielding a slew rate of 0.4 V/uS. The rule of thumb is 0.2 V/uS per volt RMS output and the E6 barely meets that requirement. So, however slow and ugly the square wave might look, the E6 shouldn’t slew rate limit with any music signal you’re likely to feed it but it’s at the bottom range of what’s acceptable:

FiiO E6 10 Khz Square Wave & Slew Rate 600 Ohms (just under clipping)

 

CLIPPING PERFORMANCE: The E6 had highly asymmetrical clipping even into a very easy 600 ohms with the negative rail clipping far sooner than the positive rail likely because the positive rail is direct from the Li-Ion battery and the negative rail is from the DC-DC charge pump built into the TPA6130. If you look close you can see the “digital phosphor” feature of the Agilent scope revealing the actual clipped waveform under all the hash on the negative side. Because this is worse than the E5 it’s a further clue FiiO may have altered the charge pump circuit—in some ways for the worse—with the E6. To be blunt, combined with all the charge pump ultrasonic garbage, this is the worst clipping performance I’ve ever seen from a headphone amp. This is only likely to be an issue if your headphones will push the E6 near its output limits. You won’t, for example, get close to clipping with most balanced armature IEM’s (Shure, Etymotic, etc.). Matching headphones to amps is further explained in my More Power article. For comparison the E11’s performance on the same test is shown in the second graph (click for larger):FiiO E6 10 Khz Sine Wave Clipping Performance 600 Ohms

FiiO E11 10 Khz Sine Wave Clipping 600 Ohms

 

TECH COMMENTS: In some ways FiiO improved on the E5 with the E6. The low frequency response is flatter, there’s significantly less low frequency distortion, a bit less noise, and there’s a special mode to allow a 2 Vrms input. But, on the negative side, the E6 has less useful EQ, even more ultrasonic garbage (at a frequency and level the FCC and/or CE might not approve of) leaking into the output, and significantly worse clipping behavior. The E6 also fails to meet FiiO’s power output specs, and assuming it really does use the TI TPA6130 chip, the datasheet indicates it’s unlikely it ever could reach their claimed power levels under the best of conditions. The E6 is not an especially clean or well behaved amplifier but in light of the very modest $30 price it’s likely acceptable for many applications for those on a budget. It would be interesting to conduct blind tests against the O2 and see how it compares and if perhaps there are some audible side effects to the large ultrasonic/RF output and other flaws.

November 16, 2011

FiiO E11 Amp

fiio e11 bench close upINTRO: Given the FiiO E11 is almost a direct clone of the AMB Mini3 I wasn’t sure it was worth my time to review it. The Mini3 has some significant problems mainly related to its 3 channel channel design. Using a similar 3 channel approach FiiO seemingly copied the Titanic complete with the big hole. But I’ve had many requests to test the E11 (and the similarly amplified but still missing-in-action E10). Plus the E11 has some significant improvements over the Mini3 including longer battery life, USB charging, two gain modes, a three position EQ switch, smaller size, lighter weight, and a much lower price of around $65 vs $150+ for an assembled Mini3. If it’s a decent amp, the E11 would neatly fill the price gap between something like the FiiO E6 and the O2 portable amp. So I gave in and decided to review it after all.

THE FiiO E11: The E11 was so light when I first held it I wondered if the battery was installed. It doesn’t have nearly the heft and solid feel of the FiiO E7. It turns out the battery was installed and the E11 is a feathery 2.3 ounces (65g)--less than half the weight of the current iPhone. If you’re going to carry the E11 in your pocket, or “rubber band” it to your iPod (it comes with a heavy rubber band), the light weight is a good thing. FiiO even protected the volume control making the E11 more “pocket friendly” helping prevent accidental volume changes. It has 3.5mm input and output jacks and a volume control that also serves as the power switch. Going beyond the Mini3, the E11 adds a mini USB port for charging, a two position gain switch, a three position EQ switch, and the battery can be swapped out without needing a screwdriver (for now at least you can get extra batteries but 2 or 3 years from now don’t count on it). It’s also smaller and thinner than the Mini3. The USB port is especially handy although it doesn’t come with a charger so you need a PC, laptop, USB charger, iPhone/iPod/iPad charger, etc. For traveling it’s one less thing you have to bring along but there’s a huge catch…

DESIGN FLAW #1 – NO DESKTOP OPERATION: Sadly the E11 cannot operate while external power is connected via the USB jack. That means it’s impossible to use it as a regular desktop amp as the battery will end up dead and you can’t listen to it while it’s charging. Ironically this might be another downside to the 3 channel virtual ground which puts the E11’s audio ground at about 4.5 volts DC while the PC’s USB ground is at 0 volts—connecting the two with the amp operating would short it out and possibly damage it. For a lot of people this is a significant limitation. For those using it at home, it also means a lot more wear and tear on the battery shortening its life. In my opinion, this makes the E11 mainly suitable only for those who want an amp to use only on the go and not at home or the office.

DESIGN FLAW #2 - THREE CHANNEL DESIGN: FiiO not only chose a similar 3 channel design but they even used the Mini3’s problematic OPA690 op amp (which isn’t even designed for audio use) for the third channel. And, incredibly, they even copied one of AMB’s false claims for the Mini3. FiiO’s website claims: “Heavy current three channels amplified circuit, effective to lower the output impedance.” In reality, using three channels raises the output impedance. Even Ti Kan at AMB has admitted as much in forum posts. This point is discussed extensively in my Virtual Grounds & 3 Channel Amps article as well as the original Mini3 review. Apparently FiiO, a Chinese company, either copied the Mini3 design without understanding its flaws (or reading my related blog articles), or they have joined forces with AMB in knowingly promoting misleading statements about 3 channel designs. The “lower output impedance” is pure myth, or purely misleading, depending on your point of view. But, in my opinion, it doesn’t reflect well on FiiO as a company.

TWO POWER RANGES: There’s a switch under the battery labeled “High Power” and “Low Power”. I did all my testing at “High Power”. FiiO lists two power specs at three impedances and they’re all 4 dB apart in the two modes. I believe the idea is to extend battery life by lowering the power supply voltage for headphones that don’t need the full output. It’s a useful feature if you plan to use it with sensitive headphones.

SUBJECTIVE LISTENING: I listened to the E11 with several headphones. It had enough output, barely, for my HD650s listening to classical music at live levels. Unlike many amps, such as the FiiO E9, the E11 sounded reasonably good with my Ultimate Ears SuperFi 5’s, but there was some hiss audible with nothing playing. Besides the hiss, I didn’t hear anything that was offensive but I didn’t spend a lot of time listening either.

BASS EQ: With some headphones, like my Etymotic ER-4 IEMs, the bass EQ was useful. On bass shy music the “2” (maximum) setting added some heft and on more normal music the “1” setting was more appropriate. But with headphones that already lean towards “boomy”, which sadly includes most of them, the EQ options just added to the boom and/or muddy bass. So it will depend on your headphones, music, and tastes, if you like the bass EQ. My personal preference is for deeper bass boost but I’ve heard worse EQ than the E11’s.

GAIN: The gains are 2.5X and 4.5X. The first is appropriate for a low gain setting, but 4.5X is on the low side for a high gain option. This is especially true for a battery-only portable amp like the E11 that’s much more likely to be used with low output portable players. Using the line out (LOD) on an iPod, or the headphone out on a Sansa Clip+, you can’t drive the E11 to full output into high impedance headphones even at the high gain setting. That’s a significant limitation for owners of headphones like the HD600/650, DT880, etc. See the Tech Section for the details. For more see: All About Gain

BATTERY LIFE: I didn’t test the battery life. FiiO rates it at 10 hours which is roughly double the real world run time of the Mini3 and slightly better than the higher output O2. It’s not clear if this is with the power switch set to the “High” or “Low” position but I’m guessing the 10 hours is at the Low Power setting.

fiio e11 benchMEASUREMENT SUMMARY: Most of the E11’s measurements are decent enough—especially for a $65 amp. The E11 gets the basics right. It has flat frequency response, a very low output impedance, is reasonably quiet, and has mostly low distortion. There are, however, some problem areas. The E11 has very odd overload/clipping behavior into 32 ohm and lower impedance loads and a bit more low frequency phase shift than is ideal. Even being generous, it fell far short of several of its claimed power measurements. It also suffers from some of the Mini3’s potentially audible Interchannel IMD distortion due to the three channel design. And it shares the Mini3’s marginal stability but, unlike the Mini3, it’s only significant with more challenging loads. There’s some RF leakage of the DC-DC converter into the audio output (but it’s much better than the E5 and E7).  Despite all these issues, the E11 still substantially outperforms the Mini3 using the same op amps and a very similar design. This is yet another example of the devil being in the details. FiiO simply did a better job with the implementation than AMB did using the same parts. For example it seems FiiO chose to use a single ground in the E11 which likely helped many of the measurements compared to the Mini3’s split ground and poor ground routing. See the Tech Section and my article on Virtual Grounds for more details. The following table compares the E11 to the O2, FiiO E9 and AMB Mini3:

.
Measurement FiiO E11 O2 FiiO E9 AMB Mini3
Frequency Response +/- 0.1 dB A +/- 0.1 dB A +/- 0.1 dB A +/- 0.1 dB A
THD 1 Khz 150 Ohms 0.004% A 0.0016% A 0.005% A 0.002% A
THD 1 Khz 15 Ohms 0.011% B 0.0023% A 0.037% C 0.017% B
THD 20 hz 15 Ohms 0.017% B 0.0023% A 0.05% Good 0.01% B
THD 20 Khz 15 Ohms 0.011% B 0.010% A 0.003% A 0.45% F
IMD CCIF 15 Ohms 0.002% A 0.001% A 0.05% C 0.043% D
IMD SMPTE 0.0079% A 0.002% A 0.002% A 0.009% B
Noise A-Wtd -101.3 dBu B -114 dBu A -94.7 dBu C -103.2 dBu B
Max Output 15 Ohms 63 mW B 337 mW A 1067 mW A 104 mW A
Max Output 33 Ohms 101 mW C 613 mW A 883 mW A 98 mW C
Max Output 150 Ohms 52 mW C 355 mW A 317 mW A 38 mW C
Output Impedance 0.5 Ohms A 0.54 Ohms A 10 Ohms C 0.9 Ohms A
Crosstalk 15 Ohms 48 dB B 65 dB A 63 dB A 40 dB C
Channel Balance 1.1 dB B 0.6 dB A 1.8 dB C 1.14 dB B
Battery Life ~10 hours ~8 hrs / ~30 hrs AC Only ~5 hours

FIRST CLASS:

  • Small and light
  • Relatively low noise
  • Flat frequency response
  • Low output impedance
  • Significantly improved performance over Mini3
  • Two gain options
  • USB battery charging

ECONOMY:

  • Cannot use the amp on external power (can only run from the battery)
  • Even the high gain setting limits output with iPod LOD input
  • Fails to meet published power specs at 1% THD both channels driven
  • Third channel creates some interchannel distortion and degrades performance
  • Some moderate low frequency phase shift
  • Shows signs of instability with difficult loads 
  • Odd clipping behavior into 15 and 33 ohms implies power supply limitations
  • Significant RF leakage from DC-DC converter into output

HEADPHONE COMPATIBILITY: For most headphones with high sensitivity the E11 should be OK. The output impedance is low enough for even balanced armature IEMs. But it’s marginal for headphones like the popular Sennheiser HD600 and HD650 which require at least about 2.5 Vrms to avoid clipping the peaks of dynamic music at typical loud listening levels. But that’s partly dependent on the type of music and how loud a person wants to listen. Forget the popular Beyer DT880-600, Audeze LCD-2, HiFiMan planars, etc. and headphones like the AKG K701/K702 will be marginal. The E11 doesn’t have enough output voltage for any of them. See my More Power article for more details on determining what headphones will work well with the E11.

BOTTOM LINE: The E11 performed better than I was expecting given it’s a virtual clone of the Mini3 and much less expensive. It’s biggest weakness for many will be only operating off battery power with the amp being unusable while charging. That makes it a very poor choice for home or office use. Many might find the adjustable gain useful although some may be left wanting for more gain—especially using an iPod LOD as the source with voltage hungry headphones. The E11 has a bit more output into high impedance loads, at least when fully charged, making it a better match for the popular Sennheiser HD600/650 and similar headphones but only if your source can produce at least 650 mV (many portables max out at 500 mV). If you don’t mind running it only from battery, and it has enough output  for your headphones, the E11 is a reasonable choice—especially if you’re on a tight budget and want a highly portable amp. Despite its several flaws it’s still a significant upgrade from the FiiO E5 and E6. If you want lower noise, lower distortion, customizable gain, and much higher output to drive nearly any headphones, the O2 portable amp is a significant upgrade but it’s larger, heavier and more expensive.

 


TECH SECTION:


HARDWARE: The E11 uses the same AD8397 op amp design as the Mini3. The AD8397 has the advantage of high current output but it’s also rather “high strung” and, in the Mini3, it runs “wide open” (with no stability compensation) on the edge of being unstable. The E11 also uses the same OPA690 op amp for the third channel virtual ground as the Mini3 which is a poor choice. It has relatively low PSRR and isn’t even specified for operation at audio frequencies. This is discussed more in the Mini3 review. The Mini3 runs from a single 8.4V NiMh rechargeable battery with the OPA690 referenced to half that creating a “virtual” bipolar +/- 4.2V power supply for the AD8397. In comparison, the E11 uses a single 3.7V li-Ion battery and a DC-DC converter stepping it up to a higher single voltage around 9 volts which is then artificially split into +/- 4.5 volts. This puts the audio ground at 4.5 volts DC which is probably why you can’t use the amp and charge it at the same time. Connecting the audio ground to your PC’s ground would short out the E11’s OPA690 and power supply if it’s not an isolated DC-DC converter.

E11 DESIGN FLAW: If you’re going to bother with a DC-DC converter, it’s not much more expensive or difficult to have a bipolar DC-DC converter that generates a true positive and negative power supply with a real ground. Such a converter would have eliminated the need for the OPA690 entirely along with it’s rather hungry power consumption, added distortion, added output impedance, added crosstalk, and limited current capability. The extra expense in the DC-DC converter is probably less than the cost of the OPA690. In short, FiiO missed the opportunity to make the E11 a better headphone amp without increasing the component costs.

ISOLATED DC-DC CONVERTER: FiiO could have used true bipolar regulated DC-DC converter with at least the negative rail being isolated from ground. That would eliminate all the third channel problems, increase run time (getting rid of the hungry OPA690), maintain the power output as the battery voltage drops, and best of all, allow using it as a desktop amp. With such a power supply the E11 also could have been used during charging making it much more suitable as a desktop amp.

WHY THE FLAW? FiiO’s choice is difficult to understand unless they were simply bent on completely copying the Mini3 and/or believed AMB’s highly misleading claims for 3 channel designs. I suspect they were just trying to seize on the Mini3’s popularity by copying it as closely as possible. But, like the Titanic referenced earlier, the Mini3’s popularity seems to be fading fast with its real performance revealed. Many have reported difficulty selling their Mini3 for a reasonable price since my review came out. See Virtual Grounds & 3 Channel Amps and the original Mini3 review for more.

FREQUENCY RESPONSE: Much like the O2, the E11 has some minor low frequency roll off. Unlike the fully direct coupled Mini3, FiiO wisely chose to add a DC blocking capacitor to the signal path. That way the E11 won’t amplify any DC present on the input. It may however, given the E11’s size and price, be an electrolytic capacitor which can degrade the audio performance. The high frequency performance was very flat and extended. The second graph shows the Mini3 for comparison into 15 and 150 ohms and the O2 comes in somewhere between two:

FiiO E11 Frequency Response 33 Ohms Both Channels Vol=Unity using Low Gain

AMB X-Audio Mini3 200mv In Vol= 50% Freq Pink-Red=150 Ohms Yel-Blu=15 Ohms (ref ~400mV)

 

PHASE ERROR: The E11’s low frequency roll off created some phase error that hits about 15 degrees at 20 hz. The goal is less than 1 degree of error from 100 hz to 10 Khz where the ear is most sensitive. The E11 manages at the high end but fails at the low end with about 4 degrees of error at 100 hz. The weird kink close to 20 hz is because the settling time was set too short on the dScope—it’s a test artifact not the E11. The second graph shows the O2 which has DC blocking and stability compensation but still manages to stay under 1 degree of error over the desired range. The Mini3’s phase response is even flatter but that’s because it lacks desirable DC blocking and high frequency compensation. The slight droop above 10 Khz is evidence FiiO wisely chose to either add compensation to the AD8397 and/or included an RF blocking filter at the input:

FiiO E11 Phase Response Low Gain 150 Ohms

O2 AC Phase Response 400 mV 150 Ohms (non-inverting)

 

1 Khz THD+N vs OUTPUT: The E11 has similar distortion to the Mini3 into 80+ ohms but with significantly higher maximum output. The E11 produced 2.7 Vrms into 80 ohms while the Mini3 could only manage 2.4 Vrms into an even easier 150 ohms. The E11 produced 2.95 Vrms into 600 ohms or 14.5 mW where the Mini3 could only manage 2.45Vrms or 10.0 mW. This gives the E11 a noticeable 1.6 dB advantage. Into 33 ohms and lower, the E11 exhibits rather odd distortion characteristics at higher outputs. Into 15 ohms (white plot) it has much lower distortion than the Mini3 up to about 800 mV then the distortion rises sharply but stops rising around 1 volt and levels off around 0.3% before finally hitting the official 1% at 1.9Vrms—much better than the Mini3’s 1.25 Vrms into the same load. But, to be fair, I would call the real “clipping point” of the E11 around 1 volt. It exhibits similar behavior into 33 ohms as seen in the green plot. My best guess is the DC-DC converter power supply in the E11 is current limiting and/or the power supply capacitors are undersized and there is excessive power supply ripple at higher output currents. The relatively poor PSRR of the OPA690 would pass a significant amount of the ripple to the audio output creating greatly increased THD+N. The second graph shows the Mini3 for comparison into 15 and 150 ohms. While it’s not shown, the E11 produced 2.85 Vrms into 150 ohms at 1% THD vs 2.4 Vrms for the Mini3 with the same load. The third graph shows the much higher output and overall lower distortion of the O2 into all loads. In summary, the E11 produces 63, 101, 91, 52 and 15 mW into 16, 32, 80, 150 and 600 ohms respectively:

FiiO E11 1 Khz THD N vs Output 15 - 33 - 80 - 600 Ohms (top to bottom) Gain=Low BW=22 Khz

AMB X-Audio Mini3 THD vs Output Blue-Yellow=150 Ohms Red-Pink=15 Ohms (ref ~400mV)

O2 V11 Battery Both Ch 1 Khz 10mV  THD N vs Output Left to Right 15 33 80 150 600 Ohms comments

CLAIMED OUTPUT vs ACTUAL OUTPUT: FiiO Claims 300 mW into 16 ohms fully charged. Using 1 volt as the clipping point, the E11 only delivers 62 mW and even using the 1.9V value it still falls short at 225 mW. It’s also interesting 300 mW is exactly the same number AMB quotes for the Mini3 and it also falls far short. In all cases, taking the higher of the clipping points the amp at least exceeded the low number provided by FiiO but it should have been much closer to the high number. I suspect, like perhaps AMB, FiiO was somewhat cheating here and probably had the battery at an unrealistically high voltage and was only driving one channel. The gain switch made no difference in these measurements. Here are all the numbers with about 15 minutes of light average use on a fully charged battery in the High Power mode:

  • 16 Ohms FiiO claims 300/120 mW (High/Low power) but it only delivers 63 mW (225 mW at 1%)
  • 32 Ohms FiiO claims 200/88 mW but it only delivers 101 mW (180 mW at 1%)
  • 300 Ohms FiiO claims 35/13 mW but it only delivers 26 mW at 1%

100hz OUTPUT IMPEDANCE & MAX POWER: Since the Mini3 review back in May, I’ve revised my impedance measurements to use 100 hz instead of 1 Khz to more closely reflect the resonance impedance of headphones where electrical damping is most needed. At 100 hz the E11 gets close to 1% THD somewhat earlier than the 1.9 Vrms seen in the earlier graph at 1 Khz  (1.62 Vrms). Removing the load resulted in the output rising only slightly to 1.66 Vrms. This works out to an output impedance of 0.5 ohms. The Mini3 measured a higher but similar 0.9 ohms at 1 Khz. Both are “good enough” but I have to wonder if FiiO chose to use a single ground for the entire amplifier vs splitting the input and output grounds like AMB chose to do. That would allow the amplifier to partially correct for some ground impedance via feedback and would explain the lower output impedance. It’s also worth noting while the E11 matches the O2 at 0.5 ohms, in the O2 thee output isolation resistor dominate that value, the actual output impedance of the amp is much lower. In the E11, however, it’s the OPA690 virtual ground channel likely dominating the measurement:

FiiO E11 100hz Max Output & Impedance 15 ohms & 100K Low Gain

 

THD+N vs FREQUENCY: Here’s the THD+N vs frequency at the same 400 mV I tested the Mini3 at. Into 600 ohms it’s respectably low around 0.003%. Into 33 ohms it’s still under 0.01%. Into 15 ohms it approaches 0.02% which is acceptable but not ideal as there are distortion products above –80 dB. The rise at low frequencies might be due to power supply capacitors that are too small. Just like the Mini3, the E11 isn’t very happy into low impedance loads. And, like the Mini3, this picture gets worse at higher output levels. I didn’t run this exact test on the Mini3 but I did measure 400 mV THD+N at fixed points of 20hz, 1 Khz and 20 Khz where the distortion was 0.01%, 0.016%, and a very high 0.45%, respectively into the same 15 ohms. The E11 performs very similarly to the Mini3 at low and mid frequencies and much better at high frequencies. The second graph shows the much lower distortion of the O2 for comparison into 33 and 150 ohms:

FiiO E11 THD N vs Frequency 600 - 33 - 15 Ohms (top to bottom) Low Gain BW=22 Khz

O2 AC V11 2.5X THD N vs Frequency Both Ch 33 & 150 Ohms Vol = 100% (BW 22 Khz)

 

SMPTE IMD 15 OHMS: I tested the E11’s SMPTE IMD at 400 mV which is generally my current standard. It had acceptably low distortion with everything below –80 dB. The Mini3 performed poorly at 400 mV into 15 ohms so it was tested at only 200 mV, and even at half the output, had slightly higher but similar IMD. The second graph shows the Mini3 at 200 mV. The third graph shows the O2 under exactly the same conditions as the E11. Note the O2’s distortion reading is over 7 times lower and how much smaller the “mountain” around the base of the 7 Khz signal is. Note also the relative lack of distortion products above 7 Khz on the O2:

FiiO E11 SMPTE IMD ~400 mV 15 Ohms Low Gain (ref ~400 mV)

AMB X-Audio Mini3 200 mV Vol=50% SMPTE IMD 15 Ohms (ref ~400mV)

O2 AC Both Channels IMD SMPTE 400 mV 15 Ohms Vol = 50%

 

CCIF IMD 15 OHMS: The E11 also does respectably well on the CCIF IMD test with everything below 80 dB. This is a much better performance than the Mini3 which had 18 times more distortion on the exact same test. The E11 has everything below –90 dB in the audio band while the Mini3 had distortion products up to a likely audible –52 dB. Considering both amps use the same op amps this is a classic example of implementation is everything. Just using high-end parts doesn’t assure high-end performance. FiiO apparently did a much better job with the PC board layout, ground scheme, stability compensation, etc. The result is 18 times less high frequency distortion from the same parts. The O2 is shown in the third graph with about half the overall IMD of the E11 on the same test:FiiO E11 CCIF IMD ~400 mV 15 Ohms Low Gain (ref ~400 mV)

AMB X-Audio Mini3 200 mV Vol=50% CCIF IMD 15 Ohms (ref ~400mV)

O2 AC IMD CCIF 400 mV 15 Ohms Vol = 50%

 

NOISE: The E11 had respectable noise levels of –101.3 dBu A-Weighted and 13.5 uV of raw noise at full volume on the low gain setting. Using high gain added a few dB more noise. The readings were similar at 50% volume. The industry standard dBu reference is 775 mV. The Mini3 noise measurement was referenced to 400 mV. To convert it’s 20*LOG(775/400) = +5.7 dB. So the Mini3’s A-Weighted noise in the second graph is 97.5 dBr + 5.7 dB = 103.2 dBu or about 2 dB better than the E11. I suspect the Mini3 does slightly better because of the lack of a noise producing DC-DC converter. It’s also possible the EQ circuitry in the E11 contributes some noise even when it’s set to “flat”. It could also be due to the Johnson noise of the resistors. Overall, for a cheap amp with a DC-DC converter, the E11 is impressively quiet. The O2 noise is shown in the third graph in dBv referenced to 1 volt. To convert to dBu it’s –2.2 dB giving 116.2 – 2.2 = 114 dBu. So, in summary the E11, Mini3 and O2 are 101.3, 103.2, and 114 dBu respectively  A-Weighted. The E11 and Mini3 are similar while the O2 has much lower noise:

FiiO E11 Noise Vol=100% Low Gain

AMB X-Audio Mini3 Muted Input Vol= 50% Noise 150 Ohms (ref ~400mV)

O2 Battery 6.5X Gain Noise 25 Ohm Source Impedance Volume=50% dBV (ref 1V)

 

CROSSTALK: The E11 had crosstalk of 53.6 dB into 33 ohms and 48 dB into 15 ohms. For comparison, the Mini3 measured 46 dB and 40 dB into 33 and 15 ohms respectively. The second graph shows the Mini3 into 15 ohms and 150 ohms. Into 600 ohms the E11 was better than 80 dB up to about 1500 hz and rising from there mostly due to coupling in the volume control. I suspect the E11 does nearly 8 dB better than the Mini3 because the ground system isn’t split between the input and output like it is on the Mini3. Ti Kan claims this improves performance, but as seen here and with the E11’s lower output impedance earlier, it actually degrades performance as it removes the ability of the op amp's to correct for ground signals. And even with an improved third channel ground scheme, the E11 is still far inferior to the O2 shown in the third graph. Into the same 33 ohms the O2 managed 65 dB vs 53.6 dB for the E11:

FiiO E11 Crosstalk 33 Ohms (blue) 600 Ohms (yellow) Vol=Unity Low Gain

AMB X-Audio Mini3 200mv In Vol= 50% Crosstalk Pink=15 Ohms Blue=150 Ohms (ref ~400mV)

O2 Crosstalk 400 mV 15 (yellow) 33 (green) 150 (blue) & 600 Ohms (red) Reading 1 Khz 150 Ohms

 

INTERCHANNEL IMD BASELINE: The most likely audible flaw of the Mini3 is Interchannel IMD. Because the two channels interact with a common third channel (the OPA690 virtual ground amp) they share any distortion produced by the OPA690 and in the ground system in general. This is a potentially highly audible form of distortion because the distortion products are not directly related to what’s playing in that channel. The Mini3 splits the internal grounds removing the ability of the op amps to correct for certain kinds of ground signals. I measure Interchannel IMD by first showing the THD spectrum for just one channel operating at 1 volt into 15 ohms. This establishes a baseline for distortion related to just that channel and signal. The E11 produced a very respectable 0.005% THD+N figure with all the distortion products below the desired –80 dB threshold. This about 6 times better than the Mini3 managed which hit 0.03% THD+N and –62 dB as seen in the second graph. The third graph shows the O2 same signal, same load with even lower distortion than the E11:

FiiO E11 THD N 1 Khz 1 Vrms One Channel Driven (ref 1 Vrms) BW=22 Khz

AMB X-Audio Mini3 THD N 1 Volt RMS 15 Ohms One Channel Driven

O2 AC 1 Khz One Channel Driven Residual (non-THD) Noise 1V 15 Ohms

 

INTERCHANNEL IMD 1 Khz & 300 hz: This is where things get interesting. Nothing changes with the channel shown above, but now the other channel is also driven to 1 volt but at 300 hz. This creates peak currents that are still within the OPA690’s specifications so it should not be overloading any part of the E11 or Mini3. Ideally one would just see a component at 300 hz at the level of crosstalk present in the amplifier. This isn’t distortion it’s linear (i.e. relatively pleasant sounding) crosstalk. But any other significant new distortion products are cause for concern. The E11 shows the expected crosstalk at 300 hz while everything else is still below about –90 dB but there’s a new “forest” of IMD products visible in the spectrum that were not present in the baseline graph above. Each spike is summed or added together to form the total amount of distortion—something I would have to write a custom script to get the dScope to calculate while leaving out the 300 hz crosstalk. In short, this is still a potentially audible amount of distortion due to the shear number of distortion products. The second graph shows the much worst result from the Mini3 where there are several distortion products above the –80 dB threshold and very likely sum to an easily audible value. See Virtual Grounds & 3 Channel Amps and the original Mini3 review for more. The third graph shows the vastly better result from using a real ground in the O2’s conventional 2 channel design. Notice the “forest” is gone in the O2 graph with mainly just the expected signal-related harmonics of the 1 Khz signal visible:

FiiO E11 THD N 1 Vrms 300 hz Left 1 khz Right (shown) 15 Ohms Ref 1 Vrms BW=22 Khz

AMB X-Audio Mini3 THD N 1 Khz Left 300 hz Right 1 Volt RMS 15 Ohms (ref 1 V) comments

O2 AC 1 Khz & 300 hz (other channel) Interchannel IMD 1V 15 Ohms

 

CHANNEL BALANCE: For a relatively inexpensive amp, the E11 had decent channel balance being just slightly over the goal of 1 dB at -45 dB. You can also see from the gap between the blue lines there was also about 1 dB of error at –22 dB. This is similar to the Mini3’s worst case error of –1.1 dB at –28 dB. The O2 was better with a worst case error of 0.6 dB at –55 dB:

FiiO E11 Channel Balance Tracking -22 -35 -45 dBr (ref 2 Vrms)

AMB X-Audio Mini3 400mv In Vol= 9 0'Clock (worst case balance error) (ref ~400mV)

 

GAIN: At the low gain setting the E11 measured 8.2 dB (2.5X) of gain (no EQ) as can be seen below with 387 mV input and 1 volt of output. This is identical to the O2’s default low gain setting. The second graph shows the high gain setting at 12.9 dB (~4.5X). This means, at high gain, with 500 mV of input from an iPod LOD (line output), and many portable players, you’ll get 2.25 Vrms out. This is well shy of the 2.9 Vrms the E11 is capable of and may not be loud enough for many headphones like the HD600/650. For comparison, the O2 has 16 dB (6.5X) of default gain at the high setting, and the default fixed gain of the Mini3 is about 14 dB (5X). These graphs also show respectably low distortion at 1 volt output into 150 ohms. For more see: All About Gain

FiiO E11 Low Gain Volume=100% 150 Ohms (ref 1 Vrms)

FiiO E11 High Gain Volume=100% 150 Ohms (ref 1 Vrms)

 

EQ: Here’s the frequency response with the EQ switch set to “0” (EQ off yellow), “1” (white) and “2” (blue). Both EQ options create a peak centered at about 50 hz with a 3.5 dB maximum boost on setting “1” and a 6.8 dB peak on “2”. both EQ options also raise the overall volume about about 0.6 dB at all frequencies. There is significant boost above 80 hz which may result in some “boomy” bass with many headphones. I personally prefer bass boost down lower that’s mostly nil by 80 hz but some may still find the E11’s options useful:

FiiO E11 Frequency Response EQ=2 EQ=1 EQ=0 Top to Bottom

 

SQUARE WAVE PERFORMANCE: The high speed scope shot below shows the E11 into Sennheiser CX300 headphones with 0.01 uF of added capacitance to simulate more difficult headphones. As was observed with the same very fast AD8397 op amp in the Mini3, there’s some ringing and signs of instability. What’s different is the Mini3 show even more ringing without any added capacitance into the same headphones. So the E11 is at least better behaved than the Mini3. The thick “fuzzy” horizontal portion of the waveform is RF noise from the DC-DC converter. The second graph shows the O2 on the same test with the same load. Note the lack of ringing, and the lack of “fuzz” as the O2 doesn’t use a DC-DC converter and uses a 2 stage design with properly compensated op amps better suited to audio use:

FiiO E11 10 Khz Square Wave Response CX300   0.01uF 2Vp-p

o2 agilent 10 Khz Sennheiser CX300 plus cap 2v p-p

SLEW RATE: As shown below, the E11 slews 4.5 volts in 710 nS which is a 6.6 V/uS slew rate. This is many times what’s required for any audio signal the amp will ever encounter. It’s worth noting it’s a bit slower than the Mini3’s measured 10.7 V/uS which is further evidence FiiO wisely applied some additional compensation, RF input filtering, or otherwise decided not run the AD8397 “wide open” as AMB chose to. The O2 has a 3.6 V/uS slew rate which is still well in excess of what’s required any recorded music for the output voltage the amp is capable of:

FiiO E11 10 Khz Square Wave Rise Time Max Output 600 Ohms

DC-DC CONVERTER NOISE: Here’s a zoomed-in view of the “fuzz” seen on the square wave response above. The scope measurement shows it has a frequency of 96 Mhz and it’s likely the DC-DC converter which is probably designed to run somewhere around a nominal 100 Mhz. You can also see it’s about 15 mV peak-to-peak or about 5 mV RMS. That’s only –43 dBu. Obviously nobody can hear 100 Mhz but if it were in the audio band it would be plainly audible. It can also be troublesome for FCC and/or CE approval for RF emissions:FiiO E11 RF Noise From DC-DC Converter at 96 Mhz and 12 mV p-p

 

CLIPPING PERFORMANCE: The E11 clipped cleanly and symmetrically at a bit over 8 Vp-p into 600 ohms. This shows the internal DC-DC power supply is likely around 9 volts (accounting for the inability of the AD8397 to swing all the way to the rails). The 100 Mhz “fuzz” is plainly visible on the entire waveform:

FiiO E11 10 Khz Sine Wave Clipping 600 Ohms

THIRD CHANNEL VIRTUAL GROUND: The large number of distortion products created by just a single sine wave in each channel is something the E11 shares with the Mini3. In the E11 the Interchannel IMD is much lower in level but the total sum of all the distortion could still be audible—especially with the complex interactions of real stereo music fighting between the channels via the shared OPA690 ground channel. The E11 likely uses a single ground and perhaps a better ground layout (the Mini3 uses a poor ground PCB layout and isolates the input and output grounds) but it still suffers from the shared OPA690—an op amp not very well suited to the task. As explained at the start of the Tech Section, the E11 would have been much better as a two channel amp with a DC-DC converter that produced two supply rails instead of just one. See Virtual Grounds & 3 Channel Amps and the original Mini3 review for more.

TECH COMMENTS: The E11, in short, measured better than I thought it would although, just like the Min3, it failed to deliver its claimed power output. When driving low impedance headphones its true clipping point is a bit hard to determine likely due to either an undersized DC-DC converter and/or wimpy power supply capacitors. Even being generous and using the 1% THD points with only 15 minutes run time on a fully charged battery it it still fell significantly short of the claimed output. It’s best to use 16 ohm headphones that need no more than 1 Vrms and 32 ohms headphones that need no more than 1.7 Vrms. I’m also concerned about the very large number of Interchannel IMD distortion products created by the third channel as discussed in the paragraph above. It’s interesting the E11 performed significantly better in nearly every area than the Mini3 despite both using a 9 volt power supply, similar 3 channel design, and the same op amps. As mentioned earlier, it’s more evidence implementation is more important than the components used.