REVIEW: iFi GO Bar - Portable DAC / Headphone Amp (Cirrus Logic DAC, Balanced & Unbalanced).

My friend AudioPhil is really into his headphone gear. Recently he acquired the product above and asked if I might want to play with this for a few days... Sure, this looks interesting!

This is the new iFi GO bar (around US$320), a portable headphone DAC/amp. The USB interface is a USB-C form factor (480Mbps USB2.0 Hi-Speed), and there are outputs for single-ended/unbalanced 3.5mm TRS and balanced 4.4mm pentaconn. Interestingly, the company (iFi, subsidiary of AMR/Abbingdon Music Research) does not specify which DAC chip is used inside. Suspicion is that it's likely the Cirrus Logic CS43198 or CS43131, both are recent chips circa 2017 capable of "pseudo-differential" outputs and low-power consumption <40mW with rated THD+N of -115dB, DR 130dB, maximum PCM 32/384, and DSD256. The specs look good but don't get too excited about these ideal numbers when implemented in a USB dongle.

I don't know what's the point of being so secretive about the DAC chip used (Crutchfield says it's the 43131).

Here's a better look at the side where we see a small switch for "iEMatch" which can be applied to either the 3.5/4.4mm outputs for sensitive IEMs:

There are +/- volume buttons. The last round button on the right is for changing digital filter settings (press down >3sec and then +/- to cycle through filter settings indicated by LED color), and to turn on/off the XSpace (3D virtualization), and XBass+ (bass boost).

There are small colored LEDs to indicate samplerate and whether XSpace and XBass+ are on.

Notice that I'm angling it a bit against the light  to highlight the text which can be hard to read due to low contrast and small fonts.

The matte surface which has a dark greenish hue depending on light conditions looks quite nice (there's an all-gold version as well if you're into precious metals). It doesn't pick up fingerprints and the metal construction feels good in the hand. At 65 x 22 x13mm (2.6 x 0.9 x 0.5"), weighing just less than 30g, it's small and light enough to carry around with no problem. Notice the 5" short cable that comes with it with a USB-C to USB-A adaptor.

Small LEDs lit to indicate in this example that I'm playing a 192kHz signal, blue XSpace and reddish XBass+ features turned on. Due to the small size, some might find the LEDs to be dim especially outdoors. The small darker text on dark background also can be a challenge to read depending on lighting and visual acuity!

Yeah, it has MQA support (not tested).

As you can see from the side view, it's got a switch to tell it when you're using an IEM so as to reduce the output level, also there's an XBass+ bass boost, let me show you what they do:

As you can see, the iEMatch feature turned down the gain by -6dB to accommodate sensitive IEMs and might increase the output impedance (presumably this is done through a voltage divider as per this ASR discussion, effectively reducing the efficiency of the amp). And the XBass+ feature basically gives a boost in the bass starting around 200Hz and up to +7dB at 20Hz; potentially a pleasant EQ for many headphones. Otherwise the DAC/amp has a nice and flat frequency response.

The XSpace feature is a spatialization effect to increase the sense of 3D imaging. The iFi marketing page suggests it's not standard crossfeed but seems to act like it (we can see an increase in stereo crosstalk - see RightMark results further down). The manual states that both XBass+ and XSpace are implemented in analogue circuitry, not as DSP.

I made sure the firmware was updated to the latest version at the time of measurements - the device was shipped with version 1.48, updated to 1.70 for these measurements:

One last important setting that can be found in the product manual and which AudioPhil pointed out to me - iFi has this "Turbo Mode". You activate it by holding down the + & - volume buttons for >2 seconds and the LEDs will show a pattern to indicate on/off. With it on, the device will push gain up another +6dB.

Since I am interested in distortion and maximum power this little device can produce, most of the results here will be with Turbo Mode "ON". I will specify when I turn it off.

I. Measurements

To start, let's have a look at the output impedance from this little device (iEMatch OFF). Using my usual procedure with 7 measurements across the 20Hz-20kHz audible spectrum, here's what the impedance curve looked like from the single-ended 3.5mm jack:

An excellent result of <0.25Ω output impedance basically across the audible frequencies (iFi lists it as just "<1Ω"). You should not have any issues with any headphone you want to pair this amp with. Regrettably, I did not check what happened to the impedance with iEMatch turned on which could increase a little - will have a closer look if I borrow the device again.

When you first power on the DAC, the volume defaults to a low 80mVrms output with a 0dBFS sine wave unbalanced or double at ~160mVrms balanced. This is good with sensitive headphones like the CCA C12 IEM, using the phone volume for control of output level (the IEM's inline remote does not work, unfortunately I did not check the microphone function). If this default output level is too low, use the hardware volume buttons, you can increase this by +34dB(!), or maximum of 4Vrms unbalanced, up to around 8Vrms balanced with Turbo Mode ON (2Vrms unbalanced, 4Vrms balanced with Turbo OFF).

Let's have a peek at the output waveform itself. Here's a 0dBFS 1Hz sine direct to oscilloscope with input impedance 1MΩ, Turbo Mode ON:

Notice with a 0dBFS 1kHz sine signal, full volume on the DAC/amp with Turbo Mode ON, the output clips. You have to dial the volume down 1 "click" or about -1dBFS to prevent clipping. Otherwise the stereo channels are well balanced which is excellent. With Turbo OFF, maximum output level will be at 2Vrms, there will be no clipping. So essentially with a high impedance headphone, this means that Turbo will give you up to +5dB extra output without clipping.

This dongle DAC provides 4 filter options - here they are with the iFi description from the manual:

Bit Perfect​ (BP) [misnomer, should be "NOS"] – no digital filtering, no pre or post ringing
​Standard (STD) – modest filtering, modest pre and posting ringing​
Minimum Phase​ (MP) – slow roll-off, minimum pre and post ringing​
Gibbs Transient Optimized (GTO) – minimum filtering, no pre-ringing and minimum post ringing​

And here are the measured characteristics using my typical "Digital Filter Composite" (DFC) graph based on the Reis test used over the years with DAC testing:

From what I can tell, the BP, STD, and MP are Cirrus Logic filters as described in the datasheet with BP being "NOS=1" mode, "STD" roll-off is steeper than "MIN" which allows a significant amount of imaging to pass through from the 19 and 20kHz tones, but both are just variations of minimum phase. Notice also the intermodulation distortion with that 19+20kHz signal. The GTO filter looks like a custom setting programmed by iFi which upsamples to 352.8/384kHz but intrinsically a type of very gentle minimum phase filter as well.

I find it curious that they described the "Standard" filter as having "modest" pre- and post-ringing so I was expecting some kind of linear/intermediate phase setting. As you can see, there is actually no pre-ringing behaviour for any of these filters on the impulse response! (As a reminder, do not be afraid of the pre-ringing on impulse response, it doesn't show up with music playback unless you're playing badly clipped or poorly filtered material, which is not a sign of high-quality digital music to begin with.)

 

Absolute phase is maintained by each filter. Other than the "BP" (NOS) setting, they all show some overloading with the 0dBFS wideband white noise signal. Notice the purple "Digital Silence" is well below the other curves. This is because when nothing is playing, the DAC will auto-mute hence the very low apparent noise level. [In the past, by implementing a mute function, some brands were able to tout excessively high signa-to-noise numbers used in marketing. For a low-power device, muting might be beneficial to save energy I suppose.]

For me, the "Standard" (STD) setting works best to remove ultrasonic imaging so I'll use that as the basis for the tests.

Using the Drok USB power meter, I noticed the dongle is quite energy-efficient (5V USB voltage): 0.11A idle, 0.18A playing 0dBFS 1kHz sine into 560Ω high impedance load, and 0.45A into low impedance 20Ω load. The metal casing gets only slightly warm after 60 minutes of playing music; enough to know that it's in use, but far from any discomfort to touch.

DAC Resolution:

Let's consider the potential resolution for this DAC. The best we can achieve is with the balanced 4.4mm pentaconn, 100% hardware output level. As a start let's just look at some RightMark results at 24/44.1, various filter settings, and also XBass+ and XSpace on or off. This was done with Turbo Mode ON:

With a 24-bit signal, notice that noise level and dynamic range show that this DAC is capable of better-than-16-bit resolution. Numbers bounce around a little bit and there are oddities (like that IMD+N for the STD filter) I think due to the auto-mute which causes some timing issues with the RightMark signal detection at times. Let's have a look at the frequency response and crosstalk graphs:

On the frequency response we see the effect of that XBass+ setting. Also the -3dB at 20kHz roll-off using the "BP" (NOS) setting. The only change I saw with the XSpace feature was in the stereo crosstalk dropping significantly which is what we would expect with some kind of crossfeed mechanism.

Staying with RightMark, we can now compare the performance of this DAC (balanced out, STD filter, 100% hardware volume) with other devices using a hi-res 24/96 signal:

Overall, not bad for a dongle DAC. The Oppo and SMSL are desktop devices with full XLR output. The Chord Mojo has its own internal battery and better filter mechanism, more expensive, less portable weight and size. As you can see, the resolution of the iFi is much superior than the Cayin RU6 R-2R DAC running in OS mode measured not long ago.

And here are some graphs to compare:

Notice the iFi GO Bar's noise graph looks odd. When the device doesn't detect a signal, it mutes the output. This muting mechanism is also why there's a larger difference between the "Noise Level" and "Dynamic Range" numbers compared to the other DACs. I'm not sure if the increase in IMD+N on the sweep might be an artifact; at least the increase starts >30kHz.

Next, let's run a 1kHz THD+N vs. Generator Level stepped sine to see the harmonic distortions across output levels:

We see that this DAC is capable of up to around -93dB THD+N at -10dBFS or so with the 4.4mm balanced output and Turbo Mode ON.

Similarly, if we look at the 3.5mm unbalanced output, we see that at best, it's capable of about THD+N -90dB at -16dBFS:

For both balanced and unbalanced, the right and left channels are well matched with basically identical distortion characteristics. Using those unbalanced stepped sine results, we can look at output linearity:

0.5dB deviation down at -111dBFS level is excellent for something like this.

Looking closer at the FFTs, here's the -60dBFS 1kHz for measurement of dynamic range:

The 4.4mm balanced output provides a boost in output levels (+6dB) but does not improve dynamic range.

At -12dBFS level; this is about the best result one will see with this DAC on THD+N:

2nd and 3rd harmonics predominate with the 3rd typically stronger at the higher output levels.

Notice that although the output level is higher in the balanced mode, the noise level is not improved (but THD is better slightly). Think of the balanced output as being able to provide more power but not better resolution although balanced might do a better job at noise removal and freedom from hum I suppose.

Let's have a look at some J-Test, jitter results:

Jitter is not a problem these days with good DACs. There's a slight "skirt" under the primary signal suggesting some low-level random jitter and we can see the noise floor isn't as low as better DACs out there. It has been years since I have seen truly "bad" jitter performance (as per jitter demo). I call BS to the guys who think they can hear jitter issues with reputable products these days, never mind the nonsense about jitter with ethernet switches and other voodoo (like this).

Headphone Amp Resolution under load:

While we could use it as a DAC, this dongle obviously is meant to act as an amplifier connected to headphones with varying impedances. Using loads at 20/75/560Ω, we can create a picture of distortion at increasing output levels (graphed in Vrms); these were done with the unbalanced 3.5mm output with Turbo Mode ON:

Notice that the lowest THD+N is achieved around 0.7-1Vrms ranging from -87 to -92dB. At 1% distortion, the amplifier is able to provide >300mW into a low impedance 20Ω load, >150mW into a medium impedance 75Ω load, and >3.8V into high impedance 560Ω. These are in line with the published specs from iFi.

You might be curious about Turbo Mode OFF:

With the maximum gain pulled down by -6dB, distortion no longer goes to 1% by 0dBFS, and maximum power has been pulled down to 200mW into 20Ω, 53mW into 75Ω, and output voltage capped at 2Vrms. Lowest THD+N is achieved at around 500-800mV, remaining the same as with Turbo Mode ON with THD+N -90dB.

It's good to also do a THD vs. frequency stepped sine to see if the harmonic levels fluctuate across the audible frequencies. Frequency-dependent harmonic levels might affect the tonality of the music. Let's use the moderate impedance 75Ω load for this one, and we'll examine the harmonics at -3/-8/-15/-20dBFS levels (24/96) with hardware volume at 100%:

Interesting. Notice that the odd 5th, 7th, and 9th harmonics increase with frequencies as observed in the -8/-15/-20dBFS graphs. For example, on the -15dBFS graph, all three of these higher-level odd harmonics rise gradually from 250Hz and are above the 2nd by >2kHz which is within the most sensitive portions of the audio spectrum for human hearing (as per Fletcher-Munson). While I'm not going to lose any sleep over this, be mindful that the best DACs tend to have relatively consistent relationships between the harmonics across the audio spectrum (for example, the recent S.M.S.L. DO100 DAC). My suspicion is that this might be a property of the Cirrus Logic chip.

I don't have dummy loads for the balanced output. There's no reason to doubt iFi's claims that it can achieve 475mW into 32Ω and 7.2V into 600Ω at 1% distortion.

Before leaving this section, let's get a triple-tone (48/960/5472Hz) TD+N at 0.5V into 20Ω (12.5mW):

Around -75dB Triple-Tone TD+N isn't bad. With this, we can compare to other headphone amps like the Drop + THX AAA 789, one of THX's Achromatic Audio Amplifier designs, scoring an excellent -93dB on this same measurement of noise and distortion (including intermodulation amounts). That's what a high performance desktop headphone amplifier can do even with unbalanced output.

And finally, a quick look at the 1/10 Decade Multitone 32 (discussed here) at 0.5V into 20Ω as well:

We're looking at better than 85dB noise/distortion-free range in the 32-multitone. Quite a bit of low-level "hash" that doesn't get cleaned up with 8-sample averaging. I turned on the peak tracing also. As you can see, even after 80 seconds of audio capture to get the 8 averages (96kHz sampling rate, 1M-sample FFT), there are no significant anomalies in the peak tracing compared to the average, demonstrating consistent playback quality with no evidence of excess spurious noise or distortion over time.

Notice that for both the Triple-Tone and Decade Multitone tests, on this unit, the right channel is a little bit cleaner.

DSD Playback:

While I don't use DSD much, it's good to at least confirm that it works without excess noise or unusual instability at the margins of output level. Here's a look at the -2.9dBFS/+3.1dBDSD 1kHz signal in PCM compared to DSD64/128/256 (discussed here); playback device was the Raspberry Pi 4 "Touch" using Volumio, DoP output:

Other than the ultrasonic noise highest with DSD64 as expected, results are equivalent, indicating basically identical performance whether PCM or DSD. I listened to a bit of Sting's Brand New Day (1999, SACD rip to DSD64 - love the track "Desert Rose") and it sounded good, no glitches or sound quality issues.

Subjective Impressions & Summary:

Listening to some Guardians of the Galaxy Awesome Mix Vol. 1.

I had this device for 6 days to run tests and listen. This was enough time to get a sense of how the DAC/amp sounds, try out a few headphones/IEMs, and see if there were any general usability issues (no issues while torture-testing!). I'll have to leave it for owners to tell me if there are any problems in longterm use. For consistency I stuck with a similar playlist of music and headphones as I did with the Cayin RU6 R-2R DAC/amp: Dali CD Volume 3, Pure Ella, Dawn FM, Blues At Carnegie Hall, Les Mis Live!, The Greatest Showman OST.

Without question, this DAC/headphone amp provided a better sonic experience than the Cayin RU6. Much less distortion (none of that haze I discussed), good clarity, punchy sound. It's able to drive and highlight the sense of space with open headphones like the AKG Q701 shown above. Vocal tracks like the Ella Fitzgerald album was rendered with excellent tonality and nuance. On Pure Ella's simple vocal + piano accompaniment, the contrast between the voice and background piano parts was very good with the impression of a genuine spatially layered performance was there even though for me headphone listening still doesn't produce the soundstage with actual depth as I would hear through good speakers in a room. The iFi is also better than the AudioQuest Dragonfly Cobalt which sounded less dynamic and clearly had lower output power for times when you might want to pair it with large circumaural headphones. The Cobalt, while still able to deliver volume, struggled sound-quality wise when paired with the Q701 tested back in 2019. This iFi GO Bar had no difficulty driving the AKG Q701 to deafening levels and likewise adept at handling sensitive IEMs without buzz/noise even without using iEMatch; this versatility IMO is impressive.

As a small USB-powered mobile device, IMO, the absolute lowest distortion is not the most important characteristic - you're unlikely going to be using this in a treated sound room with low ambient noise and full-range speakers as a DAC. What is more useful is that the device produces enough power with adequate clarity when one is out on the move and in relatively noisy environments, and versatile enough when using different headphones. In this regard, I think the iFi GO Bar succeeds.

Except when paired with the most demanding headphones, it's probably fine most of the time to keep the Turbo Mode OFF. Physically it's small, light, uses little power, barely gets warm (uncomfortable in shirt pocket if it gets hot!), achieves excellent <1Ω impedance across audible frequencies, and provides a few amenities like the bass boost (XBass+ comes in useful with some headphones) and XSpace crossfeed-type function (I found the effect subtle and not intrusive). The iEMatch feature is a useful switch to help optimize the powerful amp with very sensitive >100dB/mW headphones/IEMs; it will drop the gain -6dB and should reduce background hiss if you run into that issue.

While this DAC/amp is capable of better-than-16-bit dynamic range performance (not so the Cayin RU6 R-2R), the -90dB THD+N is not going to set any low-distortion records. If we simply ranked headphone amplifiers by this single number, this dongle probably would not rank particularly well, I suspect. However, remember that harmonic distortion is not strongly audible in music (until quite high like >0.3% or -50dB, as per our blind test awhile back which I think is in the ballpark across various studies). Furthermore, most of the device's harmonics are in low-order 2nd and 3rd components which make it even less objectionable (and might even have some euphonic character).

In terms of potential improvements for the future, clearly the THD+N can be better; something like aiming for THD+N of -100dB, at 1Vrms into a 20-30Ω load would be a nice "benchmark" of sorts. This is the first time I've tested a Cirrus Logic DAC chip in a dongle so I'm sure this plays a part in the distortion characteristics I'm seeing.

Personally I see no need for MQA these days except for the few folks who still subscribe to TIDAL HiFi+, and saving most consumer some money by dropping the license would be good IMO. Also, I think it would be nice for iFi to offer a linear phase, fast roll-off digital filter setting which would be technically more accurate than this group of NOS and minimum phase options (enough with the silly fear about "pre-ringing" already!). At least allow the user to select the standard linear phase filter available in the Cirrus Logic DAC chip. Even better, maybe perform high quality linear phase filtered upsampling to supplement the custom "GTO" setting (which upsamples to 352.8/384kHz) and put that 16-core XMOS microcontroller to better use than MQA-decoding; a kind of "Chord-lite" with even 1000 taps should be excellent already. ;-)

At the current price of US$320, the AMR/iFi GO Bar is not the least expensive USB dongle DAC/headphone amp out there, but I don't think it's an unreasonable package given the overall feature set and output power capabilities. If you want some "bling" go check out the 10th anniversary GOld edition:

iFi stock image. This special edition costs US$500.

One last thing. While I believe that for high-fidelity DACs like desktop devices, we can make a link between the DAC chip performance and output quality since the job of the device is usually to maintain the ideal performance of the converter, we cannot necessarily expect the same with these low-power USB dongles. This iFi is an example; it really doesn't matter that the Cirrus Logic CS43131 chip is rated at -115dB THD+N and dynamic range 130dB(A) in the specs sheet; in the context of the rest of the headphone amp, this is too much to ask for this USB-powered small package (maybe one day). The "sound" therefore is less a reflection of the DAC chip and more determined by iFi's amplifier circuitry.

Thanks AudioPhil for letting me borrow another one of these devices to check out and measure. ;-)

--------------------

To end off, I thought I'd take a picture of the current lineup of balanced / unbalanced headphone connectors that I have here among my devices and adaptors. I've also labeled the typical pin connections for R/L/GND and +/- for comparison:

For a balance of convenience and robustness with mobile devices, I think the 4.4mm pentaconn is a nice form factor. XLR 4-pin clearly being too large for mobile use, and the 2.5mm TRRS feels really tiny and a bit flimsy. IMO, standardization using the 4.4mm pentaconn connector would get my vote.

In the next few weeks... Let's go into a detailed look and listen at this device:

Sabaj A20d 2022 Version DAC with headphone out.