• M2Tech hiFace Asynchronous USB To S/PDIF Converter Review

    The M2Tech hiFace has received a lot of press this year. It was one of the first very inexpensive asynchronous USB to S/PDIF converters to support all sample rates from 16/44.1 kHz through 24/192 kHz. The hiFace's good specs, good technical design, support for high resolution sample rates, and $150 price tag has had users from all over the world going gaga. While there is no such thing as bad press too much good press can make it very hard for a product to live up to expectations. Such is the case with the M2Tech hiFace. I tried for several months to pull every ounce of sound quality out of the hiFace. I began to wonder if I was the only person on Earth unsatisfied with this converter. I have no qualms about saying the hiFace, through no fault of M2Tech, is overrated. Fortunately this has nothing to do with value. At $150 it's well worth the price and has a very high price to performance ratio.


    Preliminary Notes

    There is no sense in writing a confusing review that interweaves terms like good performance, disappointment, overrated, and great value only to leave readers wondering what I really think. Let me lay some groundwork before going deeper into the hiFace review. As many Computer Audiophile readers know terms like overrated and good performance are not mutually exclusive. Neither are the terms great value and disappointment. Also the conclusions reached by me in my listening room with my components don't say anything about another individual's conclusion reached in his home or even in my listening room. There are so many variables involved when judging an audio component. Readers should only use reviews and others' comments as single data points that have nothing to do with their individual opinions and conclusions.


    Got A Lot Going For It ...

    Designed and assembled by M2Tech in Italy the hiFace asynchronous USB to S/PDIF converter looks unbeatable on paper or screen. Async USB with dual crystal oscillators and support of sample rates up through 24/192 kHz are more than many manufacturers can say about their converters. Add the $150 price to this list and most of the competition falls to the wayside. On paper.

    The hiFace is nearly a self explanatory device. One end has a USB connector that can only fit into a computer's USB port. The other end has either a coaxial RCA or a BNC digital output. There are no switches or power cables to contemplate while physically connecting the hiFace to a computer and audio system. A single electrical digital cable connects the hiFace to an external DAC completing the physical setup.

    Asynchronous is currently the buzzword of all buzzwords. If a component does anything asynchronously manufactures frequently label it with the async buzzword. The hiFace is a true async USB device as it operates in asynchronous USB transfer mode. Async USB transfer mode has nothing to do with asynchronous sample rate conversion (ASRC) even though some manufacturers would like listeners to believe ASRC is an equivalent competing technology addressing jitter reduction. Some manufacturers just use the plain asynchronous label and let consumers try to decipher what that means with regard to the component in question. The bottom line is these two technologies are vastly different and can have a major impact on sound quality.

    M2Tech's async USB implementation is pretty solid on paper. The hiFace uses two separate quartz precision oscillators instead of a PLL with a single oscillator and synthesized frequency. This enables very accurate clocking with less jitter or timing errors. Two oscillators allow the hiFace to have separate clock generators for 44.1 kHz and 48 kHz sample rate families. The 44.1 kHz family consists of 44.1, 88.2, and 176.4 kHz and the 48 kHz family consists of 48, 96, and 192 kHz. Nearly every engineer I talk to about this area of HiFi suggests a PLL with synthesized frequency based on a single oscillator that cannot be a multiple of 44.1 kHz and 48 KHz will result in much higher jitter. Since high end audio components shoot for extremely low jitter measurements in the single digit picoseconds many engineers will only opt for dual oscillator configurations similar to the hiFace. One notable exception is the Weiss Engineering DAC202. It uses the Jet PLL and synthesized clock frequencies to produce excellent results.

    In addition to this very good technical design the hiFace supports every relevant sample rate. It wasn't long ago that extracting quad speed sample rates of 176.4 and 192 kHz from a laptop was nearly impossible because there weren't any acceptable devices like the hiFace. If listeners wanted the higher sample rates they had to install a card like the Lynx AES16(e) or RME 9632 into a desktop computer. The hiFace was one of the first widely accepted devices in the audiophile community to free listeners from the unsightly and frequently noisy desktop computer.

    There's no denying the hiFace has a lot going for it with its async USB transfer mode, support of all sample rates, and very inexpensive price tag. If it wasn't for the music and the fact that I want to listen to said music at the highest quality possible the M2Tech hiFace would certainly make the Olympic podium (gold, silver, or bronze).


    ... But Far From Ideal

    - Software

    Driver: Noun

    • the operator of a motor vehicle
    • someone who drives animals that pull a vehicle
    • driver (a golfer who hits the golf ball with a driver
    • a program that determines how a computer will communicate with a peripheral device
    • number one wood (a golf club (a wood) with a near vertical face that is used for hitting long shots from the tee)

    Source [Princeton University]

    Until the recent release of Apple's OS X 10.6.4 proprietary drivers were necessary to reach the highest sample rates via USB audio devices like the hiFace. Windows based computers still require proprietary drivers for playback of high sample rates (176.4 and 192) because Windows does not support Class 2 Audio. When the hiFace was released proprietary drivers were required by all operating systems and USB hardware to reach these sample rates. M2Tech had no choice but to use its own drivers for the hiFace to function with Windows and Apple's OS X. In addition M2tech designed the hiFace with specific hardware that requires proprietary drivers even if the operating system supports Class 2 Audio. For example the hiFace will not work on a Mac running OS X 10.6.4 without installation of M2Tech's driver. Whereas devices like the Wavelength Audio WaveLink work as designed on OS X 10.6.4 without proprietary drivers at all sample rates.

    M2Tech's proprietary driver is necessary but not sufficient. Several of the first iterations of the driver required the use of Foobar2000 and manually placing specific files (dll) in a certain location on the computer. Each release has drastically improved the ease of use and eventually added options like WASAPI support. Now a simple double-click -> Next -> Next -> Reboot routine is all that's required. The insufficient part of the M2Tech driver comes from two fronts. Lack of an easy uninstall without contacting M2Tech for a special command run via the Terminal application and the confusing nature of M2Tech's driver delivery.

    It's entirely possible to use a Mac without uninstalling the hiFace driver. It's benign as far as I know. But when troubleshooting an audio issue it's very nice to rule out possible causes by uninstalling software. M2Tech's current hiFace driver removal process is unacceptable.

    On several occasions hiFace users have installed the incorrect version of the hiFace driver only to suffer frustrating and time consuming consequences. Just because most people haven't had an issue with this doesn't make it OK. Identifying the correct driver on the M2Tech website isn't rocket science and has been made easier over time. However, a simple line of code in the installation process could let users know if the downloaded driver was incorrect for their operating system. For example if someone downloads the Apple OS X 10.4 version of the software when they really need the OS X 10.6 version this operating system "pre-flight" check would remove the possibility of such frustrating issues before they happened. The last thing computer based audio needs to deliver is frustration to end users. Especially when it could easily be avoided.

    The fact that drivers are required, the inability to easily remove the driver easily, and the unneeded driver confusion have caused real world problems as evidenced by the users at CA and other sites. These users have sought help with installation, uninstallation, and related issues frequently after several hours of attempting to solve the issue themselves.

    Note: The vast majority of hiFace users have not experienced the aforementioned issues. I raise the issues only because they've appeared several times in the real world and they could be avoided altogether.

    - Hardware

    The hiFace asynchronous USB to S/PDIF converter uses the Cypress Semiconductor ezUSB design. ezUSB provides the component designer (M2Tech) a Windows, OS X, and Linux base driver that operates in bulk mode. The designer then plugs in code for each operating system that creates whatever device is needed. Devices like the Wavelength Audio WaveLink and Halide Design Bridge require no proprietary device driver. These units use the driver supplied by the operating system in a true plug n' play fashion. Although the Bridge does not support quad speed sample rates and the WaveLink currently does not support quad speed on the Windows operating system. As the saying goes, there's no free lunch.

    Correction: I was just informed a Windows driver is available on the Wavelength Audio website that enables the WaveLink to support quad speed sample rates.

    Internally the hiFace uses three DCDC converters to power the Cypress USB controller, the dual oscillators, and the SPDIF converter. Unfortunately the ground of the digital output is connected via 1 kOhm to the USB ground instead of being galvanically isolated which is highly preferable on the S/PDIF output. If the digital input on a listener's DAC is not galvanically isolated either then computer's power supply will be connected to the audio system via the 1 kOhm on the digital input ground pin. This is a very good reason to use a MacBook Pro or different laptop running on battery power eliminating the direct connection to a noisy and cheap switching power supply.

    The build quality is nothing to write home about and is probably what most audiophiles expect for a $150 device that offers quite a bit of functionality. I recommend using a little USB extension cable that connects between the computer and the hiFace. The hiFace is much wider than a USB port and may block or interfere with a neighboring USB port. Also, the extension reduces strain on the USB port and hiFace itself when heavier S/PDIF cables are used or when cables must be routed awkwardly to the audio component. Frequently pulling on the hiFace isn't a good idea. Especially if connected directly to the computer's USB port.

    It's hard to definitively say if using the operating system's built-in USB drivers or different hardware design decisions would have a big impact on sound quality from the hiFace. I can say the async USB to S/PDIF converters I've used, that don't require proprietary drivers, sound better and more accurate. More on sound quality a bit later.


    Music Servers

    During the review period I used several different music servers. The two main configurations used were based on a Mac Pro and the C.A.P.S. server [Link].

    The C.A.P.S. server runs Windows 7 32-bit and J River Media Center v15. The server accesses music on a NAS drive stored in WAV, AIFF, and FLAC formats. I used the Kernel Streaming and WASAPI output modes in J River. ASIO was unavailable with the hiFace and ASIO4ALL doesn't currently support quad speed sample rates of 176.4 kHz and 192 kHz. The hiFace Windows driver in use at the end of the review period was version 1.0.3.

    The Mac Pro runs OS X 10.6.4 and iTunes with and without Amarra version 2.1 (4244). It also access music on the same NAS drive as the C.A.P.S. server and accesses some music stored locally. The hiFace OS X driver in use at the end of the review period was version 1.0.45.

    I compared the hiFace to several components. The components range from a couple hundred dollars more expensive than the hiFace to several thousand dollars more expensive. These are the asynchronous USB to S/PDOF converters on hand during the review:

    • M2Tech hiFace
    • Halide Design Bridge
    • Wavelength Audio WaveLink
    • dCS U-Clock


    Sound Quality

    I had the M2Tech hiFace here for several months. Quite a bit longer than normal component review periods. The only reason for such an extended period of time was so I could try every way I knew to squeeze the last ounce of sound quality out of the unit. Upon its arrival I immediately noticed substantial sonic differences between the hiFace and the Lynx AES16e internal digital audio output card I was using at the time. With over $500 difference between the two components, major design differences, and the fact I had just added the hiFace to my system I simply added this experience to my digital notebook as a single data point among many I would gather throughout the review period.

    A couple weeks went by and I'd used the hiFace off and on in addition to using the Halide Design Bridge. The hiFace just didn't sound as good as everyone online and in personal conversations was claiming. Since I had already tested to make sure the digital output was bit transparent I knew I wasn't' altering the bits before entering the hiFace. I wondered what was going on so I emailed a few first rate engineers with decades of high end digital audio experience. CA readers would be surprised at how many engineers from top high end audio companies purchased the hiFace to test in their own systems. I was not interested in using their opinions to influence mine whatsoever. I just wanted to compare some external data points to my personal experience. (If many groups of people are claiming a color is red but I see it as blue it's never a bad idea to talk to some people who've reached their own independent conclusion). The possibility of groupthink was ever so present in my mind. Without revealing the details of private conversations these engineers told me a bit about how the hiFace was designed and what they thought about the device. Each one of them said the hiFace sounded pretty good for $150. If I gained anything from these conversations it was a bit of knowledge about the hiFace from sources outside of M2Tech and some data points from independent thinking engineers.

    In addition to several async USB to S/PDIF converters I used a few different DACs during the review period. I used my main DAC, Berkeley Audio Design Alpha DAC, as well as the Weiss Engineering DAC202 (now sent on to the next reviewer) and the Esoteric D-07.

    Through the Alpha DAC and DAC202 the sound quality via the hiFace was lackluster and uninspiring compared to the other converters on hand. After listening to something very good it's hard to take a step down in quality. The differences tend to be accentuated. If I didn't have the other units on hand I likely would have said the hiFace sounds a bit too dark for my taste but in general offers good performance. It certainly is not a bad sounding component by any means. It just doesn't match the level of the competition in my listening room.

    A week after the Esoteric D-07 arrived, and I had listened through the DAC enough to get a handle on its sonic signature, I began comparing the converters through the galvanically isolated RCA digital inputs of the D-07. Using the hiFace did not yield positive results at all. The sound I heard was really veiled and really dead. This was evident after long term listening sessions and A/B comparisons. I usually don't gain much from quick A/B comparisons and but I tried the method anyway because the D-07 and J River made it very easy. Using the hiFace and Halide Design Bridge configured as separate zones in J River MC I synchronized both zones and sent the output to two different coaxial S/PDIF inputs on the D-07. The d-07 doesn't offer BNC digital inputs. This worked well since the hiFace I reviewed was the RCA version. Once playback commenced I was able to switch inputs on the D-07 and hear the same audio stream as the previous input. Pretty cool, but ultimately not the best or most revealing way to review components in my opinion.

    I performed much more extended listening using all the S/PDIF converts and the Esoteric D-07. In every case music through the hiFace was much more veiled and dead. No matter what type of music I played from Reference Recordings HRx 24/176.4 material (via WaveLink only) to the new single mic'd John Mellencamp album produced by T-Bone Burnet as soon as I started using a non-M2Tech converter the sound opened up and the level of clarity was wonderful. Almost like I removed cotton from my ears. At the end of the review period I really concentrated on comparing the Halide Design Bridge to the M2Tech hiFace. I used Windows, OS X, iTunes, Amarra, J River etc... to make sure I reached an accurate conclusion. Every comparison ended the same way. Using the Bridge was like removing cotton from my ears as the greater level of clarity and detail were readily apparent. I try very hard not to make unsubstantiated mountains of difference out of realistic mole hills of difference as can be the case in so many audiophile conversations. I admit I am just as guilty of hyping a component as the next guy when we are sitting around chatting. When it comes to publishing a review, that is part of my permanent record :~), I never want to mislead a reader by making a big deal out of nothing. It's bad for both of us and the manufacturers involved. That said, with the components used during this review in my listening room I state unequivocally that the hiFace did not match the performance of the other asynchronous USB to S/PDIF converters. The difference was not subtle. I urge everyone considering the purchase of a converter like the ones used in this review to give them all a shot in a familiar environment.

    Note: As shown in the measurements below the hiFace's output voltage is 2.328 Vpp. This is higher than the standard 0.5 Vpp. It is possible the D-07 does not handle higher voltages as well as the Alpha DAC or DAC202. The bottom line is readers should look at the specs of their DAC and test components in person before purchasing.



    The M2Tech hiFace entered the audiophile scene as a little known device from Italy. It soon surged to the top of several recommended lists. Groups of audiophiles on the Internet couldn't get enough hiFace-time. None of these hiFace users are wrong. It's a good component if it sounds good to the individual listener. Period. The hiFace does offer good specs and features on paper. There's no doubt the M2Tech design team had the right idea. After several months of listening and comparing I think M2Tech's implementation is a bit underwhelming. The hiFace offers good stand-alone performance and value while simultaneously disappointing me. The fact that I believe it's overrated has just as much to do with hiFace users' opinions as it does the hiFace's performance.

    I'm going to end on a positive note. I wish no ill will to M2Tech or any manufacturer. We are all part of the same industry and wonderful hobby. I hope M2Tech continues the success of the hiFace with its new upscale Evo product. At $150 the price to performance ratio of the hiFace has got to be at the top of the charts. Audiophiles used to spending tens of thousands of dollars for an extra 0.01% of performance may be a bit disoriented by the hiFace's value.

    One more time, don't take my word or anyone else's word to be the final answer. When in doubt check it out.



    __________________________________________________ ______________

    Product Measurements (Using BNC version of hiFace):

    Output Voltage [Image Link]
    The output voltage with a 75 Ohm load is 2.328 Vpp. This is a lot higher than the nominal 0.5 Vpp desired at the digital input of most DACs. Sound quality may vary depending on how well a DAC handles this higher voltage. Some digital inputs can be over driven by this 2.328 Voltage PP when they amplify the digital signal, with an HC04UB inverter, that is the regular SPDIF recommendation for an input device.

    Output Resistance [Image Link]
    An approximate 73 Ohm output resistance can be calculated using the hiFace's 4.684 Vpp (without 75 Ohm load) and 2.328 Vpp (with 75 Ohm load). This is close enough to 75 Ohm for most engineers.

    Status Bit Information [Image Link]
    The transmitter in the HiFace always sends 48 kHz sample rate information in the Status Bits, no matter what sample rate is really playing. This is not really as big of problem to consumers as it is in the professional audio world. Consumer DACs by Theta, some by Mark Levinson, and others with a frequency synthesizer as a secondary PLL or those using use what is called slaving the SPDIF receiver could have issues with this status bit error. When reading the channel bit status area to find out what the frequency is the DACs sets the frequency synthesizer and uses either a digital PLL or analog one to determine if the synthesizer should be increased or decreased. Without the correct status bit as a foundation for this method problems will likely arise.

    Jitter (Bi-Phase Signal)
    The average jitter measured on the Bi-Phase Signal from 700 Hz up to 100 kHz is about 284 picoseconds. [Image Link]
    The peak jitter measured on the Bi-Phase Signal from 50 Hz up to 100 kHz is about 1.246 nanoseconds. [Image Link]

    Jitter (Bit Cell) [Image Link]
    Measuring the Bi-Phase jitter over time shows about 1.2 nanoseconds peak (Blue line) in the data area and rises to about 2.1 nanoseconds peak, in the Staturs Bits and Frame Sync area. Thus, jitter is greatest at the Sync signals which is shown clearly in the J-Test.

    Assumed Analog Jitter FFT [hiFace Image Link] | [Bel Canto USB-Link Image Link]
    Here a 16 Bit J-Test Signal (Julian Dunn) is sent. The frequency modulation of the carrier is analyzed via FFT in the audio band and compared to a PLL slave clock. This measurement has a limitation in resolution because the PLL also has its own phase noise characteristics. But, this is the only way to evaluate this without an actual DA converter. This is in principal an assumption of what it could look like after a DA converter. The jitter in the bass area is about 100 picoseconds. This is the measurement limit of the Audio Precision. From 1 kHz on, it is about 1 picosecond. This is also the measurement limit of the AP (similar behavior as the sensitivity of the ear to detect jitter). From 100 Hz to 1 kHz it drops slowly. Clearly visible is the frame sync signal at 229 Hz and multiples of that (44.1 kHz / 192).

    This is what really happens, when you connect the HiFace to good, but typical 96 kHz PLL DAC (with a 192 kHz DAC, it would be worse, because 192 kHz PLL Receivers have higher Jitter than 96 kHz PLL Receivers). Every good design, based on the Crystal CS8414 Receiver (96 K PLL) will have similar numbers. (This is just a typical graph, every DAC acts different, regarding suppression of jitter, but for comparison, one must use a “typical” PLL receiver, in order to get some graph). Here you can see that mostly the jitter that is correlated with the sync signal, creates the most variation compared, to what the signal should look like (red line).

    [hiFace Image Link] | [Bel Canto USB-Link Image Link]




    Product Information

    • Price - RCA $150, BNC $180.00
    • hiFace Product Page - Link
    • hiFace FAQs - Link (PDF)
    • hiFace White Paper - Link (PDF)
    • Purchase hiFace (USA Only) - Link


    Associated Equipment:

    Verity Audio Fidelio loudspeakers, McIntosh MC275 amplification, Richard Gray's Power Company High Tension Wires, Berkeley Audio Design Alpha DAC, Wavelength Audio Proton, Esoteric D-07 DAC, C.A.P.S. server, Bel Canto USB Link, Halide Design Bridge, dCS Debussy DAC, dCS Puccini U-Clock, Kimber USB Cu, Kimber USB Ag, Benchmark DAC1 PRE, Kimber Select KS1011 Analog Cables, Kimber Select KS2020 Digital Cable, Kimber Monocle X Loudspeaker Cable, ASUS Xonar HDAV 1.3 Slim, Apple iPad, Sonic Studio's Amarra, M2Tech hiFace, Weiss Engineering DAC202, Lynx Studio AES16 Digital I/O Card.




    Comments 207 Comments
    1. ar-t's Avatar
      ar-t -
      In a way, that makes more sense. It is a bit harder to send 48 kHz, constantly, rather than 44 kHz, constantly. Why they made the conscious decision to do that, is beyond me.<br />
      <br />
      Nice guys, eh?
    1. stefano consoli's Avatar
      stefano consoli -
      Good morning Chris,<br />
      I’d like to have your opinion about the usb/spdif interfaces performances. <br />
      I have a Lenovo laptop running with Jriver in direct streaming a Hiface M2tech, a XLO digital cable to a counterpoint da10 converter upgraded with Mike Elliot “Rapture daccard” (Altavista audio).<br />
      I realised that my Mark Levinson 37 transport linked through spdif to the converter is more transparent than the PC/hiface combination.<br />
      In your experiences did you find an interface actually as transparent as a good transport connected through his spdif output ?<br />
      Stefano<br />
      <br />
    1. barrows's Avatar
      barrows -
      But the Wavelink is the first interface which I have tried to equal/beat a very good CD transport into the same DAC via the same SPDIF cable/input.<br />
      Hard/soft-ware shown below.
    1. stefano consoli's Avatar
      stefano consoli -
      thank you, Barrows<br />
    1. tpaxadpom's Avatar
      tpaxadpom -
      Chris, <br />
      <br />
      When you measured average and peak jitter values did you stream audio? If so, did you use j-test.wave?<br />
    1. The Computer Audiophile's Avatar
      The Computer Audiophile -
      Hi tpaxadpom - The Julian Dunn invented J-Test Signal was used.
    1. tpaxadpom's Avatar
      tpaxadpom -
      Thank you Chris.