Saturday 26 September 2015

MEASUREMENTS: Light Harmonic Geek Out V2


I know some folks have been awaiting the results for the Light Harmonic Geek Out V2. As you know, I have already provided a bit of a preview a few weeks back so no need to say much more I think. Generally my feelings about this device have not changes over the weeks... The 3D printed plastic smell has completely dissipated by now. It does sound great and I must say, going into the measurements, the impression was that this is some of the best sound I have heard from these little USB devices! Remember, this is a $235US unit I got as an "early bird" off the Indegogo promotion.

Let's see if the measurements are consistent with some of my early subjective impressions.

Part I: Digital Oscilloscope, Impulse Response, Digital Filters Test

Okay, let's start off with the usual 1kHz 0dBFS square wave off the headphone output at the default 100mW (at 16-ohms) setting:

That's encouraging. Excellent channel balance! About 875mV peak output. Clean, voltage-stable square waves. Of course we'll need to press the high-power output toggle button and see how the 1000mW setting looks like:

Voilà. An immediate increase to 3.1V peak (2.2Vrms). Or about an 11dB gain. Again, we're looking at excellent channel balance. One of the best I have ever seen whether USB "stick" type DAC like the AudioEngine D3 (excellent balance) or AQ Dragonfly (not great) and better than the ASUS Essence One from last week and some other desktop self-powered DACs I have come across! Output impedance is spec'ed at 0.47ohms. (I see the spec sheet talks about maximum output of 4Vrms, I imagine this must be using the balanced output with the proper TRRS cables with about a +6dB gain.)

How about impulse response? Remember there are 2 digital filter modes - default "TCM" (Time Coherence) minimum phase and "FRM" (Frequency Response) linear phase... Here they are:

Alright then. TCM is basically a standard sharp minimum phase filter, FRM is a weak anti-aliasing linear phase filter. Absolute polarity maintained.

And here then are the digital filter overlay graphs:


As you can see, the digital filter effects in the frequency domain are quite different between the TCM and FRM settings. TCM basically is a steep minimum phase setting with good suppression of aliasing. It doesn't "overload" even with 0dBFS wideband white noise unlike the ASUS Essence One or Samsung Note 5 shown previously - fantastic. I think this really speaks to the quality of the digital filter in the ESS SABRE in that it can handle the extremes of "intersample over" potential in oversampling digital processing. Furthermore the noise floor is really quite clean. This is essentially ideal behaviour for a sharp anti-aliasing filter for 44kHz signals.

The FRM is a different beast. It has a slow roll-off after 20kHz. Aliasing is evident with the mirror signal pairs at 24 & 25kHz. Noise floor is still excellent. Remember that companies like to show "clean" impulse responses with minimal ringing like the FRM in ads and promotional material. But there is a price to pay - less ringing generally means less aggressive filtering and higher potential for aliasing distortions; there's no free lunch. I think it's wise that Light Harmonic chose the TCM filter as default in this case and the listener is free to try the FRM mode to see if he/she prefers this alternate weaker filter.



Part II: PCM RightMark Tests & Comparisons

Setup:
Microsoft Surface 3 Pro --> LH Geek Out V2 (100 mW default setting) --> 3' shielded RCA --> E-MU 0404USB --> shielded USB --> Measurement Windows 8 computer
16/44:
As usual, let me just show you the big numerical summary graph comparing the Geek Out V2 with a bunch of other DACs:

Notice just how close they all are in 16/44. The "least accurate" of the bunch would be the Dragonfly 1.2 but we're generally talking very small differences here. As I have shown in the past, the Dragonfly's stereo crosstalk is typically higher than most DACs I have seen.

Even though the Geek Out V2 TCM and FRM digital filters do function differently, they measure almost identically between 20-20kHz in this test. Remember, the aliasing may cause audible distortion due to non-linearities in a playback system like your speakers even though the effect is not appreciated in the electrical domain here when measured.

Frequency Response. The Tascam is the one with slightly more low frequency roll-off compared to others.
Noise Floor
IMD+N
Notice in the graphs just how tight everything is! As I have said before, 16/44 poses no challenge to good modern DACs.

24/96:
Now we get a little more challenging... 96kHz is of course the peak samplerate for other small USB-powered DACs like the AudioEngine D3 and Dragonfly:

Again, no real numerical difference between the TCM and FRM settings. On the whole, we see that the noise level for these USB-powered DACs are a little higher than their larger desktop counterparts. Remember that I'm measuring the Geek Out V2 in 100mW mode... I bet the 1000mW mode would score even better in terms of dynamic range and noise level (unfortunately the high voltage level at 100% overloads my ADC).

I zoomed into the top end a bit... Nothing much to see in the low bass. At most the DACs roll off -0.5dB at 20kHz so no big deal.
Noise level. Notice the low level 60Hz hum with the Transporter.
IMD+N
The Geek Out V2 holds its own quite well at 24/96. It achieves >17.5-bits of resolution. Definitely not bad for something that plugs right into the USB connector as in the picture up top!

24/192:
Okay. Doubling the samplerate once more to 192kHz and we see this:
Again, no significance between TCM and FRM modes. Thought I'd include the PonoPlayer in here as well - Pono's showing a bit more distortion and slightly higher noise floor but generally comparable. As above, the non-USB-powered desktop DACs have an advantage in the department of noise floor. I also included the Oppo BDP-105 here... This jack-of-all-trades Blu-ray player/DAC measures very well! Impressive!

Frequency response. Again at 20kHz we're looking at most -0.5dB attenuation.
Noise floor. The Pono and USB-powered Geek Out V2 appear to have higher noise level.
IMD+N chart. Notice PonoPlayer's higher 2nd harmonic peak.

Part III: DSD64 & DSD128 Tests & Comparisons

Even though the Geek Out V2 goes up to 384kHz PCM, my ADC maxes out at 192kHz; I have yet to buy or download any music beyond 192kHz anyhow so the only value I see to this high samplerate is with software upsampling using one's own computer and disengaging the internal digital filter. But like last week, I can explore the DSD performance based on converting a 192kHz test signal to DSD64/128 and running it through the analyser to assess accuracy.

Like I did with the ASUS Essence One last week, this is with the KORG AudioGate software. Here's how it looks with DSD64:

Frequency Response
Noise Level
IMD+N
That looks excellent. Performance level completely in line with the regular PCM measurements. No penalty with DSD input to the Geek Out V2 and the only difference between the two is basically the high frequency noise inherent in DSD64 noise shaping.

Here's DSD128:

Frequency Response
Noise Level
IMD+N
So we clearly see the superiority of DSD128 compared to DSD64 above with the DSD noise pushed further out starting at 40-50kHz. The Geek Out V2 performs very well and again is equivalent to the 24/192 PCM performance (of course within the limits of my measurement equipment). Comparatively, the PonoPlayer exhibits a higher noise floor and the self-powered TEAC UD-501 performs with the lowest noise level.

Part IV: Jitter

Here are the Dunn J-Test spectra with both TCM (default) and FRM modes:


What can I say... Modern asynchronous DACs are generally immune to jitter issues as typically tested back in the day when the J-Test was constructed (and timing had to be extracted from SPDIF signals). This has been the situation for years with well engineered equipment.

Part V: Conclusions

No need to belabour the point here guys... The Geek Out V2 is a well engineered and excellent measuring audio product. It is to date the best measuring "USB stick" type DAC I have had the opportunity to examine. As far as I can tell, there are no glaring flaws or issues from an objective perspective! In fact, I believe the measurements would be even better if I could measure the 1000mW high power mode if my ADC didn't overload at the high voltage. I think that's as good a praise as I can give for a device like this, not to even mention the reasonable price. Good job LH for making this and clearly putting energies into the design; particularly I suspect the choice of the new "flagship" ESS SABRE9018AQ2M DAC chip played a huge amount in the technical performance we see. As a perfectionist, I can quibble about the choice of digital filters (ie. no traditional sharp linear phase or slow roll-off linear phase with strong antialiasing). As per results shown previously, I do not believe that filter settings make a huge difference that would change one's enjoyment unless you start eating into the audible spectrum <20kHz like with the PonoPlayer at 44kHz samplerate. Besides, as a USB DAC that can handle up to 384kHz, you can "roll your own" filter setting with computer based upsampling using software like HQPlayer with the Geek Out V2; that could be fun.

Now as for subjective impressions, the 3D printed casing could be more refined (eg. more polished, less rough at places). However, as I said in my preview, the plastic resin material does keep the device cool to the touch even running in the high power 1000mW mode for hours - warm but not hot. Heat was noticeable with the metal construction of the AudioEngine D3 and I suspect the Geek Out V1 might have been similar. Also, I'm glad they put the red cap on the TRRS balanced connector to prevent potential equipment damage. I'll have a look at the balanced output performance if I have access to the proper cables one day.

I was impressed by the sound quality already in the preview article. This is "technically accurate", "high fidelity" sound, folks. The highs are high, lows are low, details intact, tonally neutral, dynamics powerful, jitter-free, channel separation and balance excellent. As a headphone amp, there's enough power to keep the volume up even with demanding inefficient headphones. The Geek Out V2 will not "euphonize" your music but present it as is; warts, "digititis" and all (I can recommend the AQ Dragonfly if you prefer a more mellow sound). Harsh sounding pop (like my copy of Duran Duran's Greatest - still fun of course!) will be fatiguing. But well recorded and produced albums - currently listening to tenor Jonas Kaufmann's recent operatic collection The Age of Puccini or earlier today Dave Matthews Band's Crash from 1999 (though dynamically compressed, have a listen to "Say Goodbye") - will be presented in full glory. Assuming the rest of one's equipment is up to the task, this is the kind of DAC that just "gets out of the way" in terms of objective transparency, exactly the kind of result which from an objective perspective I would like to see in all my high-fidelity gear.

My feeling is that if Light Harmonic even just took the base internal circuitry used here, put it into a nice enclosure, quiet power supply, provide RCA & XLR outputs, add a few more buttons, and indicators/display up on the front, we could see a technically unbeatable desktop DAC at an excellent price point (can we get this done <$500 with 2 big volume knobs up front like the ASUS Essence One?)!

I've never had a chance to listen/measure the Geek Out V1 but I can certainly say that the "second coming" has demonstrated just how fantastic sounding (and measuring!) inexpensive DACs can be these days.

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

Looking back on this past summer, it has been a bit of a measurement whirlwind! And gauging from the hits I'm seeing to the blog from all around the world through Google and Blogspot affiliates, I believe there's a huge amount of interest in objective analysis and a desire to actually understand and at least try to "know" how things work and what level of accuracy we can expect from devices like these DACs.

I found the recent "clock" incident involving the 14-year old kid Ahmed Mohamed interesting. Beyond the social concerns brought out in a case like this (ie. racial assumptions, culture of fear, educator incompetence?), it's good to see that "tinkering" and the joy of science (even something as "simple" as a digital clock these days) can still be found (especially encouraging when it's in our young people). Understanding the science behind all that we easily take for granted I believe is a basic foundation of living in a technologically sophisticated age; science is not magic, audio technology is not magic. The laws of nature help us differentiate the likely from the improbable or illogical. IMO anyone who in 2015 markets an audio high fidelity product without clear scientific justification and evidence (and I'm obviously not talking scientific-sounding "truthy" gibberish [Comedy Central classic video]) is sadly participating in promoting mysticism. As expressed previously, just having testimony isn't generally good evidence. Some people seem to find this kind of stance "close minded". I personally do not think that discounting magical thinking when it comes to engineered products is at all unreasonable, or obdurate.

I do hope that in time the mainstream "audiophile press" can find it in themselves to transform more into a medium of education (perhaps back into a medium of scientific education since I have found many articles from years past providing so much more substance). Of course I can accept a commercial component but it really is not a good sign when not infrequently substantial portions of "reviews" are simply comments lifted from company claims or interviews with company representatives with no evidence that the reviewer has any ability to question, or perhaps even understand such claims. It would be nice to see a day when the term "audiophile" regains respectability and represents a hobby interested in "high-fidelity" as it might apply to the hardware gear and assessment of recording quality; a hobby able to anticipate and promote advancement rather than appearing to be a cohort of consumers being marketed to (as demonstrated by Sony's ridiculous "premium sound" SD memory earlier this year).

Change is in the air. Welcome to the fall... Which for me also means work demands and upcoming overseas trips. Posts will get a little spotty over the next few months as a result.

Hope you're all enjoying the music :-).

Addendum: (January 30, 2016)
I can confirm that the output impedance is <0.5-ohms measured with a 1kHz signal and 20-ohm test load. Impressive!

16 comments:

  1. Hear! Hear! I very greatly appreciate your untiring search for audio truth and your efforts to help us understand how excellence in the design of playback equipment affects fidelity to the recording.

    I know that you listen to lots and lots of music and I would be grateful for even a starter list of the albums/tracks that you believe to be well-recorded. I have learned the hard way that SACD's and Blu-Ray audio discs are often less satisfying to me than well-recorded CD's. Hats off to those engineers, producers and artists who insist on high sonic standards.

    ReplyDelete
    Replies
    1. Hi Don,
      Thanks for the note and suggestion... Indeed, just because it's on SACD and Blu-Ray with claims on the box for high quality sound does not mean anything about it being a good recording.

      Often, it's easier to point out poor recordings or examples of poor remastering especially with hyped up releases like Beck's "Morning Phase" or Dylan's "Shadows In The Night" with the hope that if people knew about them, maybe the producers and record companies would take note!

      Now as for a starter list, we could be entering into the audiophile 'canon' :-). Over the years Harry Pearson, Stereophile 'Records 2 Die For', etc. type lists have been in circulation. Many good sounding albums out there but would you actually listen to them other than for demo? Also, many new recordings are enjoyable but sadly not recorded or mastered well (eg. the recent Interstellar soundtrack).

      Here's some to try - no guarantees, not necessarily to treasure since it's mostly personal and subjective; art mixed with sounding engineering. Remember also that classical lovers are blessed with generally amazing recordings! Off the top of my head at breakfast on September 28th:
      Eric Clapton - Unplugged
      Rebecca Pidgeon - The Raven
      The Diana Krall discography - I personally prefer the recordings up to 2004's "The Girl In The Other Room"
      Was (Not Was) - What Up Dog?
      If you find a Japanese SHM-SACD of a favourite album - grab it, likely the best mastering (eg. Buggles "The Age of Plastic", Roxy Music "Avalon")
      Roger Waters - Amused To Death (new r2015 remaster and 5.1 excellent)
      Sheffield Lab releases
      Tons of MFSL, Audio Fidelity and DCC releases over the years
      Tom Petty - Full Moon Fever
      Tons of audiophile samplers! (check out the HDCD samplers, Accuphase, Chesky)
      Alexei Aigui & Ensemble - Happiness, Fame and Fortune (something foreign and funky, check out "Halla - Balla")
      Al Di Meola, John McLaughlin, Pace de Lucia - Friday Night In San Francisco (amazing guitar fingerwork!)
      Miles Davis - Kind Of Blue (of course... I like "Sketches of Spain" for demos as well)
      Charles Mingus - Mingus Ah Um
      Oscar Peterson Trio - We Get Requests
      Queen - A Night At The Opera
      Arvo Part - Tabula Rasa
      John Pryce-Jones & Northern Ballet - Dracula (Ballet in 3 Acts)
      Stan Getz & Joao Gilberto - Getz/Gilberto
      Yello - Stella
      McCoy Tyner with Stanley Clarke and Al Foster - S/T
      Frank Sinatra - Come Fly With Me (I prefer the stereo)
      Ernest Ansermet - The Royal Ballet: Gala Performances
      Eva Cassidy - Live At Blues Alley
      The Simpsons - The Simpsons Sing The Blues - no joke... pristine studio production! :-)
      Keb' Mo' discography
      Harry Belafonte - Belafonte at Carnegie Hall
      Laurie Anderson - Mister Heartbreak
      Dire Straits - Brothers In Arms
      Bobby McFerrin - Simple Pleasures

      ... Oh boy... I could go on :-). Alas. Off to work soon...

      Delete
  2. Yes, was waiting for this... patiently.

    Would always do -10dB squarewaves to 'see' the filtering. 0dBFS is interesting to see the max output voltage only.

    About 20 years ago I built the JK Acoustics 'JK CD-filter'.
    It must be connected directly after the (R2R) DAC chip of a CD player instead of the 'standard' Low Pass filter and got good reviews.
    in (blind) AB comparisons I could pick out this filter and it sounded 'fuller' yet tonally the same.
    Lots of blabla nonsense why this was the case (discrete components, audiophile parts ... etc. the familiar song)
    So ... I built the exact same filter characteristic with 'inferior sounding' opamps.
    Sounded exactly the same as the 'discrete' version and gave the same 'full bodied sound'.
    Measurements already showed me that the roll-off was gentle and only -0.5dB @ 20kHz.
    It couldn't be as simple as that, surely -0.5dB at 20kHz (could still hear up to 18kHz back then) is inaudible ?
    So I mounted the filter AFTER the original filter (which was a steep post filter designed for 4x oversampling)
    Quess what ... same 'changes'.
    To further investigate I shifted the -3dB point more into the audible range and the sound became rolled-off.
    Moved the -3dB point one octave higher but couldn't hear the 'fullness' difference any more.
    The sound was 'the same'

    So it wasn't the 'audiophile' parts but the -0.5dB @ 20kHz (or the phase shifts ?) that gave the sonic change.
    A 'fuller' sound yet not rolled off and still clear.
    At least that's the conclusion I drew from these experiments.
    For Non oversampling CD players other aspects would also have played a role but only the first generation (non Philips) CD players had horrible and steep LPF's... different story there.

    Interesting to see that for all sample rates the analog FR response is always that same 'magical' 20kHz -0.5dB.
    Even for 192/24 and DSD where this could have been much higher up.
    Most likely a fixed analog low-pass filter after the DAC chip.

    Prefer a sharper and FIR filter over both used filter types myself.

    Still ... looks like a capable and reasonably affordable dongle style DAC.

    Maybe someone should send you their 'Schiit Fulla' dongle DAC/AMP to try/test.
    (I kid you not about the name)


    ReplyDelete
    Replies
    1. Thanks for the anecdote Frans! Appreciate your thoughts and experience.

      "Fulla Schiit", eh? At <$100 looks like some classic components - AKM4396 DAC, CM6631A USB receiver, 250mW @ 16-ohms power, max 24/96. My guess is that this should measure and sound very similar to the AudioEngine D3 but with better output impedance at <1-ohm according to specs.

      Cute name I guess... Memorable for advertising purposes!

      Delete
    2. Shiit gear has a funny name, but their gear is no joke. I am sure there are many that would love for you to test some of their other decoders - Modi, Bitfrost, Gungir, Gungir Multibit, Yggdrasil.

      I would not mind sending you my Modi 1st Gen for a test, but I just got it and I am enjoying it too much. It is a USB powered DAC (no amp) with the same AKM4396 and CM6631A. It's only capable of 24/36 due to the design of not needing drivers. The new Modi 2 has an AKM4399 and can do 24/128, but needs drivers to go up to 128. It has a switch on the back to use drivers or driverless. Both of them are very capable DACs and it would be interesting to see how they compare to the likes of the Dragonfly and this Geek Out box.

      Delete
  3. Fancy a Geekout V1 1000mw to measure and compare? I've been meaning to get another one and could send it your way to measure first.

    ReplyDelete
    Replies
    1. Hi "Unknown", thanks for the offer. Unfortunately I don't think I'd be able to get a good grasp of the 1000mW version because the high power would overload my ADC. This is why I didn't test the V2 through RightMark at the higher power setting... I suppose I could lower the volume and measure but this likely would not demonstrate optimal results.

      Tell me this though. How HOT does the metal case get on the V1 after using for 2 hours of use? Does it ever get too hot to handle?

      Delete
    2. I have rather delicate fingers, they get uncomfortably warm to touch but never too hot to handle.
      My food thermometer says the case gets up to around 42-43c.

      Delete

  4. I have a sort of fundamental question for you - do electrical systems change their characteristics when the measurement chain changes? Its like saying that the very act of measurement introduces a shift in the value of the variable being measured. This might also mean that if we connect a speaker (or headphone) to an amplifier, the value shifts. Does the same happen to DACs when connected to amplifiers? Have you noticed this in your measurements?

    ReplyDelete
    Replies
    1. measurements do change when a 'real' load is connected compared to no-load measurements.
      In case of speaker amps the differences between no-load and real load can be quite high because of the currents present
      These currents can even be considerably out of phase with the supplied voltage.
      The lower the output resistance of the amplifier and the higher the current capabilities the less chance there is of specs changing significantly enough.
      It also depends on the topology used in the amplifier design and the impedance of the connected speaker.
      A resistive load used for testing is handy for repeatable and comparative tests BUT when speakers are connected the behaviour of the amp may differ.

      For headphone amps driving low impedance headphones the differences (mostly distortion) will be higher compared to no-load conditions than when a high impedance headphone is connected.
      Again due to the currents.
      Another big factor is the output resistance of the amplifier and the load impedance of the used headphone (and above all how much this impedance varies over the frequency range)

      Then there is the interaction of a moving membrane of the speaker/driver itself (back EMF) with the amplifier output in both cases above.

      For DAC's connected to amplifiers these factors are not an issue.
      There are no high currents present and there is no interaction (back EMF) coming from the connected amplifier input.

      A DAC usually has a very low output resistance and the load it sees is usually VERY high (realtive and absolute) and are purely resistive in 99% of the cases.
      Perhaps a very small amount of cable/input capacitance can also be present.
      However, due to the low output resistance of a DAC small capacitances (which are also present when measuring) these small capacitances are of absolutely no influence in the performance/measurements.

      Delete
  5. Another request to add to your long list ... since you are a big fan of Oppo (so am i), can you please measure the Oppo HA-2 DAC/headphone amplifier? It is also 300 USD and has had good reviews. I'm just curious how much of an improvement will BDP 105 over that.

    ReplyDelete
  6. I have followed and enjoyed you for quite a while, and I now have my first request: could you please test the audio quality of the dac/analogue output of the new Chromecast Audio?

    ReplyDelete
  7. Hello sir i am not able to sure to pick which USB DAC from the above to Audioengine D3 and LH V2.

    So can u please help me out which one has better audio quality with some impression review between these two??

    ReplyDelete
  8. I have recently bought an ODAC usb DAC for my home office, and I am very happy indeed. It is cheap, measures perfectly, and sounds fine to me in a pretty high end system (QUAD 405-2 amp driving Harbeth P3ESRs). Even if it is not a dongle, it is very small.

    ReplyDelete
  9. What voltage level did you test at?

    ReplyDelete
  10. Kinda disappointing that the SE output is only 2.2Vrms. So looks like if you are after this thing for power output out of this form factor, the Geekout 1k is the better deal because it is SE topology and actually puts out 4Vrms over that, and has better build quality and reliability to boot.

    ReplyDelete