Saturday 5 September 2015

MEASUREMENTS: Samsung Galaxy Note 5 Phone

As you probably know, for me, convenience and practicality are the most important elements of mobile audio. As much as I enjoy the audiophile hobby and achieving high fidelity sound, I think it's important and realistic in acknowledging that consumption of music on-the-go these days is tied in to the ubiquitous "smartphone" many of us carry around. They have become indispensable tools in so many ways that for music lovers and audiophiles to not take them seriously IMO would be denial. I think it's fascinating that audiophile magazines rarely ever mention these devices despite the fact that millions of us use them every day and gain so much enjoyment in the process!

The last phone I got was the Nexus 5 which I measured back in early 2014. As you can see, it doesn't measure too well using the standard test bench but adequate for busy environments. Not inspiring 16/44 performance and it does accept 24-bit audio but can't really perform beyond close to 16-bit dynamic range. It's also limited to standard resolution 44kHz only.

Last year, my wife got the iPhone 6 which I also examined. Up to 24/48, it performed well subjectively and with objective measures. Clearly a step above my Nexus 5 in terms of achieving >16-bit dynamic range. In terms of output impedance, Ken Rockwell measured the iPhone 6+ as 3.18-ohms and the Nexus 5 is probably around an unimpressive 10-15-ohms.

Well, after close to 2 years, here then is my phone upgrade:
Shown in Otter Box case.
That is of course one of the two new Samsung "flagship" phones. The Galaxy Note 5, 5.7" screen @ 2560x1440 AMOLED QHD resolution (gorgeous!), 4GB RAM, Android 5.1.1, Samsung Exynos 7 Octa 7420 64-bit 8-core CPU (4x2.1GHz + 4x1.5GHz, Mali-T760 MP8 302 GFLOPS GPU!), S-Pen, and all the bells and whistles which I will not repeat (a detailed review here). The other flagship phone is the Galaxy S6 Edge+ which almost has the same feature set but with the cool edge curvature (for bookmarks, quick launch icons, notifications) and missing the pen. Despite how "cool" the Edge looked, I just didn't think it added much to potential productivity. To buy these phones outright, we're looking at $750-$900USD before taxes or shipping currently... Not insignificant investments for most people.

In the picture above, I have it nestled in the excellent but bulky Otter DEFENDER Case - it provides a nice rugged secure belt clip; essential for day-to-day work. The rubberized feel of the case is also good since the body of the phone is now made of glass front and back which could be a little slippery when wet. The thickness might compromise wireless Qi charging; my more expensive Panasonic QE-TM101 charger actually fails to charge with the case on but my cheap generic charging pad from eBay actually works fine!

Here's a look at the size compared to the old Nexus 5 and wife's iPhone 6 (not Plus):
Each phone rendering this blog in Dolphin browser at full brightness.

Nexus 5 left, then iPhone 6 (a very pretty phone IMO), and Note 5 right.

An obvious and substantial handful size-wise, my friends. Five years ago you'd look like a total dork lugging this around (first Samsung Note came out 2011), but these days, it's pretty well standard around here. Especially useful I believe for those reading complex Asian fonts. You've probably seen in other reviews that although having a slightly larger screen than the iPhone 6+ (Note 5 5.7" vs. iPhone 6+ 5.5"), it is actually shorter, about the same width, and slightly thicker; overall it feels similar in my hand.

Let's get to it with the audio. Notice that there are some results out already using the Note 5 and and AudioPrecision APx525! The output impedance is 4.6-ohms. Not far off from the iPhone 6... That's an encouraging start. "Willyman" called it "an audiophile's dream"! We'll see about that.

Although I don't have such esteemed equipment, let's see what my results look like and compare the Note 5 with the iPhone 6, the previous Nexus 5, a few of the other DACs, and of course the PonoPlayer.

I. Oscilloscope Square Wave @ 0dBFS, Impulse Response, Digital Filter

As usual, let's start with a few preliminary evaluations. First up is the ol' digital oscilloscope to see the maximum voltage and channel balance:

Nice channel balance, peak voltage at ~1V, or 0.7Vrms. This is essentially the same as the PonoPlayer (single ended mode) and a bit lower than the iPhone 6 which is capable of about 1Vrms.

Impulse response:

Minimum phase filtering is being used with Apple's iPhone 6 and now I see it with the Samsung Galaxy Note 5 - seems like the "norm" with expensive cell phones these days. Absolute phase maintained.

Stereophile has been using an overlay graph with wideband measurements of white noise and a 19kHz signal over the last few years to look at filter roll off, and aliasing products based on discussions with Juergen Reis (the "Reis Test"). Here's a similar version (with 19kHz and 20kHz components to scare up some intermodulation as well), signal constructed as 24/44:

Since this is the first time I've posted something like this (beyond just wideband white noise), I've annotated it with some details. As you can see, I have 3 spectra overlaid. The green FFT is the twin 19kHz and 20kHz sine waves at -6dBFS [peak amplitude -0.02dB, no clipped samples] in the test signal but I've set the peaks at -20dB on my ADC to prevent any overloading - I'm going to use this as the standard volume calibration for these tests. The purple is with the device playing digital silence. As you can see, there's a bit of very low level noise from 20Hz to 20kHz. Notice the noise floor increases when I play the 19 & 20kHz test tones in green. The new noise peaks in the green FFT (compared to the purple digital silence tracing) represent added distortion (aliasing, harmonic distortion, intermodulation distortion, spurious noise). I've also added a solid horizontal line at -120dB as a marker; any noise products below this means the signal is less than 100dB below the peaks at -20dB. In the case of the Note 5, even though there are quite a number of peaks <20kHz, all except one at 13kHz appear to be at or more than 100dB below the 19 and 20 kHz primary signals. I think that's pretty good.

Finally, there is the yellow tracing. This is the wideband white noise tracing used to look at the steepness of the digital filter. With the Note 5, notice that the overall noise floor appears to increase with signal complexity from silence, to dual tones, to random noise. I've seen this kind of thing happen with other portable devices but had not reported it. Ideally, it would be nice to see the noise floor remain stable irrespective of what signal is being played (which is the case with the excellent DACs I've tested). This is an example of a suboptimal digital filter design. As the signal gets louder and more complex, the DAC digital filter becomes "overloaded" resulting in increased distortion and restriction of dynamic range can be seen... This makes it difficult to get a handle on the impact on playing real music since the dynamic range will fluctuate through the course of a piece; whether this is an audible problem is unclear to me but certainly this is not ideal performance.

Now, for some context, here's what the same test looks like with the PonoPlayer:

Notice again the weak digital filter (yellow) with a very gradual roll-off instead of the sharp filter with the Note 5 around the 22kHz Nyquist frequency. The result of this is the presence of strong aliasing images at 24 and 25kHz (green tracing). Compared to the Note 5, there are a number of signal spikes poking through the solid horizontal line down at -120dB (or 100dB below the sine wave peaks calibrated to -20dB)... Basically, a demonstration of more distortion products present; still lowish levels so I'm not claiming this is audible although the 24 and 25kHz aliasing peaks could cause significant problems with nonlinear speakers and headphones.

The nice characteristic demonstrated by the PonoPlayer is how the noise floor remains stable irrespective of the signal being played. I like this stability! (And as I commented on the PonoPlayer review, I wish Pono would come out with firmware that lets me turn off the Ayre filter and run a more typical steeper filter which would clean up much of the aliasing distortion with 44/48kHz signals!)

II. RightMark Tests & Comparisons

As usual, I'll be using the measurement equipment "chain" as such:
Phone/DAP/DAC Device --> 6' single-ended cable (shielded phono-RCA or RCA-RCA) --> E-MU 0404USB --> 6' shielded USB cable --> Windows measurement computer
Some devices like the PonoPlayer are of course capable of balanced operation, I will indicate this as appropriate.

RightMark 6.4.1 PRO software for acquisition and analysis; some of the older devices would have been measured with earlier versions and data saved in a compatible format.

16/44:
The big summary chart with a number of comparison DACs:

Scanning through this summary, one cannot help but be reminded that 16/44 audio is generally not a problem at all! The only device that clearly struggled with the full 16-bit resolution was the Nexus 5 phone. We see the frequency response idiosyncrasy with the PonoPlayer of course. And the Dragonfly's higher stereo crosstalk.

Frequency Response

Noise floor

IMD test tone

Stereo Crosstalk

24/48:
The summary chart:

The reason I'm doing 24/48 is mainly to compare to the iPhone 6 (with the Onkyo HF Player app) which peaks out at this bit-depth and samplerate. As you can see, the Note 5 holds its own remarkably well at this setting... For all the complex electronics inside, it actually achieves a very low noise level reading. However, I do believe this low noise / high dynamic range result is exaggerated with simple test signals like RightMark as suggested in Part I above.

Frequency Response

Noise Level


24/96:
Lots of DACs tested over the years at this almost universal "high-resolution" setting:

What can I say, the Note 5 remains very accurate compared to the others based on this test. With a dynamic range of 111dB, we're looking at slightly more than 18.5-bits of resolution (again, remember the caveat from Part I about the noise floor). Again, some excellent measured results from a compact mobile device. Interestingly the Note 5's frequency response has a gentle roll-off earlier than the others... Not a big deal since we're only looking at -0.75dB at 20kHz (much less than the PonoPlayer at 44/48kHz of course).

Frequency Response

Noise Level. You can see a bit of 60Hz mains hum with the Transporter and TEAC which are of course plugged into the wall socket.

IMD test tone

Stereo Crosstalk. There's that Dragonfly again... And the benefit of balanced operation.

24/192:
Yes ladies and gents... The Samsung Note 5 phone does indeed support and play back 24/192 files.

Like the 24/96 results above, that gentle roll-off is again notable for the Note 5 compared to the others.

Frequency Response

Noise Level. Curious that the Note 5 has a 60Hz component here which was not seen with the 24/96 test. I double checked this and it appears to be a real finding though not of concern at -120dB down.

III. Jitter - Dunn J-Test




I expected to see a little more of the jitter modulation pattern in the 16-bit spectrum. But otherwise, there's no evidence of significant "skirting" of the primary signal base nor sidebands as seen in SPDIF interfaces which are more prone to jitter (eg. TosLink, Coaxial). As I have noted before, I'm not sure we need to be concerned about jitter these days with the vast majority of devices (despite all the claims out there)...

IV. Subjective Impressions

Hey, it's a phone... A typical Android phone, so I trust nothing needs to be said about the user interface and functionality. Just use whatever music player software you like. I'm fine with the default Google "Play Music" app (despite the odd way it doesn't pick up cover images at times; it's bit perfect so long as you're not activating the DSP/EQ settings). It handled all the high-resolution music in FLAC just fine. Also, at home I have been listening to music streamed off my Logitech Media Server using SqueezePlayer and controlled with Squeeze Commander or Orange Squeeze. Works great with lossless streaming through the home WiFi.

The Note 5's 0.7Vrms output puts it at about the same maximum level as the PonoPlayer. Like I said with the Pono, even at maximum volume, my AKG Q701 headphone isn't loud enough to sound uncomfortable for music with decent dynamic range. Reasonably sensitive headphones recommended.

I've had a listen to a few of my usual test tracks over the last week to get an idea of how it sounds. Delicate details are well rendered; for example Ali Farka Touré & Toumani Diabate's "Monsieur Le maire de Niafunké" (from In The Heart of the Moon, 2005) retained the lovely intricacies of the overlapping guitar and kora.

Last week, I checked out the Beach Boys live here in Vancouver at the annual Pacific National Exhibition. Clearly, they aren't exactly "boys" any more and the original line-up is gone, but it was still a fun concert. Anyhow, I listened to the 2012 album That's Why God Made The Radio in 24/48. Sounds good, typical Beach Boys harmonies among the pop confections. Plenty of details retrieved, excellent pristine sound.

Suzanne Vega's album Solitude Standing (1987) has been an old favourite of mine for decades. No disappointment here either. Realism of the a cappella "Tom's Diner" was stunning as usual. Remember, this vocal track was used to "fine-tune" the development of the MP3 codec back in the day since it can be challenging to get the nuances of well-recorded vocals right. Channel balance is also important in this track to keep the voice central since there are no instruments to detract from the singular performance.

The Spotify app works well as expected. Nothing special to report... Sounds fine.

V. Conclusions

As a Samsung Note large-screen phone, this baby is fast, has 4GB RAM, and can handle whatever LTE Advanced (450Mbps) network you might want to throw at it. The screen is gorgeous so long as you desire the larger size of course! This screen size is great as an all-in-one "phablet". Plus the S-Pen works really well to jot the occasional note or doodle. And for those wondering, the USB data transfer speed is about 19MB/s copying music to storage (vs. 4MB/s with PonoPlayer). Remember that unlike previous Note phones, this one does away with the removable battery and microSD slot. Non-replaceable battery is typical these days but my feeling is that the loss of the SD slot is a bit unfortunate since it would be nice to store more music or movies on the machine itself with an interchangeable card. Consider getting the 64GB model if you intend to store a bunch of media. Do not underestimate the power of streaming of course! Just make sure you have a good data plan if you're going to be consuming a bunch of lossless audio or video...

Now as for the audio performance, let's say it's more than adequate for me personally as a portable music player. Remember, I'm pragmatic about portable audio. Even if the player outputs the most accurate sound possible, I do not believe I'm actually going to be able to appreciate this with headphones sitting on the subway, taking the bus, or walking down the street - 256kbps MP3 sounds great. Lossless is overkill for me in these settings (no need to even mention 24-bits or 88+kHz sampling rate)! I guess I might appreciate lossless better in the quiet of my own home at night with the kids asleep but frankly speaking, as much as I've invested in headphones over the years, no headphone can take the place of real speakers recreating the soundstage in front of me for high-fidelity enjoyment.

Personal preferences aside, the objective results from the headphone output of this phone is excellent with standard measurements. It can handle high-resolution audio all the way to 24/192 and is capable of >16-bit audio resolution. However, if we only use standard test signals, we would miss out on the issue with the noise floor and effective loss of dynamic range as shown in the digital filters FFT's in Part I. Unfortunately we see limitations of the digital filter (fixed-point overload/overflow?). You can read more about the technical details and pitfalls of digital filters here (BTW: for DIY guys, here's an interesting blog post with lots of filter settings/coefficients to test out with the Soekris R-2R DAC). Honestly, I'm not sure I can hear this effect through my headphones with the music selection I listened to. I e-mailed JR_Audio about this and he likened the effect to a lossy codec - the more complex and higher the signal level the more low level details will lose definition... Again, not the kind of thing I want to see with "high fidelity" audio in my main system, but it seems fine for my portable needs. I suppose listening tests with real music would be interesting (like maybe equalizing the volume and using a headphone amp with multiple inputs doing quick A/B switching between this device and something that has better filters that do not overload). If anyone has more technical info / experience with this kind of effect, let me know!

Audiophiles... Here is an audio device that's modern with all the bells and whistles, sounds good subjectively, measures reasonably well overall (with the digital filter limitation caveat of course), and doubles as your PDA / personal communication device :-). If the iPhone 6 and now the Note 5 are examples of objective improvements in devices yet to come, I'd say that the needs of audiophiles will be well taken care of in the days ahead.

Enjoy!

PS: I'm now very curious about the digital filter FFTs and will try running a few of these tests with the iPhone, Nexus 5, etc...

Addendum: January 30, 2016
Measured the output impedance - using 1kHz signal and 20-ohm load. 1.3-ohms - good.

9 comments:

  1. Hi Archimago,

    Your measurement of noise floor of digital silence and 2 test sinewaves shows worsening of the noise floor of about 20 dB, or 3.5 bits.
    Does this imply that the resolution of the phone's DAC dives from 18.5 to 15 bits when playing test signals and even lower when playing real music?
    What about the 16/44 FLAC files, does it goes from 16 to 12.5 bits?

    Aside from this, can you please help me understand how do you evaluate if the Google Play Music app is bit-perfect? I am using VLC Media Player and want to know if it - or any other app for that matter - is bit-perfect.

    Keep doing this great job! This blog educated me more than anything else I read in the last 2 decades!

    ReplyDelete
    Replies
    1. Hi Gok.

      Yes, clearly the 2-tone signal is demonstrating a rise in the noise floor to about -135dB at the base. As an extremely gross "back of the napkin" calculation... The peak 19&20kHz tones are -6dB themselves so there is a (135-20)+6 = 121dB (20-bit) spread maximal and if you look at including the noise peaks, saying that -120dB is the actual base noise floor, then the spread becomes (120-20)+6 = 106dB (~17.5-bit)...

      The RightMark DR measurement is done with a single tone I believe and that's why we're seeing a relatively high 18.5-bit result which makes it look better than the ~17.5 with the 2 tones.

      As for real music, this would be hard to estimate! We can say that with a highly complex 0dBFS white noise, the noise floor goes as high as (95-20)+6 = (-)81dB (~13.5-bits) as an estimate based on that FFT graph... This is the worst case scenario though.

      I did run some 16-bit signals through and noise floor looks about the same so when playing 16-bit audio, it is capable of full 16-bit resolution and down to 13.5-bits with loud white noise.

      I believe Google Play Music is bit perfect with EQ and DSP off. Reason I believe this is because in the 16-bit Dunn J-Test, I can still make out the jitter modulation pattern out of the noise floor (hard to see in the picture above but easily seen in "real-time" when looking at the FFT on the computer). This indicates that the 16th bit LSB is being retained by the player software.

      Have not tested VLC on my phone. Will let you know if I do!

      Delete
  2. 3 Points

    I am a bit in hurry today, so only very short additions / explanations:

    1. RMAA does calculate the dynamic range with a - 60 dBFS 1 kHz sine wave signal and this does not stress the digital filter at all and so the noise level is very close to the digital silence noise level.

    2. When reading the numbers / value of level of the white noise in FFT graphs, you have to take the bin width (and fft window) into account. So reading noise levels at around - 130 dB at the 16 bit RMAA measurements are similar to - 97 dBFS RMS values (about 32 dB shift (as a rule of thumb with this FFTs here).

    3. Bit True: I am sorry to say, but looking at the 16 Bit LSBs in the J-Test is by far not "good enough" to tell if something is bit true or not (it is only valid to tell if 16 bit are transfered). Better would be the "Dolby" or "DTS" or "HDCD" test (play back those files natively and look if the decoder to indicate those files correctly).

    Juergen

    ReplyDelete
    Replies
    1. Thanks for the details & tips Juergen :-).

      About point 3, I'll see if I can run the digital output through my receiver to decode DD/DTS using an OTG cable and my CM6631A USB to SPDIF device. If at 100% volume and Android is bitperfect, I should see a 16-bit DD/DTS signal decoded properly...

      (Ref to the CM6631A: http://archimago.blogspot.ca/2013/03/measurements-adaptive-aune-x1.html)

      BTW: I will have to wait a little bit for this test first though. My Onkyo TX-NR1009 developed a faulty HDMI board. Apparently quite common with the 2009-2012 receivers they make and they're aware and extending warranties to 2018 on these:
      http://www.avsforum.com/forum/90-receivers-amps-processors/1652514-onkyo-acknowledges-failed-units-extending-warrranties-until-2018-a.html

      Appreciate "manning up" on the issue... Despite the hassle...

      Delete
    2. Just a quick update on this.

      Got the CM6631A device to work via an OTG USB cable, using USB Audio Player PRO on the phone and trying to send a FLAC DTS signal to my old Denon receiver thru TosLink. So far, no luck on DTS playback... Hmmm, perhaps Juergen is right, bit-perfect isn't something achievable at this time with the the Android OS'es. It is processing the 16-bit signal hence I see the modulation pattern in the last bit, but it's still going through a high precision mixer in the process.

      I've looked around and have not been able to find much on this with Android OS'es...

      Anyone know if "bit true" output is achievable in the sense of sending out the unadulterated 16/24-bit "integer" data?

      Delete
    3. Cross Checked with iOS?
      Hi Archimago. Have you / could you cross-check your method / setup with an iOS device? With newer iOS Devices (iPhone, iPad, ...) and Onkyo HD Player, I can play up to DXD (PCM 352,8 kHz 24 Bit) and 2xDSD (DSD128) natively via CCK (camera connection kit) into an external DAC with ease.
      Juergen

      Delete
    4. Hi Juergen,
      I'll give this a try when I can :-). Problem is I gave away the old iPhone 4, I don't have a Lightning CCK cable for my wife's iPhone 6, and my ancient 1st gen iPad doesn't run the latest Onkyo HD Player!

      Delete
  3. FWIW on Window devices you need a player that can access ASIO or WASAPi output modes in order to bypass the Windows mixer and pass bit-perfect streams (i.e., to pass DTS or Dolby Digital bitstreams to an AVR for decoding). Don't know what would work on an Android device.

    ReplyDelete
  4. Don't give up your Nexus 5 yet. Still some optimization left for noise/THD at lower gain settings:
    http://forum.xda-developers.com/android/software/qualcomm-gain-control-t3253957

    ReplyDelete