S7 Edge - Headphone output measurements (SD version)

Search This thread

Willyman

Senior Member
Aug 26, 2015
414
162
Hi guys,

As I did with my Note 5, I performed some AudioPrecision measurements on the headphone output of the S7 Edge (Snapdragon version, SM-G9350).

UPDATE 1: I made an error in the max output level. It's -9 dBV/360mV into a 16 Ohm load, not -13 dBV. Still... it's low. Same as Mate 8/Z5/, lower than V10 (-6 dB), a LOT lower than iPhone (-1 dBV).

UPDATE 2:
  • Redid the Dynamic range measurement using the more convention AES617 standard. Shows 100 dB now.
  • THD is really quite poor... It raises as a function of frequency and output level. My colleague suspects some form of capacitive loading. I would say this is probably the worst metric of this device, the other elements are very acceptable/good.


Setup
  • Audio Precision APx525
  • Loaded (unless specified): 16 Ohm

Result summary

  • Supports high resolution playback up to 192 kHz.
    Flat frequency response: <0.08 dB delta upto 96 kHz
    Very good dynamic range (AES17): 100 dB
    LOW output level: -9dBV (360mV using a 0dBFS signal) into a 16 Ohm load (for reference, Asian iPhone 6 = -1 dBV)
    Output impedance: 5.6 Ohm
    Crosstalk: didn't have to means to this measurement, sorry
    THD+N: at nominal level = max 0.05%, at maximum level = max 0.7%

Some of these results are quite disappointing. Especially the very low output levels and high THD+N at max level. Some graphs below for illustration.

Subjective assessment
Been comparing the quality to my reference setup at work (PC to USB DAC Sony PHA-3 line-out to Talisman T-3H full Class-A amp). Used a number of sources (44/16 FLAC, 192/24 FLAC, Tidal HiFi).

I would say the sound quality is really quite good for a phone, but not exceptional. There's nothing really wrong with it. Good spatial quality, extended in the highs and the lows. It just doesn't sound that WOW and lacks somewhat in its dynamic delivery and instrument seperation. My guess is that the limited amplification power is letting it down somewhat in this regard. It's a pity I don't have my Note 5 anymore for direct comparison, but from memory I believe that was a better performer.

Now, to call it "incredibly disappointing" is a gross overstatement... (see some other threads on here).

Dynamic Range (AES17)
0oakTs7.jpg


Noise floor measurement loaded (multi-tone signal)
vlAlC2v.jpg


Frequency response (loaded)
lnYbM3P.jpg


THD+N spectrum at maximum output level
tcum5CF.jpg
 
Last edited:

Willyman

Senior Member
Aug 26, 2015
414
162
Updated the data, made an error in my output level measurement. Apologies...
 

johanbiff

Senior Member
Apr 11, 2015
501
146
"Audio output doesn't disappoint
The Samsung Galaxy S7 almost perfectly matched the output of its predecessor with an active external amplifier. This is to say the smartphone delivered flawlessly clean output with just above average volume levels for an excellent showing.

Stereo crosstalk did increase a bit more compared to what the Galaxy S6 had, but output remained very solid when headphones came into play. Volume didn't drop either, so it was all very good and worthy of a modern-day flagship.

Anyway, here go the results so you can do your comparisons.

Test Frequency response Noise level Dynamic range THD IMD + Noise Stereo crosstalk
Samsung Galaxy S7 +0.01, -0.04 -92.5 92.6 0.0027 0.0078 -92.7
Samsung Galaxy S7 (headphones) +0.05, -0.05 -91.9 92.1 0.0044 0.063 -73.4
Samsung Galaxy S6 +0.01, -0.04 -95.6 92.8 0.0024 0.0094 -94.5
Samsung Galaxy S6 (headphones) +0.02, -0.05 -92.6 91.9 0.0025 0.042 -83.4
Sony Xperia Z5 +0.01, -0.04 -95.5 89.5 0.0033 0.012 -94.8
Sony Xperia Z5 (headphones attached) +0.22, -0.24 -95.1 89.5 0.0057 0.212 -59.8
Apple iPhone 6s +0.03, -0.04 -93.5 93.5 0.0016 0.0075 -73.2
Apple iPhone 6s (headphones attached) +0.10, -0.06 -93.8 93.9 0.0030 0.101 -68.2
LG G4 +0.04, -0.07 -93.4 93.3 0.0021 0.050 -92.6
LG G4 (headphones) +0.93, -0.13 - 91.4 91.9 0.013 0.244 -50.4
Nokia Lumia 950 +0.01, -0.03 -91.3 91.2 0.0036 0.012 -91.6
Nokia Lumia 950 (headphones attached) +0.59, -0.03 -92.0 91.9 0.011 0.316 -63.9"

http://www.gsmarena.com/samsung_galaxy_s7-review-1408p7.php

seems like exynos is the way to go :)
 

Willyman

Senior Member
Aug 26, 2015
414
162
"Audio output doesn't disappoint
The Samsung Galaxy S7 almost perfectly matched the output of its predecessor with an active external amplifier. This is to say the smartphone delivered flawlessly clean output with just above average volume levels for an excellent showing.

Stereo crosstalk did increase a bit more compared to what the Galaxy S6 had, but output remained very solid when headphones came into play. Volume didn't drop either, so it was all very good and worthy of a modern-day flagship.

Anyway, here go the results so you can do your comparisons.

Test Frequency response Noise level Dynamic range THD IMD + Noise Stereo crosstalk
Samsung Galaxy S7 +0.01, -0.04 -92.5 92.6 0.0027 0.0078 -92.7
Samsung Galaxy S7 (headphones) +0.05, -0.05 -91.9 92.1 0.0044 0.063 -73.4
Samsung Galaxy S6 +0.01, -0.04 -95.6 92.8 0.0024 0.0094 -94.5
Samsung Galaxy S6 (headphones) +0.02, -0.05 -92.6 91.9 0.0025 0.042 -83.4
Sony Xperia Z5 +0.01, -0.04 -95.5 89.5 0.0033 0.012 -94.8
Sony Xperia Z5 (headphones attached) +0.22, -0.24 -95.1 89.5 0.0057 0.212 -59.8
Apple iPhone 6s +0.03, -0.04 -93.5 93.5 0.0016 0.0075 -73.2
Apple iPhone 6s (headphones attached) +0.10, -0.06 -93.8 93.9 0.0030 0.101 -68.2
LG G4 +0.04, -0.07 -93.4 93.3 0.0021 0.050 -92.6
LG G4 (headphones) +0.93, -0.13 - 91.4 91.9 0.013 0.244 -50.4
Nokia Lumia 950 +0.01, -0.03 -91.3 91.2 0.0036 0.012 -91.6
Nokia Lumia 950 (headphones attached) +0.59, -0.03 -92.0 91.9 0.011 0.316 -63.9"

http://www.gsmarena.com/samsung_galaxy_s7-review-1408p7.php

seems like exynos is the way to go :)

From my measurements, it's not so clear cut... The only relevant measurement is the one with headphones attached (i.e. loaded)
Dynamic Range: SD > Exynos
Freq response: SD > Exynos
THD: Exynos > SD
 

pyrokinetic

Member
Sep 16, 2012
29
3
Sofia
Hi guys,

As I did with my Note 5, I performed some AudioPrecision measurements on the headphone output of the S7 Edge (Snapdragon version, SM-G9350).

UPDATE 1: I made an error in the max output level. It's -9 dBV/360mV into a 16 Ohm load, not -13 dBV. Still... it's low. Same as Mate 8/Z5/, lower than V10 (-6 dB), a LOT lower than iPhone (-1 dBV).

UPDATE 2:
  • Redid the Dynamic range measurement using the more convention AES617 standard. Shows 100 dB now.
  • THD is really quite poor... It raises as a function of frequency and output level. My colleague suspects some form of capacitive loading. I would say this is probably the worst metric of this device, the other elements are very acceptable/good.


Setup
  • Audio Precision APx525
  • Loaded (unless specified): 16 Ohm

Result summary

  • Supports high resolution playback up to 192 kHz.
    Flat frequency response: <0.08 dB delta upto 96 kHz
    Very good dynamic range (AES17): 100 dB
    LOW output level: -9dBV (360mV using a 0dBFS signal) into a 16 Ohm load (for reference, Asian iPhone 6 = -1 dBV)
    Output impedance: 5.6 Ohm
    Crosstalk: didn't have to means to this measurement, sorry
    THD+N: at nominal level = max 0.05%, at maximum level = max 0.7%

Some of these results are quite disappointing. Especially the very low output levels and high THD+N at max level. Some graphs below for illustration.

Subjective assessment
Been comparing the quality to my reference setup at work (PC to USB DAC Sony PHA-3 line-out to Talisman T-3H full Class-A amp). Used a number of sources (44/16 FLAC, 192/24 FLAC, Tidal HiFi).

I would say the sound quality is really quite good for a phone, but not exceptional. There's nothing really wrong with it. Good spatial quality, extended in the highs and the lows. It just doesn't sound that WOW and lacks somewhat in its dynamic delivery and instrument seperation. My guess is that the limited amplification power is letting it down somewhat in this regard. It's a pity I don't have my Note 5 anymore for direct comparison, but from memory I believe that was a better performer.

Now, to call it "incredibly disappointing" is a gross overstatement... (see some other threads on here).

Dynamic Range (AES17)
0oakTs7.jpg


Noise floor measurement loaded (multi-tone signal)
vlAlC2v.jpg


Frequency response (loaded)
lnYbM3P.jpg


THD+N spectrum at maximum output level
tcum5CF.jpg

Great research! I have the Exynos version but lack the dedicated $20K analyzer. What I have is the xonar u7 external adc/dac , which spec sheet states the following:

Input Signal-to-Noise Ratio (A-Weighted) :
110 dB
Frequency Response (-3dB, 24bit/96KHz input) :
10 Hz to 46 KHz

Can I contrive something to do a adequate measurement (with some 3rd party software, or even code my own) and what will the error be, in your professional opinion? Thanks.
 

Willyman

Senior Member
Aug 26, 2015
414
162
Great research! I have the Exynos version but lack the dedicated $20K analyzer. What I have is the xonar u7 external adc/dac , which spec sheet states the following:

Input Signal-to-Noise Ratio (A-Weighted) :
110 dB
Frequency Response (-3dB, 24bit/96KHz input) :
10 Hz to 46 KHz

Can I contrive something to do a adequate measurement (with some 3rd party software, or even code my own) and what will the error be, in your professional opinion? Thanks.

Well, given that the SNR is only barely better than the S7 and the FR is actually worse, I don't think there is much you can do with that device :D
There's a reason why those AP's are so ridonculously expensive :'(
 

jwhatts

Senior Member
Mar 3, 2013
50
15
St. Paul, Minnesota
I understand only some of these words... Any chance you could dumb it down a bit for me? And is there a solution to get back the WOW factor without an update from Samsung or an external DAC?

Thanks for the research, I find this stuff fascinating!
 

NaimC

Senior Member
Apr 23, 2008
150
173
From my measurements, it's not so clear cut... The only relevant measurement is the one with headphones attached (i.e. loaded)
Dynamic Range: SD > Exynos
Freq response: SD > Exynos
THD: Exynos > SD

Where are you getting the dynamic range comparison from? Also, seems to me like frequency response goes to the Exynos too. There's a more notable drop off with the SD below 200Hz, whereas the Exynos's doesn't kick in till 30Hz. Likewise with the upper end, with the SD tailing off after 5k, whilst the Exynos seems consistent all the way through.

lnYbM3P.jpg


gsmarena_a001.png


Granted, neither changes in frequency response are likely to be audibly perceivable, as the db values are so miniscule.
 
Last edited:

pyrokinetic

Member
Sep 16, 2012
29
3
Sofia
Well, given that the SNR is only barely better than the S7 and the FR is actually worse, I don't think there is much you can do with that device :D
There's a reason why those AP's are so ridonculously expensive :'(

Thanks. Second question, what is the frequency step during FR measurement? I presume it is measured by variying the frequency of some test tone, please correct me if wrong. I am thinking of borrowing an osciloscope with some samples capture capability and post processing afterwards.
 

ethan3686

Senior Member
Aug 19, 2009
1,168
135
India
Ok this is all alien language to me :p
Can you explain simply? Is the Exynos edge better quality and louder sounding than the Qualcomm through headphones?
 

Willyman

Senior Member
Aug 26, 2015
414
162
Where are you getting the dynamic range comparison from? Also, seems to me like frequency response goes to the Exynos too. There's a more notable drop off with the SD below 200Hz, whereas the Exynos's doesn't kick in till 30Hz. Likewise with the upper end, with the SD tailing off after 5k, whilst the Exynos seems consistent all the way through.

lnYbM3P.jpg


gsmarena_a001.png


Granted, neither changes in frequency response are likely to be audibly perceivable, as the db values are so miniscule.

I was going by the numbers you posted compared to my findings. The graph you post shows a flatter response for the Exyones however. But like you rightfully point out... for all intents and purposes, both are ruler flat. Anyone who says they can hear a 0.1 dB difference is flat out lying.
 

Willyman

Senior Member
Aug 26, 2015
414
162
Ok this is all alien language to me :p
Can you explain simply? Is the Exynos edge better quality and louder sounding than the Qualcomm through headphones?

It's impossible to judge sound quality based on these figures alone. I have not heard the Exynos implementation yet so I cannot comment.

What I can say about the objective results of this test is that the Snapdragon performs well on all counts except for the THD (harmonic distortion). It's still less than 1% though even at max level, so inaudible (there's plenty of research on the audibility of THD...). The iPhone 6 Plus generated a whopping 8% of distortion at max level! It's essentially clipping the signal at max level. You don't see anyone losing their **** over that.
 
  • Like
Reactions: tsm106

Maxha

New member
Mar 15, 2016
1
0
S7 snapdragon version and exynos version has totally dofferent sound chipset
Snapdragon version use chipset called wcd9335 that is implented most of the snapdragon 820 powered smartphone.
Exynos version use custom chipset CS42L73 that was slightly enhanced from galaxy s6 version.
 

dodgeboy

Senior Member
Dec 6, 2006
255
39
Detroit, MI
I'd love to see how the THD differs between the top 3 or 4 volume settings. I can hear the EQ changing as I increase the volume near the top end. If I keep the volume at about 75%, it sounds acceptable (albeit very quiet). When I go beyond that, all kinds of weird things happen to the sound.
 

Willyman

Senior Member
Aug 26, 2015
414
162
I'd love to see how the THD differs between the top 3 or 4 volume settings. I can hear the EQ changing as I increase the volume near the top end. If I keep the volume at about 75%, it sounds acceptable (albeit very quiet). When I go beyond that, all kinds of weird things happen to the sound.

I will do so tomorrow the measurements are at work. I recall the THD does increase significantly during the last few volume steps. Still, I sincerely doubt the level is high enough to cause any Audible side effects. I believe that what is happening is that the exponential volume steps result in timbre changes because that's just how our hearing works. Look up equal loudness curves to understand better what I'm trying to explain.
 

Top Liked Posts

  • There are no posts matching your filters.
  • 9
    Hi guys,

    As I did with my Note 5, I performed some AudioPrecision measurements on the headphone output of the S7 Edge (Snapdragon version, SM-G9350).

    UPDATE 1: I made an error in the max output level. It's -9 dBV/360mV into a 16 Ohm load, not -13 dBV. Still... it's low. Same as Mate 8/Z5/, lower than V10 (-6 dB), a LOT lower than iPhone (-1 dBV).

    UPDATE 2:
    • Redid the Dynamic range measurement using the more convention AES617 standard. Shows 100 dB now.
    • THD is really quite poor... It raises as a function of frequency and output level. My colleague suspects some form of capacitive loading. I would say this is probably the worst metric of this device, the other elements are very acceptable/good.


    Setup
    • Audio Precision APx525
    • Loaded (unless specified): 16 Ohm

    Result summary

    • Supports high resolution playback up to 192 kHz.
      Flat frequency response: <0.08 dB delta upto 96 kHz
      Very good dynamic range (AES17): 100 dB
      LOW output level: -9dBV (360mV using a 0dBFS signal) into a 16 Ohm load (for reference, Asian iPhone 6 = -1 dBV)
      Output impedance: 5.6 Ohm
      Crosstalk: didn't have to means to this measurement, sorry
      THD+N: at nominal level = max 0.05%, at maximum level = max 0.7%

    Some of these results are quite disappointing. Especially the very low output levels and high THD+N at max level. Some graphs below for illustration.

    Subjective assessment
    Been comparing the quality to my reference setup at work (PC to USB DAC Sony PHA-3 line-out to Talisman T-3H full Class-A amp). Used a number of sources (44/16 FLAC, 192/24 FLAC, Tidal HiFi).

    I would say the sound quality is really quite good for a phone, but not exceptional. There's nothing really wrong with it. Good spatial quality, extended in the highs and the lows. It just doesn't sound that WOW and lacks somewhat in its dynamic delivery and instrument seperation. My guess is that the limited amplification power is letting it down somewhat in this regard. It's a pity I don't have my Note 5 anymore for direct comparison, but from memory I believe that was a better performer.

    Now, to call it "incredibly disappointing" is a gross overstatement... (see some other threads on here).

    Dynamic Range (AES17)
    0oakTs7.jpg


    Noise floor measurement loaded (multi-tone signal)
    vlAlC2v.jpg


    Frequency response (loaded)
    lnYbM3P.jpg


    THD+N spectrum at maximum output level
    tcum5CF.jpg
    1
    Do this for the eyxnos model if you can
    1
    Do this for the eyxnos model if you can

    Sorry, I only have the SD version.
    1
    Ok this is all alien language to me :p
    Can you explain simply? Is the Exynos edge better quality and louder sounding than the Qualcomm through headphones?

    It's impossible to judge sound quality based on these figures alone. I have not heard the Exynos implementation yet so I cannot comment.

    What I can say about the objective results of this test is that the Snapdragon performs well on all counts except for the THD (harmonic distortion). It's still less than 1% though even at max level, so inaudible (there's plenty of research on the audibility of THD...). The iPhone 6 Plus generated a whopping 8% of distortion at max level! It's essentially clipping the signal at max level. You don't see anyone losing their **** over that.
    1
    May I ask a favor: Could someone with the Qualcomm model please post the mixer_paths.xml file? It sits in /system/etc/
    Just looked at the kernel driver for the WCD9335 and I'm interested in how Samsung sets it up. I noticed two settings in the codec that may contribute to sub-optimal audio reproduction.

    Edit: There may be another 6 dB of gain that can be applied via the mixer. @Willyman measured the maximum output voltage into a 16 Ohm load as 360 mV (around -9 dBV) while the output impedance of the codec/amp is 5.6 Ohm. As a result, the maximum "unloaded" (very high impedance load) output voltage of the headphone jack is close to -6 dBV. Depending on how exactly the amp built into the WCD9335 codec is configured, the maximum achievable "unloaded" output voltage should actually be very close to 1V. That's why I think there is another 6 dB of gain on the table.

    Here's my copy of mixer_paths.xml. However, since I've used the adapt sound feature, I'm not sure if these values are default. Also, since I don't have root (SD 820 in the US), I can't modify this file even if it could be modified to fix this problem.