The RMAA (RightMark Audio Analyzer) Source and Audio Device Measurement Thread [Overview of Measurements in Post #3, Tutorial in Post #2]

Discussion in 'Sound Science' started by hifichris, Mar 1, 2016.
5 6 7 8 9 10 11 12 13 14
Page 15 of 17
16 17
  1. HiFiChris Contributor
    Yeah, it could also be lower for my very personal standards that prefer 0.5 Ohms and below - but it' not that 1.2/1.3 Ohms are a high number at all, and while my inner output impedance perfectionist is slightly paranoid seeing that, my common sense tells me that it's still a value that is a good bit lower than what other devices have in terms of OI.

    Most multi-BA models that aren't as much diva-like as the Triple-Fi 10 shouldn't show much deviation from the unloaded measurement at all.
    Last edited: Dec 14, 2017
  2. hakuzen
    agree. i do like the results you got from FiiO Q1 MkII. good value indeed. ..but i suffer of same perfectionism/paranoia :wink:
    HiFiChris likes this.
  3. yuriv
    Google USB-C audio adapter with USB-C to USB-A adapter and box.jpg
    The Google USB-C to 3.5mm headphone adapter for the Pixel 2 and Pixel 2 XL used to be $20--a fair price, I thought, because it can do more than the Apple Lightning to 3.5 mm headphone adapter. For example, it works with MacOS and Windows 10 Creators Update with the default drivers. But after some backlash from the public, Google lowered the price to $9, the same as the Apple audio dongle.

    Google USB-C audio adapter + USB-A adapter.jpg

    Over at xda-developers, they already have RMAA measurements when running it from Android. This post is about how it does in Windows, which shouldn’t be that much different.

    When you plug the adapter into a PC, Windows doesn’t show it in its list of audio devices until something is plugged into the 3.5mm jack. If it’s a very high impedance load, Windows will say that it’s a line out. I measured 1.88 Vrms for a full-scale 1 kHz sine wave, which is the same as for the unit measured over at xda-developers. That’s a good 5.5 dB more than the Apple adapter’s maximum voltage. With Windows volume at 100, there’s how the line out did in RMAA:

    Google USB-C audio adapter - line out RMAA summary.PNG
    Test results from left to right: 1644, 2444, 2448, 2496

    Google USB-C audio adapter, line out - 16 vs 24 bit 2444 noise.PNG
    Relative noise level, 1644 and 2444

    Google USB-C audio adapter, line out - 16 vs 24 bit 2444 DR.PNG
    Dynamic range, 1644 and 2444

    Google USB-C audio adapter, line out - 16 vs 24 bit 2444 1k THD.PNG
    1 kHz THD, 1644 and 2444

    Google USB-C audio adapter, line out - 16 vs 24 bit 2444 IMD.PNG
    SMPTE IMD, 1644 and 2444

    Google USB-C audio adapter, line out - 16 vs 24 bit 2444 stereo crosstalk.PNG
    Stereo crosstalk, 1644 and 2444

    The frequency response in Windows 10 Creators Update seems to be limited: RMAA 2496 frequency response measurement shown.

    The adapter doesn’t switch from line out mode to headphone out mode until the 3.5mm plug is disconnected from the adapter’s jack. It’s possible to get the higher voltage mode for a headphone with these steps: plug in something like a splitter or a 3.5mm to 6.3mm adapter with no headphones attached; the adapter sees an open circuit so switches to line out mode. Then plug in the headphones into the splitter or adapter. The device stays in line out mode. This could be useful for high impedance headphones like an HD600:

    Google USB-C audio adapter <-- headphone splitter or 6.3mm adapter​

    This switches it to line out mode (max. 1.88 Vrms). Then plug in the headphones, for example:

    Google USB-C audio adapter <-- headphone splitter or 6.3mm adapter <-- HD600​


    When the adapter is plugged into the PC and it sees a lower-impedance load, it switches to headphone out mode. Here’s how I determined the maximum voltage available:

    Google USB-C audio adapter <-- headphone splitter <-- 16-ohm dummy load or earphones​

    This switches the adapter to headphone out mode. When the dummy load is unplugged, it stays in headphone out mode. It doesn’t switch to a line out. I measured around 400 mV open-circuit voltage in headphone out mode this way. At xda-developers, they got two different headphone out modes, with different voltages. I didn't test to see if a third level was available in Windows or MacOS.

    Here’s how the adapter performed with the 16-ohm dummy load:

    Google USB-C 2444 into 16 ohms RMAA summary.PNG
    Headphone out mode RMAA test results at Windows volume 100 (left) and 68 (right)

    Google USB-C 2444 into 16 ohms RMAA frequency response.PNG
    Frequency response: Headphone out mode into 16 ohms vs Line out mode

    Google USB-C 2444 into 16 ohms RMAA noise level and dynamic range.PNG
    Relative noise level and dynamic range: Headphone out mode into 16 ohms vs Line out mode

    Google USB-C 2444 into 16 ohms RMAA 1k THD.PNG
    1k THD into 16 ohms at Windows volume 100 and 68.

    For me, comfortable listening level with most dynamic-driver IEMs doesn't exceed Windows volume 60. It's usually closer to 50.

    Impulse response.png
    Impulse response

    High gain, max volume, no load.png
    Line out mode, no load, Windows volume 100

    High gain, max volume, 16-ohm load on both channels.png
    Line out mode, 16-ohm load on both channels, Windows volume 100

    High gain, Windows volume at 70, 16-ohm load on both channels.png
    Line out mode, 16-ohm load on both channels, Windows volume 70

    High gain, Windows volume at 68-2, 16-ohm load on both channels.png
    Line out mode, 16-ohm load on both channels, Windows volume 68, onset of clipping

    High gain, Windows volume at 66, 16-ohm load on both channels.png
    Line out mode, 16-ohm load on both channels, Windows volume 66

    Low gain, max volume, no load 2.png Low gain, max volume, 16-ohm load on both channels.png
    Headphone out mode, Windows volume 100: no load (left), 16-ohm load on both channels (right)

    See also 407 * 16.2/(R + 16.2) = 393

    16-bit J-test.png 16-bit J-test signal

    24-bit J-test.png
    24-bit J-test signal

    I may have to double-check the J-test results.

    My subjective take on the noise level with a UE600: It’s a very soft background hiss. It doesn’t produce irregular-sounding noises like an iPhone 6S does. At least not when moving around windows or when the SSD is busy.

    So far, this Google adapter has outperformed the built-in audio of every computer that will recognize it. I tried it with a 2016 Lenovo Carbon X1, a few Microsoft Surface tablets, and a Chuwi mini PC. Macs usually have clean audio, but they can’t output 1.88 Vrms like this adapter.

    The headset mic input works well with my headphone measurement rigs and REW, both in Windows and MacOS. It also works with my 2013 Nexus 7, which doesn’t have great built-in audio:

    Google USB-C audio adapter, Nexus 7 (2013), IEC711 clone, Etymotic ER4PT.jpg

    Edit 12/20: The mic input has a significant amount of roll-off in the bass. I compared it to the Startech ICUSBAUDIO2D audio interface, and it's around 5-6 dB down at 20 Hz. It probably has a smaller coupling capacitor. So if you're planning to use this adapter in a measurement rig, you'll need to make a calibration curve for the mic/mic input system.

    Sadly, I couldn’t get the adapter to work with my Amazon Fire tablets. Maybe a FireOS update will fix it in the future.

    Edit 12/30: It works on the 2017 Fire 10 HD with Fire OS! Still no luck on the 2017 Fire HD 8 or the 2014 Fire HD 6, both also running Fire OS
    Last edited: Dec 30, 2017
    hakuzen, bartzky and HiFiChris like this.
  4. castleofargh Contributor
    didn't know you had a QA400. ^_^ (I'm secretly jealous of anybody who owns a 400 or 401).
  5. HiFiChris Contributor

    You don't happen to have an output impedance spec for the adapter, do you?
  6. yuriv
    I've had it for a while. There was a photo of it when I posted the Apple Lightning audio dongle measurements. But I would have gone for the 401 if it were available at the time.

    It's in the post. I guess the link to Wolfram Alpha is easy to miss: 407 * 16.2 / (R + 16.2) = 393

    They're getting a lower figure over at xda-developers: 0.43 ohms. I'm more conservative and try to make it so that if I mess up the measurement, I err on the high side.
    HiFiChris likes this.
  7. SilverEars
    I just saw this. This explains the response up top. Since it's not flat, it tells me there is a bit of a reactive load.


    Looking at the TF10 impedance response and the Mojo's impedance response, it's apparent why the response looks like that.

    Last edited: Dec 22, 2017
    tarhana and HiFiChris like this.
  8. HiFiChris Contributor
    Yeah, I'm fully aware of that - the Mojo's output impedance starts climbing in the middle highs, hence this behaviour. It's good to finally see that on a publicly available measurement graph, since the one that I am already aware of and that also confirms an impedance rise in the highs is in a copyrighted print document that I cannot share for obvious reasons - thanks for posting!
  9. Arpiben
    I would be curious to see Impedance curves of Hugo2 or Dave for comparison. Not yet found.. :L3000:
  10. SilverEars
    We all know that headphone out impedance is not purely resistive, and can be reactive, particularly capacitive to roll off the low end significantly with a capped output, but this is the first I've seen of a source output impedance response, and paritcularly the top end impedance vary in impedance specifically. This goes to show how much we can see from source output impedance response.
    Last edited: Dec 23, 2017
  11. headwhacker
    The deviation is only about 0.5dB, at worse it's about 1 dB a 10k I don't think it's significant enough to be noticeable
  12. castleofargh Contributor
    on one hand, it's obvious that it doesn't matter. picking any IEM over any other IEM will result in so much more changes. and just the way we place the IEM from one time to the next can generate a more significant variation in sound.
    on the other hand, it's for a device where the designer talks about silly small stuff as if they were audibly significant. so by my standards it's irrelevant, by the typical Chord standard it's a disaster. ^_^
    headwhacker, hakuzen and HiFiChris like this.
  13. bigshot
    If you level match speakers, you'll still get variations if you turn your head one way or the other too. There is such a thing as "close enough for government work".
  14. bartzky
    Interesting point of view :-D This made an interview with Watts come into my mind in which the talks about the huge benefits of 350 dB noise shaping performance...
  15. hakuzen
    by my hearing standards it's ok as well, but by multi-ba iem measuring standards, it made me search for another dac+amp source.. pissed off
5 6 7 8 9 10 11 12 13 14
Page 15 of 17
16 17

Share This Page