[yuriv]

   
Hey Yuriv, it's Lachlan! Nice work on the measurements, I'm looking forward to seeing more results. We're discussing the EMI noise and your measurements in this thread, so it's nice to get some actual numbers.

 
Hi Lachlan and everyone else. I tried to measure the iPhone 6S noise on an analyzer with a lower noise floor than the Focusrite, but I didn't get anything conclusive yet. The noise is definitely there, but the voltages involved are very low. I'm currently evaluating a QA400 analyzer, and here's what it gets:
 

 
 
If I hadn't read about the the iPhone 6S's noise problem, I probably wouldn't have noticed it with my usual headphones and IEMs, and I wouldn't have dug up my UE600vi. In fact, I had to look up which IEMs on Inner Fidelity had the highest measured sensitivity to figure out which one to use. Their headphone measurement data sheets list how much voltage it took to produce 90 dB SPL. As far as I can tell, none of their headphones got a measurement lower than 11 mV rms, which is what Tyll got for the UE600 and a few others models. I suspect it takes even less than that because 11 mV could be getting close to the limits of their measuring system, the way it is set up. My scope can see a fuzzy waveform for a clearly audible 1 kHz tone that is in the hundreds of microvolts range. When I increase the volume, it's still well under 11 mV before it gets pretty loud on the UE600vi.
 
When I measured the noise on the Focusrite Saffire 6 USB last week, I boosted the instrument input gain to amplify the iPhone 6s's noise as much as possible to make it easier to measure. The 6S is very quiet when it's almost idle and the screen is off. If you short the inputs on the Focusrite, it gets the same measured noise as the iPhone. So the Saffire 6's noise covers up the true noise level of the iPhone 6S. Not so when it's on the app switcher.

Uncalibrated measurements on the Focusrite Saffire 6 + level meter readout
 
I included the Sandisk Sansa Clip Zip for comparison. On the UE600vi, I can hear noise from the Clip Zip, but it's a soft, constant background hiss, which is not as objectionable. Also, notice on the graph that the higher noise on the Clip Zip happens at lower frequencies, where human hearing is less sensitive. On the iPhone 6S's app switcher, the noise increase is where our hearing is much more sensitive.
 
Maybe I'll try a better preamp next time. Or I'll leave it to folks with more sensitive equipment to show how big the noise increase really is.
 
 
 
Band aids for those pesky IEMs
 

The humble voltage divider
 
 
I built the voltage divider I wrote about in a previous post and put it between the iPhone and the UE600vi. As expected, the frequency response going to the UE600vi is now much flatter, and the voltage going to it is down by at least 23 dB. The noise at the app switcher screen is all but gone. The down side is that you have to crank up the volume on the iPhone. But the UE600vi is so sensitive that I can get a satisfying level even without going to maximum volume. Maybe for another IEM, a 22+2 split works better than 15+1: less attenuation, but higher (2 instead of 1 ohm) effective output impedance. The humble voltage divider is a useful tool. It can even make a Pono or an AK100 sound good, but that's a subject for another thread.

About five years ago, when we had the iPhone 4 and Antennagate, Apple gave us free bumper cases so that we could get away with "you're holding it wrong." This time, I don't think they're going to give free voltage dividers to a tiny percentage of their customers. Such an adapter can be small, like the P-to-S adapter for the ER4PT. It would be nice if some outfit like Monoprice or Fiio would build one with four conductors so that the headset still works.
 
Another band aid is to use a headphone amp like a FiiO E5 or E6 for the attenuator instead of a voltage divider. Increase the headphone output volume to maximum on the iPhone 6S to maximize the signal-to-noise ratio (essentially using it as a line output) then cut the level on the amp. Here, the amp really isn't amplifying the signal voltage, but rather attenuating it. In the case of the FiiO E6, the UE600vi picks up a little hiss but it's not nearly as bad as the one on the app switcher. It's quieter if you use the not-so-portable O2. This will sound strange to some, but an HD600 makes the iPhone behave at its best, while you need an amp to make an SE535 sing.  
 

6S, E6, UE600

 

[elfary]

Thanks for your thorough assessment Yuriv.

As i don't believe in voltage dividers since they skew the intended fr of the balanced armatures i use (Shure SE535 and Westone UM Pro 50), i'll stick to the last year iPhones which are pitch black without the need for band aids. Hope that Apple clean their act up next year.

And they take out the status bar off the art-work


If the tonality it's not right, then nothing else really matters.

 

[davidsh]

Thanks for your thorough assessment Yuriv.

As i don't believe in voltage dividers since they skew the intended fr of the balanced armatures i use (Shure SE535 and Westone UM Pro 50), i'll stick to the last year iPhones which are pitch black without the need for band aids. Hope that Apple clean their act up next year.

And they take out the status bar off the art-work


If the tonality it's not right, then nothing else really matters.

No it won't. Rather the opposite

 

[elfary]

Thanks for your thorough assessment Yuriv.

As i don't believe in voltage dividers since they skew the intended fr of the balanced armatures i use (Shure SE535 and Westone UM Pro 50), i'll stick to the last year iPhones which are pitch black without the need for band aids. Hope that Apple clean their act up next year.

And they take out the status bar off the art-work


If the tonality it's not right, then nothing else really matters.

No it won't. Rather the opposite

Yes it will. And it does.

Believe it or not in ear monitors are not meant to have a flat fr. They are engineered to provide exactly the fr that they show when driven by zero ohm output impedance. That you might rather other tone it's another thing. But engineers do not design the fr thinking that end user will add a 30 ohm resistor. Actually Westone even bundled Fiio E6s for a while along with their iems.

I'll look into my email for the exchange about that matter that i had with a Westone engineer in 2011.


If the tonality it's not right, then nothing else really matters.

 

[davidsh]

The effective output impedance as seen by the headphone is significantly lowered in the configuration shown earlier (voltage divider) from what I hear.

 

[yuriv]

Yes it will. And it does.

Believe it or not in ear monitors are not meant to have a flat fr. They are engineered to provide exactly the fr that they show when driven by zero ohm output impedance. That you might rather other tone it's another thing. But engineers do not design the fr thinking that end user will add a 30 ohm resistor. Actually Westone even bundled Fiio E6s for a while along with their iems.

I'll look into my email for the exchange about that matter that i had with a Westone engineer in 2011.


If the tonality it's not right, then nothing else really matters.

davidsh is right. The voltage divider that was shown earlier is not the same as adding a 30-ohm resistor in series. Here's what happens when you omit R₂:
 


The 15-ohm resistor is now in series with the IEM, increasing the effective output impedance by 15 ohms. If we do this, the frequency response at V_Out will, indeed, be skewed. But if we keep R₂, the iPhone 6S will see a nearly constant 16-ohm load, and the IEM will see 1 ohm in parallel with (15 ohms + the iPhone 6S's output impedance), which comes out to slightly less than a 1-ohm effective output impedance.
 
I recorded the voltage at the UE600vi's terminals while an iPhone played an impulse. Then I imported the recorded impulse into Room EQ Wizard to derive the frequency response:
 

Magnitude response from Room EQ Wizard's import of impulse response, normalized, 1/6 octave smoothed
 
It'll look a lot worse if these were a Pono or an AK100. A voltage divider might help them deliver a flatter response and lower noise. The disadvantage, as noted in a previous post, is that the volume must be set higher on the player.

 

[elfary]

So i got your trick wrong. I thought your idea was just a resistor. Where do a regular guy get a 1 ohm voltage divider?



If the tonality it's not right, then nothing else really matters.

 

[castleofargh]

   
I recorded the voltage at the UE600vi's terminals while an iPhone played an impulse. Then I imported the recorded impulse into Room EQ Wizard to derive the frequency response:

 now I feel stupid for not thinking about it before. I worked around the problem by recording the IEM with a microphone and adding resistors in series, but still thanks for the idea of importing impulse responses from outside. it opens up a lot of possibilities I'll have to try. the obvious one being that I could use my 2I2 instead of the crap usb stuff I need to make my little mic work, and get a cleaner ADC in the process for stuff more demanding than FR.
so thank you for turning my brain ON ^_^
 
 

@elfary as mentioned, the main problem is that in yuriv example, he ends up 23db lower, with a starting point of 1V max that's a max of about 0.07V.
with the UE600 http://www.innerfidelity.com/images/UltimateEarsUE600.pdf  0.011Vrms seems to get us 90db loud(at 1khz).  so 0.07Vrms would go 90+16=106db. more than enough for my use as I never listen loudly, but add to that something like replaygain/soundcheck, or the use of EQ, and power might become a problem. we tend to clear a source for power when it can do 110 or 115db. not the case here with the voltage divider.
and the UE600 is mighty efficient, it would be quieter on almost any other IEM. so yuriv's warning about other values being more beneficial to other IEMs is very rational and important. it's not worth it getting 1 ohm, if the power becomes an issue.
so it might be a good idea to compromise on the resistors values, or just get an external amp with low impedance. that could solve all problems at once(except that you have to carry it and it costs a good deal more than 2 resistors :'(  ).
 
(as always, I hope I didn't mess up with the numbers)

 

[elfary]

Double amping would not the degrade quality?

I own a Fiio E6 and an iBasso T3. But always amp via line out dock.


If the tonality it's not right, then nothing else really matters.

 

[castleofargh]

well ideally an external DAC/amp would be best, but then we don't care about the phone's audio performance at all ^_^.
double amping is not audiophile, but when used with the phone volume pretty high for better SNR, the output into a few thousand ohm of another amp should be pretty nice I would guess, perhaps not line-out nice, but still. I doubt we would get much audible reason to complain. I used an amp plugged into a laptop or on the move into the HO of a DAP for a long time when I started my audiophool adventure. it was a great way to remove the hissing on the laptop and I was very happy about that.
only later on did I learn that I was doing something evil ^_^.

 

[yuriv]

Quote:

   now I feel stupid for not thinking about it before. I worked around the problem by recording the IEM with a microphone and adding resistors in series, but still thanks for the idea of importing impulse responses from outside. it opens up a lot of possibilities I'll have to try. the obvious one being that I could use my 2I2 instead of the crap usb stuff I need to make my little mic work, and get a cleaner ADC in the process for stuff more demanding than FR.
so thank you for turning my brain ON ^_^
 
...

 
Last night was the first time I tried doing this with REW. What you see in the graph are the results of one impulse. I recorded  a 1 Hz impulse train in Audacity, then I selected the region surrounding one impulse and exported it into a .wav file to be imported into REW. It gets better resolution if you leave a lot of time after the impulse and you record at the highest level possible without clipping. REW can also average the results; you might get something more reliable if you have many recorded impulses for averaging.

I also tried to import a whole train of 60 recorded impulses, spaced 1s apart. This will start to look like a periodic 1 Hz signal to the FFT, with components at 1 Hz, 2 Hz, ..., but I'm lazy and I don't want to average things by hand. The results seem to be fine as long as you don't look at really low frequencies.

About the voltage divider: I described in a previous post my experience measuring signals going to the UE600vi while listening to it. I gave my reasons for suspecting that it actually takes less than 11 mV rms to produce 90 dB SPL. So I might be getting close to 110 dB SPL at the maximum volume even with the 15Ω/1Ω voltage divider. In any case, the noise is gone and it gets loud enough for me without going to the maximum volume on the iPhone.
 
 
 
Quote:

Double amping would not the degrade quality?

I own a Fiio E6 and an iBasso T3. But always amp via line out dock.


If the tonality it's not right, then nothing else really matters.

 
The objections to "double amping" usually include the following:
 

• The line ins of many amplifiers will clip with a signal that's only a little higher than 2V. So if you amplify the signal before feeding it into the second amplifier, it might clip.
• The second amplifier will amplify the noise of the first one.

First, the headphone output of the iPhone is only 1V rms with a full-scale sine wave at maximum volume. It's not going to overload the input of the headphone amplifier. Second, if we set the iPhone at maximum volume, then the signal reaching the amplifier will have the maximum signal-to-noise ratio possible. If the headphone amp's load is a sensitive IEM, then we don't want to further amplify the voltage coming from the iPhone; we want to attenuate it. So everything coming from the iPhone is attenuated by the amp, including the noise.

Head-fiers usually complain when see iPod and iPhone headphone output measurements with no load attached. The numbers are simply too good compared to what happens in real-world use--with actual headphones attached. But those unloaded measurements are exactly what's going on here. The iPhone headphone output has very good numbers when it's not driving a headphone, but rather the line in of an amplifier, with impedance in tens of kilohms, if not more. It's a very easy load for a Phone.

iDevices, despite having only 1V available at he headphone output, have better noise and distortion performance than the line outs of many consumer devices. If you're using a FiiO E6, then it, instead of the iPhone headphone output, is the one limiting the sound quality, as it is the one that has to drive the difficult load.

Of course, if you max the gain on the FiiO E6 and keep the iPhone's volume low, then you're not just amplifying the iPhone's noise; you're adding the FiiO's noise on top. But if you keep the iPhone volume at maximum, then it will be way too loud for an IEM if the FiiO keeps it at that level. The FiiO must attenuate the signal from the iPhone, incoming noise included.

Essentially, you're trading the iPhone's noise and output impedance for the FiiO E6's.

As for the voltage divider, I don't know of any that you can buy already built. It would be nice if an outfit like, say, Monoprice built 15Ω/1Ω or 15Ω/2Ω voltage divider adapters, especially ones with four conductors: TRRS. But they don't, so you pretty much have to build one yourself. Experiment with the resistor values to find the ones that work best for the IEM at hand.

 

[castleofargh]

I never got to know exactly what this was, but I suppose it might be a voltage divider http://www.head-fi.org/t/695086/ifi-nano-ican-name-that-attenuator-competition-winner-no-1-winner-of-ican-nano-is
 
 

  Quote:

   now I feel stupid for not thinking about it before. I worked around the problem by recording the IEM with a microphone and adding resistors in series, but still thanks for the idea of importing impulse responses from outside. it opens up a lot of possibilities I'll have to try. the obvious one being that I could use my 2I2 instead of the crap usb stuff I need to make my little mic work, and get a cleaner ADC in the process for stuff more demanding than FR.
so thank you for turning my brain ON ^_^
 
...

 
Last night was the first time I tried doing this with REW. What you see in the graph are the results of one impulse. I recorded  a 1 Hz impulse train in Audacity, then I selected the region surrounding one impulse and exported it into a .wav file to be imported into REW. It gets better resolution if you leave a lot of time after the impulse and you record at the highest level possible without clipping. REW can also average the results; you might get something more reliable if you have many recorded impulses for averaging.

I also tried to import a whole train of 60 recorded impulses, spaced 1s apart. This will start to look like a periodic 1 Hz signal to the FFT, with components at 1 Hz, 2 Hz, ..., but I'm lazy and I don't want to average things by hand. The results seem to be fine as long as you don't look at really low frequencies.

About the voltage divider: I described in a previous post my experience measuring signals going to the UE600vi while listening to it. I gave my reasons for suspecting that it actually takes less than 11 mV rms to produce 90 dB SPL. So I might be getting close to 110 dB SPL at the maximum volume even with the 15Ω/1Ω voltage divider. In any case, the noise is gone and it gets loud enough for me without going to the maximum volume on the iPhone.
 

 thanks, I'll try a few sizes for the impulse and see how it goes.
and about my warning for max voltage with the divider, it was of course a warning in general for others, to say that there is no such thing as a free lunch.
I'm not against the idea of a voltage divider. someone did that on a sony A15 recently that is about 4ohm, then I was against the idea because the A15 is maxed out at 0.4v(even less on my euro version), so losing any more voltage would really limit the possible usage of the DAP.  but I'm sure there are many instances where it is very useful, and not just for impedance purpose.

 

[elfary]

Thanks a lot Yuriv for such a detailed explanation about double ampling. Which is the route i might go if at some point i feel the urge to ditch my iPhone 6 to get a 6s.
 
anyway too bad that Apple took such a step back. Never ever heard a noise off my iPhone 6 or 5s headphone out. No matter the headphones, screen, wifi, bluetooth or modem.

 

[yuriv]

  I never got to know exactly what this was, but I suppose it might be a voltage divider http://www.head-fi.org/t/695086/ifi-nano-ican-name-that-attenuator-competition-winner-no-1-winner-of-ican-nano-is
 
...

 
Well, well, well. I bet they are. 1Ω output impedance for their -24 dB attenuator very much sounds like a 15Ω/1Ω voltage divider. 3Ω output impedance for their -12 dB attenuator sounds like 3Ω + 9Ω. If you use that one, the iPhone will see a 12Ω load, which is low enough to make the phone send a clipped signal if you have the volume high enough. So we have to be careful if we're using that one with portables. Heck, the 5S went into clipping with a 16Ω load at maximum volume. More sensible would have been 4Ω + 12Ω for -12 dB and 4Ω output impedance, or even 3Ω + 21Ω (-18 dB). 
 

 
The pic comes from this post: http://www.head-fi.org/t/695086/ifi-nano-ican-name-that-attenuator-competition-winner-no-1-winner-of-ican-nano-is/450#post_11667445

 

[SilverEars]

   
The objections to "double amping" usually include the following:
 

• The line ins of many amplifiers will clip with a signal that's only a little higher than 2V. So if you amplify the signal before feeding it into the second amplifier, it might clip.
• The second amplifier will amplify the noise of the first one.

First, the headphone output of the iPhone is only 1V rms with a full-scale sine wave at maximum volume. It's not going to overload the input of the headphone amplifier. Second, if we set the iPhone at maximum volume, then the signal reaching the amplifier will have the maximum signal-to-noise ratio possible. If the headphone amp's load is a sensitive IEM, then we don't want to further amplify the voltage coming from the iPhone; we want to attenuate it. So everything coming from the iPhone is attenuated by the amp, including the noise.

Head-fiers usually complain when see iPod and iPhone headphone output measurements with no load attached. The numbers are simply too good compared to what happens in real-world use--with actual headphones attached. But those unloaded measurements are exactly what's going on here. The iPhone headphone output has very good numbers when it's not driving a headphone, but rather the line in of an amplifier, with impedance in tens of kilohms, if not more. It's a very easy load for a Phone.

iDevices, despite having only 1V available at he headphone output, have better noise and distortion performance than the line outs of many consumer devices. If you're using a FiiO E6, then it, instead of the iPhone headphone output, is the one limiting the sound quality, as it is the one that has to drive the difficult load.

Of course, if you max the gain on the FiiO E6 and keep the iPhone's volume low, then you're not just amplifying the iPhone's noise; you're adding the FiiO's noise on top. But if you keep the iPhone volume at maximum, then it will be way too loud for an IEM if the FiiO keeps it at that level. The FiiO must attenuate the signal from the iPhone, incoming noise included.

Essentially, you're trading the iPhone's noise and output impedance for the FiiO E6's.

As for the voltage divider, I don't know of any that you can buy already built. It would be nice if an outfit like, say, Monoprice built 15Ω/1Ω or 15Ω/2Ω voltage divider adapters, especially ones with four conductors: TRRS. But they don't, so you pretty much have to build one yourself. Experiment with the resistor values to find the ones that work best for the IEM at hand.

That's a good explanation.  I can see the reason why the voltage divider is not used is because it lowers the output voltage for headphones that require it.  That's 1/16 the voltage output.
 
The voltage divider is clever, and it does indeed change the output impedance to 1ohms as David pointed as the output impedance is what the headphone's sees.  This could be a popular mod for those that are using the AK100MK1 with iems.  
 
I think you hit everything that people should know when it comes to double amping.  Like the amp used for iems are not really for ampling, but it's actually attenuating the signal, and also it lowers the output impedance.  Kinda acts like a buffer and noise attenuate rather than what it's called, an amp.