[Dannemand]

@csglinux and others:

I got myself into an interesting discussion with member DooMLoRD77 over on XDA (starting around here) about the noise/artifacts caused by Android Mixer when playing Redbook 16/44. I was lecturing at first about how it's the upsampling causing it. But he is making a good case that it isn't the upsampling, but rather the Mixer applying Digital Volume Control (DVC) and thus losing resolution (sample bits).

He tested config/policy changes forcing the Mixer to output 24-bit, which eliminated the noise for him, presumably due to the higher resolution being less affected by DVC. And he tested forcing the Mixer to output 44.1KHz (no upsampling) which still produced noise, presumably because of the DVC. I asked him to share the config/policy changes he made, but have not gotten them yet.

Meanwhile, we all agree that bypassing the Mixer (DIRECT path) is what we really want, rather than merely "improve" the Mixer to molest our music less. He is now trying to achieve that.

Feel free to join the discussion if you want!

 

[csglinux]

@csglinux and others:

I got myself into an interesting discussion with member DooMLoRD77 over on XDA (starting around here) about the noise/artifacts caused by Android Mixer when playing Redbook 16/44. I was lecturing at first about how it's the upsampling causing it. But he is making a good case that it isn't the upsampling, but rather the Mixer applying Digital Volume Control (DVC) and thus losing resolution (sample bits).

He tested config/policy changes forcing the Mixer to output 24-bit, which eliminated the noise for him, presumably due to the higher resolution being less affected by DVC. And he tested forcing the Mixer to output 44.1KHz (no upsampling) which still produced noise, presumably because of the DVC. I asked him to share the config/policy changes he made, but have not gotten them yet.

Meanwhile, we all agree that bypassing the Mixer (DIRECT path) is what we really want, rather than merely "improve" the Mixer to molest our music less. He is now trying to achieve that.

Feel free to join the discussion if you want!

Very interesting post - thanks for your continued efforts in all this @Dannemand! IMHO, the errors have to come from a combination of resampling interpolation error with 16-bit integer amplitudes and insufficient bits to accurately represent those amplitudes at low volumes. I recently noticed that the new Amazon "HD"(?) music service appears to use its own floating-point representation for its up-sampling and it doesn't have these audible nasties at low volumes:
https://www.head-fi.org/threads/mus...s-for-the-lg-v30.868978/page-51#post-15220713

There'll still be errors there with 24-bit int or 32-bit float interpolation, but they will be much lower. (Right now, to my ears, Amazon's "Ultra HD" doesn't sound as good as a Tidal Master, but it's still much better than 44/16 PCM on the V30 at low volumes through the Android mixer.)

A while back I spent several days with a rooted V30, trying to find some useful global/system-wide change for the mixer. I'm afraid I failed and gave up Both the UAPP and Neutron devs told me they thought it was an impossible task. So if this individual has really figured this out and can tell us how they forced the mixer to 24-bit output, that would be awesome! How much do we need to bribe them?

 

[Dannemand]

Very interesting post - thanks for your continued efforts in all this @Dannemand! IMHO, the errors have to come from a combination of resampling interpolation error with 16-bit integer amplitudes and insufficient bits to accurately represent those amplitudes at low volumes. I recently noticed that the new Amazon "HD"(?) music service appears to use its own floating-point representation for its up-sampling and it doesn't have these audible nasties at low volumes:
https://www.head-fi.org/threads/mus...s-for-the-lg-v30.868978/page-51#post-15220713

There'll still be errors there with 24-bit int or 32-bit float interpolation, but they will be much lower. (Right now, to my ears, Amazon's "Ultra HD" doesn't sound as good as a Tidal Master, but it's still much better than 44/16 PCM on the V30 at low volumes through the Android mixer.)

A while back I spent several days with a rooted V30, trying to find some useful global/system-wide change for the mixer. I'm afraid I failed and gave up Both the UAPP and Neutron devs told me they thought it was an impossible task. So if this individual has really figured this out and can tell us how they forced the mixer to 24-bit output, that would be awesome! How much do we need to bribe them?

Hey Buddy,

Oh I remember all your work with the config and policy files. And several others tried as well (including myself) even if not as extensively as you did. What we tried (or at least what I tried) was to configure the audio_flinger to use DIRECT path when playing 16/44. What DooMLoRD77 said he'd done was configure the Mixer to output 44.1KHz or 24-bit (although not in combination yet, as I understand). That's still not ideal, as we're still sending our music stream through a meat grinder, even if the grinding is less damaging.

What we want is ideally to use DIRECT path and bypass the Mixer altogether. But if that keeps proving impossible, then maybe it's time to accept this improvement with the Mixer over an ideal solution that keeps eluding us.

I see your point that it might be a combination of the upsampling with the meager 16 bits resolution that produces the artifacts, and not DVC as DooMLoRD77 contends (although kind of the same problem, just worse if indeed the Mixer applies DVC). I took note of your reporting about Amazon HD being able to resample without artifacts (or at least less audible). Resampling with PowerAmp also doesn't produce the artifacts. So we have some indication there.

Let's see what DooMLoRD77 comes up with. I know we've been disappointed before.

Once again feel free (anybody) to join the discussion on XDA, even if just for encouragement. I think that's the bribe we need!

 

[charleski]

This is about the LG G7ThinQ, but since it has the same DAC as the V30 I thought I'd post in this thread to keep all the info collected.

I've been looking for a new DAC and found a G7 on eBay for £cheap which seemed like a better option than an external DAC/Amp which would cost the same (actually a little more) and be an extra box to lug around. The only thing that worried me was whether I'd get sufficient current output to drive my low impedance (16Ω) headphones, especially after seeing eio's review of the V20, which showed a lot of clipping on a low impedance load when tricking it into aux mode. The V20 uses an earlier version of the ESS chip found in the V30/G7, and I wanted to see if there was any difference.

My measurement system is very primitive and really geared towards measuring frequency response for EQing, but good enough to compare the different modes. I measured the output through my low-distortion Sony MH755s (16Ω) using a -0.3dB 1kHz sine wave. The following results are only applicable to similar low-impedance headphones, though you'll probably see similar results with a 32Ω set.. They also are not reflective of the actual distortion and noise output of the phone, which will be several orders of magnitude lower as shown by the audioscience review. This is just an attempt to get some real-world measurements so I could decide if it's worth bothering about the impedance switching and might be useful to others.

Using the RTA built into REW I compared the distortion results from the 1kHz sine wave in both normal mode (as shown by the Hi-Fi Status app) and aux mode (tricked by first plugging in an extension cord then plugging the headphones into that). All tests were done using Neutron as the player (outputting at 32bits and native frequency), which remaps Android's 16-step volume scale to a proper 0-100 range rather than the 0-75 used natively, so I report the volume setting as 0 for maximum, then -1, -2, -3, etc for each step down below that. THD+N measures weren't very helpful, with the results dominated by the noise floor (mostly external environmental noise). But it's clear that full volume in aux mode drives the output amp into some nasty clipping, though below that it seems fine.


Here's the damage at full volume in aux mode into 16Ω:


Looking at the THD numbers alone is a bit more useful in this case:


This is distortion from the full playback and recording chain (phone output+headphones+recording circuitry on my motherboard). So while max volume in aux mode is unusable, stepping the volume down 1 step yields a distortion of only 0.5%, which is quite acceptable.

The real question is whether it's worth bothering with aux mode on low impedance headphones at all. At max volume in normal mode I was getting the same volume output from the headphones as when I set the volume to -3 in aux mode. So basically aux mode only gives you an extra 2 steps of usable volume, which comes out at 5.7dB ([aux mode vol -1] - [normal mode vol 0]).

Summary
Aux mode isn't worth bothering about, unless you absolutely have to squeeze the maximum volume out of the phone. If you really need more volume into low impedances it's probably best to go with an external amp that can supply the extra current. I suspect that reports of it sounding better in aux mode are mostly placebo effect.

 

[rjoudrey]

This is about the LG G7ThinQ, but since it has the same DAC as the V30 I thought I'd post in this thread to keep all the info collected.

I've been looking for a new DAC and found a G7 on eBay for £cheap which seemed like a better option than an external DAC/Amp which would cost the same (actually a little more) and be an extra box to lug around. The only thing that worried me was whether I'd get sufficient current output to drive my low impedance (16Ω) headphones, especially after seeing eio's review of the V20, which showed a lot of clipping on a low impedance load when tricking it into aux mode. The V20 uses an earlier version of the ESS chip found in the V30/G7, and I wanted to see if there was any difference.

My measurement system is very primitive and really geared towards measuring frequency response for EQing, but good enough to compare the different modes. I measured the output through my low-distortion Sony MH755s (16Ω) using a -0.3dB 1kHz sine wave. The following results are only applicable to similar low-impedance headphones, though you'll probably see similar results with a 32Ω set.. They also are not reflective of the actual distortion and noise output of the phone, which will be several orders of magnitude lower as shown by the audioscience review. This is just an attempt to get some real-world measurements so I could decide if it's worth bothering about the impedance switching and might be useful to others.

Using the RTA built into REW I compared the distortion results from the 1kHz sine wave in both normal mode (as shown by the Hi-Fi Status app) and aux mode (tricked by first plugging in an extension cord then plugging the headphones into that). All tests were done using Neutron as the player (outputting at 32bits and native frequency), which remaps Android's 16-step volume scale to a proper 0-100 range rather than the 0-75 used natively, so I report the volume setting as 0 for maximum, then -1, -2, -3, etc for each step down below that. THD+N measures weren't very helpful, with the results dominated by the noise floor (mostly external environmental noise). But it's clear that full volume in aux mode drives the output amp into some nasty clipping, though below that it seems fine.


Here's the damage at full volume in aux mode into 16Ω:


Looking at the THD numbers alone is a bit more useful in this case:


This is distortion from the full playback and recording chain (phone output+headphones+recording circuitry on my motherboard). So while max volume in aux mode is unusable, stepping the volume down 1 step yields a distortion of only 0.5%, which is quite acceptable.

The real question is whether it's worth bothering with aux mode on low impedance headphones at all. At max volume in normal mode I was getting the same volume output from the headphones as when I set the volume to -3 in aux mode. So basically aux mode only gives you an extra 2 steps of usable volume, which comes out at 5.7dB ([aux mode vol -1] - [normal mode vol 0]).

Summary
Aux mode isn't worth bothering about, unless you absolutely have to squeeze the maximum volume out of the phone. If you really need more volume into low impedances it's probably best to go with an external amp that can supply the extra current. I suspect that reports of it sounding better in aux mode are mostly placebo effect.

I have paired a Cayin C5 Amp with my V30 and found it hits all the areas (especially low frequency) I thought were slightly lacking with my HD660s headphones. None of the Fiio amps I tried made any difference. I highly suggest this amp.

 

[csglinux]

This is about the LG G7ThinQ, but since it has the same DAC as the V30 I thought I'd post in this thread to keep all the info collected.

I've been looking for a new DAC and found a G7 on eBay for £cheap which seemed like a better option than an external DAC/Amp which would cost the same (actually a little more) and be an extra box to lug around. The only thing that worried me was whether I'd get sufficient current output to drive my low impedance (16Ω) headphones, especially after seeing eio's review of the V20, which showed a lot of clipping on a low impedance load when tricking it into aux mode. The V20 uses an earlier version of the ESS chip found in the V30/G7, and I wanted to see if there was any difference.

My measurement system is very primitive and really geared towards measuring frequency response for EQing, but good enough to compare the different modes. I measured the output through my low-distortion Sony MH755s (16Ω) using a -0.3dB 1kHz sine wave. The following results are only applicable to similar low-impedance headphones, though you'll probably see similar results with a 32Ω set.. They also are not reflective of the actual distortion and noise output of the phone, which will be several orders of magnitude lower as shown by the audioscience review. This is just an attempt to get some real-world measurements so I could decide if it's worth bothering about the impedance switching and might be useful to others.

Using the RTA built into REW I compared the distortion results from the 1kHz sine wave in both normal mode (as shown by the Hi-Fi Status app) and aux mode (tricked by first plugging in an extension cord then plugging the headphones into that). All tests were done using Neutron as the player (outputting at 32bits and native frequency), which remaps Android's 16-step volume scale to a proper 0-100 range rather than the 0-75 used natively, so I report the volume setting as 0 for maximum, then -1, -2, -3, etc for each step down below that. THD+N measures weren't very helpful, with the results dominated by the noise floor (mostly external environmental noise). But it's clear that full volume in aux mode drives the output amp into some nasty clipping, though below that it seems fine.


Here's the damage at full volume in aux mode into 16Ω:


Looking at the THD numbers alone is a bit more useful in this case:


This is distortion from the full playback and recording chain (phone output+headphones+recording circuitry on my motherboard). So while max volume in aux mode is unusable, stepping the volume down 1 step yields a distortion of only 0.5%, which is quite acceptable.

The real question is whether it's worth bothering with aux mode on low impedance headphones at all. At max volume in normal mode I was getting the same volume output from the headphones as when I set the volume to -3 in aux mode. So basically aux mode only gives you an extra 2 steps of usable volume, which comes out at 5.7dB ([aux mode vol -1] - [normal mode vol 0]).

Summary
Aux mode isn't worth bothering about, unless you absolutely have to squeeze the maximum volume out of the phone. If you really need more volume into low impedances it's probably best to go with an external amp that can supply the extra current. I suspect that reports of it sounding better in aux mode are mostly placebo effect.

Interesting post there @charleski

Some thoughts... It might be better to look at absolute numbers, rather than percentages, as the rise in signal amplitude could mask a constant level of noise or harmonic distortion. I've measured these devices before (V30 and V40 at least, which share the same DAC hardware) and haven't seen a difference in THD+N levels, but I was always matching output levels. If you push the output power higher in aux mode, that might not really be a sensible comparison anymore if normal device mode can't output those levels to being with. Regardless of output mode, THD+N of everything (DAC/amp and headphones) can be expected to rise with increasing output power, But at the same output levels, there's basically no difference between the three modes (look for the tab on the V30 and V40 measurements):

http://soundexpert.org/articles/-/blogs/audio-quality-of-high-end-portable-players

You're including the effect of your headphones here, which might(?) be a valid test for you and your headphones, but you've also then got the effect of your sound card and recording device. Your coupler microphone will be contributing to the distortions, so I think it would be difficult to conclude that this is really what your headphones experience. Even if it was, it wouldn't necessarily imply that others (with harder-to-drive headphones) would experience the same increase in distortion with aux mode at max volume.

 

[charleski]

Interesting post there @charleski

Some thoughts... It might be better to look at absolute numbers, rather than percentages, as the rise in signal amplitude could mask a constant level of noise or harmonic distortion. I've measured these devices before (V30 and V40 at least, which share the same DAC hardware) and haven't seen a difference in THD+N levels, but I was always matching output levels.

Yes, obviously my measurements aren't really any good if you’re interested in the actual noise output, I really just included that for completeness. The THD+N increases with decreasing volume because the signal goes down but the external noise remains the same.

But at the same output levels, there's basically no difference between the three modes (look for the tab on the V30 and V40 measurements):

Yes, I didn't mention it, but also found the THD numbers were basically unchanged between the two modes when the player was producing the same volume. As you can see from the THD graph, the if you shift the normal mode curve 3 steps to the left it's basically the same as found in aux mode.

You're including the effect of your headphones here, which might(?) be a valid test for you and your headphones, but you've also then got the effect of your sound card and recording device. Your coupler microphone will be contributing to the distortions, so I think it would be difficult to conclude that this is really what your headphones experience. Even if it was, it wouldn't necessarily imply that others (with harder-to-drive headphones) would experience the same increase in distortion with aux mode at max volume.

Absolutely. This was specifically aimed at looking at how the phone worked with low impedance headphones and plugging in a 300Ω set will present an entirely different picture. While uncorrelated noise from the phone, mic and recording interface won't sum, the correlated harmonic distortion will and the only thing we can say is that the actual level of distortion products being produced by the phone is lower (almost certainly appreciably lower) than presented here. Since everything else remains the same, though, I think it's a valid means of comparing normal mode and aux mode, which was all I was really interested in.

 

[csglinux]

Yes, obviously my measurements aren't really any good if you’re interested in the actual noise output, I really just included that for completeness. The THD+N increases with decreasing volume because the signal goes down but the external noise remains the same.


Yes, I didn't mention it, but also found the THD numbers were basically unchanged between the two modes when the player was producing the same volume. As you can see from the THD graph, the if you shift the normal mode curve 3 steps to the left it's basically the same as found in aux mode.


Absolutely. This was specifically aimed at looking at how the phone worked with low impedance headphones and plugging in a 300Ω set will present an entirely different picture. While uncorrelated noise from the phone, mic and recording interface won't sum, the correlated harmonic distortion will and the only thing we can say is that the actual level of distortion products being produced by the phone is lower (almost certainly appreciably lower) than presented here. Since everything else remains the same, though, I think it's a valid means of comparing normal mode and aux mode, which was all I was really interested in.

It's probably valid for your headphones, since the distortion and self-noise from the mic will typically be lower. Just curious - what was your mic/coupler/pinna setup?

 

[charleski]

It's probably valid for your headphones, since the distortion and self-noise from the mic will typically be lower. Just curious - what was your mic/coupler/pinna setup?

My mic is just an electret measuring mic that came with my AV receiver. I calibrated it by measuring the response from an Etymotic ER4 and it actually did fairly well, with about -5dB at 20Hz and -8dB at 20kHz but pretty much flat across most of the range. The coupler is just a piece of 8mm tubing with the earphone inserted so it sits 25mm from the mic. As I said, very primitive, but it produces results that reasonably match the FR found by those with better equipment and they’re consistent, which is what really matters for me as I use it to check the results of the EQ I set in the player.

Back to the phone, while the datasheet for the ES9218P isn’t available, I found this Japanese DIY audio site where a guy plays around with the chip and does some measurements. If you scroll down and do some automated translation he confirms a 2VRMS output unloaded and shows some output traces when the chip’s loaded with 47ohms. It looks like he’s getting a clean output at 1.4VRMS with clipping seen on pushing the output further. While eyeballing a sine wave is a pretty poor way of estimating distortion the rise in THD seems quite gentle before clipping sets in. So if we assume the sine wave at 1.4VRMS is under 1% THD that would mean the chip can produce just under 42mW into 47ohms and source just under 30mA. The value measured by Amirm into 50ohms is about 12.5dB lower, so it might be worth tricking the gain limiter if your headphones fall into this range.

 

[csglinux]

My mic is just an electret measuring mic that came with my AV receiver. I calibrated it by measuring the response from an Etymotic ER4 and it actually did fairly well, with about -5dB at 20Hz and -8dB at 20kHz but pretty much flat across most of the range. The coupler is just a piece of 8mm tubing with the earphone inserted so it sits 25mm from the mic. As I said, very primitive, but it produces results that reasonably match the FR found by those with better equipment and they’re consistent, which is what really matters for me as I use it to check the results of the EQ I set in the player.

Back to the phone, while the datasheet for the ES9218P isn’t available, I found this Japanese DIY audio site where a guy plays around with the chip and does some measurements. If you scroll down and do some automated translation he confirms a 2VRMS output unloaded and shows some output traces when the chip’s loaded with 47ohms. It looks like he’s getting a clean output at 1.4VRMS with clipping seen on pushing the output further. While eyeballing a sine wave is a pretty poor way of estimating distortion the rise in THD seems quite gentle before clipping sets in. So if we assume the sine wave at 1.4VRMS is under 1% THD that would mean the chip can produce just under 42mW into 47ohms and source just under 30mA. The value measured by Amirm into 50ohms is about 12.5dB lower, so it might be worth tricking the gain limiter if your headphones fall into this range.

Interesting article, made even more interesting by Google translate Thanks for the link.

I'm just going to play Devil's advocate here, because, well, that's what I do on headfi I don't doubt there's a scenario at some output power level into some load where the V30 will start to clip, but I think it's premature to blame aux mode. After all, one can play aux mode at an output level <= that of normal device mode. I think the caution should simply be: don't over-drive your headphones or your ears. The example our Japanese friend uses seems to be theoretical max output power (1.4 Vrms into a 47 Ohm load), which you'd only get from maximum volume in high-impedance mode.

I'm not taking anything away from the tests that you ran. They're clearly valid for your scenario (your V30->your headphones->your mic->your soundcard), but we know that dynamic/electret mics are fairly bad for generating their own noise and harmonic distortions (they're fine for measuring FR, but THD+N is a different story), and we also know your headphones will inevitably show an increase in distortion (even if driven from the world's most perfect source) once driver excursion gets too high. IMHO, it would be a better test to measure the V30 at max aux mode volume, isolated from any other hardware.

 

[charleski]

Interesting article, made even more interesting by Google translate Thanks for the link.

I'm just going to play Devil's advocate here, because, well, that's what I do on headfi I don't doubt there's a scenario at some output power level into some load where the V30 will start to clip, but I think it's premature to blame aux mode. After all, one can play aux mode at an output level <= that of normal device mode. I think the caution should simply be: don't over-drive your headphones or your ears. The example our Japanese friend uses seems to be theoretical max output power (1.4 Vrms into a 47 Ohm load), which you'd only get from maximum volume in high-impedance mode.

I'm not taking anything away from the tests that you ran. They're clearly valid for your scenario (your V30->your headphones->your mic->your soundcard), but we know that dynamic/electret mics are fairly bad for generating their own noise and harmonic distortions (they're fine for measuring FR, but THD+N is a different story), and we also know your headphones will inevitably show an increase in distortion (even if driven from the world's most perfect source) once driver excursion gets too high. IMHO, it would be a better test to measure the V30 at max aux mode volume, isolated from any other hardware.

Ok, I think I see what you're worrying about. Electret mics actually have pretty good distortion characteristics, but their dynamic range is quite limited. You're worrying that the increased volume is pushing either the headphone or the microphone into distortion, which shouldn't be ascribed to the phone output.

To test this I hooked up a Topping NX1 headphone amp to the phone. The phone was in aux mode for all these runs and all were recorded using the same input gain on the recording interface (and the same volume set on the external amp where applicable). With the extra amp I was certainly able to push the headphones into distortion at maximum gain, but turning the volume knob down a bit yielded a clean signal. I've installed the latest REW beta which has broken the RTA analyser, so I'll just show the waveforms, which are clear enough for this purpose.

Aux mode vol 0 amped

Aux mode vol -1 amped

Aux mode vol -2 amped


Direct output from phone into MH755
Aux mode vol 0 direct

Asymmetrical distortion at a volume level that was clean when going through the external amp. Clearly this distortion is coming from the phone, which is hitting its current limit when asked to drive too much voltage into a set of low-impedance headphones, but can happily supply far more voltage into a high-impedance line stage.

The results I found will be applicable to any headphones with a similar 16Ω impedance.

Having said all this, I'm quite happy with the phone and have no problems using it with my MH755s in Normal Mode, usually running with the volume 3 to 4 steps below maximum. I just wanted to find out what the actual limits were. With the MH755's 100.5dB/mW sensitivity the phone has enough headroom to satisfy my needs.

 

[csglinux]

To test this I hooked up a Topping NX1 headphone amp to the phone. The phone was in aux mode for all these runs and all were recorded using the same input gain on the recording interface (and the same volume set on the external amp where applicable). With the extra amp I was certainly able to push the headphones into distortion at maximum gain, but turning the volume knob down a bit yielded a clean signal. I've installed the latest REW beta which has broken the RTA analyser, so I'll just show the waveforms, which are clear enough for this purpose.

Aux mode vol 0 amped

Aux mode vol -1 amped

Aux mode vol -2 amped


Direct output from phone into MH755
Aux mode vol 0 direct

Asymmetrical distortion at a volume level that was clean when going through the external amp. Clearly this distortion is coming from the phone, which is hitting its current limit when asked to drive too much voltage into a set of low-impedance headphones, but can happily supply far more voltage into a high-impedance line stage.

The results I found will be applicable to any headphones with a similar 16Ω impedance.

Having said all this, I'm quite happy with the phone and have no problems using it with my MH755s in Normal Mode, usually running with the volume 3 to 4 steps below maximum. I just wanted to find out what the actual limits were. With the MH755's 100.5dB/mW sensitivity the phone has enough headroom to satisfy my needs.

I see now, thanks. Yes, it does looks like it's a bit unhappy going into 16 Ohms at aux/max volume. Though in fairness: 1) I suspect that's not an aux issue per se - I'm pretty sure that would apply to normal-device-mode too if only its output V_rms went that high, and 2) most of the time you wouldn't need to go to aux mode or push the volume very high in normal-device mode with a 16 Ohm headphone. I'm not saying it wouldn't be relevant for some users (planar headphone owners who like to push the volume up?), but for most people I can't see this being an issue.

BTW, somebody on another thread recently linked to the site below. It has measurements for an LG G7, but that has the same DAC hardware as the V30. You might find it interesting. You can select different modes (high impedance or normal device - doesn't look like aux was included) and put different loads on it to see the effect:

https://reference-audio-analyzer.pro/en/title-report.php?id=1892

I don't understand what their threshold is, but they're apparently looking for some cut-off where the distortion pushes the device out of Class A classification. The slope shows the cut-off V_rms lowers as the load impedance drops.

 

[csglinux]

A quick follow on... I ran a 500 Hz sine wave via high-impedance mode at max volume through a 32 Ohm resistive load into REW's scope. No problem at all:




I then used my SE846 as the load. Now we get something interesting:



But curiously, I see no problem at all, even with the SE846 as load, until the very last volume step. (Second-to-last still looks perfect.) @charleski - at what volume step (in aux mode) do you first see any obvious distortion or clipping? I'm a bit surprised that the MH755 would show issues like this, because its impedance curve is pretty flat. I'll try to test with some other headphones and resistive loads when I get time.

 

[jagwap]

A quick follow on... I ran a 500 Hz sine wave via high-impedance mode at max volume through a 32 Ohm resistive load into REW's scope. No problem at all:




I then used my SE846 as the load. Now we get something interesting:



But curiously, I see no problem at all, even with the SE846 as load, until the very last volume step. (Second-to-last still looks perfect.) @charleski - at what volume step (in aux mode) do you first see any obvious distortion or clipping? I'm a bit surprised that the MH755 would show issues like this, because its impedance curve is pretty flat. I'll try to test with some other headphones and resistive loads when I get time.

It is likely the SE846 is less than 32 ohms at 500Hz, or has large phase angle there. The negative voltage rail is usually created by a charge pump, while the positive comes from the battery. This means that it is nearly always the negative voltage that clips first. So the result further up the thread is odd.

 

[csglinux]

It is likely the SE846 is less than 32 ohms at 500Hz, or has large phase angle there. The negative voltage rail is usually created by a charge pump, while the positive comes from the battery. This means that it is nearly always the negative voltage that clips first. So the result further up the thread is odd.

Correct - the SE846 is around 8 Ohm @ 500 Hz.