E-MU Teak Review (Headphone)

[Robbo99999]

Oups

(Useful post/conversation )

 

[PeteL]

What do you mean by better and what would come in 2nd for you ?

Hard to say, but it's important, to me, with my various use case that a headphone sounds good without EQ and has you can see in the frequency response these rolls off quite low, around forty HZ or so, and they manage to give you lots of bass while not sounding Boomy to me, because the response in the bass, despite being elevated is relatively flat, where most bassy headphone have a .bump around 100Hz, maybe. In the end tough it's a subjective assesment, they sound better to me in the bass region at least than the headphones I've own, which are: HD-6xx, He400i, MSR7B, Focal Elex. I did listen to most of the usual suspect as well at trade shows, but of course not all of them and sometimes not ideal listening conditions, so yeah to me they are the one to make me happy with the bass response, but some opinion may differ, there are no absolutes, and preference matters.

 

[PeteL]

Yes I am sure. The Qudelix 5K is equally as excellent as a wired DAC/amp as it is wireless. The DSP of the IL-DSP only functions with PCM sampling rates at 96 kHz and below (not DSD) anyway, so that's functionally the same as the Qudelix when using PEQ (the whole point of these devices). The IL-DSP has a max power of just 30 mW (max voltage 0.5 Vrms, less than most laptops/cellphones) into 32 ohms whereas the Qudelix provides 85 mW (max 2 Vrms, single-ended 3.5 mm), and 210 mW (max 4 Vrms, balanced 2.5 mm) into 50 ohms. The Qudelix comes with a fantastic, intuitive app with a huge array of customization options in addition to the 20 custom PEQ profiles savable (e.g. choosing among 8 DAC aliasing filters). The IL-DSP in contrast can only save one PEQ profile at a time, only via a PC plugin, and even then the PEQ implementation is faulty, sometimes resulting in digital clipping even with an EQ curve that does not go above 0 dB. There's simply no contest, the Qudelix 5K is by far the better device, without even considering the optional bonus of its (LDAC) Bluetooth capabilities.

Are you certain that Qudelix DSP processing is at 96KHZ? I know that the Kalimba framework was at 48kHz on the CSR8675, but this use a newer SOC and Qualcomm don't freely broadcast their datasheets without NDA. I don't doubt that Qudelix has a lot to offer, plus the power to drive headphones is no match, and I honestly haven't experienced neither. I do however have experience with both Qualcomm CSR Architechture and the XMOS environment, and if nothing else, purely as a USB receiver, In the CSR SOCs USB is a side tought, It's not up to the performance of any modern DAC in the USB side of thing, Where XMOS is the reference in receiving USB audio Stream. The DSP flexibility is no match neither, but it does require a bit more work on XMOS, so of course if their implementation is broken, if the user experience is poor, that's all points for Qudelix. Now we would need to see measurments for the MiniDSP of course, but you can see right away in the measurments that as a USB DAC The Qudelix it's not anywhere near todays standard, I doubt many people use it that way anyway, and Bluetooth can sound very good, and me personally I would rather have the BT amp, but that's why I said "It depends" IF DSD is important to you, if you are just looking for a USB DAC, maybe the choice is not as obvious and one sided, that's all I meant.
Edit: from Amir measurments, it seams to only support the 96k sample rate, looks like qualcomm doubled their operating sample rate, it use to be only 48k, so a 44.1 file would need to be resampled each time. It's not Ideal of course, but it don't mean it cannot sound good.

 

[IVX]

Not sure what the SINAD of the powerdac 2.1 is, perhaps @IVX or @staticV3 can comment. I have the v2 and it's good kit but only for balanced headphones. The v2.1 can do single ended too with adapter.

PowerDAC is not about spectacular numbers at all, and it has THD level like Dragonfly Cobalt i.e. marginal vs 9038S/D. PDv2 is the USB audio bridge + digitally generated PWM which is filtered by passive LC filters. No opamps, no feedback pure-digital approach, THD and noise performances are sacrificed.

 

[Jimbob54]

PowerDAC is not about spectacular numbers at all, and it has THD level like Dragonfly Cobalt i.e. marginal vs 9038S/D. PDv2 is the USB audio bridge + digitally generated PWM which is filtered by passive LC filters. No opamps, no feedback pure-digital approach, THD and noise performances are sacrificed.

Yes, I thought similar hence not wanting to say it has great metrics, whereas the D/S dongles are the measurement champs but different beasts.

 

[GaryH]

Are you certain that Qudelix DSP processing is at 96KHZ?

I was talking about the input file max sampling rate supported when using DSP, which is the same (96 kHz) for both devices.

Edit: from Amir measurments, it seams to only support the 96k sample rate, looks like qualcomm doubled their operating sample rate, it use to be only 48k, so a 44.1 file would need to be resampled each time. It's not Ideal of course, but it don't mean it cannot sound good.

The initial measurements were borked. Here are the corrected measurements. The Qudelix works perfectly with all supported sample rates when correctly set to autodetect 44.1/48/88.2/96 kHz input in the app settings. It's clearly stated in the IL-DSP's specs that its internal DSP rate is 48 kHz and will resample anything but that, so your misgivings about the Qudelix being worse there are unfounded. The Qudelix's 87 dB SINAD is beyond audible transparency - that being "not anywhere near today's standard" is not only hyperbolic, but irrelevant to the user's aural experience of the product. Oh and I (and many others) use it almost exclusively as a USB DAC/amp. The Bluetooth is just a nice bonus for me that I only use very occasionaly.

 

[nyxnyxnyx]

Fantastic review and overall conclusion Amir.
Now since you have already measured E-MU Teak - which is considered one of the best in the whole "fostex/emu/denon" family, are there any chances you will be able to measure those Denon headphones?

 

[phoenixsong]

Fantastic review and overall conclusion Amir.
Now since you have already measured E-MU Teak - which is considered one of the best in the whole "fostex/emu/denon" family, are there any chances you will be able to measure those Denon headphones?

"Those" Denons are incredibly hard to measure now as due to their age, many of their pads are no longer in fair judging conditions. There is a new line of Denon headphones replacing the old one though- the D5/7/9200 series

 

[phoenixsong]

Measurement's report Fostex TH610 Natural pressure test and graphs - Reference Audio Analyzer (reference-audio-analyzer.pro)
Measurement's report Denon AH-D5200 test and graphs - Reference Audio Analyzer (reference-audio-analyzer.pro)

Measurement's report Denon AH-D7200 test and graphs - Reference Audio Analyzer (reference-audio-analyzer.pro)

Measurement's report Fostex TH-X00 test and graphs - Reference Audio Analyzer (reference-audio-analyzer.pro)
Measurement's report Denon AH-D9200 test and graphs - Reference Audio Analyzer (reference-audio-analyzer.pro)

Lack of distortion measurements on this site has always saddened me though

 

[PeteL]

I was talking about the input file max sampling rate supported when using DSP, which is the same (96 kHz) for both devices.


The initial measurements were borked. Here are the corrected measurements. The Qudelix works perfectly with all supported sample rates when correctly set to autodetect 44.1/48/88.2/96 kHz input in the app settings. It's clearly stated in the IL-DSP's specs that its internal DSP rate is 48 kHz and will resample anything but that, so your misgivings about the Qudelix being worse there are unfounded. The Qudelix's 87 dB SINAD is beyond audible transparency - that being "not anywhere near today's standard" is not only hyperbolic, but irrelevant to the user's aural experience of the product. Oh and I (and many others) use it almost exclusively as a USB DAC/amp. The Bluetooth is just a nice bonus for me that I only use very occasionaly.

OK but I was not talking about supported sample rate, Yes we sure hope that they are all supported, but it will resample everything, not only for DSP, for playback too, well everything, once it's in it's 48kHz. That's just how Qualcom SOC works, I did not comment on the relevance in aural experience, I've never listened to it, I am talking about measured performance, which is mainly what this site is about, As a matter of fact I like to never comment on the audibility, to me it's too arbitrary. If you want to start a debate on if a 110 SINAD dB DAC is not audibly better than a 87 dB SINAD DAC, I'm not your guy for that, I don't know, if you say so, cool. But If mentioning that it's not competitive in today's standard is irrelevant, then pretty much all Amir's do in DAC measurements is irrelevant, this website is irrelevant, because pretty much every DACs today are above that treshold, why even measure them?

 

[dsnyder0cnn]

Thanks for another great review, @amirm . I have a slightly newer version of the E-MU Teak with non-detachable cable (which many folks have said is a bit better due to design issues with the previous connections). Paired with the THX AAA 789 amp and Khadas Tone2 Pro DAC, I could not be happier. The new iFi Audio ZEN Stream really brings it all together as a fairly stunning Roon headphones rig without spending a mint.

 

[nyxnyxnyx]

"Those" Denons are incredibly hard to measure now as due to their age, many of their pads are no longer in fair judging conditions. There is a new line of Denon headphones replacing the old one though- the D5/7/9200 series

Yeah I was thinking about the D*200 series, they seem to have great build, comfort, the drivers as far as I know are made by them as well.
Would be interesting to see how good those fostex variants/biodynamic headphones can score on an objective test!

 

[phoenixsong]

Yeah I was thinking about the D*200 series, they seem to have great build, comfort, the drivers as far as I know are made by them as well.
Would be interesting to see how good those fostex variants/biodynamic headphones can score on an objective test!

InnerFidelity Headphone Measurements | Stereophile.com
If it helps, they have both the old Denon, Emu Teak and Fostex TH610 measurements here for a cross-comparison. Might be easier to use the Emu Teak as a base reference since there are already measurements here that you can utilise

 

[nyxnyxnyx]

InnerFidelity Headphone Measurements | Stereophile.com
If it helps, they have both the old Denon, Emu Teak and Fostex TH610 measurements here for a cross-comparison. Might be easier to use the Emu Teak as a base reference since there are already measurements here that you can utilise

Thanks for the URLs. I did check the content but it's not truly the way I want like how Amir does his measurements (although, their content are still very helpful nonetheless).
I'm interested in those older denon/fostex but my main curiosity remains on D5200/D7200/D9200 models. Hope that at one point someone'll send Amir a sample!

 

[GaryH]

OK but I was not talking about supported sample rate, Yes we sure hope that they are all supported, but it will resample everything, not only for DSP, for playback too, well everything, once it's in it's 48kHz.

You have provided zero evidence the Qudelix 5K does this. Even if it does however, as I said in my previous post, miniDSP clearly states that their IL-DSP's internal DSP rate is 48 kHz and so will resample to this rate, so the Qudelix would be no worse than the IL-DSP in this regard with PEQ active, which is the whole use case and point of these devices. I doubt anyone will be spending $100 on the (underpowered) IL-DSP to not actually use it with DSP engaged.

I did not comment on the relevance in aural experience, I've never listened to it, I am talking about measured performance, which is mainly what this site is about

No, this site is about characterizing and measuring objective parameters of audio devices that correlate with audible fidelity. One of the biggest criticisms of this site from outsiders is that some members are obsessed with figures that go well beyond audibility and so have no relevance to the consumer. Don't be a part of the problem and add fuel to their fire.

But If mentioning that it's not competitive in today's standard is irrelevant, then pretty much all Amir's do in DAC measurements is irrelevant, this website is irrelevant, because pretty much every DACs today are above that treshold, why even measure them?

The measurements are relevant and useful in telling us whether a DAC/amp has any functional issues and is transparent within audibility. Anything beyond that is irrelevant to the end user experience. It would be like obsessing over the measured but invisible UV output of a TV screen - irrelevant and nonsensical.

 

[PeteL]

You have provided zero evidence the Qudelix 5K does this. Even if it does however, as I said in my previous post, miniDSP clearly states that their IL-DSP's internal DSP rate is 48 kHz and so will resample to this rate, so the Qudelix would be no worse than the IL-DSP in this regard with PEQ active, which is the whole use case and point of these devices. I doubt anyone will be spending $100 on the (underpowered) IL-DSP to not actually use it with DSP engaged.



No, this site is about characterizing and measuring objective parameters of audio devices that correlate with audible fidelity. One of the biggest criticisms of this site from outsiders is that some members are obsessed with figures that go well beyond audibility and so have no relevance to the consumer. Don't be a part of the problem and add fuel to their fire.



The measurements are relevant and useful in telling us whether a DAC/amp has any functional issues and is transparent within audibility. Anything beyond that is irrelevant to the end user experience. It would be like obsessing over the measured but invisible UV output of a TV screen - irrelevant and nonsensical.

I feel you are taking this personally, I already said that out of the 2 I would probably choose the Qudelix and that I am sure it sounds good. I am not sure what more evidence you need than the fact that I have worked with both platform and knows how it works, I know the architecture, know the difference between a USB audio class one and UAC 2 environment, and know the limitations of each. You believe that anything below 87 dB SINAD is transparent, Again, as I said, me I believe you. Amir don't think that it's beyond treshold of proven transparency,, many in this community think that it's not irrelevant to target better performance than that, It's irrelevant to you, and it's OK, I'm just saying that it's not competitive in term of measured performance as a USB DAC. I'm not obsessing about this, I state the obvious, the numbers, and not only on SINAD.

 

[GaryH]

I feel you are taking this personally, I already said that out of the 2 I would probably choose the Qudelix and that I am sure it sounds good. I am not sure what more evidence you need than the fact that I have worked with both platform and knows how it works, I know the architecture

Haha I'm not taking anything personally at all. Just putting some things straight. You yourself said right here:

I know that the Kalimba framework was at 48kHz on the CSR8675, but this use a newer SOC and Qualcomm don't freely broadcast their datasheets without NDA.

So no, you do not know this specific architecture, and you do not know the above 48 kHz internal rate holds true for the QCC5124 in the Qudelix 5K, which has an embedded "dual-core 32-bit KALIMBA DSP" as Qudelix state, in contrast with the single-core 24-bit KALIMBA in the CSR8675.

You believe that anything below 87 dB SINAD is transparent, Again, as I said, me I believe you. Amir don't think that it's beyond treshold of proven transparency,, many in this community think that it's not irrelevant to target better performance than that, It's irrelevant to you, and it's OK, I'm just saying that it's not competitive in term of measured performance as a USB DAC. I'm not obsessing about this, I state the obvious, the numbers, and not only on SINAD.

When the lowest THD ever measured in a headphone is -80 dB, it's obvious that a device such as the Qudelix 5K with a SINAD of 87 dB will not add an audible amount of distortion to consumers' headphones. If this is not obvious, then the bottom of this web page should make this clear. And this is not even considering hearing thresholds and the psychoacoustics of masking which will make this audible transparency threshold even more lenient. I'm just stating the obvious facts given by the numbers, based purely on the mathematics and physics.

 

[PeteL]

Haha I'm not taking anything personally at all. Just putting some things straight. You yourself said right here:


So no, you do not know this specific architecture, and you do not know the above 48 kHz internal rate holds true for the QCC5124 in the Qudelix 5K, which has an embedded "dual-core 32-bit KALIMBA DSP" as Qudelix state, in contrast with the single-core 24-bit KALIMBA in the CSR8675.



When the lowest THD ever measured in a headphone is -80 dB, it's obvious that a device such as the Qudelix 5K with a SINAD of 87 dB will not add an audible amount of distortion to consumers' headphones. If this is not obvious, then the bottom of this web page should make this clear. And this is not even considering hearing thresholds and the psychoacoustics of masking which will make this audible transparency threshold even more lenient. I'm just stating the obvious facts given by the numbers, based purely on the mathematics and physics.

You know that Sinad is a best case scenario right? meaning listening close to 0 DBFS, or at a 2V unbal 4V Bal. the link you send me seems to suggest you did not take this in consideration, you have to know the real SINAD at your listening volume to be able to say that "It will not add an audible amount of distortion" or noise. With sensitive earphones it is not rare to use quite large amount of digital attenuation.

 

[GaryH]

You know that Sinad is a best case scenario right? meaning listening close to 0 DBFS, or at a 2V unbal 4V Bal. the link you send me seems to suggest you did not take this in consideration, you have to know the real SINAD at your listening volume to be able to say that "It will not add an audible amount of distortion" or noise. With sensitive earphones it is not rare to use quite large amount of digital attenuation.

You know that -80 dB headphone THD is a very best case scenario right? The vast majority of headphones will have THD much higher than this, typically -50 to -60 dB, even for high quality headphones. This is significantly higher than the Qudelix's highest distortion (before clipping) at low but still audible volumes, so it will not add anything audible over the headphone's distortion even at low levels. Then there's the aforementioned distortion hearing thresholds, masking and psychoacoustics that means this is even more of a non-issue. You're worrying about figures that will have no audible consequence in practice. I suggest you download the excellent Distort software and experiment with blind listening of different levels of THD for yourself. I think you'll be surprised at how high the threshold of distortion audibility is in practice when listening to music.

 

[amirm]

The vast majority of headphones will have THD much higher than this, typically -50 to -60 dB, even for high quality headphones.

SINAD has both THD and noise. Speakers/headphones have no self-noise so they are much better than any electronics there. On the distortion-front, the measurements represent the sum of microphone and headphone and no way to separate the two. In addition, it takes a very special setup to measure headphones at very low levels as noise pollutes the measurements. Net, net, headphones can be incredibly good in this regard.