Review and Measurements of Melokin DA9.1 ES9038PRO DAC and Headphone Amp

[Patrick1958]

Pity, another dac to remove from my wishlist.

 

[JustIntonation]

Thank you so much Amirm!
I have no hope whatsoever that my unit will not have these problems / that it is a production error in your version.
I'll be going for a refund through Ebay.

You can still clearly see the channel imbalances of the DAC through the headphone output.
I'm guessing the headphone amplifier is itself perhaps better than the measurements show but the DAC errors are limiting the THD+N. I see it reaches a low of 0.0017% THD+N (about -97dB THD+N). I'm guessing this is with the volume control not at max? If not and these measurements are made with the volume control at max this means the output is actually more distorted at the RCA outputs than at the headphones out? Which would be even stranger.

 

[amirm]

I'm guessing the headphone amplifier is itself perhaps better than the measurements show but the DAC errors are limiting the THD+N. I see it reaches a low of 0.0017% THD+N (about -97dB THD+N). I'm guessing this is with the volume control not at max?

The THD curves are definitely at max volume. I just double checked.

And yes, something is strange at the line out since it gets distorted depending on the volume control position.

 

[JustIntonation]

The THD curves are definitely at max volume. I just double checked.

And yes, something is strange at the line out since it gets distorted depending on the volume control position.

Oh really at full volume.
Am I right in thinking that this means the headphone output THD+N measurements can probably be a bit better when turning the volume control down a bit as it's an analog volume control (to my knowledge, could be mistaken in this)? The result should be best when the DAC is at full output internally and the headphone output outputting around 1.5 vrms. Don't know at which position the volume dial should be for this.
Not asking you to do anymore measurements though This wouldn't fix the channel imbalances and strange behavior of the line output of course. Which is indeed so strange my technical knowledge is insufficient to imagine how to accomplish by design fault that the headphone volume control affects the distortion of only the line output..
But thanks again for catching this.

 

[amirm]

Am I right in thinking that this means the headphone output THD+N measurements can probably be a bit better when turning the volume control down a bit as it's an analog volume control (to my knowledge, could be mistaken in this)?

Not in that graph. There, I show the performance at lower level signals ramping up to max. Distortions in amps are level sensitive and not the position of the volume control (although there can be a small effect there).

 

[JustIntonation]

Not in that graph. There, I show the performance at lower level signals ramping up to max. Distortions in amps are level sensitive and not the position of the volume control (although there can be a small effect there).

Yes, but what it shows is a cumulative effect of both the DAC and amp distortion+noise. Since with the volume at max there is more amp gain the amp distortion is reached before the DAC has full output resolution so the point of lowest distortion of the amp (~1.5 vrms) does not coincide with the point of lowest distortion of the DAC (-0dBFS I assume).
So the graph should show the amp clipping points (it does) but not the lowest THD+N of the DAC + amp combo, this could be much lower? (could also stay the same if there is almost no THD+N from the DAC chip contributing at 1.5 vrms output from the amp in the graph you posted. Perhaps the 1.5 volt amp output is reached at about -12dBFS if I have to guess? If so this would perhaps not make a big difference.)

 

[JustIntonation]

Btw does anybody know which software to use for testing THD+N?
I'd like to test my Melokin DA9.1 when it arrives just to make sure it has the same problems and I'm not sending back a perfectly good DAC (fully expect to find the same problems though)
I'll be using a USB DAC/ADC which has optical out and feed the optical out to the DA9.1 and the RCA outputs of the DA9.1 back into my ADC. (this way I won't have to use 2 USB audio connections at the same time or use 2 computers)

 

[amirm]

I think rightmark does that.

 

[SIY]

Btw does anybody know which software to use for testing THD+N?

For casual use, ARTA and REW work fine. For more serious work, Virtins MI.

 

[Ron Texas]

In the thread title shouldn't it be 9038 and not 3098?

 

[Blumlein 88]

For JustIntonation, Are you using Mac or Windows or Linux?

But yeah, REW would work on either.

 

[JustIntonation]

For JustIntonation, Are you using Mac or Windows or Linux?

But yeah, REW would work on either.

Using Windows 10.
Already have REW for room measurements, didn't realize it's suited for DAC measurement as well.

Thanks for all the tips! Have plenty of software ready now. Though think I'll just start with a 1kHz sine and look at a real time spectrum analyzer from DA9.1 RCA output with headphone volume at min and then at max. Should be able to see the distortion increase clearly despite my potential bad measurement resolution. If I also see that one of my ADC's is clean enough (have 3 all very old and probably bad, never use them anymore) I'll look into doing more measurements.

 

[amirm]

In the thread title shouldn't it be 9038 and not 3098?

Thanks for catching that. I fixed it.

 

[maxxevv]

Out of curiosity went to the QLS Hifi page to look at the Melokin.

The US$360 price does not include the USB interface. An Amanero module from them costs an additional US$80/-. That makes the base price US$440/- for most of the people out there who will be using USB.

Such a waste that its implementation wasn't that great despite so much promise. Wonder if they used separate single channel opamps (with requisite circuitry ) instead of dual-channel ones, might results look a little different?

Great connectivity options for LAN and I2S though.

 

[JustIntonation]

I've received the DA9.1 and measured it a bit.
It has the same problem of added harmonic distortion from the RCA line outputs when the headphone volume knob is turned to max. But.. this only happens when it's at max volume or very near max volume. At about 4/3 volume the added harmonic distortion is completely gone. So while this is a strange thing, it's not really a problem to me.
It may have something to do with the headphone output and RCA line outputs always being active. Usually when you plug in headphones the line output mutes, this does not happen with the DA9.1

I also measured the THD+N best as I could. Used an E-MU 0404USB for this. Could only get a decent measurement when I used the optical output of the E-MU into the DA9.1 and the analog output of the DA9.1 into the line inputs of the E-MU. When I used the USB input of the DA9.1 and then line out into E-MU line in the measurement was lifted in the middle, probably a timing error? Had to use the E-MU for both input and output in the software. Used VIRTINS Multi-instrument for the measurements.
I measured about -105dB for the 3rd harmonic distortion. The other harmonics were lower. Noise was well below this.
Also measured a Sabaj D4 which I loaned from my little brother for a day. It has about all harmonics at about -100dB and a slightly higher noisefloor than the DA9.1

I didn't figure out how to measure channel imbalances and linearity yet. However I did a subjective test with headphones and mono music using digital volume control down to -90dB and headphone volume on the DA9.1 at max and I could not hear any drifting to the left or right, the music stayed dead center.
I did however hear severe drifting left and right at the very minimum settings of the analog headphone volume control of the DA9.1 however this was only at the very minimum for any normal low volume use the stereo image was stable.

I also found that the Amanero Directsound drivers are not great on my computer (windows 10). At time there were pops and ticks, about once a minute perhaps more often even. Also could see this in the measurement window when I tried out USB through directsound.
However the Amanero ASIO driver is rock solid for me. I see in the review that this driver wasn't used but ASIO4ALL was used instead?
The official drivers can be downloaded here: https://amanero.com/drivers.htm

And now for the subjective part..
I was fully expecting to send back the DA9.1, untill I started listening and also comparing to the Sabaj D4 (AKM chips) and HifiMe 9018D (ESS 9018 mobile chip).
I can put it very simple. To me the DA9.1 is the best DAC I think I have ever heard
It is better sounding than the Sabaj D4 and HifiMe 9018D. The Sabaj D4 has less detail especially in the trebble / with fast transients and it sounds less natural. The Sabaj appears to have more bass, at least mid bass. But it doesn't have a fuller sound overall, the DA9.1 does to me. The DA9.1 also appears to have more sub bass, and more contrast throughout the frequency range. The Sabaj D4 doesn't sound harsh at first to me but it is somewhat harsh compared to the DA9.1 don't know how to describe it. Sabaj sounds dynamic but with perhaps some exaggeration through distortion on the dynamics or something. Really don't know the words to describe the difference in sound..
And the 9018D can sound a bit thin and harsh as well in comparison the the DA9.1. Again I find it hard to describe the differences in sound.
The biggest difference to me is the transients with a lot of high trebble. I'm noticing certain percussive sounds in records I'm very familiar with that I didn't notice before. The contrast is just bigger there.

So I'm actually not sending it back and am happy with it despite the bad measurements here..
Btw, about -105dB for the 3rd harmonic isn't that bad is it? I'm reading it's about -115dB (edit: -125dB) for the OPPO UPD-205? That's only a 10dB (edit: 20dB!) difference?
What I'm also thinking is that perhaps the harmonic distortion and channel imbalances are at low enough levels that they do not degrade the sound quality in any significant way.
Intermodulation distortion and crosstalk are perhaps worse types of distortion and the DA9.1 meaured pretty well in those areas?
And I can't help but talk about what is not being measured. The what I've come to suspect from experience is a fairly audible difference between DAC chips. Perhaps there is a part of the sound of a DAC chip that is not measured in linearity, harmonic distortion, intermodulation distortion and crosstalk etc. I suspect DAC chips do something to transients perhaps.
The DACs I've owned in the past: Lavry DA10, and old Mytek (can't remember the type), Violectric V800, Lehmann (forgot type and not even sure about the brand, expensive and didn't like it), Yulong D200 and many other cheaper DACs. Most of these are old and surely measure bad even though they were expensive in their day. And at least the Violectric should measure well. But the one that stuck in my mind was the Yulong D200. It must have measured quite poor I think it had a weird output stage and you could hear the music was "excited" as in having distortion in a pleasant tube like manner (but strangely enough also rolled off trebble). But still there was something in its sound that was great, it had a lot of detail and contrast in a very musical way. Not "overly dynamic" yet "empty" in a way like most DACs. It has an ESS 9016 chip (full version not mobile). And I hear a cleaner version of the D200 back in the DA9.1

Perhaps it is worthwhile to think of an extra measurement for transients or something in that area? And if not transients then what else could be it? A lot of people seem to think they hear differences between DAC chips. Of course implementation matters a lot and of course the measurements as they're being done here matter a lot. But I can't help but feel something is still being missed. Especially now again when listening to the DA9.1

Oh and btw, one last thing about the DA9.1. When I first plugged it in it had a bit of noise from the headphone out at max volume and no music. Was inaudible on my DT1990 with the volume turned a bit back so no real practical problem yet I still didn't like it. And strangely enough it was completely gone after an hour or so and hasn't returned since (about a day now). Now it's pitch black with the volume turned to max. Still a little bit worrysome.. Hope it doesn't return.

And one last thing. Can really recomend the DT 1990 Pro headphones from Beyerdynamic. They need some EQ though, at minimum to lower the resonant peak at about 8kHz by a min of 6dB. Still experimenting with EQ curves in general to fit my preference. But so far I like the EQ-ed DT1990 more than the HD800 overall (without any EQ the HD800 wins though for me).

 

[Wombat]

I've received the DA9.1 and measured it a bit.
It has the same problem of added harmonic distortion from the RCA line outputs when the headphone volume knob is turned to max. But.. this only happens when it's at max volume or very near max volume. At about 4/3 volume the added harmonic distortion is completely gone. So while this is a strange thing, it's not really a problem to me.
It may have something to do with the headphone output and RCA line outputs always being active. Usually when you plug in headphones the line output mutes, this does not happen with the DA9.1

I also measured the THD+N best as I could. Used an E-MU 0404USB for this. Could only get a decent measurement when I used the optical output of the E-MU into the DA9.1 and the analog output of the DA9.1 into the line inputs of the E-MU. When I used the USB input of the DA9.1 and then line out into E-MU line in the measurement was lifted in the middle, probably a timing error? Had to use the E-MU for both input and output in the software. Used VIRTINS Multi-instrument for the measurements.
I measured about -105dB for the 3rd harmonic distortion. The other harmonics were lower. Noise was well below this.
Also measured a Sabaj D4 which I loaned from my little brother for a day. It has about all harmonics at about -100dB and a slightly higher noisefloor than the DA9.1

I didn't figure out how to measure channel imbalances and linearity yet. However I did a subjective test with headphones and mono music using digital volume control down to -90dB and headphone volume on the DA9.1 at max and I could not hear any drifting to the left or right, the music stayed dead center.
I did however hear severe drifting left and right at the very minimum settings of the analog headphone volume control of the DA9.1 however this was only at the very minimum for any normal low volume use the stereo image was stable.

I also found that the Amanero Directsound drivers are not great on my computer (windows 10). At time there were pops and ticks, about once a minute perhaps more often even. Also could see this in the measurement window when I tried out USB through directsound.
However the Amanero ASIO driver is rock solid for me. I see in the review that this driver wasn't used but ASIO4ALL was used instead?
The official drivers can be downloaded here: https://amanero.com/drivers.htm

And now for the subjective part..
I was fully expecting to send back the DA9.1, untill I started listening and also comparing to the Sabaj D4 (AKM chips) and HifiMe 9018D (ESS 9018 mobile chip).
I can put it very simple. To me the DA9.1 is the best DAC I think I have ever heard
It is better sounding than the Sabaj D4 and HifiMe 9018D. The Sabaj D4 has less detail especially in the trebble / with fast transients and it sounds less natural. The Sabaj appears to have more bass, at least mid bass. But it doesn't have a fuller sound overall, the DA9.1 does to me. The DA9.1 also appears to have more sub bass, and more contrast throughout the frequency range. The Sabaj D4 doesn't sound harsh at first to me but it is somewhat harsh compared to the DA9.1 don't know how to describe it. Sabaj sounds dynamic but with perhaps some exaggeration through distortion on the dynamics or something. Really don't know the words to describe the difference in sound..
And the 9018D can sound a bit thin and harsh as well in comparison the the DA9.1. Again I find it hard to describe the differences in sound.
The biggest difference to me is the transients with a lot of high trebble. I'm noticing certain percussive sounds in records I'm very familiar with that I didn't notice before. The contrast is just bigger there.

So I'm actually not sending it back and am happy with it despite the bad measurements here..
Btw, about -105dB for the 3rd harmonic isn't that bad is it? I'm reading it's about -115dB for the OPPO UPD-205? That's only a 10dB difference?
What I'm also thinking is that perhaps the harmonic distortion and channel imbalances are at low enough levels that they do not degrade the sound quality in any significant way.
Intermodulation distortion and crosstalk are perhaps worse types of distortion and the DA9.1 meaured pretty well in those areas?
And I can't help but talk about what is not being measured. The what I've come to suspect from experience is a fairly audible difference between DAC chips. Perhaps there is a part of the sound of a DAC chip that is not measured in linearity, harmonic distortion, intermodulation distortion and crosstalk etc. I suspect DAC chips do something to transients perhaps.
The DACs I've owned in the past: Lavry DA10, and old Mytek (can't remember the type), Violectric V800, Lehmann (forgot type and not even sure about the brand, expensive and didn't like it), Yulong D200 and many other cheaper DACs. Most of these are old and surely measure bad even though they were expensive in their day. And at least the Violectric should measure well. But the one that stuck in my mind was the Yulong D200. It must have measured quite poor I think it had a weird output stage and you could hear the music was "excited" as in having distortion in a pleasant tube like manner (but strangely enough also rolled off trebble). But still there was something in its sound that was great, it had a lot of detail and contrast in a very musical way. Not "overly dynamic" yet "empty" in a way like most DACs. It has an ESS 9016 chip (full version not mobile). And I hear a cleaner version of the D200 back in the DA9.1

Perhaps it is worthwhile to think of an extra measurement for transients or something in that area? And if not transients then what else could be it? A lot of people seem to think they hear differences between DAC chips. Of course implementation matters a lot and of course the measurements as they're being done here matter a lot. But I can't help but feel something is still being missed. Especially now again when listening to the DA9.1

Oh and btw, one last thing about the DA9.1. When I first plugged it in it had a bit of noise from the headphone out at max volume and no music. Was inaudible on my DT1990 with the volume turned a bit back so no real practical problem yet I still didn't like it. And strangely enough it was completely gone after an hour or so and hasn't returned since (about a day now). Now it's pitch black with the volume turned to max. Still a little bit worrysome.. Hope it doesn't return.

And one last thing. Can really recomend the DT 1990 Pro headphones from Beyerdynamic. They need some EQ though, at minimum to lower the resonant peak at about 8kHz by a min of 6dB. Still experimenting with EQ curves in general to fit my preference. But so far I like the EQ-ed DT1990 more than the HD800 overall (without any EQ the HD800 wins though for me).

PSOOOC.

 

[JustIntonation]

PSOOOC.

?

 

[Wombat]

?

Personal, singular, opinion of course.

 

[JustIntonation]

Btw, I think both my channels measure about the same as Ch1 of the measurements by Amirm. I can only tell from the harmonic distortion and low frequency noise now which are about the same for both my channels. Will perhaps try to do IMD measurement later.
Ch2 of Amirm's measurements has extra low frequency distortion and extra harmonic distortion (especially the higher 2nd harmonic) that's not present in my Ch2.

edit: nevermind. Did some further testing and there is indeed more distortion in one channel than the other. 2nd harmonic is higher than the third in one channel and 6dB or so lower in the other amongst other differences.
Had only one channel of the E-MU 0404USB input to work with the other was too noisy. Used USB ASIO out to the DA9.1 this time with a 1000Hz sine and different program with a good spectrum analyzer to see the results.
I think I'm getting the exact same results as were measured by Amirm with the exception of the low frequency distortion rise in one channel as measured by Amirm.
Oh btw the big distortion increase from the RCA line outs is only there with a 0dBFS signal and headphone volume at max. There is no added distortion with for instance a -3dBFS signal and headphone volume at max. So it isn't really audible under realistic usage.

I also did a lot of back and forth listening at matched levels between the Sabaj D4 and the DA9.1 and to me the DA9.1 sounds more "analog" for lack of better words. More relaxed where the D4 can sometimes be a little bit "edgy" sounding in dynamics.
But.. the D4 has more precise imaging. The DA9.1 is a touch hazy with imaging for me at least with close testing from the headphone out. And this is the final dealbraker for me.
I guess this shows in the measurements by Amirm where the distortion is not equal in left and right channel except at very specific volumes. I couldn't detect any drifting with mono music but this is apparantly audible with music in the imaging. (and pinpoint imaging is one of the reasons I like my DT1990 better than the HD800 which has a wider soundstage but much less precise imaging for me)
So I'm sending the DA9.1 back. Seems like there is no budget ES9038Pro DAC that performs precise enough. Will probably go for a Topping D50 + good headphone amp.

 

[amirm]

So I'm sending the DA9.1 back. Seems like there is no budget ES9038Pro DAC that performs precise enough. Will probably go for a Topping D50 + good headphone amp.

Ah, I thought you were going to keep it. Regardless, thanks for confirming some of my findings.

I think as the DAC chip gets better, it gets even harder to get the best performance out of them since the distortion and noise of the rest of the DAC (box) needs to improve correspondingly. The Oppo UDP-205 pulls this off but I suspect they had direct help from ESS to get there.