Speaka USB DAC vs Samsung Galaxy S10

[bobbooo]

Hi, I'm not great at interpreting the audibility of measurements. Would the Speaka / XtremPro X1 USB DAC be audibly better or worse than Samsung's (Snapdragon) Galaxy S10?

Here are the Speaka's measurements:
https://www.audiosciencereview.com/...-speaka-usb-dac-and-headphone-amplifier.2423/ (along with THD on this page: https://www.audiosciencereview.com/...a-usb-dac-and-headphone-amplifier.2423/page-4)

And here are some measurements of the S10 by Soomal:
https://translate.google.com/transl...n&u=http://www.soomal.com/doc/10100008405.htm
(The second set of meaurements - the first are for LDAC playback. And I'm mainly concerned with 44.1kHz and 48kHz sample rates.)
And by Anandtech:
https://images.anandtech.com/doci/14072/SnapdragonAudio.png

 

[pozz]

There's not enough information in the S10 measurements. At minimum you would also want to know its output impedance and output power for easy vs. difficult to drive headphones.

 

[bobbooo]

There's not enough information in the S10 measurements. At minimum you would also want to know its output impedance and output power for easy vs. difficult to drive headphones.

I believe the output impedance is around 2 ohms, with a max voltage output of 1Vrms. And here's the codec it uses: https://www.qualcomm.com/products/wcd9341

 

[pozz]

There wouldn't be much difference, unfortunately. The Speaka is marginally better, but I wouldn't go out of my way to get it.

 

[bobbooo]

There wouldn't be much difference, unfortunately. The Speaka is marginally better, but I wouldn't go out of my way to get it.

Ok thanks. How about E1DA's PowerDAC V2 (https://e1dashz.wixsite.com/index/pdv2), would that offer any audible improvement over either of those (discounting its EQ options)? Or are the S10 and Speaka already transparent within the limits of audibility?

 

[pozz]

The E1DA 9038S measured really well, so it's likely that the PowerDAC V2 would be similarly well-designed.

The main thing that would make an audible difference is low output impedance and suitable power delivery. Unless you're running 300 ohm headphones it would be fine. Just keep in mind its output jack is the 2.5mm kind.

I assume you're looking for a portable DAC? I don't really follow that area of audio but Topping NX4 DSD measured really well before. Others would be able to give better advice.

 

[bobbooo]

The E1DA 9038S measured really well, so it's likely that the PowerDAC V2 would be similarly well-designed.

The main thing that would make an audible difference is low output impedance and suitable power delivery. Unless you're running 300 ohm headphones it would be fine. Just keep in mind its output jack is the 2.5mm kind.

I assume you're looking for a portable DAC? I don't really follow that area of audio but Topping NX4 DSD measured really well before. Others would be able to give better advice.

Thanks. Measurements for the PowerDAC V2 are towards the bottom of that page I linked to (I trust they are accurate as E1DA's measurements for the 9038S are in accordance with the ones on here). I have the Hifiman HE4XX, which are 49 ohms but with an efficiency of just 88.2 dB/mW so are relatively hard to drive (measurements by Innerfidelity). I'm not necessarily looking for a portable DAC, although that would be a nice bonus as I could then also use it with my IEMs while out. Really I'm just looking for the cheapest, completely transparent within audibility DAC I can get, that will power my HE4XX, or failing that, a transparent stack of a DAC and amp (e.g. the JDS Labs Atom?). This all assumes that the Samsung S10 / Speaka are not completely transparent (are they?). If they are, then I'd only need an amp to increase power output. I'd appreciate any advice.

 

[pozz]

Just FYI, I had looked at the measurements E1DA provided, but they aren't as comprehensive as Amir's.

I use the Atom, and it's pretty much an endgame headphone amp until I can afford the Benchmark HPA4. I think if you get it to supplement the Samsung you're sitting near the top of what you can reasonably expect.

Many on ASR have recommended the Neutron app for bitperfect playback (I don't have experience with it but you can look up other threads), so try that out.

If you really want to get into technical transparency you'll end up paying more money. Under most circumstances you won't notice the difference. Maybe at some point consider getting a dedicated DAC for your desktop setup when funds are a bit freer to experiment.

 

[bobbooo]

Just FYI, I had looked at the measurements E1DA provided, but they aren't as comprehensive as Amir's.

I use the Atom, and it's pretty much an endgame headphone amp until I can afford the Benchmark HPA4. I think if you get it to supplement the Samsung you're sitting near the top of what you can reasonably expect.

Many on ASR have recommended the Neutron app for bitperfect playback (I don't have experience with it but you can look up other threads), so try that out.

If you really want to get into technical transparency you'll end up paying more money. Under most circumstances you won't notice the difference. Maybe at some point consider getting a dedicated DAC for your desktop setup when funds are a bit freer to experiment.

I'm using Neutron on my S10 at the moment, amazing app. It not only has bitperfect playback, but also a max 30-band parametric EQ with oversampling, dither, noise-shaping etc. which is great for EQing the HE4XX to the Harman target using the AutoEQ tool on GutHub. Actually, one of the things that's tempting me to get the PowerDAC V2 is that it has a parametric hardware EQ built in – it's only 7-band, but that should just be enough, and it will allow me to EQ to the Harman curve outside of Neutron e.g. while using streaming services like Spotify. You control the EQ via an app over Bluetooth, all pretty amazing for the $50 asking price really. The only problem is, the HE4XX needs a 13.5dB bass boost to match Harman, which means I would need to run the PowerDAC at -13.5dB volume to avoid clipping, but at that level I’m not sure playback would be loud enough through the fairly inefficient HE4XXs…

 

[bobbooo]

There's not enough information in the S10 measurements. At minimum you would also want to know its output impedance and output power for easy vs. difficult to drive headphones.

Hi, I have just one more question on this. Looking at Anandtech's measurements of the S10's audio output (https://images.anandtech.com/doci/14072/SnapdragonAudio.png), it says "N: 0.17%" and "THD+N: 0.17%" at the top-left, but the graph doesn't seem to reflect this, showing a noise floor of about -130dB, and the loudest harmonic at -120dB, both below audibility according to this page: https://www.audiosciencereview.com/...-thresholds-of-amp-and-dac-measurements.5734/, which says -120dB noise/THD is the equivalent of 0.0001%. Am I completely misunderstanding something?

 

[pozz]

A few things about the Anandtech measurement:

• It's set to -10.2dBV or about 0.3 volts output, which is very low. Peak signal of the measurement is -26dBFS. All of the distortion numbers would rise once the signal is increased to 0dBFS (maximum level before clipping). The very minimal contribution of THD to THD+N is due to that.
• 0.17% is -55dB. This is a very high and unlikely number. Notice that the bandwidth is 91.2kHz (upper left-hand corner), so the measuring system's picking up ultrasonic hash beyond the standard 20kHz from the S10's circuitry (see the upwards climb of the graph on the right-hand side around 50kHz). It's also unclear (at least to me) what measurement system the reviewer is using, and what its noise and distortion contribution would be (you can't assume the measuring gear is perfect).
• The FFT gain is set to 512k points, which is around 54dB of (visual) noise reduction for the noise floor. This was done to reveal the distortion products buried under the noise floor.
• Despite that, the dynamic range measurement is dBA-weighted, rolling off both the low and high ends and inflating the figure.

Given all these things it's not reliable.

To compare measurements you have to understand the boundary conditions, otherwise the figures are more or less useless.

Check out https://www.rane.com/note145.html

 

[bobbooo]

A few things about the Anandtech measurement:

• It's set to -10.2dBV or about 0.3 volts output, which is very low. Peak signal of the measurement is -26dBFS. All of the distortion numbers would rise once the signal is increased to 0dBFS (maximum level before clipping). The very minimal contribution of THD to THD+N is due to that.
• 0.17% is -55dB. This is a very high and unlikely number. Notice that the bandwidth is 91.2kHz (upper left-hand corner), so the measuring system's picking up ultrasonic hash beyond the standard 20kHz from the S10's circuitry (see the upwards climb of the graph on the right-hand side around 50kHz). It's also unclear (at least to me) what measurement system the reviewer is using, and what its noise and distortion contribution would be (you can't assume the measuring gear is perfect).
• The FFT gain is set to 512k points, which is around 54dB of (visual) noise reduction for the noise floor. This was done to reveal the distortion products buried under the noise floor.
• Despite that, the dynamic range measurement is dBA-weighted, rolling off both the low and high ends and inflating the figure.

Given all these things it's not reliable.

To compare measurements you have to understand the boundary conditions, otherwise the figures are more or less useless.

Check out https://www.rane.com/note145.html

Thanks for the detailed answer. So does this mean the true noise floor is actually -130dB (graph) + 54dB = -76dB ? This seems very high. Soomal's measurements (see link in my original post) report a noise floor of -95.7dB and THD of 0.0007%, and their THD graph looks to be in general agreement with Anandtech's (considering it's at -3dB output), see here: https://translate.googleusercontent...700271&usg=ALkJrhjunNl-Etzq0Ziw9ULs4vdhRuHe3Q

It seems strange they would both be incorrect by about the same amount, or am I misinterpreting again?

 

[pozz]

Soomal's noise measurement is A-weighted: https://en.m.wikipedia.org/wiki/A-weighting

I can't say for sure unless another measurement is made. Safe to say, if you can't hear noise with headphones or IEMs, it's probably not that high.

 

[bobbooo]

Soomal's noise measurement is A-weighted: https://en.m.wikipedia.org/wiki/A-weighting

I can't say for sure unless another measurement is made. Safe to say, if you can't hear noise with headphones or IEMs, it's probably not that high.

Is Soomal's THD graph also A-weighted? I just want to get a rough idea from those and Anandtech's measurements, what the numbers would be if the S10 was measured with Amir's method, to compare with his other measurements for DACs/amps.

 

[pozz]

No. If the graph had A-weighting applied it would look like this: https://www.audiosciencereview.com/...rn-pluto-and-art-djpre-ii-phono-preamps.3457/

 

[pozz]

I just want to get a rough idea from those and Anandtech's measurements, what the numbers would be if the S10 was measured with Amir's method, to compare with his other measurements for DACs/amps.

Have you seen the analyzer reviews?

• (24 August 2019) QuantAsylum QA401
• (9 June 2018) Audio Precision AP2522 & APx555 & (31 August 2019) APx555 4V Output & (20 July 2019) APx555 5V Output

Most other reviewers won't be able match the level of precision. Plus the settings of an APx555 will affect what shows up, as will the rig for simulated headphone loads.

So there's no procedure to match numbers from disparate setups, but you can get some idea of what to expect once you're clear on the measurement conditions.

The Anandtech and Soomal measurements aren't very useful overall, honestly.

 

[bobbooo]

Have you seen the analyzer reviews?

• (24 August 2019) QuantAsylum QA401
• (9 June 2018) Audio Precision AP2522 & APx555 & (31 August 2019) APx555 4V Output & (20 July 2019) APx555 5V Output

Most other reviewers won't be able match the level of precision. Plus the settings of an APx555 will affect what shows up, as will the rig for simulated headphone loads.

So there's no procedure to match numbers from disparate setups, but you can get some idea of what to expect once you're clear on the measurement conditions.

The Anandtech and Soomal measurements aren't very useful overall, honestly.

That's a shame. Unfortunately, it seems that compared to the multitude of measurements of niche and expensive audio equipment, there's a distinct lack of accurate measurements of the audio quality of smartphones around, which is puzzling, considering a large majority of people listen to their music primarily through these devices. I really think it should be the priority of any pragmatic, scientifically minded audiophile to first confirm whether our most commonly used sources are in fact audibly transparent or not, before moving on to measure more niche/expensive 'audiophile-grade' equipment, instead of just taking it as given that they're not transparent.

 

[bobbooo]

No. If the graph had A-weighting applied it would look like this: https://www.audiosciencereview.com/...rn-pluto-and-art-djpre-ii-phono-preamps.3457/

If that's the case, it seems to me Soomal's graph does then show that the S10's THD+N is below audibility at -3dB, no? (With the 2nd harmonic maybe just audible.)

 

[pozz]

2nd harmonic is at -108dB while 3rd is at -98dB or so. I wouldn't say either is audible if using the S10 as a DAC.

And keep in mind that THD+N contains the amplitude of all harmonics as well as noise, so noise audibility is more of a question given the a-weighting.

Amp power is probably the only significant factor. As long as your headphones aren't hard to drive or current hungry it's unlikely that you'll detect a difference.

 

[bobbooo]

2nd harmonic is at -108dB while 3rd is at -98dB or so. I wouldn't say either is audible if using the S10 as a DAC.

And keep in mind that THD+N contains the amplitude of all harmonics as well as noise, so noise audibility is more of a question given the a-weighting.

Amp power is probably the only significant factor. As long as your headphones aren't hard to drive or current hungry it's unlikely that you'll detect a difference.

Thanks, but I'm confused again...I can see the 2nd harmonic at -108dB, but I don't see a 3rd harmonic at -98dB (or any significant others at all) in Soomal's THD measurement. We're looking at the same graph, right? This one: https://translate.googleusercontent...700271&usg=ALkJrhjunNl-Etzq0Ziw9ULs4vdhRuHe3Q