Okto dac8 stereo DAC Review

[MadMan]

The DAC doesn't come with its own power cable, would've been nice if this was explicitly mentioned somewhere (prior to ordering).

You're supposed to use your own cable that costs at least 2x the DAC, silly!

 

[amirm]

I can't think of a quicker, easier measurement with as much utility as that. Well worth a try at least I reckon.

Then what are you waiting for? Do the test.

 

[pozz]

As I said, SoundExpert's previous measurements have shown the null difference metric using this program simulation noise to be consistently just ~1-2 dB higher (worse) than the same metric calculated using the 2-hour real music suite, over all devices tested (45 so far), and using different ADCs.

What kind of dBs are we talking about? Referenced to what value?

If it's the df metric, that compresses differences between devices into a few dB after calculation. This is parallel to the issue we see in speaker preference scores, where the range above 4 starts getting crowded and problematic.

 

[Coffee_fan]

@Amir, after the review, did you have time to set a Roon endpoint and stream? I ask because when I was streaming using a Raspberry Pi outputting to USB, I would always hear a click-click every so often which seems was related to the PI using a common bus for Wifi, Eth and USB.

 

[EchoChamber]

@Amir, after the review, did you have time to set a Roon endpoint and stream? I ask because when I was streaming using a Raspberry Pi outputting to USB, I would always hear a click-click every so often which seems was related to the PI using a common bus for Wifi, Eth and USB.

I had some issues with Volumio on my Pi3b+... Overall I find Volumio to be a little buggy and the experience is not great either. Have you tried RoPieeeXL instead? I don’t use it all the time, but I haven’t had issues with it so far...

 

[Feyire]

@Amir, after the review, did you have time to set a Roon endpoint and stream? I ask because when I was streaming using a Raspberry Pi outputting to USB, I would always hear a click-click every so often which seems was related to the PI using a common bus for Wifi, Eth and USB.

RPi4 doesn't have this issue (i.e. no shared bus).

 

[Tks]

I'm not proposing to use the full 2-hour test suite, for the vey reasons you mention. I'm proposing to use the already standardized and widely used BS EN 50332-1 'Program Simulation Noise' test signal (described toward the bottom of this page), which was originally devised to have spectral content representative of music and speech for testing audio devices. This test signal is just 30 seconds long. As I said, SoundExpert's previous measurements have shown the null difference metric using this program simulation noise to be consistently just ~1-2 dB higher (worse) than the same metric calculated using the 2-hour real music suite, over all devices tested (45 so far), and using different ADCs. This is strong evidence that the program simulation noise would make an ideal analogue to worst-case real music (i.e. music that would produce the worst distortion through the DUT) to use in the null difference measurements instead of the 2-hour test suite. So that's just a 30-second recording and a few mouse clicks in the software to measure all possible sound degradation of any part of the electronic audio reproduction chain. I can't think of a quicker, easier measurement with as much utility as that. Well worth a try at least I reckon. (The non-coherent distortion metric mentioned in the video I linked is not related to null difference measurements by the way - I just mentioned it as it also uses music as the test signal, and to demonstrate that the multitone test is poorly correlated to sound quality.)

I see there are several owners of an APx555(B) in this thread - would any of you be willing to try out these null difference measurements using your AP analyzers? Preferably recording a device already measured here on ASR so correlations with SINAD can be made, but any data would be useful. A loopback null test would also be interesting to see the upper limit of the APx555 in terms of the null difference metric.

Ah I see now. That's fine I guess, though the whole thing with with the video as I'm sure you also see is, no one here claims any of those tests in isolation are to be used as a gauge for "sound quality" as perception in the levels we're talking about here are beyond audibility anyway. It's the same with jitter for example, you're not going to hear jitter artifacts 130dB down, it just demonstrates the engineering prowess, more than any metric of audible sound quality. Usually devices that follow suit with good performance of some of these synthetics, can be reliably predicted as to have good performance in the rest of the metric stack most of the time. Like you don't usually see devices with sprays of jitter artifacting go on to produce clean SINAD graphs for example.

All the tests you see in Amir's reviews are mostly now just an engineering flex fest. I'm not convinced anyone here is able to hear differences between any of the devices in this 120db tier (unless of course there are anomalies which a test like this may be able to reveal to some degree however doubtful). And for that I think this standard is worth running I suppose now that I gave it more clear thought.

Sergei (the guy who runs the site) was here a while back if I recall, while I was still learning and reading about audio (still reading). I am unable to parse the detailed information he provided even to this day, so I doubt I can comment in good faith. If you want to see how that all went down (seeming problems philosophically speaking too nuanced for a simple summary) check it out. Though statements like in post #206 aren't reassuring overall to his approach. Then again, much of the discussion is way over my head even today.

 

[bobbooo]

Then what are you waiting for? Do the test.

I would if I had access to the right equipment. As with SINAD, obviously an ADC with lower noise and distortion than the DUT is needed to get accurate results. Serge has shown the proof of concept with his Df metric calculations (via headfi member csglinux's measurements) using an RME Babyface Pro of some portable devices here. As can be seen from this list of loopback Df measurements, the Df value of the Babyface Pro is about the same as the top few devices in the previous link (~-35 dB), so it's likely even these portable devices are pushing the Babyface Pro to its limits, and something with much less distortion and noise is required to continue the project, such as an AP Analyzer. Unfortunately I'm not in a position to afford that right now, so I was asking people who already have one if they would mind doing a few quick tests.

 

[amirm]

@Amir, after the review, did you have time to set a Roon endpoint and stream?

No, but if I get a chance I will do so.

 

[amirm]

I would if I had access to the right equipment. As with SINAD, obviously an ADC with lower noise and distortion than the DUT is needed to get accurate results. Serge has shown the proof of concept with his Df metric calculations (via headfi member csglinux's measurements) using an RME Babyface Pro of some portable devices here. As can be seen from this list of loopback Df measurements, the Df value of the Babyface Pro is about the same as the top few devices in the previous link (~-35 dB), so it's likely even these portable devices are pushing the Babyface Pro to its limits, and something with much less distortion and noise is required to continue the project, such as an AP Analyzer. Unfortunately I'm not in a position to afford that right now, so I was asking people who already have one if they would mind doing a few quick tests.

Why don't you do the test with any ADC you have. Post those results and then we can talk about what is "quick and easy." In the process you will get first hand experience with what you are asking others to do.

 

[Labjr]

Why don't you do the test with any ADC you have. Post those results and then we can talk about what is "quick and easy." In the process you will get first hand experience with what you are asking others to do.

 

[bobbooo]

Why don't you do the test with any ADC you have. Post those results and then we can talk about what is "quick and easy." In the process you will get first hand experience with what you are asking others to do.

I already did a while ago. Couldn't find the files, so I just did it again, literally takes a couple of minutes. Loopback Df value of my PC's Realtek HD Audio soundcard using DeltaWave:

 

[bobbooo]

What kind of dBs are we talking about? Referenced to what value?


If it's the df metric, that compresses differences between devices into a few dB after calculation. This is parallel to the issue we see in speaker preference scores, where the range above 4 starts getting crowded and problematic.

It's the Df metric Serge of SoundExpert uses i.e. the ratio of the RMS level of the difference signal to the RMS level of the original test signal, as defined in his AES paper:

Wouldn't any 'compression' of the metric at better (lower) Df values also occur for SINAD? As they're both effectively signal to 'noise' (in its most general sense of unwanted sound) ratios, or a 'noise' to signal ratio in the case of the Df metric. The latter just seems a more generalized version of SINAD to me, taking into account all signal degradation instead of just THD+noise (just actual noise), and applicable to any input signal, including real music.

 

[Billy Budapest]

Better DACs won't help with that unfortunately. We are still talking about DACs, right?

Yes.

 

[pozz]

It's the Df metric Serge of SoundExpert uses i.e. the ratio of the RMS level of the difference signal to the RMS level of the original test signal, as defined in his AES paper:
View attachment 73849
Wouldn't any 'compression' of the metric at better (lower) Df values also occur for SINAD? As they're both effectively signal to 'noise' (in its most general sense of unwanted sound) ratios, or a 'noise' to signal ratio in the case of the Df metric. The latter just seems a more generalized version of SINAD to me, taking into account all signal degradation instead of just THD+noise (just actual noise), and applicable to any input signal, including real music.

I think the way you used the word "compression" for SINAD refers to diminishing returns past some semi-arbitrary value which refers to "good" performance. But what I meant was that the Df metric compresses measured values into a smaller scale. It's still a static test of signal fidelity in the end.

Say it was implemented as part of the regular test suite. It would still have to be supplemented with all of the various electrical measurements, like level, power or gain. That implies that it would even translation into a graph: level vs. Df, or power vs. Df. At which point the metric starts to collapse since it's not referenced to SPL or psychoacoustic considerations.

I'm probably not being completely fair. My thinking here is that it seems like there is no reason to apply the Df to electrical signals given the kind of performance possible today. I would see it being, however, very useful for analyzing production plug-ins for DAWs, especially ones that are supposed to be transparent. That's one area of measurement that is still completely underdeveloped and which, I think, would cause a lot of people to pay attention.

 

[Billy Budapest]

True... but the place with the most room for improvement isn't the DAC, or even the amp. It's the microphones used to record the sound, and the headphones or speakers used to play it back.

That could very well be. Still, I anticipate manufacturers will continue to produce better performing DACs. We reached inaudible levels of distortion years ago, yet here we are in 2020 with a DAC capable of reproducing sound with 122 SINAD levels of noise and distortion. Will that figure be surpassed? Of course it will. And Okto itself stated they are exploring using the ES9038 architecture, just as I had posited.

 

[SDX-LV]

I wish my ears were even close to such SINAD...
Then I would start thinking about spending more than 250$ for a stereo DAC.

I think people are way better off spending money on miniDSP SHD series, or just less perfect DACs.

 

[Vasr]

I already did a while ago. Couldn't find the files, so I just did it again, literally takes a couple of minutes. Loopback Df value of my PC's Realtek HD Audio soundcard using DeltaWave:

View attachment 73844

Can you send me by DM or better post on a separate thread the simplest way to do this - what to connect and what to install, etc (without having to read a lot of links and theory)? I am just curious. While the absolute value may not be that accurate, I can measure the onboard audio vs sound card audio vs Optical out to AVR and loop back via its pre-out vs optical-out to optical-in loopback on PC. Thanks.

 

[Czech6]

@Kal Rubinson, would Stereophile review the Okto dac8 Stereo? I bet many would be interested in both John's take on the measurements and the subjective reviewer's impressions vis-à-vis car/house priced competition!

 

[Labjr]

I wish my ears were even close to such SINAD...
Then I would start thinking about spending more than 250$ for a stereo DAC.

I think people are way better off spending money on miniDSP SHD series, or just less perfect DACs.

DACs are getting very good. But it took a long time to get there. Bad digital distortion is much more offensive than bad analog. So going to the extreme is a good thing to ensure audible distortion is never possible with normal use.