Ok, it seems you are not tired anymore, I have rested as well, let’s continue. Melokin (let’s call him using his brand’s name, I don't remember the guy’s name) neither designs the USB to i2s convertor nor writes the drivers for it. He designed the hardware and provided an option to install second party’s USB to i2s convertor named Amanero. It is designed by an Italy based company Amanero Technologies, which also provides Windows drivers for their boards - for free. As you have figured out by yourself, the problem with linearity is not associated with the hardware design itself - using SPDIF port, which doesn’t require a driver, the linearity looks good enough for $300 DAC. It is obvious that the problem is related to two things - the USB input/Convertor (and the standard Windows driver), and also with your 0 d FS input level. I will talk about the input level a bit later, but tell me why didn’t you even try to use a native Amanero Windows driver for your test?You haven't read a fraction of the testing I do here to associate IC selection with fidelity. It is the implementation that by far determines the final performance of an audio device, not what IC is used internally. The best DAC chip in the wrong hand produces horrible performance as I have shown in countless reviews. Here is the Melokin 9.1 which you said you like:
Second thing - your graphs can be used for YOUR Melokin DAC ONLY, they can not be true for my Melokin. The reason is simple - I don’t use either USB input or standard Amanero convertor, I use an advanced version of the convertor (if necessary) and SPDIF input (usually). There is another reason - 0 dB FS measurements can not be approximated to other similar devices, period. Why? I will tell you a bit later.
There is nothing to do with ESS9038PRO - see above.That ES9038PRO is not doing you a darn bit of good, is it? Or maybe you are a fan of it saturating so early?
ONLY the devices you have tested, period. It can not be said about other Okto DAC8 or Melokin devices being tested at 0 dB FS. See below - why.The top performing DAC in THD+N measurements is the Okto DAC8. It uses ES9028 and gets a SINAD of 118 dB. The Melokin 9.1 despite using ES9038Pro stops at 100 dB for a whopping 18 dB higher noise+distortion.
First - I have spent about half of that price. Second - my level of demand is pretty high, and majority of stuff I have auditioned (and I auditioned a lot of devices at all levels) fits in the category “just acceptable”. Exceptional devices like dCS or EMM Labs are not acceptable for me due to their price level."Just acceptable" after spending $4,500. There are plenty of just acceptable multi-channel products out there at far lower prices.
This is Major misunderstanding of the DAC functionality and the standard. Look what happens when people try to test commercial products at 0 dB FS: https://service.tcgroup.tc/media/Level_paper_AES109(1).pdfAES standards are created by the industry so at times you want to be careful in accommodations they make for companies. -1 dBFS to allow filter overruns and such is one. So in this regard, my standard is more rigorous.
Check out those players measurements. 7 tested, none passed. Are they all “bad players”? Are those players which had been tested, bad ones?
Chris Tham published a perfect article regarding the issue: https://www.audioholics.com/audio-t...dbfs-levels-on-digital-audio-playback-systems
And finally, there is an approximation of 0 dB FS to analog level https://www.av-iq.com/avcat/images/documents/pdfs/measdigaudlvls.pdf
Not all professional products can output 24 dBu, just the best ones.A 0 dBFS measurement unit is to be the highest audio level. Assuming this is to be the highest audio level before clipping occurs, this corresponds to an analog level of 24 dBu.
Now let me tell you what is happening in real life devices at 0 dBFS. Let’s say DAC outputs this level properly to the analog stage, but such as it is the last step before clipping occurs, we have to take in consideration the tolerance of analog components used in that stage. Usually it is +/-5% for commercial devices, so some devices will start clipping and we will get Nielsen’s results. Some of them with no tolerance (or negative tolerance) will pass. Can we say that all measured devices are good or bad? No we can not. We can say it only about the tested device.
But there is another problem with a DAC. They all use a reference voltage to create analog output. What if that voltage also has a tolerance (either the reference voltage source or a resistive divider)? We will also get the clipping ruining the test results, and some people after such test even start blaming the tested devices ...
Now let’s get back to Marantz 8805. Per data sheet AK4490 DAC has maximum analog output voltage min +/-2.65V, typical +/-2.8V, and max +/-2.95V, according to note 10 it is p-p value. Those levels equal to 1.87V RMS, 1.97V RMS and 2.08V RMS accordingly. Let’s check what is used after the DAC:
https://hometheaterhifi.com/reviews/receiver-processor/processors/marantz-av8802-processor-review/
OK, it is doubled, but it is still within the max output limits, where is 4V RMS?The AV8801 output stage had a gain of two. The AV8802 output stage is now unity gain. ... The signal at the output of the DAC, which is before the volume control, is doubled in the AV8802 to compensate for the gain reduction at the output.
The design is based on the output capabilities of the DAC chip, connected to the unity gain op amp. Is it wrong? No, the signal part from the DAC to the output is maximally simplified to avoid added distortions, like for example in Denon 5803, which had 7!!! Op amps after DAC per channel.
Now to the voltage reference level for the DACs, it is formed by PQ050DNA1ZPH chip, per it’s data sheet the output voltage can be min 4.875V DC, typ 5.0 V DC, and max 5.125V DC, it is 2.5% tolerance. What if that chip will have a bit higher voltage, but still within the tolerance limit? This DAC will accept 0 dB FS signal, but due to the 2% DC reference level shift the output stage will start clipping. Is it a design problem? No, it it 0 dB FS problem / tolerance of the components problem:
Ref. Chris Tham
- 0dBFS+ levels cannot be generated by recording from an analog source. No matter how high the recording gain level is set at, the digital recording will never contain 0dBFS+ levels. Of course, if the recording gain is set too high, digital "clipping" will occur, but this is not the same thing as 0dBFS+ levels.
- 0dBFS+ levels can be created through subsequent manipulation or processing of a digital recording. For example, if a digital recording is "amplified" (by multiplying each digital sample by a constant amount) it may then subsequently lead to 0dBFS+ levels.
- Therefore, 0dBFS+ levels are technically "illegal" states and represent anomalies or artifacts created by processing signals in the digital domain.
[*]If a player does not handle 0dBFS+ levels it is NOT strictly speaking the fault of the player. There is no "requirement" in the CD specification for players to be able to handle 0dBFS+ levels and arguably the ability to handle 0dBFS+ levels are a violation of specifications.
Now what exactly is -1 dB FS level per AES 17 standard? It is enough to cover all component’s tolerances. (-1dB=-12% decrease). Is it “too old” and 0 dB FS testing is “more reliable”?
May be, but not for commercial products. 24 dBu is used for professional production, as well as 1% components tolerance. Thus if you test a commercial product be ready that one of them will fail, and another one will pass. It doesn’t give a right to either blame or praise such product basing “conclusions” just on one test. Make them statistically correct and then call such test results “a science”, otherwise they are equal to some subjective opinion and will always be different.
Last edited: