[goodvibes]

  This is not true form many reasons:
- DX90 have SW volume control. It mean, that on lower SW volume with higher amp gain You lost more music information.
- Circuits before amplifier have any noise. This noise is independent on SW volume. Higher gain of amp = You got higher this noise.
- "Gain swith" set gain of opamp (I think swith between different resistors in loop). Higher gain of opamp can produce more distortion, noise, etc. But this can not be say generaly, it is more difficult (any opamp are more stable at higher gain).

That's not correct. It has a proper digital volume control that that I think ( site with pdf is down right now) uses a 32 bit floating word and probably a bit of dither to be able to change volume while maintain a bit perfect stream with proper tracking. There is no loss of low level information at reduced volume settings.

 

[goodvibes]

   
Lurker0, can you explain what you mean here regarding "distortions"? 

Any change to signal, wanted or not, will measure as distortion when compared to the source. He's saying that if you opt to use crossfade etc in rockbox, not rockbox itself.

 

[Lurker0]

Thank you Lurker0 for the international version firmware. So does this version display Chinese or Japanese characters properly? Thanks..

I believe it does

I called it Intl becuase the previous one (distributed in Russian forum) had no Asian languages characters at all. Here they are, but whether they are at right places and in full qualtity - that's what I don't know.

 

[Lurker0]

 

   
Lurker0, can you explain what you mean here regarding "distortions"? 

Any change to signal, wanted or not, will measure as distortion when compared to the source. He's saying that if you opt to use crossfade etc in rockbox, not rockbox itself.

You took it right generally, but not in details. I mean also Rockbox itself, because it re-samples any input to its internal repsentation. Only format equal to its internal representation can be output without distortions. For Android builds, this format is usually 16/44.1, though it is possible for devices like DX90 to build a version with higher parameters with the device-specific output, avoiding Android common path.
 
Because I mostly listen hi-res records, RockBox is no go for me.

 

[goodvibes]

I didn't consider resampling as I only used appropriate formats when I had rockbox in in the past. I don't even like ASRCs

 and if needed with increasing the word.

 

[Lurker0]

 

  This is not true form many reasons:
- DX90 have SW volume control. It mean, that on lower SW volume with higher amp gain You lost more music information.

That's not correct. It has a proper digital volume control that that I think ( site with pdf is down right now) uses a 32 bit floating word and probably a bit of dither to be able to change volume while maintain a bit perfect stream with proper tracking. There is no loss of low level information at reduced volume settings.

You can't find a PDF, because DX90 utilizes DAC digital volume control, and Sabra detailed datasheets are not available.
 
But in DACs volume control is usually implemented after oversampling and before delta-sigma processing. It means, information loss can't be avoided, and here Wishmasterx is right in general sense.
 
By the way, this means that no a player with digital volume control is able for direct DSD playback, regardless of manufacturers claims: CPU may send DSD stream directly to DAC, but DAC has to convert DSD to PCM for volume control and then pass the result to delta-sigma. I'm not aware of any real world DAC that has volume control after delta-sigma processing, right before voltage or current analog output.

 

[borrego]

  That exactly is. More CPU load means higher temperature, and more power consumption by CPU, which, in turn, affects output amplifiers. Also, I'm curious, may it lead to bigger jitter, but I'm lazy (or busy?) enough to start investigating this question

 
You might notice that all the discussions around different firmware versions are mostly concentrated on bass. Temperature and power supply have more impact exactly on bass.
 
And this is exactly the reason that I like removing unneeded libraries and services: it makes the system more stable and predictable.

 
Since you mention jitter, I would like to share my experience with the ESS9018 DAC chip.
 
The DAC of my speaker setup is the audio-gd NFB-1 using the ESS9018 DAC chip. The version I have was the very first batch, first version of the NFB-1 which I pre-ordered it with special introductory price at the time. This very version of the NFB-1 did not use a TCXO quartz chip for the DAC chip clock. After I had used the NFB-1 DAC for about 3-4 months, I read from audio-gd site that they were offering TCXO quartz chip as an upgrade for all newly built NFB-1. I emailed audio-gd and asked them whether the TCXO would make an obvious improvement. The reply from audio-gd  was inconclusive: Kingwa (the chief designer and owner of audio-gd) said the TCXO used was just for the clock of the ESS9018 chip and was not in the signal path, so theoretically it should not make much difference except may be better operation stability under temperature change. But Kingwa said he "thought" the sound signature of the DAC did change after they switched using the TCXO.
 
Kingwa said it was not worth the money sending back the DAC just for the TCXO upgrade, but if I didn't mind doing the soldering work, they could sell me the TCXO chip for me doing upgrade myself. They even agreed honoring the 1 year warranty of the DAC despite my own soldering.
 
So I did the TCXO upgrade myself. My Avatar you see right now, was the photo of the NFB-1 ESS9018 digital section board, with all the wires dangling when I de-soldered it  from the analog board section of the NFB-1.
 
And after the TCXO upgrade, the immediate sound signature change I noticed was weaker bass, but less "sterile" sound reproduction. I contacted Kingwa about it to verify if it was supposed to be result expected. He told me yes but the TCXO would again need another 100 hours to settle down, and by then the bass would become stronger again. I could never verify it myself as the burn in period was simply too long for me to remember the original sound signature.
 
So does my experience mean anything with the DX90 sound signature explanation? Probably totally irrelevant. But my experience tells me the ESS9018 chip is extreme sensitive to clock jitter, even with its "jitter elimination" technology. So it is likely the 9018K2M chip could behave similarly, when its clock non-TC XO is operating at different temperature uncompensated.
 

 

[nmatheis]

Just flashed the alternate 2.1 firmware, and it is indeed really easy to tell if you succeeded by the font change within the UI - much cleaner looking!

 

[pegasus21]

It's starting to sound like the DX90 might be working a bit like a tube amp in that it has to warm up and settle down before listening.
 
I have some question about rockbox. So internally it resamples all audio to 44.1KHz/16bits? So in this case, it might be better to avoid it if one prefers a less processed audio source or wants high res audio (ie 2.1.0)?

 

[Marleybob217]

  Pretty looking, useless stuff removed DX90 firmware based on 2.1.0, International version
 
Brief description:

• The two fonts used by iBasso are replaced with one, which is Roboto Condensed, with the missing characters merged from Arial Unicode MS.
• Useless background services are removed from the startup procedure.

 
Benefits over the stock firmware:

• More informative display due to semi-bold condensed font (European languages), and correct character widths used (any language).
• Slightly faster ready from power on.
• Less RAM and CPU usage (does it really matter? See below).
• Theoretically, less battery consumption (can it even be measured?).

 
Download link
 
Detailed description
1. Fonts
iBasso uses Microsoft YaHei for Latin-based and Chinese characters, and Nimbus Sans Global Bold for the rest (Cyrillic, Japanese, Korean, Thai etc). It also uses pre-calculated character width tables, which (confirmed by calculations!) do not correspond to any, or any combination, of those fonts. Regardless of the reason iBasso went this way, the resulting texts look awful for me, and it was the primary reason for the firmware modification.
 
Note: still sometimes character spacing may look odd, but this is due to character widths are rounded to integer number of pixels, and because iBasso does not use kerning in text rendering.
 
2. Useless services
The full list is:

1. netd (internet servers and services).
2. debuggerd (gdb debugger interface, also creates crash logs of native applications).
3. rild (controls wireless hardware).
4. rild3 (controls wireless hardware).
5. mediaserver (indexes media files, provides access to them for Android applications).
6. installd (required to install Android applications from APK).
7. servicemanager (Android applications interface to hardware and resources).

 
It is obvious that none of them have any use in DX90. Note for the last two items: DX90 has removed Dalvik JVM, so there is no a way to install and run Android apps!
 
Speculations
It should be obvious that none of changes made may impact on sound streams. That's why I, like iBasso programmers, claim that there are no changes made sound wise.
 
But some of you claim that you hear the difference between different firmware versions, and even iBasso confirmed that there is a "sound signature" of each and every change of the software. From this point of view, you may notice a difference with this build as well.
 
Please note, that, although the listed services almost do nothing, OS kernel wastes CPU ticks on serving them, and, which is more important, OS kernel does this small additional job with interrupts disabled. The latter mean, those changes may reduce number of dropouts, which may be caused by sound output buffer underruns.
 
Anyway, I feel much better when I look into the list of running processes and see only MangoPlayer, along with a few system processes I know exactly why are they needed

 

Thanks! I would really like it if someone could measure the 'sound signatures' between the firmwares. 
I mean, I sometimes have issues telling the difference between dac/amp setups and that's when I can AB them within seconds. A firmware update takes quite a bit longer, and I doubt anyone can really remember such small differences for such a long period of time.

 

[Lurker0]

Originally Posted by borrego /img/forum/go_quote.gif
 
So does my experience mean anything with the DX90 sound signature explanation?

Of course, yes. But I'd tell you some other thoughts that you might don't think of.
 
"Deeper bass" or "wider sound stage" don't always mean "better quality". It's our brain that "explains" the differences such a way. People tend to like some distortions, and, because they like them, they think about them in positive terms. Nobody who prefer tube amps express like "I love those even harmonic distortions!". Instead, they speak about tone balance, sound stage etc.
 
Back to DX90. It utilizes balanced current outputs from the DAC. Balanced outputs mean there are wires with inverted left and right channel signals. So, if, under some conditions, there is a crosstalk between inverted left and direct right, or/and vise versa, which words would be used to explain the result? I bet it is "wider sound stage" and not "crosstalk on balanced outputs", because stereo width controller acts exactly the same way

 
On the other hand, it can be crosstalk between left and right channels somewhere further in the path, and this crosstalk may depend on some conditions as well.
 
So, when you notice sound difference in sound stage, how to tell what is the reason: less/more crosstalk on balanced outputs, or more/less crosstalk on unbalanced paths? In other words, is it a distortion or an improvement?
 
In everyday life, I don't care. I prefer to listen to music, not to hardware, firmware, formats...
 
For many years, listening to the same records, from different sources, in different quality, on different hardware, I trained my brain to search for similarities, and this way to tune my internal DSP to compensate all those differences. The quality of hardware used I feel as a level of tiredness (load on DSP). Same with the quality of records. With a new hardware, I need some time to listen to a number of tracks I'm familiar with from the previous experience. Then I can enjoy listening new music on the new device.
 
It does not mean I can't hear differences. It means I don't concentrate on them without a real need. Frankly... I listen DX90 on my trip from home to office at mornings, and back at evenings. And every day, when I turn the player on at the evening, I do notice that sound stage is wider and bass is deeper than they were at the morning. But I don't care about the reasons and I'm not dissatisfied the next morning

 
I pity those who waste the time seeking for differences, which always can be found!

 

[SilverEars]

Why is the channels inverted?  What purpose does it serve?
 
I know Tyll measured United Airlines phones and they were purposely out of phase for some application according to Tyll and they can provide a sense of space because of the phase non-linearity of the channels.  There are other ways to create stage I'm sure.  Bascally stage is result of timing of the sound to reach your auditory sensing system.  Cans provide more since the sound entering from the outside provide sense of direction of the sound.
 
I know each pair of output per channel on DACs are for differential mode to subtract out picked up noise in the path.  I saw this with the Wolfson, but I don't think the ESS is detailed enough to show it.
 
I know what you mean by perception of sound changing from time to time.  One day it may sound good and later like total crap.  This depends on out sensory condition at the time of listening.  What is odd is, other DAP thread do not go into this, and I'm sure the experienced members are aware of this perception affect. Yes, perceptual sound memory is quite short, but if something is significant, your mind notes it conceptually, not perceptually.  That can be recalled.

 

[goodvibes]

Here we go. http://www.esstech.com/PDF/digital-vs-analog-volume-control.pdf I was thinking 40 bit would be needed for worst case HiRes at lowest volume settings but for all intents the ESS V control should be noise free in this application where we are reaching our own hearing noise floor in use, especially when in a lower than max gain setting.
 
Inverting has nothing to do with effects. It's used along with a noninverting in differential mode to lower the noise floor and lower distortion. Using an inverting output along with the noninverting in differential or balanced mode reduced the noise floor (increasing dynamic range) along with the 2 D2A converters in each chip being used in the same fashion. Lowering the noise floor should aid the digital V control operation as well.

 

[Wyd4]

I have really enjoyed reading the last few pages of posts.
Very interesting stuff indeed.

I dont understand any of it... lol... But it does interest me.

New fonts are much clearer.  Cant speak for the sound changes if any at the moment as I only have Senn Momentum On ears on hand, so not the best headphone for trying to listen critically.
 
Very much enjoying the DX90

 

[zeppu08]

Just a question before I do an upgrade.. Is the Intl version 2.1 stable or still on process of testings and all? Thanks!