SONY NW-WM1Z / WM1A
Status
Not open for further replies.
Jul 30, 2020 at 11:55 AM Post #42,121 of 45,723
Well...it has always been my questions.....how exactly does AI play the role in upscaling LOL?
Neutron can do a heck lot more....I guess the algorithms count everywhere but where does AI play the role ?
I read in Sony's interview once, sorry can't source it anymore. The AI tries to recognize the content of your music with the records databases it learned, for example, at the intro when drums are pounding, it uses percussion mode to process, and when female vocal comes in, it switch to female vocal mode. Basically it dynamically apply to different DSEE algorithms within one song, without your interference and trying to be smart.
i found the interview somewhere on headfi, but can’t source it right now. it’s in Japanese from a Japanese engineer
 
Last edited:
Jul 30, 2020 at 11:57 AM Post #42,122 of 45,723
I read in Sony's interview once, sorry can't source it anymore. The AI tries to recognize the content of your music with the records databases it learned, for example, at the intro when drums are pounding, it uses percussion mode to process, and when female vocal comes in, it switch to female vocal mode. Basically it dynamically switch to different DSEE algorithms within one song, without your interference and trying to be smart.
i found the interview somewhere on headfi, but can’t source it right now

It's here:
https://www.sony.jp/feature/products/dseehx/


I think the translation isnt perfect.

Sony is able to apply the 5 modes all at once. It is able to separate out the percussion Instruments to apply the percussion effect and separating out the vocals to add the vocal effect.
Hence they talk about scenes.

Yamamoto: Yes, that's right. Also, regarding that mode, it is not as simple as the conventional five operation modes, but optimal upscaling is performed every time for various scenes. It may be overstated to say that "there are infinite modes", but in the vocal scene it is not as simple as vocal mode.
 
Last edited:
Jul 30, 2020 at 12:03 PM Post #42,123 of 45,723
The treble glare can probably be tuned out with a custom firmware revision from mrwalkman.
The treble glare can probably be tuned out with a custom firmware revision from mrwalkman.

I'm going to play around with some of the regions to see what affect that might have (those that are non-veiled vs.veiled on top) with DSEE AI
 
Jul 30, 2020 at 12:06 PM Post #42,124 of 45,723
Is the dsee hx Ai on zx507 the same performance and sound quality as the one in dmp-z1?

By memory it seems that the dmp-z1 dsee hx ai in our 1a and 1z sounds better then on zx507.
 
Jul 30, 2020 at 12:13 PM Post #42,125 of 45,723
Is the dsee hx Ai on zx507 the same performance and sound quality as the one in dmp-z1?

By memory it seems that the dmp-z1 dsee hx ai in our 1a and 1z sounds better then on zx507.

I think the zx507 you heard is on the firmware 1.x
Sony was trying to make the zx507 almost like the ZX300 with a warmish laid back analog-like sound signature. Maybe back then they were just convincing zx300 users to switch over.

Somewhere in mid 2020, Sony suddenly added ZX507 to their signature website, which likely means they have a refocus/rebranding on this Walkman.

I feel that with the new firmware 2.01.00, ZX507 seems to have taken on an almost new sound signature that is more focus on resolution and dynamics.

Its time for tiger ears to revisit ZX507 on the new firmware.
 
Last edited:
Jul 30, 2020 at 12:23 PM Post #42,126 of 45,723
I think the zx507 you heard is on the firmware 1.x
Sony was trying to make the zx507 almost like the ZX300 with a warmish laid back analog-like sound signature. Maybe back then they were just convincing zx300 users to switch over.

Somewhere in mid 2020, Sony suddenly added ZX507 to their signature website, which likely means they have a refocus/rebranding on this Walkman.

I feel that with the new firmware 2.01.00, ZX507 seems to have taken on an almost new sound signature that is more focus on resolution and dynamics.

Its time for tiger ears to revisit ZX507 on the new firmware.


Thats so unfortunate cause I sold my zx507...
I had it on fw 1.x maybe third fw release. Definitely not the fw 2.x

I liked the zx507 except it lacked on that sound stage a lot. It was very good with vocals and amazing bass! I even find bass to be deeper and more rumbly then 1a or 1z.

Sounds like I may like a lot the new fw.
But I had a huge downside with zx507 its the small screen 🙁
I have no such issues with 1a....
 
Jul 30, 2020 at 12:24 PM Post #42,127 of 45,723
I've worked for a company that did machine learning / AI. So my question is that for this to work then need to do real time what is called feature extraction to sample the audio then compare it to their existing models and then apply any change. This has typically been CPU and memory intensive.

So how can they do this in a DAP with limited cores and RAM. I would be interested to know the details on the Mediatek SOC they use.
 
Jul 30, 2020 at 12:28 PM Post #42,128 of 45,723
I've worked for a company that did machine learning / AI. So my question is that for this to work then need to do real time what is called feature extraction to sample the audio then compare it to their existing models and then apply any change. This has typically been CPU and memory intensive.

So how can they do this in a DAP with limited cores and RAM. I would be interested to know the details on the Mediatek SOC they use.


Sony is a software master!
Why do you think 1a/1z are still so good!
They can even tune it to sound like chord dave lmao
 
Jul 30, 2020 at 12:28 PM Post #42,129 of 45,723
Well, the thing is that each filtering techniques, noise shaping and algorithms have a special Favor upon a specific frequencies. This can be observed upon desktop system with high quality chains and using HQplayer.

So this is probably the case ? The AI guess upon the dominant frequencies across the processing for a given amount of time ? Because upscaling do need filters and noise shaping techniques. So, it just is a matter of which options will be best ?

I tried on the HQplayer desktop system And it is a PITA to switch to which options for which to best perform the tracks or album or genres that I am listening to
I've worked for a company that did machine learning / AI. So my question is that for this to work then need to do real time what is called feature extraction to sample the audio then compare it to their existing models and then apply any change. This has typically been CPU and memory intensive.

So how can they do this in a DAP with limited cores and RAM. I would be interested to know the details on the Mediatek SOC they use.
 
Last edited:
Jul 30, 2020 at 12:29 PM Post #42,130 of 45,723
Got an used WM1Z here in Dubai, they are awesome with Z1r. Feels like having desktop sound quality in my pocket.

I have two question.

1. All my playlist are organized in Folders, each folder is playlist for me. After I copy them using Music Center, each Album name becomes a folders and I am kind of lost. Wanted my folder names to same as after coping.
2. Wanted to try WM1A/Z+ firmware, if I do not like sound can I roll back to factory setting?

Please suggest.
 
Jul 30, 2020 at 12:37 PM Post #42,131 of 45,723
I've worked for a company that did machine learning / AI. So my question is that for this to work then need to do real time what is called feature extraction to sample the audio then compare it to their existing models and then apply any change. This has typically been CPU and memory intensive.

So how can they do this in a DAP with limited cores and RAM. I would be interested to know the details on the Mediatek SOC they use.

Hence they have designated different levels of DSEE HX for different devices depending on available processing power and battery capacity.
11244041.jpg
 
Jul 30, 2020 at 12:54 PM Post #42,132 of 45,723
Hence they have designated different levels of DSEE HX for different devices depending on available processing power and battery capacity.11244041.jpg
Then to think that DMP doesn’t even have DSEE Ultimate is like a cold can of sewer water splashed in the face
 
Jul 30, 2020 at 1:03 PM Post #42,133 of 45,723
Just three years ago you have this video about Sony:


I feel that Sony in 2020 is a much more technologically advanced company than most of us would imagine it to be.

They have alot more successful products than ever with their xm3 headphones and TWS and as well as their PS4 and upcoming PS5.

Their ccd sensor imx500 is also gaining AI
https://www.cinema5d.com/sony-imx500-worlds-ai-image-sensor-announced/

It going to be exciting times in 2021 as a Sony fan.

I guess what we have to do now is save up money and wait for Sony to release the next flagship walkman.
 
Status
Not open for further replies.

Users who are viewing this thread

Back
Top