The only measurements I've seen was done by Jussi Laako with DSC 2.5. His recommendations, based on measurements, was DSD256 as a "sweet pot" for DSC 2.5.
With HQP I've tried DSD256 and DSD512. My opinion (quite subjective):
1. DSD512 is better then DSD256 with the same filter+modulator. The difference is not big, but quite noticeable, even in blind tests. More "analog" sound, I'd say.
2. Full-length filters in HQP (poly-sinc-xtr (-mp), closed-form, poly-sinc (-mp) ...) is MUCH better then -2s or -shrt (short) filters. The difference is audible at once and very big.
So, the reasonable solution is in using poly-sinc-xtr (-mp), closed-form, poly-sinc (or other full filters) in DSD256. Very high SQ.
My old I7-8700 with fast DDR4 did it (computer cost ~ 1500 USD). Don't forget SoC as NAA. I am with Odroid C2 for now.
To handle DSD512 with poly-sinc-xtr (-mp) and closed-form you'll need much more power. Cost will be 3 times more.
If you won't invest to a new computer, try offline upsampling by using Saracon (up to DSD128) or AUL converter (up to DSD1024). I slightly prefer the last one.
Aleksey Galkin says that has got a very good results with special version of foobar + sox resampler + ASIO Proxy.
But bear in mind - all are very subjective
Hello Vit,
what about the modulators? any subjective perceived difference between then??
Yes, modulator choiсe is also affect. Not so big as filters, but quite noticeable.
Mostly I use DSD7 256+fs, sometimes AMSDM7 512+fs.
"I saw a few people talking about memory speed being important for this project and dsd. I just learned the other day that formatting memory sticks and hard drives with the largest block size increases read and write speeds and after trying it (64kb block size instead of standard 4kb) it has a significant effect. Minimum file size is 64kb however, or whatever block size you use."
that was my quote earlier. Well I dont want to be responsible for anyone else having to go through what I did, So that is the purpose of this update. I do have compassion when it comes to others loking their life's music collection or even one album.
Well it certainly increases speed, using max block size and since the block size determines the min file size, and redbook files are large, it sure makes sense not to divide them up into 4kb blocks with metadata surrounding each 4kb block, using space and adding innumerable operations to file reading, writing, search, processes.
OS files, often <4kb, are best stored on an HD partion preformatted to standard block size format due to block size being min filesize, hence only my two user partitions used NTFS 64kb block size.
So as I often install linux to the system partion, I installed Windowmaker on Debian.
Even after setting the Hard drive info and both ntfs (64kb block size) were detected, they could not be read.
Took ownership, edited internal HD addresses, even installed more NTFS utilities.
So I invoked testdisk, a lower level scan tool, and it complained of missing sectors.
With Block size higher, a partition of music will require less space on the hard drive and read and write quicker.
So even testdisk was built to locate files of standard block size.
Its error cited missing sectors for each user partition. (since big files stored in 64kb block size use less space, the data content did not match the amount of space it was expected to occupy.
I tried recuva after reinstalling windoughs and that probably looks for 4kb blocks because it found nothing.
Win did recognise the other music partition which uses 64kb block size after reinstall, as that had never been scanned and fixed by any linux program
So 64kb block size works best in a windows only machine if NTFS is the file structure you use.
I have not tried raising the fat32 blocksize to 64 kb but will soon.
Also need to explore ext4, btrfs, and their structures.
Well, TTL for Audiophile linux.
I built that OS via shell with netw,,, Arch Linux.
Throughout the years having lost a lot of data due to hd failure and viruses and mistakes I made, my arch audiophile linux HD with all my losslesss music has always retained every file. Lost my collection five or seven times on various mainstream distros, even Avlinux. Maybe its bc during install I only build it out of a couple dozen programs and its only a local jukebox, never online. How would one accidentally erase everything on a computer which lacks absolutely any software which can perform that function, and does not use the net?
Btw, Audiophile linux really seems to be a world apart from others with its realtime kernel and absence of power hungry processes, which can affect the audio section and create voltage fluctuations, produce rfi within the computer, which can be picked up and carry through to the audio stream it outputs.
My laptop remains at 2-5% of max state at all times on audiophile linux.
I can hear a massive difference and will never repurpose that lcd-less laptop again
...
...
that was my quote earlier. Well I dont want to be responsible for anyone else having to go through what I did, So that is the purpose of this update. I do have compassion when it comes to others loking their life's music collection or even one album.
Well it certainly increases speed, using max block size and since the block size determines the min file size, and redbook files are large, it sure makes sense not to divide them up into 4kb blocks with metadata surrounding each 4kb block, using space and adding innumerable operations to file reading, writing, search, processes.
OS files, often <4kb, are best stored on an HD partion preformatted to standard block size format due to block size being min filesize, hence only my two user partitions used NTFS 64kb block size.
So as I often install linux to the system partion, I installed Windowmaker on Debian.
Even after setting the Hard drive info and both ntfs (64kb block size) were detected, they could not be read.
Took ownership, edited internal HD addresses, even installed more NTFS utilities.
So I invoked testdisk, a lower level scan tool, and it complained of missing sectors.
With Block size higher, a partition of music will require less space on the hard drive and read and write quicker.
So even testdisk was built to locate files of standard block size.
Its error cited missing sectors for each user partition. (since big files stored in 64kb block size use less space, the data content did not match the amount of space it was expected to occupy.
I tried recuva after reinstalling windoughs and that probably looks for 4kb blocks because it found nothing.
Win did recognise the other music partition which uses 64kb block size after reinstall, as that had never been scanned and fixed by any linux program
So 64kb block size works best in a windows only machine if NTFS is the file structure you use.
I have not tried raising the fat32 blocksize to 64 kb but will soon.
Also need to explore ext4, btrfs, and their structures.
Well, TTL for Audiophile linux.
I built that OS via shell with netw,,, Arch Linux.
Throughout the years having lost a lot of data due to hd failure and viruses and mistakes I made, my arch audiophile linux HD with all my losslesss music has always retained every file. Lost my collection five or seven times on various mainstream distros, even Avlinux. Maybe its bc during install I only build it out of a couple dozen programs and its only a local jukebox, never online. How would one accidentally erase everything on a computer which lacks absolutely any software which can perform that function, and does not use the net?
Btw, Audiophile linux really seems to be a world apart from others with its realtime kernel and absence of power hungry processes, which can affect the audio section and create voltage fluctuations, produce rfi within the computer, which can be picked up and carry through to the audio stream it outputs.
My laptop remains at 2-5% of max state at all times on audiophile linux.
I can hear a massive difference and will never repurpose that lcd-less laptop again
...
...
Last edited:
We talked about the speed of memory - RAM.
The speed of the storage system (HDD, etc.) does not matter.
One reason I stopped visiting computeraudiophile.com is computer black magic like this.
Hello all, i have been listening carefull to my dac.. on the last 3 days and i have say i am really impressed with the quality of the music i m earing..
I m using signalyst and rpi3 as naa with lps.
I ve listened to a lot of dacs. Akm, cirrus, ad1965, pcm1702/4.. with valves, with opamps (almost every opamp) etc..
This dac is very good.. unfortunaly i need a better pc to use better filters and to do dsd 512.
The sound is analog, clean an rich full of details...
this is my subjective opinion ....but i would like to share it.
My fully respect for pavel and all involved in the production and fine tunning of this "no dac".
I m using signalyst and rpi3 as naa with lps.
I ve listened to a lot of dacs. Akm, cirrus, ad1965, pcm1702/4.. with valves, with opamps (almost every opamp) etc..
This dac is very good.. unfortunaly i need a better pc to use better filters and to do dsd 512.
The sound is analog, clean an rich full of details...
this is my subjective opinion ....but i would like to share it.
My fully respect for pavel and all involved in the production and fine tunning of this "no dac".
The authorship and elegance of the solution of this DAC belongs to Jussi Laako (Signalyst).
I just managed to bring the project up to those indicators that correspond to my notion of good sound.
We can say that the DSC2 project finally got a stable release in the form of v2.5.2 and v2.6.2. These versions proved to be very good not only in SQ but in good cloning. The spread of THD in the boards is 0.0004-0.0007% at 1000Hz -10db.
Maybe I'll be working on new versions of DSC2. But already in a lower priority.
Last edited:
Has any of you ever tried an board from Acmesystems as a NAA with Amanero?
Acqua - System On Module
I tried long time ago with a JLSounds I2S board without success (I wasn't able to compile kernel). Looks like pre-built image for the board is: Kernel Linux 4.4.13 - Debian Jessie 8.5
Acqua - System On Module
I tried long time ago with a JLSounds I2S board without success (I wasn't able to compile kernel). Looks like pre-built image for the board is: Kernel Linux 4.4.13 - Debian Jessie 8.5
Looks like you will need to do a lot of stuffs to make it work as a NAA: add ethernet jack, add usb jack, and upgrade your linux.
It is more expensive than RPi with less computational power.
IIRC, HQPlayer NAA requires higher version of libc, which means you will need to upgrade Jessie to Stretch to make it work.
Ahhh ... That's the tricky part
I did all hardware stuffs, but Linux upgrade for arm it's out of my knowledge of linux caveman.
Ahhh ... That's the tricky part
I did all hardware stuffs, but Linux upgrade for arm it's out of my knowledge of linux caveman.
Not that hard:
You will need to edit /etc/apt/sources.list
Change all jessie to stretch
apt update
apt dist-upgrade
Ten you will be good to go.
Has anybody tried this Amanero pin compatible XMOS interface from SingXer with DSC-2.x?
Singxer C-1 XMOS Digital Interface XU208 U8 Upgraded Femtosecond TCXO Audio | eBay
SingXer is fairly reputable. Will mute signal work well?
Singxer C-1 XMOS Digital Interface XU208 U8 Upgraded Femtosecond TCXO Audio | eBay
SingXer is fairly reputable. Will mute signal work well?
I tried xmos u208 with dsc2.6.2. At start i makes HUGE pop sound, so be carefull.
For sound quality to me Amanero is much better.
I've tried SingXer X1 with MUTE pin. Works well - no pop, stable at DSD256 (didn't test DSD512, but Pavel says it works well at DSD512)
As for me, it sounds a little diffrerent from Amanero (not worser or better).
Still prefer Advanced Amanero.
Thanks Vit...
I ask because the XMOS approaches typically play much more consistently with linux and high rate native dsd streams. I am listening to Amanero/Cronus on a Buffalo ESS 9038 DAC right now at DSD 256 with 1099c, and there are still some issues, and I guess 2004be has even some more issues, whereas XMOS in my experience has been perfect via linux.
Also the XOs on the SingXer C-1 are skosh better than the Crysteks, but the advanced Amanero looks to have a bit more sophisticated design in terms of power regulation. My main concern with the SingXer, is wanting to know if mute signaling worked as with Amanero, and that seems to be the case here.
I ask because the XMOS approaches typically play much more consistently with linux and high rate native dsd streams. I am listening to Amanero/Cronus on a Buffalo ESS 9038 DAC right now at DSD 256 with 1099c, and there are still some issues, and I guess 2004be has even some more issues, whereas XMOS in my experience has been perfect via linux.
Also the XOs on the SingXer C-1 are skosh better than the Crysteks, but the advanced Amanero looks to have a bit more sophisticated design in terms of power regulation. My main concern with the SingXer, is wanting to know if mute signaling worked as with Amanero, and that seems to be the case here.