Thanks Ray, this has been very useful. Looks like the back-end stuff is firming up nicely! I have just ordered the latest version of AudioLinux. Not sure if this has DSD256 support already.
I will also wait until the next generation Skylake-based SBCs hit the stores. Until then, I will see if my old tablet PC could run this OS with NAA
Last edited:
I'm not sure if DSD 256 is supported with the Amanero on Linux. You might have to try the Jlsounds interface. Some say it's better than the Amanero anyways.
There's gonna be a new Skylake I7 NUC coming out soon. I'm thinking that would make an awesome HQplayer server running Audiolinux in realtime mode. There's a way to optimize the realtime settings when Audiolinux is used for both server and client in NAA mode. It's supposed to be awesome.
Translate this to English:
http://www.nexthardware.com/forum/p...tivo-per-hqplayer-ma-non-solo.html#post930705
Last edited:
Here's a complete turnkey NUC system using pretty much the same SBC I'm using for my NAA. Power this with a good linear supply, disable everything unnecessary in the Bios, slap a cheap 4gb stick of ram in it, install Audiolinux on USB 3 stick, and you have an NAA that is out of this world. And all 100% Intel quality. This SBC in the box is built to industrial computer standards.
http://www.amazon.com/Intel-Thin-Canyon-NUC-DE3815TYKHE/dp/B00LPLQD04
On the server end, use the new Skylake I7 NUC when it comes out running Audiolinux and powered by Linear supply. The i5 is probably fast enough for that matter but I'd wait for the i7.
Here's info on the i3 and i5 versions. I'm expecting 1-2 months and we will see the i7 version:
http://www.digitaltrends.com/computing/intel-shows-off-four-new-nucs-sporting-skylake-hardware/
http://www.amazon.com/Intel-Thin-Canyon-NUC-DE3815TYKHE/dp/B00LPLQD04
On the server end, use the new Skylake I7 NUC when it comes out running Audiolinux and powered by Linear supply. The i5 is probably fast enough for that matter but I'd wait for the i7.
Here's info on the i3 and i5 versions. I'm expecting 1-2 months and we will see the i7 version:
http://www.digitaltrends.com/computing/intel-shows-off-four-new-nucs-sporting-skylake-hardware/
Last edited:
nautibuoy & bavmike,
Audiolinux does support DSD native (non-DoP) and DSD256. Perhaps one needs to update it.
On Linux Amanero supports only DoP.
JLSounds is superior to straight Amanero. I have a "special" Amanero with low noise xo, separate ground, isolation, separate PS. It's night and day compared to straight one, but I have not compared with new JLSounds.
Audiolinux does support DSD native (non-DoP) and DSD256. Perhaps one needs to update it.
On Linux Amanero supports only DoP.
JLSounds is superior to straight Amanero. I have a "special" Amanero with low noise xo, separate ground, isolation, separate PS. It's night and day compared to straight one, but I have not compared with new JLSounds.
Yes I'm listening to Native DSD 256 right now with Audiolinux. Sounds spectacular! Acko has a very nice reclocker board for the Amanero. But it still won't enable native DSD on Linux. Perhaps his reclocker will also work with the Jlsounds board.
Yes that's right. I suppose reclocking twice might not be all that beneficial. The best upgrade for the JL sounds board would be to have the NDK clocks measured beforehand to ensure your not getting duds on your board.
This is how about 1% on a reel of 500 measure. However up to 5% on the same reel can be 15db worse in 10hz phase noise.
An externally hosted image should be here but it was not working when we last tested it.
These are the 22 and 24Mhz clocks. One disadvantage of the reclocking on the Jlsounds board is 44 and 48Mhz clocks must be used in order to support DSD 256. They have much higher phase noise. But perhaps the benefit of reclocking outweighs this factor.
Ok, if with Linux and DSD256, I will have to abandon Amanero. Amanero stock version is built to sell so choice of parts accordingly to meet the price point that will appeal to many. Amanero- S03 reclocker is heaps better without any mods to Amanero.
Most likely, will use the JLsounds OEM board that has no clocks and this will go into the S04 reclocker with its own premium clocks plus options for connecting even better external clocks.
Yes, agree the 45/49Mhz NDK as used on the JLSounds USB has higher noise than the 22/24 types. I guess when you try to cover all bases PCM/DSD you will loose some advantages compared with some thing that is specifically build.
Good idea. Only the reclocked version uses the 45 and 49's. I heard you need twice the MHz when reclocking.
It is the clock that counts and all the output timing aligns to it. If you clock or reclock w.r.t. a noisy ref clock you get noisy outputs. Reclocking is just the means to tie the output to a better reference clock especially after going through an isolator.
Another reason for getting AudioLinux is to also test the case that after reclocking it may not make any difference to the SQ regardless whether Windows or hot-rod Linux
So the phase noise of the clocks on the USB board have no bearing on the end result, as long as the clocks on the reclocker board are premium?
I think there may be more than just the jitter factor that makes lighter weight OS's sound better. This is why even people with mega buck super low jitter DAC's can hear a difference with different source computers/ OS's and media software. Memory management as well.
Yes, looks like it! I 'lost' my 9018 test set to a Vinylphile friend after upgrading the S03 with Pulsar Clock. He said it is the closest sound to his top flight analog gear! Cheap Amanero back end. So I am now without any DAC, Christmas is coming and the natives are making noise! Hopefully Sonny's DAC to the rescue until the DSD DAC is ready
Needs further investigation but if something is streamed out bit-perfect (as claimed) exactly as what is stored on disk why would things other that jitter affect SQ?
I don't know all the details about every little thing that affects sound along the audio chain. I think the best way of isolating the influence of a general purpose computer from the DAC is via fibre optic Ethernet. But then you need a very low noise fibre to copper converter connected into a renderer device located inches from the DAC chip, that can spit out I2S/DSD direct.
I am under the understanding that jitter can't be introduced until the data gets clocked. If the data is clocked for the 1st time by the master clock of the DAC, the only jitter that should be introduced is the jitter of the master clock, and what's introduced in the short PCB trace between the master clock and DAC chip. In this situation, async or reclocking, shouldn't even be required. Just 1 very good master clock.
Last edited:
I was running the latest version of AudioLinux and whilst I don't deny it is native DSD capable, the kernel did not seem to have the updates to support the JLSounds board specifically. I checked with Piero and he replied that whilst the JLSounds patches had been submitted but hadn't made it into the Kernel being used by AudioLinux. This was late August.
Anyway, it is academic now as I have native DSD256 with my Debian Stretch core installation.
Ray
Chaps, I'm gonna stick my neck out!
There's been a lot of interesting discussions over the last few pages but isn't a lot of it largely conjecture and opinion at this stage, given that the direct drive DAC doesn't actually exist yet? I think an initial baseline needs to be established then iterative improvement/development can proceed in an objective manner.
There's no need to wait for the latest shiny SkyLake processors and the like. I'm running HQPlayer, upsampling from 44.1KHz PCM to DSD256 ,on a four year old Xeon processor-equipped workstation and it isn't being overly stressed, CPU loading is around 30%. The processor benchmarks at around 8000, so that gives an indication of the capability required. BTW, Xeon processors are server oriented and better at 'grunt' processing than the i5, i7 processors.
For HQPlayer NAA I'm using an even older Intel Atom processor/board that benchmarks at under 700 and it doesn't even break into a sweat; running Audiolinux and playing DSD128 CPU back along utilisation was around 5-7%. I'm confident that I could use something like a Beaglebone Black to put together a DSD256 capable NAA (assuming output to a USB board rather than the Botic approach).
https://www.cpubenchmark.net/
Ray
There's been a lot of interesting discussions over the last few pages but isn't a lot of it largely conjecture and opinion at this stage, given that the direct drive DAC doesn't actually exist yet? I think an initial baseline needs to be established then iterative improvement/development can proceed in an objective manner.
There's no need to wait for the latest shiny SkyLake processors and the like. I'm running HQPlayer, upsampling from 44.1KHz PCM to DSD256 ,on a four year old Xeon processor-equipped workstation and it isn't being overly stressed, CPU loading is around 30%. The processor benchmarks at around 8000, so that gives an indication of the capability required. BTW, Xeon processors are server oriented and better at 'grunt' processing than the i5, i7 processors.
For HQPlayer NAA I'm using an even older Intel Atom processor/board that benchmarks at under 700 and it doesn't even break into a sweat; running Audiolinux and playing DSD128 CPU back along utilisation was around 5-7%. I'm confident that I could use something like a Beaglebone Black to put together a DSD256 capable NAA (assuming output to a USB board rather than the Botic approach).
https://www.cpubenchmark.net/
Ray
Of course you don't need the latest Skylake processor computer for the task. But if your gonna buy something new, you can't beat the power and form factor of these units.
As far as for NAA, I have compared the sound of that little NUC 3815 board against 5 computers so far. Nothing comes close. And it's a solid and reliable unit as well with a 3 year warranty from Intel. Also fanless. If you know of a better fanless system for the application of running Audiolinux/NAA, Jriver etc for under $133, please share. I've been looking for a while. And of course it has to sound as good as well. I like the fact that it has both a USB 3.0 bus and 2.0 bus. This way the USB stick for the OS can be on a separate bus/controller chip than the USB audio output. You can also buy SD card adapters that can fit in the mini PCI-E slot on the board, if you don't want to use a USB stick. Then you also have the on board 4gb e-MMC for lightweight Linux installations. That's what I'll be trying next. I couldn't get the networkaudiod going on Debian Jessie from that drive, but I'll try again with stretch.
Last edited:
- Status
- Not open for further replies.