Tír Na HiFi

Hi Chaps

In terms of the sound - the instant improvement was dynamics and musical flow. Like going from Windows Media Player to MQN or JLP, only going from JLP to Volumio/RPi. Detail and ease of presentation are very much better. Bass is hugely more articulate and punchy. Hihats and cymbals are stunning. Vocals are clearer and more human sounding.

It just sounds like the music flows out of the dac rather than being forced through it.

Not a small change in any area and shows just how much muddle the whole pc/usb signal path introduces.

You could use USB, but I suspect that brings some of that muddle back and if you use the RPi and a USB drive for your music library you can hit issues with the hardware implementation on the RPi. I have a Beaglebone Black on order for next round of experimentation. Using the Rpi as server and Beaglebone to stream I2S seems like a viable plan.

jkeny wrote:Interesting, F
Can I ask why you chose Gentoo? I know Fran said that it is a tailored Linux that you only compile what you will need into the kernel but did you look at many of the other Linux kernels? There appears to be a vast range of kernels which are apparently optimised in different ways. For instance Linux micro-kernels seem to also be cut-down versions of the full OS & seem interesting. DaveF also mentioned real-time Linux (with some qualifiers) which seems interesting also.

Gentoo is not about the building customized Linux kernel. That you can do on any distribution and that would be either easy or little more complicated - depending on the distro. Gentoo is about building customized entire distribution (kernel + user space programs). It gives you a choice about everything and it is extremely easy to do that after you get few simple concepts. Of course the price is that it takes little bit longer as all software is built from source code.

My choice of Gentoo reaches couple of years back, when I started using it personally and at work. I found it very flexible and quite easy to maintain. The decision of using Gentoo for X2 came naturally. For approx 1.5 year I also heve been using ArchLinux at work. It is also very light and flexible. It's a binary distribution and it has freshes versions of software available. Not necessarily super stable, but I did not have big troubles with that so far.

In terms of Linux kernel - I've been always compiling it from sources. I also tried -rt patches for Linux in the past. I'd call it generally stable, but it crashed few times. I will try it on X2 as well.

One more thing: the term 'micro-kernel' has a very specific meaning. It describes the OS kernel architecture. Linux does not have such architecture - it is a monolithic beast.

Linux->I2S - I am not saying no, but I am little skeptical about that. I have not seen the Volumio source code, but I am quite curious how they guarantee low latencies at the kernel level. Personally I would use some small real-time OS here, but the trade-off is TCP/IP, USB, filesystems. All that stuff is in Linux for free.

F

Im going to try to repost some of the points I made yesterday just before the forum crash and it was all lost.

The general idea of this thread is to come up with a solution that moves away from the PC environment completely. Although the work done on MQN and the Comp Audio build thread have been very impressive, it would seem that a lot of the effort and tweaks made are trying to reduce the affects of noise within the PC environment. The sheer number of tweaks have put me off experiementing much with MQN and I'm sure I'd be left wondering if some of the tweaks are worthwhile or if some do anything at all.

So lets start from the ground up with a solution that eliminates the Windows bloated OS altogether. Real time OS's were mentioned yesterday such as VX Works, QNX and Real Time Linux but I question as to whether an OS is really needed at all.
If I were to hand this project(which is basically to get audio samples from an SSD/SD card etc, pack them into I2S format and send to a DAC) to the team here at work , I'd bet anything that no OS would be involved. A CPLD or FPGA would be more than up to the job with the added benefit of being able to control internal clocks within the FPGA to your liking and being able to generate, control timings/skew, current drive strengths on the I2S lines. A very low speed external crystal feeds the FPGA and it will generate its own internal clocks as required. In this way you can keep activity and power draw down compared to an MCU running an OS.

Having said all that, I'm not sure if an MCU with an OS would have much impact on noise anyway. In some of the boards here at work, we're controlling very very sensitive circuits that are designed to be ultra low noise in the first place. A combo of an MCU and FPGA are at work here. The software on the MCU changes all the time yet it never has any impact on the noise on the sensitive circuitry. So in a tightly controlled and optimised embedded solution, an MCU running software might be not be that big a deal.

The question is if an OS is really needed? Granted, an all-FPGA solution requires specific skills, experience and proper simulation tools to do the job properly.

Another thing touched on yesterday is the control interface from a user point of view. How is the user going to prep the music before playback or control the device whilst in operation. This is where software might come into play and give you plenty of options in this regard.

A softcore processor (ARM based) can also be embedded within an FPGA too so that's another option. Any kind of software be it an OS or a simple scheduler is always going to add another layer above the hardware. It will give a lot more flexibility but *might* impact on noise in the system. My experience to date in embedded systems says it wont but not impossible.

Another aspect is a nice proper PCB design with good layout techniques, isolating analog from the digital, good ground planes and using shielding over the power circuitry.

Anyway...just a few thoughts, not the most coherent post I've written but I'm at work so... :-)

I'm already out of my depth, but I happened across this elsewhere and I thought I might as well share the link here. No idea whether it's worth looking at or not, but it claims to be a Linux distro designed for audiophile playback first and foremost:

http://www.ap-linux.com/

Diapason wrote:I'm already out of my depth, but I happened across this elsewhere and I thought I might as well share the link here. No idea whether it's worth looking at or not, but it claims to be a Linux distro designed for audiophile playback first and foremost:

http://www.ap-linux.com/

Hi Diapason

Audiophile Linux works well enough but is ultimately an alternative to a Windows type environment, running on a x86 intel architecture.

What we are discussing here is dumping the PC and using a stripped down, cheap, low-power platform that allows (in some instances) direct access to I2S audio signals. The benefits here are that you don't introduce clock issues by converting to SPDIF and you don't have the packetisation/unpacking overhead of using a USB connection. It's cutting out many superfluous processes and drivers, and in my recent experience, it's impact is very beneficial.

LowOrbit wrote:

Diapason wrote:I'm already out of my depth, but I happened across this elsewhere and I thought I might as well share the link here. No idea whether it's worth looking at or not, but it claims to be a Linux distro designed for audiophile playback first and foremost:

http://www.ap-linux.com/

Hi Diapason

Audiophile Linux works well enough but is ultimately an alternative to a Windows type environment, running on a x86 intel architecture.

What we are discussing here is dumping the PC and using a stripped down, cheap, low-power platform that allows (in some instances) direct access to I2S audio signals. The benefits here are that you don't introduce clock issues by converting to SPDIF and you don't have the packetisation/unpacking overhead of using a USB connection. It's cutting out many superfluous processes and drivers, and in my recent experience, it's impact is very beneficial.

Any pictures
Mark

very tempted

LowOrbit wrote:Hi Chaps

In terms of the sound - the instant improvement was dynamics and musical flow. Like going from Windows Media Player to MQN or JLP, only going from JLP to Volumio/RPi. Detail and ease of presentation are very much better. Bass is hugely more articulate and punchy. Hihats and cymbals are stunning. Vocals are clearer and more human sounding.

It just sounds like the music flows out of the dac rather than being forced through it.

Not a small change in any area and shows just how much muddle the whole pc/usb signal path introduces.

You could use USB, but I suspect that brings some of that muddle back and if you use the RPi and a USB drive for your music library you can hit issues with the hardware implementation on the RPi. I have a Beaglebone Black on order for next round of experimentation. Using the Rpi as server and Beaglebone to stream I2S seems like a viable plan.

Very heartening to hear that it is such an improvement in SQ
Are you running the RPi from a Linear Supply?

john
is there many switchers on the rpi

farnell has beagleboards

Thanks DaveF - I agree that an FPGA looks like the most optimal solution but also requires some amount of expertise & experience to achieve results.
RPi (or similar) is a good first evaluation step along this path & seems to be a worthwhile step.

nige2000 wrote:john
is there many switchers on the rpi

I'm not sure there are any - I think they are linear regs? Mark could probably answer that?

farnell has beagleboards

Really? I thought they were sold out everywhere? No, Nige just checked & for Ireland it says awaiting stock