Overall Isolation - network, USB, and power

[jabbr]

In this logic maybe it would be better to use into for isolation, then regen for clean power and reclocking and no isolation on the dac board ?

 

 

Sent from my iPhone using Computer Audiophile

 

Isolate I2S after USB input then FIFO buffer then reclock. See @Iancanada on DIYAudio

[jabbr]

So Ian's board with the boards necessary for the NO-DAC chip DAC is a good, up-to-date project to do?

I'd do Ian's board with DSC1

[jabbr]

Hacking an FMC to replace the power supply is very cool, though I'd consider:

 

1) direct optical SFP input.

2) if you are going to hack, why not replace the PSU on something like the Q1900M motherboard?

[jabbr]

Since Krtical's sabertooth jigglybits - sorry, I'm bad with names - was cancelled, what are our options?

 

Taking an SOC board with an FPGA and perhaps an ARM chip or HDL version and hacking the SFP input ourselves?

 

Solid-Run's Clearfog is an option.

 

I'm also looking at one of the Zedboard's with a carrier card that does SFP+ I/O but that's more of a development environment.

 

The Zynq might do essentially the whole kit and caboodle on a single chip ... long term could create a custom board

[jabbr]

I know it's Miska's design, but... too many Opamps - I'm heading for chip-less and a tube output

 

I have a discrete balanced design that only uses chips for the shift logic -- no opamps -- and you could create the filters using tubes

[jabbr]

Sounds good to me.

 

PCB designs are shared on OSHpark under "JAB.R"

[jabbr]

 

You can add :

 

- Optical isolation (optical ethernet or tosslink) is the only way to get perfect electrical isolation on signal (but it adds noise or jitter by itselves).

 

Optical isolation does not add noise nor jitter. Of course everything is implementation dependent but the eye-pattern diagrams from fiberoptic Ethernet typically exceed copper Ethernet. All Ethernet switches and FMCs necessarily reclock the Ethernet signals (part of the SERDES process). High quality fiber optic Ethernet equipment is made with particular attention to noise and jitter. 10gbe, 40gbe, 100gbe and up signals would not work otherwise!!

 

I agree that inexpensive battery supplies for FMCs are an excellent option. The lowest noise and jitter options use direct fiber Ethernet and avoid the FMC altogether.

[jabbr]

Optical Ethernet can go directly to either an SFP cage on the motherboard or to a PCIe card. No copper Ethernet receiver need exist.

[jabbr]

The high quality electronics in the SFP receiver (PHYS) have less power requirement and less noise than the RJ45 Ethernet PHYS and so this will have less noise (never claimed it would be zero just less and certainly not more than copper Ethernet)

 

 

Sent from my iPhone using Tapatalk

[jabbr]

Sound interesting, but can plug them elsewhere than in a PC ?

 

SFP modules are used in pro grade (and some consumer) switches, servers, certain pro grade audio products e.g. Ravenna

[jabbr]

[ATTACH=CONFIG]31453[/ATTACH]

It's a 2"x2"x2" cube, cost me $96. (It's called a Cubox. These mini-computers have such cute names.)

 

The same company (Solid-Run) makes something newer called the "ClearFog" which has an SFP input (and so can take direct fiberoptic Ethernet). Some folks are looking at whether it can run the NAA software (including myself though I have other projects). I bet it will be able to once the Linux support matures (fairly new Armada chip).

[jabbr]

[Emphasis added.]

 

Yup, interesting, though the lack (AFAIK) of current availability of the relevant Linux distro builds for this chip means this is a future consideration.

 

Well I will forgo my daily afternoon Don Perignon today and splunk for a ClearFog (ordered) -- let's see if this can work

[jabbr]

 

However with an NAA you have to effectively double the effort for a DIY solution as you now have two PC's to isolate and two OSes to tweak. You can do these things once and get the same or better results.

 

 

There are different ways to do thing. The way I isolate my PC & NAA is with fiber. Both have x520 cards -- the advantage is that my NAA - an Asrock Q1900M is also discless and the NIC handles the iSCSI boot. I don't do anything else special in the PC -- the NAA does have LPS.

[jabbr]

Hi Jabber, yes different ways for sure. I remember your configuration. You make my point.

 

Where is your music stored? On a NAS?

 

Well I've played around with a number of configurations, and much of my chain is hand built.

 

Music stored on NAS -- a Linux server that doubles as iSCSI target and Vbox host.

[jabbr]

Well I've played around with a number of configurations, and much of my chain is hand built.

 

Music stored on NAS -- a Linux server that doubles as iSCSI target and Vbox host.

 

At some point HQPlayer renderer might be separate Linux machine but for now the HQPlayer machine runs OS X and the OS is not at all optimized for audio.

[jabbr]

Curious if you had a chance to experiment with the ClearFog? I totally missed this item being mentioned, but this would seem to be an awesome product for an NAA!

 

I'm running fiber from switch to media converter then eth to mini. Would be cool to go direct sfp w/o media converter.

 

What are the issues/limitations as of right now (does there have to be a new image created as opposed to using the cubox naa image)

 

Ordered the ClearFog base version and waiting for it to arrive -- need to be sure a version of Linux runs on it that supports NAA/networkaudiod.

 

It is different than the CuBox. Uses a Marvell Armada A388 SoC chip.

[jabbr]

Fiber NICs need not be expensive -- that said our bandwidth usage is so low there is no need for two NICs. Really.

 

 

Sent from my iPhone using Tapatalk

[jabbr]

Also for clarification, the clearfog info states the SFP modules must must conform to "I2C fast mode" (assuming on both ends) - does this imply the "typical" 1GB SFP modules will not work?

 

I think typical SFP modules should use I2C at the copper end i.e. that how the module communicates with the board its plugged into. Don't see why that would be necessarily also on the other side of the fiberoptic cable. Probably would work with "typical" modules but I haven't yet had the opportunity to test.

[jabbr]

Batteries can be great but also tricky to work with. Supercaps have become dirt cheap. What is needed is a simple and inexpensive dual bank controller that auto switches between two banks of either batteries or supercaps and keeps the mains supply disconnected from the battery's or supercaps which are discharging. This project may be of interest: http://www.diyaudio.com/forums/power-supplies/302881-dual-bank-floating-supercap-supply.html

[jabbr]

Thank you.

With one 6A battery i have 7 hours charge so the switch between two battery is not really needed (at this time i load it two times a week at night with a cheap programmable switch). The relay switching noise migth be a problem with supercaps.

 

Trade offs. MOSFETs have less switching noise (certainly not the audible kind but have a very small leakage current, perhaps 10-100nA but usually <1 uA.

 

Supercaps less ESR than batteries but all these issues can be dealt with. Point being that great mains isolation is achievable with just a little bit of work.

 

 

Room treatments for headphone users

[jabbr]

Jabbr,

 

Thanks for the design and schematics over at DIYAudio. I have several places I could use this in my setups,, mostly for battery packs and less-critial gear

 

Is this a tested design and are boards available? Any pictures of finished projects using it.

 

Look for a good starting point on this.

 

This will be usable with the most critical gear -- nothing preventing a low impedance regulator such as the Salas Shunt, or chip of choice being used after this.

 

The other issue is that precise voltage regulation is often vastly overrated. Ripple can be removed with capacitance and filtering which a wideband cap multiplier can provide. This approach reduces/eliminates the need for AC mains cleanup/filtering etc.

 

Looking at 10nA leakage MOSFETs which should do the trick