Building a DIY Music Server

[MarcelNL]

1 hour ago, di-fi said:

My understanding is that with consistent higher temperatures at the long term (months) MLC and TLC chips get errors (to be verified @ what temps) that would not happen to pure SLC.
Also after all the efforts to get passive CPU cooling at around 40 degrees, you do not want a little heater of 50-60+ degrees next to it. A heatsink seems beneficial to lower the SSD temperature but this hot air needs to get evacuated from the case also.

I was picking my battles, the NVME is not drawing that much power that it'll affect my passive cooling anytime soon (300W heatsink)

The femto nvme gets up to ~45'C, the storage NVME to 50'C or so (that is using the 'heatsink' provided with the MB...the femto drive does not fit most regular heatsinks.)

 

Battleground for the future for sure, I'm more keen on a permanent solution for the ultra low jitter Solarflare 2522 that gets REAL hot...

 

[MarcelNL]

22 minutes ago, di-fi said:

I understand, but I was talking about the NVMe and CPU as heat source themselves, not how much power they use. My NVMe heatsink has on average 10 °C higher readings than the CPU heatsink (measured on the top surface) therefore  an important heatsource. CPU heatsink evacuates heat out of the case through 6 heatpipes (FC5). The NVMe heatsink (or not) can only exchange heat with the air stuck in the case.

But we can close the subject :-), it’s an observation.

The power it consumes is all the NMVE SSD can convert into heat, so if the NVME SSD only draws a few watt it will not generate enough heat energy to affect a well proportioned heat sink of the CPU in a major way even when the surface temp of the NVME is higher than that of the CPU heatsink. It's all relative, that said, the NVME should indeed be prevented from overheating. 

 

[MarcelNL]

1 hour ago, Marcin_gps said:

 

I compared to Apacer SLC which killed it. There is no point in using a scale in this case

Do you care to specify which one? All I did find was Pesudo SLC called Lite-x using 3D TLC..just curious.

[MarcelNL]

1 hour ago, Gavin1977 said:

Actually built by Troels himself - they’re very good, but I would like something larger now so they’re up on eBay.

If you go big go BIG and high efficiency ;-)

 

 

[MarcelNL]

I can report that the Taiko rectifier and 220.000uF Mundorf plus new topology sound very good, had to tinker with some bypasses, the Vit Q I had on the first cap appears as essential as the bespoke bypass caps on the EPS.

In hindsight I do not think the common mode rejection circuit I was using had anything to do with the issue of sagging voltage, and I will go back to it to evaluate the merits of both circuits.

 

Stupidly good is a great summary!

 

Still, does anyone know how the ATX module is switched on? I want to test two of those buggers but suspect I may need to rig something up to fire them up in sync for the MB to boot,

 

 

[MarcelNL]

as a switch I presume?

[MarcelNL]

How are people's experiences with breakin in the 220.000uF Mundorf HC and Taiko active rectifier?

 

I'm starting to think 200 hours is just the beginning, so far I like what I'm hearing but am curious to hear what others think....

 

Lots of low level detail, also in the low end, nice flow in rythms, great separation between musical lines, no harsh top end etc...

The great thing is that my new amp (single EL519 providing a full 2 Watt) does much the same in it's early days, low end that is seemingly without limits and dynamics to die for coupled with detail and sweetness. Cannot wait to join the two...

 

[MarcelNL]

7 minutes ago, ciccio1112 said:

Hi, Marcel, Have you got to compare the saligny power and the Taiko rectifier?

I did not, I had to rebuild the LPSU as I thought the ATX was defect...turned out a molex pin had broken in the ATX loom....

At the moment I'm breaking in the new LPSU but probably will go back to compare the rectifiers and chokes in common mode once everything is stable.

 

My initial impression is that low end is better defined, has more extension and there is less 'noise', however there are three changes at one time making it impossible to attribute the improvement to anything.

[MarcelNL]

I find that the bypass caps are all important, wildly exotic stuf gives great results.

I'm now at 300hrs break in, and SQ is still becoming more detailed and refined, I explain the apparent lesser low end but better definition exactly with that...more definition makes the low end appear less voluminous.

 

@Tatomek7 I can recommend the Saligny power, but that only makes sense if the wait for the Taiko rectifier is loooong..

[MarcelNL]

perhaps someone should ask Saligny to create an 'Audio Power' version, the regular power recitifier has connectors that are a bit dodgy to work with.

[MarcelNL]

@Nenon allow me a short summary; EVERYTHING matters ;-)

 

(completely agree with what you wrote, and there is even more esoteric stuff that matters)

[MarcelNL]

I should read up on power factor correction, what I know now is apparently not enough since it's pretty much similar to this ;

 

Power factor correction is the process of compensating for the lagging current by creating a leading current by connecting capacitors to the supply. A sufficient capacitance is connected so that the power factor is adjusted to be as close to unity as possible.

 

I have been 'correcting' PF by adding some exotic capacitors to the 230Volt side of things, with success, but addinf enough capacitance in a cable of which nobody knows how much any user will be requiring sounds like voodoo to me

 

Anyone with a hint to reading material?

[MarcelNL]

On 10/8/2022 at 5:17 PM, alsterfan said:

@MarcelNL

Mark Cole, Sales and Support Manager @ Antipodes Audio wrote:
30.08.22 "Indeed the Samsung PM893 SSD is the new standard for best playback from local files."

and: "The PM893 has been tested at our facility at proven to be both very reliable and the best sounding we have used to date."

04.09.22 "Hi there, we have compared the QVO, EVO and PM SSD’s in our listening room by switching between OLADRA’s.
The QVO and EVO are still great, the best out there, until we heard the PM.
We can only advise our findings, your ears, your room, your server, your source material may yield a different result."

https://antipodes.support/t/ssd-recommendation-for-new-k-models/359/57

 

Best,

Uwe

thanks for this info, I tried to read up on the SSD disc, but I'm not sure if ANtipodes is seriously using SSD using SATA?

My experience is that SATA SSD, on an external linear PSU sounds worse than NVME in an M.2 slot without linear supply.

[MarcelNL]

I'm underwhelmed by the components of most music streamers/servers, sound being OK is not goal, I want to be on the cutting edge of digital. For that you need to take a few more steps and invest more $$ than what most manufacturers seem to be willing.

[MarcelNL]

all three appear to be a sort of INtel NUC type of mini computer, with a basic linear PSU.

I started off with a laptop, changed to a laptop running Daphile (headless), swapped that for a NUC on a Linear PSU and the improvements in sound quality were incremental and noticeable. That lead to the full on computer build, which went pretty much the same escalating CPU power and PSU brings serious improvements.

[MarcelNL]

that may well be an SMPS, power requirement gets in the way of a linear PSU pretty rapidly...the Taiko ATX (SMPS) does a wonderful job that I do not expect to be trumped by a true linear PSU

[MarcelNL]

2 hours ago, di-fi said:

 

 

From Antipodes website:

Three separate processing engines for each of Server apps, Player apps and Reclocking.

 

Unique, innovative high-capacity cascaded switched/linear power supplies. Featuring galvanically-isolated super-capacitor-smoothed oven-controlled reclocking of synchronous outputs.

 

User installable SSD storage, one-click to select from the world’s best playback solutions, with a full range of digital audio outputs.

 

 

Antipodes minimises the trade-offs between clocking, noise and bandwidth by:

• designing its circuits from the bottom-up for low noise, which reduces the need for noise filters
• designing very fast low-noise power supplies to support the bandwidth requirements
• designing around optimisation of a four step process.

 

https://antipodes.audio/design/#digital_audio

 

_ ----------- _

 

It's not about who is doing things better, hopefully their design approach will also inspire some here. It did for me.

 Sounds to me like a tech approach I'd expect from an EE, who did not listen to anything build other than to confirm it works. Just read up on how f.e. Taiko approaches development, it should IMO be a mix of design and thoroughly auditioning components and any cahnges in the system  in order to understand what makes digital sing.

[MarcelNL]

Every digital data trasport system comes down to implementation, I'm sure USB can sound as good as I2S or AES but it totally depends on how it is implemented at both ends of the chain. In my system I2S is the clear winner, also; using a different cable has far less impact using I2S than USB or SPDIF in my system. Testing by ear is IMO the best way for any given system.

[MarcelNL]

to do the full breadth of voltages a MB requires using a linear PSU is difficult, and expensive.

IMO the Taiko ATX is the best thing since sliced bread, add a good unregulated linear DC supply and active rectifier like the Saligny Power and you're done...it's a costly exercise but IMHO you'll struggle to improve upon that.

[MarcelNL]

looks OK to me!, keep inmind that the female square molex pin slots around a round male pin the orientation of the female pin makes no difference,

[MarcelNL]

I'm using 14 AWG Neotech solid core, it can double as suspension for the ATX module.

bending should be done with the cables not connected and you need to allow for extra length to make a turn without losing lots of wire length as the bending radius is large.

[MarcelNL]

suggest to stick to cotton or silk, pref both. Additional benefit is that you can buy ready made 'sleeves' to wrap yourself made of those materials,

 

[MarcelNL]

1 minute ago, di-fi said:

In order of performance:

1)best is a mix of both on bare wire?

2) silk OR cotton on bare wire

3) teflon/ptfe

4) pvc

basically a mix on bare wire is best IMHO,

 

order of preference and practicality;

 

silk/cotton mix

silk

cotton

 

gap

 

PVC

 

silk has the added benefit of adding some EMI shielding, it was not without reason much used in vintage wire.

It's a pain to do but for short stretches like the ATX and EPC well worth it even when there are many wires.

Cotton wire sleeve can be bought readymade if I'm not mistaken.

Forgive me but I'll not respond to the inevitable dsicussion that will probably go like ; bollocks, PTFE is best because it's dielectric constant is highest etc.. ;-)

 

 

 

loong gap

 

PTFE

;-)

[MarcelNL]

49 minutes ago, di-fi said:

Correct, I bought ready made cotton tubes/piping. https://www.hificollective.co.uk/search/site?text=cotton

 

And I just see you can also buy silk tubing: https://www.hificollective.co.uk/search/site?text=silk

 

Just asking your advise, based on your experience. plus it is super easy to change tubes and compare, and I will!

Great!

 

PS for safety reasons; if you twist the wires with silk or cotton is (gap+)PVC or (gap+)PTFE essential? 

I do not use the sleeves, but wrap manually using thin strips of silk, PAINFUL work, so please check but I suspect it will not be needed...vintage silk wrapped tinned copper wire was used in HT for tube amps. DO check continuity between the wires after some wiggling. 

 

[MarcelNL]

14 minutes ago, di-fi said:

Imagine the bad luck if a tiny area is exposed on one cable and the same happens at exactly the same spot on the second cable and  creates a short. 0.0001% chance I would say. Classic recipes will last. 

 

Less hardcore but probably someone here tried readymade Jupiter silver or copper wire in silk sleeving. 

The Jupiter wire is nice, I prefer UPOCC, strip it bare and wrap in a.o. silk, again...lotta work

Never had a short, but it can happen, same as having a Molex pin break...(had that happen)