Building a DIY Music Server

[Nenon]

23 minutes ago, Gavin1977 said:

Hi @Nenon - that is very timely feedback. 

 

So I'm using the HDPlex 800w DC-DC, with the same original silver cables, for troubleshooting, since this worked with the Nenon V3 and these components perfectly fine for the past year (my problems started and have remained since the Taiko ATX was inserted into my system).

 

Components in use: Asus Z390 Apex, 8086k, Apacer ram, Optane H10.

 

Step 1: Change GPU

Since bios indicated A2 and white light (GPU fault).  I thought this would cure my issue.  So I just tried a different GPU... Sapphire Radeon HD6450 1GB GDDR3 Low profile (<30w).  Cheap as chips off eBay and it's the one Pink Faun use in their server... its a good choice and much easier to work with.  Anyway...  I insert this and A2 bios error persists - but now with green LED light (SATA) rather than white (GPU). 

 

Step 2: HD6450 + switch out SSD

So I switch out the m2 SSD... Optane H10 for a H20 I have.  A2 error remains. green LED light.

 

Step 3: HD6450, plus new Optane H20 AND switch to powering HDPlex DC-DC using the HDPlex SMPS I used to use.

A2 error remains. green LED light.  CMOS reset doesn't help.  Switching to alternative bios doesn't help.  Now if I'd fried the choke by pulling too much current then you'd have thought that switching to an SMPS would have fixed it.

 

This build always ran pretty cool, so I can't believe it'd be motherboard VRM related (failed cap for example)...

 

I've built many PCs in my time, but this one has me stumped.  No logical reason why the issue occured in the first place, or why is persists.

 

I can understand your logic on power supply, I do have the Taiko Rectifier and cap ready to be installed, and was planning to upgrade the Hammond... but obviously can't progress with that currently.

 

Next step remove DC-DC and try with regular 500w ATX power supply (which I'd have to buy, as I sold my Corsair SFX a long time ago)...?

 

Could a failed/saturated choke cause excessive ripple to enter DC-DC converter and cause damage to both DC-DC converter and motherboard...

 

Hard to tell what happened and how. It's unlikely that sagged voltage / triggering protection would cause problems. The protection is in place to prevent problems, not to cause one. 

 

I received a call from a friend recently to tell me that his tweeter stopped working (exactly) when his wife started the microwave. It turned out one of the wires inside his speaker wasn't soldered properly and that had nothing to do with his microwave. But sometimes weird things can happen. 

 

As with anything else, the key to good troubleshooting is to isolate exactly where the issue is. If your problem is happening with the HDPlex DC-DC ATX and the Taiko DC-DC ATX, it does make sense to try with a regular computer ATX. And from there, just keep changing components until the issue goes away and you determine what components are causing it. I know I am not saying anything you don't already know, but that's the best advice I can give you short of having your server and going over the same troubleshooting steps myself. 

[littlej0e]

55 minutes ago, di-fi said:

Hi Alex (🍳)🥚🥚🥚,

 

So replacing a fibre optic cable with a Copper Twinax SFP Cable (or DAC cable) between ER and a Fiber network card in the server could improve sound due to less power needed resulting in the lowest possible noise?

 

I noticed some DAC cables are made for a specific brand but there are also general DAC cables. Anything to avoid for the ER?
You use 7m because the ER is next to the router I suppose? My ER is next to my server so 1M should do.

 

But I will have to give up both selected SFP (Finisar FTLF1318) modules on each side.

 

 

Also when using A and B side in ER there is still galvanic isolation between ER and server.
 

Please let us know your results.

Thanks,

Paul 

 

FYI, I tumbled down a similar rabbit hole a while back. The lowest power consumption + galvanic isolation always tended to yield the best results. These fiber cards consume a maximum of 1.5w (make sure you get version 2): https://www.trendnet.com/store/products/product-detail?prod=310_TEG-10GECSFP&status=view 

 

You can probably find them even cheaper from Amazon.

 

 

[Exocer]

I’ve had an instance where the first pci slot on my motherboard stopped working after a bout of “sparks”. One of my diy power cables shorted due to an oversight on my part, and that was the result.

 

To my surprise, swapping motherboards didnt fix the issue. So I ended up with an extra motherboard 🤣. It was incredibly perplexing to me. 
 

So, I installed my spare i7 8700t in place of the cpu that was installed at the time and the slot worked on both the new and old motherboards. Go-figure, the cpu itself was the issue. I had never experienced anything like this. Had I swapped CPUs earlier on I may have saved myself a couple of hundred bucks.

 

Luckily Intel RMA’d the i9 9900k and refunded full retail price. I was able to find a replacement for less…

 

Moral of the story, check every part of the build. You may be shocked at the cause 😎.

[MarcelNL]

Agree, check EVERYTHING, i ended up with two Taiko ATX modules....

 

Not a bad tbing as i plan to put both to use, I need to know what it brings...

[Dev]

2 hours ago, Nenon said:

 

 

 

Discussed with Emile this morning. The root cause of this is the ULPS not being to deliver all the current the motherboard needs at cold boot. This is most likely caused by the 5A Hammond choke. You need a bigger choke and perhaps more capacitance (although if you use the Taiko cap, capacitance is probably fine but that also depends on the choke you pick). Higher voltage also helps as long as it does not exceed the Taiko DC-DC ATX specs. 

 

I was pretty sure that was the issue, but when I asked what CPUs you are using and @rico25 said Ryzen 7 3700X, that confused me. My guess is there are some peripherals/components in addition the the Ryzen that draw power. Maybe a GPU that draws less power is all @rico25 needs?

 

When the motherboard powers on it draws too much too power. That saturates the ULPS (5A Hammond choke in this case), and the voltage sags. When the voltage sags, the Taiko DC-DC ATX turns off to protect itself. That usually happens so quickly that's nearly impossible to see with a multimeter connected to the ULPS. Then you have to wait till everything resets...

 

Why that did not happen with the HDPlex DC-DC ATX for @Gavin1977? Most likely the protection circuit on the HDPlex is not as sensitive as the one on the Taiko DC-DC ATX. I managed to fry the first prototype of the DC-DC ATX Taiko sent out to test (it was fried about 2 hours after Fedex delivered it), and that forced Emile to adjust the protection to make the unit more durable / DIY friendly. 

 

So, the solution is to use a more powerful choke (and potentially add more capacitance). I have an old post about some options here: 

 

 

 

I have some interesting (but confusing to me at least) data for this. I now have two ULPS - one built sometime back by SoulAnalogue using a custom hand wound large 10A choke. This is further modified with Saligny HC and other things. The other ULPS is the v3 version that I completed (roughly put together) this past weekend using Taiko parts - transformer, rectifier, monster caps and the Hammond 5A choke. I had these parts for a while but just couldn't find the time to put it together.

 

 

 

 

 

 

Interestingly both has the same issue of the server not powering immediately on power on but with a difference. I am still not sure that its an sole issue with the ULPS alone. I noted earlier that if you jump the PS_ON pins on the Taiko DC-ATX, the server will boot right away from cold.

 

The 10A choke version needs more time to warm up while the v3 version take less. I don't know what cap values are being used in SA ulps. I would need to disconnect and take a reading as the values are not printed on the cap itself. Maybe the 10A choke with larger cap might solve the issue.

 

 

2 hours ago, Nenon said:

However, when it comes to higher current chokes, there might be some better options than the Hammond. 

@MarcelNL  has some locally made chokes that were reported to be exceptionally good. I haven't tried them, but that would be my first choice to try, and I may do in the near future.

@Exocer had a great feedback with a Mundorf MCoil ZeroOhm inductor, but I am not sure what the max rating on this would be. These were designed for speaker crossovers but Mundorf tested them for @Exocer and confirmed they are good as a 5A replacement. I just don't know if they are capable to go beyond 5A safely. 

 

I have been thinking of getting the Mundorf MCoil ZeroOhm inductor for the ulps v3 as well. Can you also check with Emile what choke spec be recommends ?  Can two 10A chokes work with ulps v3 spec using Taiko monster cap ?

 

 

[MarcelNL]

Using a Toroidy 500VA transformer and two Lundahl ll2273 chokes(off the top of my mind) I never had to wait for booting using an AMD R9 CPU .

These chokes are capable of 10A., I had them in common mode and in parallel, no difference

[Nenon]

20 minutes ago, Dev said:

 

 

 

I have some interesting (but confusing to me at least) data for this. I now have two ULPS - one built sometime back by SoulAnalogue using a custom hand wound large 10A choke. This is further modified with Saligny HC and other things. The other ULPS is the v3 version that I completed (roughly put together) this past weekend using Taiko parts - transformer, rectifier, monster caps and the Hammond 5A choke. I had these parts for a while but just couldn't find the time to put it together.

 

 

 

 

 

 

Interestingly both has the same issue of the server not powering immediately on power on but with a difference. I am still not sure that its an sole issue with the ULPS alone. I noted earlier that if you jump the PS_ON pins on the Taiko DC-ATX, the server will boot right away from cold.

 

The 10A choke version needs more time to warm up while the v3 version take less. I don't know what cap values are being used in SA ulps. I would need to disconnect and take a reading as the values are not printed on the cap itself. Maybe the 10A choke with larger cap might solve the issue.

 

 

 

I have been thinking of getting the Mundorf MCoil ZeroOhm inductor for the ulps v3 as well. Can you also check with Emile what choke spec be recommends ?  Can two 10A chokes work with ulps v3 spec using Taiko monster cap ?

 

 

 

@Dev - the first time I heard about this issue was by someone who used the same SoulAnalogue's design that you have. That was switched to a v2 ULPS with a 5A Hammond choke, and that did not fix the issue, but just like in your case the behavior changed. So far, that's exactly what has happened to you, so no surprise there. We never tried a v2 or a v3 with a higher current choke on his computer. 

 

However, someone else with the exact same CPU (and similar motherboard) picked up the experiment. He could not boot with the 5A choke at all until we changed some BIOS settings (to reduce the power consumption). Then he experienced the several delayed boot issue. Upgrading the chokes to 10A chokes fixed all his issues. 

 

While I can't speak of SA's design (if you ask my opinion it does not have enough capacitance), I am confident that replacing the choke on the v2 / v3 design and adjusting the capacitance accordingly will resolve the slow boot issue. 

 

The only reason I was not so confident earlier was because I could not understand why the Ryzen 7 3700X was running into this (VGA card drawing too much power?) and I also could not understand why the same ULPS acted differently with the HDPlex. But after chatting with Emile today, I am confident it's all ULPS voltage sagging related. 

 

It's a fairly simple experiment to unbolt the terminal rings on one or two chokes and connect the higher current choke(s) instead. I am 99% sure that will solve the issue of slow boot! 

 

BTW, once you upgrade the chokes, the transformer will become the bottleneck. But the Noratel or 400VA Toroidy transformers should be fine with most of the hardware used here. After all, the v2 and v3 design was for lower TDP CPUs. If you need more power or you do heavy upsampling, then a 500-600VA with 10A-15A chokes will be the way to go. That would not sound the same as the v2 and v3 I have provided, and so some experimentation with parts might be a good idea. 

[Dev]

9 minutes ago, Nenon said:

But after chatting with Emile today, I am confident it's all ULPS voltage sagging related. 

 

If its the ULPS voltage sagging issue, how does it explain the fact that if you jump the PS_ON it boots right off the cold ?

[Nenon]

23 minutes ago, Dev said:

 

If its the ULPS voltage sagging issue, how does it explain the fact that if you jump the PS_ON it boots right off the cold ?

 

That's a question for Taiko as it's related to exactly how the protection circuit on the Taiko DC-DC ATX works and exactly what changes when you use the jumper. I don't have any documentation of the programmed logic and can't answer this.

The bottom line is that if the voltage on the ULPS does not sag, the Taiko DC-DC ATX will not turn itself off, and you would not be in this situation. 

[Dev]

28 minutes ago, Nenon said:

That's a question for Taiko as it's related to exactly how the protection circuit on the Taiko DC-DC ATX works and exactly what changes when you use the jumper.

 

its quiet possible that the protection circuit doesn't kick in when PS_ON jumper is on. However, the motherboard still might be happy with the sag and keep continuing the boot.

 

The 10A choke didn't work for me and I am not sure if its due to less capacitance (or why more capacitance is needed in cold boot but not otherwise) in the SA ulps. I think it might be worthwhile to try a 10A choke in the v3 version for those who are unhappy with this issue.

 

Anyway, I am not too concerned about the delayed boot issue as I hardly turn off the server. What matters is how it sounds at the end 🙂

[di-fi]

On 1/11/2023 at 3:11 PM, Exocer said:

You just reminded me, there were a few I did not try yet! Will report back as well.

 

These would be between my individual ER and Server.

Yeah look forward to hear back. You tried these two years ago? So you must have made up your mind?

I understand a longer passive DAC cable at a certain length (7m?) has to be replaced by an active DAC to make up for losses (audible or not). 

A look inside: nothing in the signal path, a chip on the bottom of the board.

 

 

[Dev]

I am curious, how many of you ordered the Taiko chassis in this batch ? I am still on the edge as I am unsure how much of improvement I would get from a i9-10900k/Asus ROG to a dual Xeon Silver/Asus Sage. Anyone changed setup like this and can comment on the sound quality ?

[Exocer]

16 minutes ago, di-fi said:

Yeah, look forward to hear back. You tried these two years ago? So you must have made up your mind?

I understand a longer passive DAC cable at a certain length (7m?) has to be replaced by an active DAC to make up for losses (audible or not). 

A look inside, nothing in the signal path. Only a chip on the bottom of the board.

I had them sitting around for two years, tried another one that I did not source from Amazon, and lost patience with it as I just wanted to enjoy my listening. I didn’t try either of the cables pictured above. They are brand new in package. Will definitely try them at some point and update this thread.

[di-fi]

25 minutes ago, Exocer said:

I had them sitting around for two years, tried another one that I did not source from Amazon, and lost patience with it as I just wanted to enjoy my listening. I did try either of the cables pictured above. They are brand new in package. Will definitely try them at some point and update this thread.

I quickly tried one last Friday, but had other issues. It did not seem to connect or possibly only at 100 Mpbs? So I wonder if every DAC cable will work.
Would be great to ditch both SFP modules since they have their “colour” and get a gain in sound quality. Even though at Taiko processes are now going through a similar (?) DAC cable in their (new) switch instead of being handled by the server to establish a gain. (and they recommended RJ45 from router to switch and SFP DAC twinax copper cable from switch to server).That’s not exactly happening here, it’s only a cable change instead. You never know. Will try again later.

[lwr]

53 minutes ago, Dev said:

I am curious, how many of you ordered the Taiko chassis in this batch ? I am still on the edge as I am unsure how much of improvement I would get from a i9-10900k/Asus ROG to a dual Xeon Silver/Asus Sage. Anyone changed setup like this and can comment on the sound quality ?

I have built a i9-20900k/Asus ROG with the Nenon v3 LPS and Taiko DC/ATX, Apacer industrial RAM, all Mundorf Angelique wiring, and JCAT XE network usb, and Femto M.2 memory cards.  The SQ is quite superb in every aspect, and I do no have any sense whatsoever that it could be any better in any way, until I then substitute the dual Xeon/ASUS Sage that I bought from Nenon.

 

The dual Xenon/Sage is the DIY Extreme that Nenon built for himself and was documented extensively, earlier in this thread.  I have moved it into a Taiko DIY chassis, powered by the latest iteration of the Nenon v2 LPS with Taiko rectifier board, Taiko sourced Mundorf caps, Taiko DC/ATX, and all Mundorf Angelique wiring.  At this point I am waiting on Taiko USB and network cards, and until they arrive, I am simply using the USB and network connections on the Sage motherboard itself.

 

Even though the i9-10900k server has the distinct advantages of the JCAT XE  network, USB, and Femto M.2 memory cards, the SQ of the Xenon/Sage is remarkably more natural, broader, deeper, and more convincingly realistic in every aspect.  It is not a night and day difference in any single aspect, but it is a slight but unmistakable further refinement of every aspect of SQ, which all adds up to a much better overall musical experience.

 

It is my understanding that the eventual inclusion of the Taiko USB and network cards will take the DIY Extreme to remarkably further refinements of SQ.

 

I can’t even imagine how superb a factory built Taiko Extreme with its USB, Network, and Taiko’s own operating system must sound.  Someday I would love to hear one in my system.

[Dev]

Thanks @lwrthat’s exactly the feedback I was looking for. What are you using to power the JCAT cards ? What are you using as OS and music player ? Also if I understand correctly, you are using the native Ethernet and USB out of the Asus Sage ? What prevented you from not using the JCAT cards in this build ?

 

did you preorder the Taiko Network card ? 

[lwr]

10 minutes ago, Dev said:

Thanks @lwrthat’s exactly the feedback I was looking for. What are you using to power the JCAT cards ? What are you using as OS and music player ? Also if I understand correctly, you are using the native Ethernet and USB out of the Asus Sage ? What prevented you from not using the JCAT cards in this build ?

 

did you preorder the Taiko Network card ? 

I power the JCAT cards with the Optimo LPS.  
 

I use the Euphony / Stylus v4 beta on the i9-10900k. The DIY Extreme has the tweaked Windows Server / HQPlayer that Nenon installed.  It is quite possible that the different OS / Player systems contribute to the differences I am hearing between the two servers, but I don’t think so.  I expect that the use of Euphony / Stylus on the i9-10900k is to its advantage, as it is a much more compact and efficient combination.  On the other hand, the DIY Extreme has roughly twice as many resources available for the relatively less compact and less efficient Windows Server / HQPlayer.  

I can’t comment on the SQ of Euphony / Stylus running on a DIY Extreme, but I believe others may be able to do so.

 

I had the JCAT cards in the DIY Extreme, but like others have reported here, after I made some modifications to the LPS, I began having boot delay issues.  My solution was to simply remove the JCAT cards and wait to deal with it further when the Taiko USB and network cards arrive.  There are USB and network ports on the Sage motherboard that are very very good, whereas moving the JCAT cards to the i9-10900 server produced a very noticeable improvement in SQ, especially in PRAT.

[Rovo]

see the picture what I use to power a HQPlayer upsampling PC. The PC mainly consists of an Intel 10900K CPU and a RTX A4000 16GB GPU. Power draw when in use approximately 180W. 

 

This is the setup:

22000uF - Hammond 195J10 - 22000uF - Hammond 195J10 - 22000uF+220000uF

 

I have no problems with startup even from cold (swithing on the power and after a couple of seconds pushing the button to start up the motherboard). 

 

In the reply I got from Mundorf it is mentioned that based on the winding  - dimension incl. the core facts the VN180 and VN300 will be loadable at 5A (I requested no more than 5A!). And they also mentioned 4A are possible for each qmm. 

 

The people having the problem with the cold startup of the Taiko DC-ATX, I have the impression the GPU is the problem. I specifically selected the RTX A4000 16GB because of the lower power draw. 

 

Robert

 

 

[Tatomek7]

I have Asus Sage with JCat Femto LAN and JCat USB powered externally by SJ DC3 & DC4. Serwer is powered by Nenon’s ULPS and Taiko ATX.

 

I went trough ULPS V2 -> V3 -> V2 with Hammond, Ogonowski and SoulAnalogue (SA) chokes with different total capacitance applied, but rather on the bigger side ( min. 130K uF with 5A and min. 170K uF plus with 10A choke).

Hammmond and Ogonowski chokes are rated with 5A and SA 10A. Currently I have Mundorf HC 22 000uF, SA choke and Taiko monster cap. I never had any problem with booting computer.

 

Recently I’ve changed the Windows power management settings and immediately computer started to behave differently -> slow down booting / repeated booting  and problems with loading OS, so I returned to original power management Windows settings. Also changing settings in BIOS ( CPU & RAM setting fix voltage ) had in my case an impact on the way, how computer was booting, so I returned to automatic settings.

 

My second server is based on AsRock Taichi X99 Mobo with strong 10 core Xeon and also no problems with booting.

 

Booting problems can be also caused by too tight screwed CPU.

[Exocer]

4 hours ago, Rovo said:

on the winding  - dimension incl. the core facts the VN180 and VN300 will be loadable at 5A (I requested no more than 5A!). And they also mentioned 4A are possible for each qmm. 

Thanks for confirming. My previous post is incorrect. For those who want a Mundorf choke with support for more than 5A, please reach out to Mundorf directly and get their take. 

[Rovo]

From the 4A for each qmm (quadratic/squared mm) I take it that the VN180-10 (1,8mm diameter wire) and definitely the VN300-10 (3mm diameter wire) can be loaded to more than 5A. But best to get confirmation directly from Mundorf if you need more than 5A. 

[MarcelNL]

Core saturation is another limitation

 

 

 

[Nenon]

Keep in mind that Inductors behave differently with AC and DC. The Mundorf chokes are designed for speaker crossovers (i.e. AC). We use them for DC.  

@MarcelNL - is your guy willing to make some more chokes? I heard excellent feedback about them, and they might be the best option for ULPS exceeding 5A.

[Exocer]

15 minutes ago, Nenon said:

Keep in mind that Inductors behave differently with AC and DC. The Mundorf chokes are designed for speaker crossovers (i.e. AC). We use them for DC.  

@MarcelNL - is your guy willing to make some more chokes? I heard excellent feedback about them, and they might be the best option for ULPS exceeding 5A.

Yeah, I would be very interested in joining a group buy. But @MarcelNLthere is one thing you mentioned that I've always wanted to ask about.

 

Do you recall mentioning your output voltage dropping to 24v with the chokes mentioned above? Can you explain if that is expected behavior or if something else was the cause of the voltage drop?

 

[MarcelNL]

He is willing, yet availability of core material is an issue.

He used his last cores and ordering new comes with a lead time of like a month or two.

It's probably best to group order to allow him to order enough in one go, he is a hobby coil winder and will not be able to keep much cores in stock, just in case.

 

I discussed the voltage drop with him and he explained it to me , the explanation sounded plausible but my knowledge of coil behaviour etc does not go far enough to fully grasp it....I'll reread and to convey it here later (do remind me as I'm in the middle of moving house)