Building a DIY Music Server

[Exocer]

6 minutes ago, realDHT said:

The PSU box for my music server (although the CPU choke isn't really visible in this image (it's hidden below the softstart board in back righthand side) 

Very clean work. Are you comfortable sharing information about the specs and topology?

 

Cheers,

-Rob

[Mags]

I have a Hammond 195 series choke in the DC PSU for the filament of the 833C. It takes 100W just to light the filament (10V, 10A) of that beast. 

In a choke input supply be sure to make the first choke current rating about 2X the steady-state current draw or it will saturate on peaks (hence my 20A choke for a 10A supply). 

[Mags]

Another trick is to put a small value cap in front of the first choke and vary its value to fine-tune the output voltage. The cap must be small enough not to change the supply into a cap input supply. This is typically known as cLCLC, for example, and is still considered choke input.

[realDHT]

59 minutes ago, Exocer said:

Very clean work. Are you comfortable sharing information about the specs and topology?

 

Cheers,

-Rob

Thanks Rob. 

 

Sure, I did plan to describe my server project in more detail on this forum before but just have not taken the time yet. I probably start a new thread then, my server was built for both music and home cinema. 

 

Most inspiration and ideas comes from this forum, and Nenons previous build recommendations in particular. 

 

In short it is a 2-box build based on

Hardware: 3700x CPU, Mainboard Crosshair VII Hero, RAM Apacer ECC, OS drive Optane 900p, USB card JCAT femto, Music storage on External HDD via USB

 

Software: Windows Server 2019, Audiophile optimizer, Process Lasso, BIOS optimizations, JRiver MC

 

What differs from most builds is that I included on/off switchable fanless graphics and that most PSU rails are choke-input and shuntregulated.

[RickyV]

5 minutes ago, realDHT said:

Thanks Rob. 

 

Sure, I did plan to describe my server project in more detail on this forum before but just have not taken the time yet. I probably start a new thread then, my server was built for both music and home cinema. 

 

Most inspiration and ideas comes from this forum, and Nenons previous build recommendations in particular. 

 

In short it is a 2-box build based on

Hardware: 3700x CPU, Mainboard Crosshair VII Hero, RAM Apacer ECC, OS drive Optane 900p, USB card JCAT femto, Music storage on External HDD via USB

 

Software: Windows Server 2019, Audiophile optimizer, Process Lasso, BIOS optimizations, JRiver MC

 

What differs from most builds is that I included on/off switchable fanless graphics and that most PSU rails are choke-input and shuntregulated.

Please do describe your server in detail your power supply looks really impressive.

[mikicasellas]

Hello Guys,

 

I updated the BIOS of my ROG_MAXIMUS_XIII_HERO_Z590, it was purchased in April 2021 and have been using the Bios that came with it in that time, until this month that I updated the Bios to Version 0902 but sound became harder, has anyone experienced this ?

 

Is there a way to return to previous BIOS Version ?, I have Bios Rollback enabled but don't know where to "roll it back"

 

Thanks !

[MarcelNL]

is it not just a matter of flashing the previous version again?

 

Did you check your bios settings? usually a new bios means starting from scratch with all the tweaking of settings for SQ.

[mikicasellas]

1 minute ago, MarcelNL said:

is it not just a matter of flashing the previous version again?

 

Did you check your bios settings? usually a new bios means starting from scratch with all the tweaking of settings for SQ.

Yep, I re-tweaked all the config to be as equal as posible as my last version, which it was  to the same but i could at least check on the similar.

 

I was trying to avoid the "download the file to USB" path and I thought it was as plain as pushing the Flash back button 

[ASRMichael]

27 minutes ago, mikicasellas said:

Hello Guys,

 

I updated the BIOS of my ROG_MAXIMUS_XIII_HERO_Z590, it was purchased in April 2021 and have been using the Bios that came with it in that time, until this month that I updated the Bios to Version 0902 but sound became harder, has anyone experienced this ?

 

Is there a way to return to previous BIOS Version ?, I have Bios Rollback enabled but don't know where to "roll it back"

 

Thanks !

Just go to website & download old version & put on usb drive. Then go to bios & load from usb. 

[mikicasellas]

1 minute ago, ASRMichael said:

Just go to website & download old version & put on usb drive. Then go to bios & load from usb. 

Yes, that was the only way !

Thanks !

[Exocer]

14 minutes ago, mikicasellas said:

Yep, I re-tweaked all the config to be as equal as posible as my last version, which it was  to the same but i could at least check on the similar.

 

I was trying to avoid the "download the file to USB" path and I thought it was as plain as pushing the Flash back button 

How long has your system been online since updating the bios and reverting to your previous settings?

[mikicasellas]

1 minute ago, Exocer said:

How long has your system been online since updating the bios and reverting to your previous settings?

In August the 5th I updated, sound immediately became harder, but i have been turning server off every night and ON first time in the morning every day, cause temperatures being higher than I want when air conditioning is off during the whole night...It is now re-updating to previous with the USB method

[Soul Analogue]

On 8/14/2021 at 5:59 AM, Mags said:

"From what I hear that first cap before the choke is equally important for sound as the choke, well everything matters as mentioned before by others."

 

Even better is putting the choke first after the rectifier. I use choke input supplies everywhere, much smoother power delivery than cap input. Try it some time.

 

I am also into Choke input PSU..

But for a CAS PSU, the choke for choke input requires really BIG investments....both cost, weight and space...

I have got big name chokes from USA and from Israel... they are big and bulky...especially for the Israel double C...

at 8a specifications clearly stated with choke input application...

However they still fail to operate quietly for the PSU for my CPU

 

So I now I need to assemble chokes by myself up to my own requirements...

In summary...the bigger choke = bigger and better sound 😁

 

One question...is the Hammond 20A for choke input silent at 10A load?

[MarcelNL]

Now THAT are chokes. Hum can often be cured by ; seriously overdesigning a choke, getting rid of input DC, clamping the core, vacuum coating with lacquer/resin (that last one is quite a chore)

[Mags]

6 hours ago, Soul Analogue said:

 

I am also into Choke input PSU..

But for a CAS PSU, the choke for choke input requires really BIG investments....both cost, weight and space...

I have got big name chokes from USA and from Israel... they are big and bulky...especially for the Israel double C...

at 8a specifications clearly stated with choke input application...

However they still fail to operate quietly for the PSU for my CPU

 

So I now I need to assemble chokes by myself up to my own requirements...

In summary...the bigger choke = bigger and better sound 😁

 

One question...is the Hammond 20A for choke input silent at 10A load?

Yes, the 20A Hammond is silent at 10A. I still usually mount chokes with a little vibration-absorbing rubber under the mounts, just in case.

[Soul Analogue]

20 minutes ago, Mags said:

Yes, the 20A Hammond is silent at 10A. I still usually mount chokes with a little vibration-absorbing rubber under the mounts, just in case.

I see .. good to know

But you are only dealing with low voltage… for 10v 10a… i think 16vac secondary from transformer would be sufficient..

For taiko atx module, we need to deal with around 38vac for choke input… the ac ripple after rectifier is some order of magnitudes higher… i expect a choke of much much bigger size is needed for silent operation..

vibration damping rubber does virtually nothing if the choke rattles….

[Peter Avgeris]

Thanks for the invitation.

I wrote you, it may take between a moment and a lifetime to realize how bad is choke input for sound quality. I do not want to conflict with you.

The differences I am talking about are relative. Switching between L input and C input, all else being same and equal, leads to straightforward results. I do not want to pressurize you, you do not need to do so.

With kind regards

Peter Avgeris

 

PS: the web site is under construction, as photoshooting has not been completed yet.

[Mags]

2 hours ago, Soul Analogue said:

I see .. good to know

But you are only dealing with low voltage… for 10v 10a… i think 16vac secondary from transformer would be sufficient..

For taiko atx module, we need to deal with around 38vac for choke input… the ac ripple after rectifier is some order of magnitudes higher… i expect a choke of much much bigger size is needed for silent operation..

vibration damping rubber does virtually nothing if the choke rattles….

Actually, the output tube anode supply is choke input at ~2500V, and the driver tube anode supply is choke input at ~450V. Of course tubes run at lower current. You just need good chokes. In my case the output supply uses a potted Monolith magnetics choke and the driver supply uses a Lundahl C-core. 

[Soul Analogue]

9 minutes ago, Mags said:

Actually, the output tube anode supply is choke input at ~2500V, and the driver tube anode supply is choke input at ~450V. Of course tubes run at lower current. You just need good chokes. In my case the output supply uses a potted Monolith magnetics choke and the driver supply uses a Lundahl C-core. 

Ok this is a thread for CAS 

and you were talking about 20a hammond choke runs at 10a.. so i was referring to that

[Soul Analogue]

2 hours ago, Peter Avgeris said:

Thanks for the invitation.

I wrote you, it may take between a moment and a lifetime to realize how bad is choke input for sound quality. I do not want to conflict with you.

The differences I am talking about are relative. Switching between L input and C input, all else being same and equal, leads to straightforward results. I do not want to pressurize you, you do not need to do so.

With kind regards

Peter Avgeris

 

PS: the web site is under construction, as photoshooting has not been completed yet.

It is not easy for someone to “switch” between L input and C input…. If simply put in and take out the C from psu.. the resultant voltage will be largely different..

and even given the same resultant voltage… the design and construction of the choke for L input choke and C input are very different….

so there is no easy switch… or the test result will be less than optimal in either cases

 

i believe it is a matter of taste for choke input and cap input psu…(talking about properly designed and executed) 

the presentations are quite different

 

In my diy venture… i first started with cap input… and later tried on choke input and could never go back….so L input for me as far as practicable

[Exocer]

1 hour ago, Soul Analogue said:

It is not easy for someone to “switch” between L input and C input…. If simply put in and take out the C from psu.. the resultant voltage will be largely different..

and even given the same resultant voltage… the design and construction of the choke for L input choke and C input are very different….so there is no easy switch… or the test result will be less than optimal in either cases

 

i believe it is a matter of taste for choke input and cap input psu…(talking about properly designed and executed) 

the presentations are quite different

 

In my diy venture… i first started with cap input… and later tried on choke input and could never go back….so L input for me as far as practicable

Thanks for sharing. It would be great to hear about the differences you've perceived, going from C input to L input LPS's in your experience.

[Nenon]

I enjoy reading what people do or have done. A lot of good stuff. There are probably people reading all these comments about chokes who are getting confused. So, let me clarify some things. 

 

1. There are some claims that the 5A choke I picked does not provide enough current and restricts the dynamics on the dual Xeon server. 

That is not the case with my build. I utilize about half of the current capabilities of this choke. And this is by far the most dynamic server I have ever built. 

There are two things to keep in mind here. First, this is an unregulated LPS + GaN based ATX, which is extremely efficient. Note the 95W vs. 270W comment in the post I am quoting below. Second, Windows (LTSC in my case) the way I have configured it is more efficient than Euphony. As a result of this I am using about 95W from nearly 200W available with this choke. 

More details here:

On 8/11/2021 at 12:01 PM, Nenon said:

 

Just did some measurements on my dual CPU server, which also has a PCIe Optane 900P card, a USB PCIe card, and an Asus Hyper card with two Intel 665p SSDs installed. 

It draws 65 Watts from the wall when idle.

It draws 95 Watts from the wall when playing music.

 

I saw it close to 200 Watts on a couple of occasions while booting... pretty sure that choke was saturated at those times. If that was what I was seeing while playing music, I would be very concerned about the choke I am using. But at 65/95 Watts, I am not concerned at all. 

 

That's not to say the more power, the better. I like everything overkill in audio :). The problem is that every choke has a different sound. If we change it with a more powerful one, we'll get more power but the sound will change. The question is will that give us a net positive result? I am quite happy with the two versions of the unregulated LPS. I spent quite a few months testing and listening. There were a lot of problems to solve (i.e. ringing, overshooting, etc.) and a lot of parts to test. And the best part - all shared for people to do as a DIY project. 

 

I remember that dual Xeon server drawing 250-270W from the wall when I was using a regulated linear power supply and running Euphony. Talk about efficiency improvement! 

 

I would be interested to try your design too. Quite like the idea of building a custom choke for this. 

 

Unless you do heavy DSD upsampling with 120W TDP CPUs, chances are this choke will work great. As a general rule, if you can passively cool down your CPU with a HDPlex/Streacom chassis, you probably won't have any issues with this choke/design.

 

2. There are some claims that the 5A choke I picked will be very noisy.

6 hours ago, Soul Analogue said:

For taiko atx module, we need to deal with around 38vac for choke input… the ac ripple after rectifier is some order of magnitudes higher… i expect a choke of much much bigger size is needed for silent operation..

This is also not the case in my system. Not at all! There are people here who have followed the DIY recipe I provided and can comment about their experience. No noise in my power supply. Please do share if your experience is different.

 

3. Very important - you SHOULD NOT just replace the choke with another part unless you know exactly what you are doing. You can potentially kill your Taiko ATX if you simply replace the choke with a much lower DCR one and don't make any other changes.

For example, modeling some of the bigger Hammond chokes turned out they will require 470,000 uF of capacitance instead of the 66,000 uF in V2. In some cases, with some chokes, you may reach the need to use 700,000 uF. 

 

4. A higher current choke does not sound better in my experience. It sounds different. There are always tradeoffs. What I have provided for DIY is a very well balanced power supply design. It's a good tradeoff between staging/envelopment and dynamics/naturality. It looks like a very simple design that I put together during my lunch break, but it actually took many months of development, testing, and most importantly - listening. It employs a lot of things I've learned from Emile and Sean, some of which I can't really share. But it doesn't really matter how I came up with this design. It's the end product that matters... and that's shared for anyone who wants to build it. I've shared details here (the cost of parts keeps going up, so my spreadsheet is already out of date):

On 5/16/2021 at 10:44 AM, Nenon said:

Unregulated LPS update

 

I am pretty much set on the 3 designs. As I explained in an earlier post, we simply add more parts to each design. You can start with v1 and then go to v2 and then to v3. As you are going up the ladder no parts remain unused. 

 

The first version (aka v1) is very simple. You have an IEC inlet --> Fuse --> Soft start --> Transformer --> Mosfet rectifier --> Output connector. The version gives you very clean, fast, and transparent sound, but it does not have the body and the full expansive sound as the other two versions. Can too transparent and too fast be a problem? I think it's way too fast and transparent, hence the additional components in the other two versions. But it's all about personal taste. Some may like that version more. 

 

The second version (aka v2) consists of: IEC inlet --> Fuse --> Soft start --> Transformer --> Mosfet rectifier --> Hammond choke --> 3 x Mundorf HC 22,000 uF (each bypassed with VCAP ODAM 0.1 uF film cap) --> Output connector. This is an improvement over v1 in my opinion. It gives you a bit more of everything and is the sweet spot. This is what I have been running for months (until I tried v3). You have about 5mV ripple with this version. You get about 2.3 mOhm output impedance but that's influenced by the wiring you are using and could be higher if you use thin wires or bad connectors.

 

The third version (aka v3) consists of: IEC inlet --> Fuse --> Soft start --> Transformer --> Mosfet rectifier --> Mundorf HC 22,000 uF (bypassed with VCAP ODAM 0.1 uF film cap) --> Hammond choke --> Mundorf HC 22,000 uF (bypassed with VCAP ODAM 0.1 uF film cap) --> Hammond choke --> 3 x Mundorf HC ( 2 x 47,000 uF + 1 x 22,000 uF, each bypassed with VCAP ODAM 0.1 uF film cap) --> Output connector. This version gives away some of the "clean" sound you have in v1 and v2 but is fuller, bigger, massively expensive. Amazing 5uV ripple and nearly half the output impedance compared to v2 - ~1.2 mOhm + the impact of your wiring. 

 

I don't have the prices of the Noratel transformers and the Taiko rectifier, but just to get some idea, let's assume they are $200 each. Here is what I have for the bill of materials with somewhat rounded prices:

 

This is just ballpark pricing to get an idea. Leave it to me, and it would get much more expensive as I would add thick pure copper busbars, multiple runs of expensive wire, exotic connectors, footers, etc. I would also have Modushop fabricate the cases for me, so it's easier to install everything. That also adds to the cost. However, if you see how massive v3 is, it would easily compare to the power supplies of very very expensive amps. If this thing was a commercial product (or a part of a commercial product) in the dealer's network I'd expect it to be over $20K. This is the beauty of DIY :). And as far as R&D goes, the amount of time spent here is on par with many commercial products to say the least. 

 

If you are impatient and don't want to wait anymore, you can order a 24VAC 400VA Toroidy Supreme transformer, get a rectifier of your choice, and get this going. Not my favorite rectifier but something like the Vishay GBPC3504-E4/51 could do the job for the time being, assuming it would be replaced with something better in the future. 

 

That was the easy part of the post. It gets quite complicated from here. As with anything else in life, there will be people with all kinds of different preferences. Some would like cheaper. Some would like better where cost is no object. Some would prefer easier to make. Some would prefer someone else to make it for them. Some would have existing parts they would like to use. And so on... It's hard to cover all possible cases, but here are some random notes to begin with. 

• If you are on a budget, the above can be made a lot cheaper. You can use a cheaper chassis, cheaper IEC inlet, cheaper connectors, cheaper transformers, cheaper capacitors, etc. You can really make this a budget LPS. It would probably still be better than most linear power supplies on the market (most of them use cheap components too). 
• If you are on the opposite end of "on a budget", you can make this even better. You can use even better chassis. You can add good feet to the chassis (i.e. the Gaia I use). You can fabricate copper busbars to screw in the Mundorf caps to (something I am looking to do for myself). You can use the Mundorf copper terminal rings. 
• If you are not handy with drilling/milling we can have Modushop predrill all the holes on the rear panel and the bottom panel. That adds to the price - from their price list guessing around 25 Euro for the rear panel (less than 10 holes) and another 35-45 Euro (more the 20 holes for v3) for the bottom panel. BTW, I have not received this chassis from Modushop yet, so I am just speculating that v3 would fit in it at this point. Pretty sure v2 would fit. But you can also use a smaller chassis for v1 and v2.
• Taiko does not have many mosfet rectifiers, so we have to decide if another batch is needed, if Emile is even willing to run another batch. There are a lot of complications with that... We need to see what the lead times would be. We need to know how many to order. Or we can look for some other alternatives. There is something else I have in mind but would require a PCB. 
• There is a guy in the US and a guy in the EU who are willing to make these unregulated LPS's for other people. The labor fee negotiated was $500, given that the chassis has all the holes pre-drilled. I will make a couple myself just to get the process started and documented.  
• We haven't discussed wiring. I am still making some tests but so far I am leaning towards busbars for the capacitors and a custom OEM copper wire I managed to procure with the Mundorf lugs (which are quite pricey by the way, but I like them). 

 

There is a global shortage of materials. Some of the products mentioned above may be hard to obtain. I mentioned doing a potential group order in the past. I can try to organise that and get the entire BOM. If anyone is interested in that, please contact me on PM before the end of May indicating if you want v2 or v3 and if you want someone to build it for you.

 

Okay, let's keep it at that for now. I can probably keep writing this post the whole day but have other things I need to do now. 

 

 

Combine this unregulated LPS (v2 or v3) with the Taiko ATX to power up your computer. It hardly gets any better than that.

Once you hear what this sounds like, it will change your perspective on fully regulated linear power supplies for high current draw applications. You still can't beat a good LPS for low current draw applications (i.e. 1A-2A-3A), but for high current CPUs this is my new gold standard. 

 

Lastly, I encourage you to build this unregulated LPS (both v2 and v3) and try it with the Taiko ATX (powering up the EPS and ATX) before going into arguments or criticising a particular part I chose. No part I picked was just a random choice. They all work together and create a synergy. Once you do that, by all means feel free to experiment and if you find something that works better, please share. I am done with the unregulated LPS development and stopped tinkering with it (for now). No upgraded versions are to be expected by me for a while. But I hope other people will pick this from where I left it and share their improvements. 

 

@Soul Analogue  I admire your work. Feel free to post details and pricing about your unregulated LPS with handmade chokes. It might be a good alternative for people who are not comfortable building a DIY power supply. I would personally love to test it myself. Maybe someone in the US will buy one and will be kind enough to send it to me to try it :). 

 

[Exocer]

31 minutes ago, Nenon said:

I enjoy reading what people do or have done. A lot of good stuff. There are probably people reading all these comments about chokes who are getting confused. So, let me clarify some things. 

 

1. There are some claims that the 5A choke I picked does not provide enough current and restricts the dynamics on the dual Xeon server. 

That is not the case with my build. I utilize about half of the current capabilities of this choke. And this is by far the most dynamic server I have ever built. 

There are two things to keep in mind here. First, this is an unregulated LPS + GaN based ATX, which is extremely efficient. Note the 95W vs. 270W comment in the post I am quoting below. Second, Windows (LTSC in my case) the way I have configured it is more efficient than Euphony. As a result of this I am using about 95W from nearly 200W available with this choke. 

More details here:

Unless you do heavy DSD upsampling with 120W TDP CPUs, chances are this choke will work great. As a general rule, if you can passively cool down your CPU with a HDPlex/Streacom chassis, you probably won't have any issues with this choke/design.

 

2. There are some claims that the 5A choke I picked will be very noisy.

This is also not the case in my system. Not at all! There are people here who have followed the DIY recipe I provided and can comment about their experience. No noise in my power supply. Please do share if your experience is different.

 

3. Very important - you SHOULD NOT just replace the choke with another part unless you know exactly what you are doing. You can potentially kill your Taiko ATX if you simply replace the choke with a much lower DCR one and don't make any other changes.

For example, modeling some of the bigger Hammond chokes turned out they will require 470,000 uF of capacitance instead of the 66,000 uF in V2. In some cases, with some chokes, you may reach the need to use 700,000 uF. 

 

4. A higher current choke does not sound better in my experience. It sounds different. There are always tradeoffs. What I have provided for DIY is a very well balanced power supply design. It's a good tradeoff between staging/envelopment and dynamics/naturality. It looks like a very simple design that I put together during my lunch break, but it actually took many months of development, testing, and most importantly - listening. It employs a lot of things I've learned from Emile and Sean, some of which I can't really share. But it doesn't really matter how I came up with this design. It's the end product that matters... and that's shared for anyone who wants to build it. I've shared details here (the cost of parts keeps going up, so my spreadsheet is already out of date):

 

Combine this unregulated LPS (v2 or v3) with the Taiko ATX to power up your computer. It hardly gets any better than that.

Once you hear what this sounds like, it will change your perspective on fully regulated linear power supplies for high current draw applications. You still can't beat a good LPS for low current draw applications (i.e. 1A-2A-3A), but for high current CPUs this is my new gold standard. 

 

Lastly, I encourage you to build this unregulated LPS (both v2 and v3) and try it with the Taiko ATX (powering up the EPS and ATX) before going into arguments or criticising a particular part I chose. No part I picked was just a random choice. They all work together and create a synergy. Once you do that, by all means feel free to experiment and if you find something that works better, please share. I am done with the unregulated LPS development and stopped tinkering with it (for now). No upgraded versions are to be expected by me for a while. But I hope other people will pick this from where I left it and share their improvements. 

 

@Soul Analogue  I admire your work. Feel free to post details and pricing about your unregulated LPS with handmade chokes. It might be a good alternative for people who are not comfortable building a DIY power supply. I would personally love to test it myself. Maybe someone in the US will buy one and will be kind enough to send it to me to try it :). 

 

Thanks for the clarification. My 2nd Hammond choke arrives today. If time permits, I will be completing the wiring for V3 but will likely wait at least a week to connect it to the circuit. Before I expand to V3, I am hoping to conduct a small non-scientific shootout of 3 USB cards. Hoping to conduct this test in the next week or so. Looking forward to sharing my impressions.

 

Then, testing of V3 will proceed (which I am very excited to hear). Once V3 settles in i'll re-test the PH SR7T/V2 feeding the Taiko-DC to ATX and do a nice write-up on what I hear. Hopefully folks will find it useful! 

 

Cheers,

-Rob

[Darryl R]

3 hours ago, Exocer said:

I am hoping to conduct a small non-scientific shootout of 3 USB cards. Hoping to conduct this test in the next week or so. Looking forward to sharing my impressions.

 

Looking forward to seeing how that Core Audio stacks up.  Will you be posting your results in the "Best Audiophile USB Card" thread?

[Exocer]

6 minutes ago, Darryl R said:

 

Looking forward to seeing how that Core Audio stacks up.  Will you be posting your results in the "Best Audiophile USB Card" thread?

Yes indeed. Looking forward to testing!