Building a DIY Music Server with custom made parts

[elan120]

Just now, RickyV said:

Wondering what would happen if you put 2 regulators in series

Quadrupled regulated power supply with very high dropout voltage.  👍

[RickyV]

12 hours ago, elan120 said:

Quadrupled regulated power supply with very high dropout voltage.  👍

 

Now that would be really over killing it 😀. 

But anyway at the moment I am doing that but with two 5A newclassd regulators, one in the external power supply, 24V. And one, 19V, inside my nuc case 10cm away from the internal connection point of the nuc.

Now with the more quality regulators I can do several configurations. 

 

Use SJ on the external power supply and still use the NCd as a second.

 

Use SJ external and use the ext power barrel connector on the nuc board. One regulator.

 

Use the NCd on the external power supply and build the SJ in the nuc case and use the int power connector. Best regulator last.

 

Build SJ in the nuc case and feed it with unregulated DC. So transformer, rectifier and caps external. Raw dc. 

 

I think the SJ fits inside the nuc case I have only 3 cm hight max.

 

What configuration would you prefer? @nenon too?

[Gavin1977]

@RickyV  Same idea I had, following your post on the Ideal Rectifiers.  Let me know how that Sean Jacobs regulators perform in series and parallel and how they share load.

[RickyV]

56 minutes ago, Gavin1977 said:

@RickyV  Same idea I had, following your post on the Ideal Rectifiers.  Let me know how that Sean Jacobs regulators perform in series and parallel and how they share load.

 

Not sure yet what configuration it's gonna be but the last one is preferable so I can use the second one for the ER. 

Of cause I'll experiment something, not sure yet.

i'll let you know.

 

[elan120]

5 hours ago, RickyV said:

What configuration would you prefer?

I have not tried NCd regulator in the past, so not sure how they compare.  SJ regulator module is set up in dual regulation, and setting the dropout voltage around 5 to 7VDC or slightly higher would be ideal, with that in mind, to use SJ regulator module in series with another regulator will easily bring total dropout voltage to 9VDC or higher, so if I have to pick, I will start with your last configuration by sending very good quality unregulated raw DC to SJ regulator and see how you like the result.

 

 

[Nenon]

18 hours ago, RickyV said:

Wondering what would happen if you put 2 regulators in series 🤪

 

I haven't tried that for two reasons.

1. Sean told me he has tried that and while the double regulation makes a significant difference vs. single regulation, he said that multiple double regulated regulators in series did not make an audible difference. I wanted to try that too and evaluate in my system, because I may hear something different, but the comment from Sean decreased the priority of that test (and my desire to perform such test) to very low. 

2. Heat dissipation might be a problem. A 6-7V drop at let's say 3A is already about 20W of heat that needs dissipation. Doubling that is not fun.  

Having said that, I don't know the details of where and how multiple regulators in series have been tried. Sean encouraged me to try it. So, I think it is worth a go. Please let us know @RickyV.

 

BTW, it's interesting how much better the LPS1.2 becomes when fed by a DC3 regulators. Give that a try if you can. 

 

[beautiful music]

On 6/12/2020 at 12:56 PM, RickyV said:

 

Now that would be really over killing it 😀. 

But anyway at the moment I am doing that but with two 5A newclassd regulators, one in the external power supply, 24V. And one, 19V, inside my nuc case 10cm away from the internal connection point of the nuc.

Now with the more quality regulators I can do several configurations. 

 

Use SJ on the external power supply and still use the NCd as a second.

 

Use SJ external and use the ext power barrel connector on the nuc board. One regulator.

 

Use the NCd on the external power supply and build the SJ in the nuc case and use the int power connector. Best regulator last.

 

Build SJ in the nuc case and feed it with unregulated DC. So transformer, rectifier and caps external. Raw dc. 

 

I think the SJ fits inside the nuc case I have only 3 cm hight max.

 

What configuration would you prefer? @nenon too?

 

Looking forward to know your experiments.

[Peter Avgeris]

On 6/10/2020 at 6:04 AM, Hauser said:

Still busy researching but have some ideas for the LGA3647 heatsink: sketch attached.  Studied Taiko's version and I'm sure its weight exceeds Intel's spec. of 600g.  Suspecting its supported from underneath.  For those with a lower DIY interest, maybe me included, I'm also considering a simpler version.  Cooltex on Aliexpress sell a skived copper heatsink.  Removing the fins provides a suitable base.  Using solder paste to attach a 58 x 94 x 1/4 inch copper plate would only leave some cutting, drilling, tapping and U slotting for the upper parts; typing this, still doesn't sound that simple.  Peter may have some more ideas.

 

Martin.

Hello Martin. 

I have received the board and 2x CPUs. Unfortunately the Ram chips are not compatible. I am waiting for the new ram chips so I will be able to fire up this unit. Regarding the specs of 600gr,judging from my preliminary look, this might not seem quite feasible if made with copper. Copper is surely preferable for smaller bricks but for such monstrous bricks, my preliminary calculations lead me to aluminum bricks. What matters is mass, along with conductivity, nothing else. Aluminum vs copper is 136:231 in terms of conductivity and 271:894 in terms of weight. Copper might be better for getting heat off the cpus but at seriously increased weight. Bricks do not necessarily need to be made from copper. I am studying the option of building smaller plates out of copper and then bigger bricks made of aluminum on top of them. 

I think the 600gr limit is impossible to be followed, but I don't thing that this is a serious limitation because the force applied by the locking mechanism is already excessive. If you put the mobo flat this limitation is absolutely irrelevant but if you put it vertical, then IMHO the weight of 600gr is already too big. In our case I would dare to write that the limit of 600gr must *not* be followed. But you should not even dare of putting the mobo in vertical position! 

 

For sure this is not a simple task, much more complicated than what I was initially considering. I plan to make some drawings after I have received the ram chips. 

[Downtheline]

4 hours ago, Peter Avgeris said:

Unfortunately the Ram chips are not compatible

I found the same problem with the X11SPM-TPF supermicro mobo. Needed registered ram. Using this:

AQD-D4U8GR26-SE, from advantech.

[Hauser]

13 hours ago, Peter Avgeris said:

Hello Martin. 

I have received the board and 2x CPUs. Unfortunately the Ram chips are not compatible. I am waiting for the new ram chips so I will be able to fire up this unit. Regarding the specs of 600gr,judging from my preliminary look, this might not seem quite feasible if made with copper. Copper is surely preferable for smaller bricks but for such monstrous bricks, my preliminary calculations lead me to aluminum bricks. What matters is mass, along with conductivity, nothing else. Aluminum vs copper is 136:231 in terms of conductivity and 271:894 in terms of weight. Copper might be better for getting heat off the cpus but at seriously increased weight. Bricks do not necessarily need to be made from copper. I am studying the option of building smaller plates out of copper and then bigger bricks made of aluminum on top of them. 

I think the 600gr limit is impossible to be followed, but I don't thing that this is a serious limitation because the force applied by the locking mechanism is already excessive. If you put the mobo flat this limitation is absolutely irrelevant but if you put it vertical, then IMHO the weight of 600gr is already too big. In our case I would dare to write that the limit of 600gr must *not* be followed. But you should not even dare of putting the mobo in vertical position! 

 

For sure this is not a simple task, much more complicated than what I was initially considering. I plan to make some drawings after I have received the ram chips. 

Hi Peter,

I looked at this using aluminium as the top clamp and weight was much closer to the limit.  Seeing the heatsinks in the Taiko, they are definitely heavier than 600g.  For this area to be covered, exceeding the weight limit with heatsinks is inevitable because copper does a much better job at collecting/transferring heat.  The answer may be in properly supporting the underside of the board directly under the socket.

 

Martin.

[Peter Avgeris]

13 hours ago, Hauser said:

Hi Peter,

I looked at this using aluminium as the top clamp and weight was much closer to the limit.  Seeing the heatsinks in the Taiko, they are definitely heavier than 600g.  For this area to be covered, exceeding the weight limit with heatsinks is inevitable because copper does a much better job at collecting/transferring heat.  The answer may be in properly supporting the underside of the board directly under the socket.

 

Martin.

Hi Martin!

No, the answer is to mount the board horizontally. Or alternatively, DON'T YOU DARE LIFTING THE BOARD VERTICALLY 🙂

Vertical stress (horizontal position) onto the CPU has no bad impact as the mounting mechanism already applies excessive force. Applying lateral force (vertical position) onto the CPU will be catastrophic.

So, copper bricks might not be harmful at all, even outside of specs, but the end user must be *EXTREMELY* cautious not to lift the machine in vertical position. Not even during transport, something that of course is bad design specifications.

To be in line with specs, the user must stress test the machine against failure. If it fails, then he should not have done this 🙂

Anyway, I think that copper bricks might not be that harmful. Power dissipation in our case is not that monstrous as the TDP of the CPUS is relatively low.

I am waiting for the RAM chips to begin.

[Ben-M]

I don't want to crap on the direction of the thread, and it's possible that across one of my multiple readings of the thread I may have missed it, but why not consider a Takachi Enclosures case? 

 

They are a professional case Maker from Japan and they make the chassis for the UpTone JS-2, the Taiko servers, and many other audio products. They have top notch quality, they take custom orders, and if you start with one of their existing base options and modify it, you can save a lot of money while still getting basically what you want. 

 

Another benefit is that they make the same case in multiple, compatible configurations. 

 

Back about 2 years ago I found out about them and requested pricing on their audio component standard size aluminum cases, which I have attached below. This line is called "HY", then the other number is the height.

 

Anyway, as you'll see in the pictures, you can get a full size ATX case laying in the horizontal format (think Taiko's size and style) for $300. 

 

Then you can go far beyond with alterations like material thickness, screw holes, or maybe even changing out materials like how Taiko swapped the side panel aluminum heatsinks for copper. 

 

Anyway, I thought it worth asking because it full achieves the flexibility in unit height, while sharing things like the top panel, bottom panel, footers, and internal mounting plate (if wanted). It opens the door for different buyers needs, all the way from 177mm high down to like 60mm. 

 

 

What're your guys' thoughts about this angle? Or what are the comparisons to the H5? 

 

I think almost any work that's been done so far is almost fully transferable to the Takachi HY177, since it focuses around designs to meet the Mobo and CPU placement, not the case manufacturer. 

[Hauser]

I started with the same idea as you having already used their boxes in a couple of other projects.  After I posted about them, it was pointed out that the heatsinks would not be up to the job.  I checked and the base thickness is only 5mm so when you carve out a 3 or 4mm slot for heatpipes there is not much material left.  I agree they are very versatile with the range of sizes available but not suitable for dealing with much heat.  I am working through making my own case.  The latest obstacle is the Marston heatsinks I ordered have a concave base.  I'm getting Element14 to check whether the Fischer ones are any better.

 

Martin

[zerung]

2 hours ago, Hauser said:

I started with the same idea as you having already used their boxes in a couple of other projects.  After I posted about them, it was pointed out that the heatsinks would not be up to the job.  I checked and the base thickness is only 5mm so when you carve out a 3 or 4mm slot for heatpipes there is not much material left.  I agree they are very versatile with the range of sizes available but not suitable for dealing with much heat.  I am working through making my own case.  The latest obstacle is the Marston heatsinks I ordered have a concave base.  I'm getting Element14 to check whether the Fischer ones are any better.

 

Martin

Could you not add a copper serrated spreader block as a receiver for the heat pipes?  Should this not do?

 

[Hauser]

That looks like a cunning plan.  I haven't seen these for sale.  I'm still struggling to cut copper but have bought some router bits to see if I can mill a slot.  Copper flat bar is expensive in New Zealand.  Regarding thermal resistance, the Marston heatsink (96CN) has less than half that of a heatsink of a similar profile to the Takachi.  Probably best to go with heatsink sizes similar to those already tested as in the Hdplex case to get some confidence of success.

 

Martin

[Ben-M]

11 hours ago, Hauser said:

I started with the same idea as you having already used their boxes in a couple of other projects.  After I posted about them, it was pointed out that the heatsinks would not be up to the job.  I checked and the base thickness is only 5mm so when you carve out a 3 or 4mm slot for heatpipes there is not much material left.  I agree they are very versatile with the range of sizes available but not suitable for dealing with much heat.  I am working through making my own case.  The latest obstacle is the Marston heatsinks I ordered have a concave base.  I'm getting Element14 to check whether the Fischer ones are any better.

 

Martin

I see your point, but I wouldn't count them down and out so quickly. We could easily request thicker stock. 

 

Besides that, it'd only be a 1-2mm groove, not 3-4mm. The Takachi side panel heatsinks are similar to the Streacom FC10, which handle up to 95W TDP CPUs. If we could use 1 panel/side per CPU, that may be enough. And if not, we could explore thicker panels or copper. 

 

I am in China now, so I will see if I can source a copper drop-in heatsink replacement for whichever case we end up with. I don't think it'd be too much here, but I could be totally wrong. 

[Peter Avgeris]

Hello from Athens, Greece.

I have been reading your comments with utmost interest.

In brief, a depth of 1-2 mm of the grooves is absolutely insufficient.

Due to the incompatibility of current RAM chips I have, I cannot use them for powering up the server board with the Intel Xeon CPUs. As soon as I am able to power up, I can measure the current drawn through these specific CPUs and IMHO this is one of the most important aspects that, among others, affect the calculations and the design of the chassis.

I am waiting to receive the compatible RDIMMs from Germany and next step is the fine tuning of the server board. Tuning has great impact on power consumption. Each and every aspect is interconnected to the other. So, I think we need to step aside from everything until I have the exact power figures handy and then make the first move.

As a preliminary assumption, I am expecting not to be able to avoid using the German Junior / Fischer heat sinks. Whenever I'm thinking of heat sinks my mind always go to Junior/Fischer.

 

[Hauser]

9 hours ago, Ben-M said:

I see your point, but I wouldn't count them down and out so quickly. We could easily request thicker stock. 

 

Besides that, it'd only be a 1-2mm groove, not 3-4mm. The Takachi side panel heatsinks are similar to the Streacom FC10, which handle up to 95W TDP CPUs. If we could use 1 panel/side per CPU, that may be enough. And if not, we could explore thicker panels or copper. 

 

I am in China now, so I will see if I can source a copper drop-in heatsink replacement for whichever case we end up with. I don't think it'd be too much here, but I could be totally wrong. 

Just as a reference; Onlinemetal price for 1.5 x 6 x 12inch copper flat bar USD 380 plus freight plus machining.  It becomes easier to understand why Taiko only did copper on one side and how much their case is likely to cost.

 

Martin

[Ben-M]

I'm not trying to say I'm the authority on this, but I have used this Streacom FC10 chassis in 4 different builds and it's always delivered effective cooling results. 

 

As you can see from the internal picture, there is 0% milling into the side panel to improve conduction or heat transfer from heatsink to heat pipes:

 

 

The heat pipes have been flattened where they interface with the side panel, riser, and CPU block.

 

I think milling grooves is cool, but from my experience so far, it seems unnecessary in order to deliver cooling up to 95W TDP.

 

And since it seems our target CPU's TDP is beyond that, I don't think milling heat pipe grooves into the side panel is the guaranteed approach to gain more heat dissipation capacity. Different material or thicker material is needed. 

 

It's interesting hashing out ideas while we wait for his real world results, but as Peter just mentioned, no point in letting the cart get ahead of the horse just yet.  

[LTG2010]

On 7/9/2020 at 5:02 PM, Peter Avgeris said:

Due to the incompatibility of current RAM chips I have

Hi Peter, if you continue to have problems finding compatible ram, I can send you 2 x 16gb Samsung modules from UK. Send me a PM. It would be great to find a cooling solution.

Generally they need to be registered dimms, ecc.

When using new ram there is a clear rtc jumper on the board which helps.

[Peter Avgeris]

On 7/10/2020 at 5:36 PM, Ben-M said:

I'm not trying to say I'm the authority on this, but I have used this Streacom FC10 chassis in 4 different builds and it's always delivered effective cooling results. 

 

As you can see from the internal picture, there is 0% milling into the side panel to improve conduction or heat transfer from heatsink to heat pipes:

 

 

The heat pipes have been flattened where they interface with the side panel, riser, and CPU block.

 

I think milling grooves is cool, but from my experience so far, it seems unnecessary in order to deliver cooling up to 95W TDP.

 

And since it seems our target CPU's TDP is beyond that, I don't think milling heat pipe grooves into the side panel is the guaranteed approach to gain more heat dissipation capacity. Different material or thicker material is needed. 

 

It's interesting hashing out ideas while we wait for his real world results, but as Peter just mentioned, no point in letting the cart get ahead of the horse just yet.  

Yes! Flat cooper heat pipes is an interesting solution that could be beneficial in the way that we can avoid digging groves onto the lateral heatsinks.

Big heat pipes are required for sure. Taiko Audio has used copper heat sink for both CPUs. Aluminum heat sink requires increased dissipation area between copper and aluminum structure.

Let's wait for a couple of days more.

The RAM chips are on their way.@LTG2010, thank you for your willingness to help. I think the ones supplied are already checked that they are fully functional.

[Hauser]

If you are trying to dissipate heat using heatpipes, larger diameter is better, fewer bends is better, round is better, more contact surface is better.  ATS site offers good information.

 

Martin

[OAudio]

Hi All,

 

This is a rendering of one of the parts I'm making for packaging of my audio server development. There are a few final details to iron out with my engineerings but they will start cutting them from billet in the next couple of weeks with luck (waiting to get a production slot agreed). 

 

The cooler is designed to handle up to 200 watts of dissipation from an LGA3647. This is far higher than is required even for heavily loaded CPU. At 200w you would need a < 0.2C/w rated heat-sink which is a very large passive sink, but this is not really necessary for an audio server  ).

 

Care has been taken to make the heat pipes exit the CPU's footprint area 40mm above the motherboard. This means that  heatpipe routing remains simple and most sensible sized memory DIMMS should be insertable and removable without moving either the cooler or heat pipes. It can be reversed by 180 degrees when fitted meaning cooling pipes can exit over either bank of memory. I dont need this feature but it just seemed a good thing to do. End point clamps for the far end of the heatpipes are also being produced. Material will be engineering grade aluminium with clear hard anodizing to keep it looking pretty as time passes.

 

 

OAudio. 

 

[Tatomek7]

Hi All,

 

I refresh the thread, as I just joined the C621 camp with two LGA3647 processors motherboard and I’m highly interested to exchange the current coolers ( Dynatron B14 ) with passive solution. I was following with big hope @Peter Avgeris involvement, but no news recently. @OAudio, do you think, your cooler has chance to become real? I was looking here around to find the local artisan with right CNC , but with no success at the moment. It seems more complicated, I thought before.

[Peter Avgeris]

The thread is still active. My problem is that although I have the MoBo & the processors in my hands, I can't power up the boards because the RAM chips I have are incompatible. I am waiting for the new chips for months so as to be able to power it up.