Building a DIY Music Server

[LTG2010]

34 minutes ago, matthias said:

I dare to say when Emile puts all his genius into designing a superlative SMPS for the Extreme it might be outperforming his best LPS.

From what I can see of the pictures of the Insides of Extreme published, it's already using buck converters and not linear regulators and I'm assuming this DC - ATX is based on that system?

High powered linear regulators need high heat dissipation and there do not appear to be any regulators fitted onto the heatsinks. placing them below would mean they heat up the base and the rest of the components above. Well I'm just guessing I would be interested to know.

I'm currently using a C-L-C Filter with HDplex 800 DC-ATX. An 'Audiopile' DC - ATX would be perfect.

 

[matthias]

14 minutes ago, RickyV said:

You could try the smps that Hypex uses for their ncore amplifiers.

 

Emile mentioned on WBF that the switching frequency is too low.

 

Matt

[RickyV]

3 minutes ago, LTG2010 said:

From what I can see of the pictures of the Insides of Extreme pulished, its already using buck converters and not linear regulators and I'm assuming this DC - ATX is based on that system.

High powered linear regulators need high heat dissipation and there do not appear to be any regulators fitted onto the heatsinks. placing them below would mean they heat up the base and the rest of the components above. Well I'm just guessing I would be interested to know.

 

We could never see what was underneath the the big capacitor bank where probably the atx regulation is located. My guess now would be their is indeed switching regulations underneath. The rest of the power supply of the extreme is transformer, rectifier, choke, capacitor bank, atx regulators.

[RickyV]

3 minutes ago, matthias said:

 

Emile mentioned on WBF that the switching frequency is too low.

 

Matt

Ah, okey.

[Nenon]

5 minutes ago, matthias said:

@Nenon

 

"So anyway, there are definitely merits in exploring Switch Mode Power Supplies for audiophile applications, the benefits over Linear Power Supplies are obvious, if implemented correctly, it can be much lower noise then a Linear Power Supply for higher current applications, have a much smaller footprint and no massive heatsinking requirements, BUT they should be designed to not have any switching "noise" anywhere near the audio frequency range, or even better yet no measurable residual whatsoever."

 

This quote from Emile mirrors exactly what I think about SMPS, and I mean audiophile SMPS up to the highest standards like the Taiko DC-ATX.

We are at the beginning of real audiophile SMPS which IMHO will surpass the best LPS in the end.

The core of my question on WBF was not addressed to upsampling but to sound quality, I dare to say when Emile puts all his genius into designing a superlative SMPS for the Extreme it might be outperforming his best LPS.

 

Matt

 

 

Matt - getting offtopic, but what are the benefits of a SMPS in your opinion? 

 

To me it's the smaller size, efficiency, speed, and less heat dissipation. The Taiko ATX is 99% efficient already, and it's crazy fast. It is so fast that we actually need to take measures in the supply feeding it to slow it down. Yes, that's right. That's a part of the voicing I am currently doing. It could be so fast that it would make my server sound too clinical, and I would not enjoy it. It's also switching at a very high frequency, so there is no switching noise anywhere near the audio frequency noise. 

I personally don't feel like putting an AC SMPS in front of the Taiko ATX would be a step in the right direction. I think we get the best of both worlds the way this is. I am strictly talking about the DIY server I am working on.  From my experience with the Taiko ATX you can easily make or break things by changing just a small part (or even a wire) on the supply. That's how sensitive it is.

 

I have no idea what shortcomings the Extreme has (if any) and how it can or cannot be improved. I don't know the details around its power supply either, so I can't comment on that. But the point I am trying to make is that the Extreme is so carefully tweaked that you can easily make or break things. Small changes I make on my server have a huge impact. I imagine that the impact of such changes on the Extreme is magnified by a lot. That does not answer your question, though... I'll leave that to Emile :). 

 

And here is me generalizing now :):

The last comment I want to make is that no matter how advanced the SMPS technology becomes, to me a well designed LPS is still the best option for powering low current devices. I do power my USB card on my server with a Sean Jacobs DC4 LPS and that makes a hell of difference. I find it best to use the right technology where it matters. For a dual CPU server that can easily draw 300W, this SMPS technology makes sense and works great. But for low current applications, nothing can beat a good old linear power supply. That's my 2 cents. I imagine that we can draw a line somewhere and say for everything under this line LPS is the way to go; for everything over that line SMPS is the way to go. I can easily see powerful amplifiers going over this line. Sorry about the generalizing paragraph :). 

[ASRMichael]

Can see a year from now this thread over taking “A Novel way to massively improve the SQ of computer audio streaming! Only 710 pages behind and counting down..... 

[matthias]

13 minutes ago, LTG2010 said:

From what I can see of the pictures of the Insides of Extreme published, it's already using buck converters and not linear regulators and I'm assuming this DC - ATX is based on that system?

High powered linear regulators need high heat dissipation and there do not appear to be any regulators fitted onto the heatsinks. placing them below would mean they heat up the base and the rest of the components above. Well I'm just guessing I would be interested to know.

I'm currently using a C-L-C Filter with HDplex 800 DC-ATX. An 'Audiopile' DC - ATX would be perfect.

 

MULTI-RAIL ULTRA LINEAR POWER SUPPLY

Our 400W linear power supply is a new inhouse development, especially created for the Extreme and its special needs. (from Taiko HP)

 

Matt

 

[LTG2010]

50 minutes ago, RickyV said:

The rest of the power supply of the extreme is transformer, rectifier, choke, capacitor bank, atx regulators.

Hi Ricky, I think it's a C-L-C filter, transformer- rectifier bridge, capacitor, choke, capacitor, regulators.

There appears to be a large mundorf capacitor and bypass capacitor connected to the bridge before the choke, the choke migt be a dual choke, ie connected both to positive and 0V, well I suppose the aim is not to mimick the Extreme, but that's what I would suggest.

[RickyV]

18 minutes ago, LTG2010 said:

Hi Ricky, I think it's a C-L-C filter, transformer- rectifier bridge, capacitor, choke, capacitor, regulators.

There appears to be a large mundorf capacitor and bypass capacitor connected to the bridge before the choke, the choke migt be a dual choke, ie connected both to positive and 0V, well I suppose the aim is not to mimick the Extreme, but that's what I would suggest.

oh yes you are right, I forgot the first C. 
Mmmm I don’t know, mimicking could be useful but you still need to listen, voicing like Nenon is doing.

[Darryl R]

On 1/9/2021 at 11:23 AM, AME said:

Hello Nenon,

 

Thanks!

 

It was a lot of handwork, but I quess a lot can also be done with an computer controlled milling machine.

I will drop an email to Larry from HDPlex to see if he is interested.

 

I will definitely wait with the power supply decision.

 

A picture of the Farad ATX PSU next to 400W HDPlex PSU

 

 

 

 

I'd asked Mattijs about this a few times.  He said he'd been too busy with Super3 sales to finalize his design, so thanks for posting it.  However I was most curious about Pachanko's new lines, which look beautifully made, but I've only seen one customer post on them (and it was very good).

 

Nenon announced this project just in time to lay on the brakes and wait.

[LTG2010]

10 minutes ago, RickyV said:

Mmmm I don’t know, mimicking could be useful but you still need to listen, voicing like Nenon is doing.

I wasn't suggesting mimicking :) I was suggesting a C-L-C filter with dual choke, that's what I ended up with after L-C and single choke, The increased supply line noise rejection was very easy to hear, anyway look forward to seeing what @Nenon ends up with.

[Exocer]

Up until now, I had assumed all linear supplies used regulators. Found this interesting:

 

https://mcitransformer.com/power-supply-basics-part-1-unregulated-linear-regulated-linear/#:~:text=Unregulated Linear Power Supply&text=Apart from that%2C the basic,voltage through a filter capacitor.

"

Unregulated Linear Power Supply

As mentioned earlier and as the name suggests, regulation of voltage isn’t involved in an unregulated linear power supply. Apart from that, the basic functions of an unregulated power supply are similar to regulated linear power supplies. The AC input voltage, just like a regulated power supply, is initially processed and then transformed into rippled DC voltage through a filter capacitor. But since the unregulated power supplies don’t have a voltage regulator circuit, the ripples in the DC output voltage cannot be done away with. This leads to voltage sags, which are often reflected in the output.

Recap: Regulated Linear vs. Unregulated Linear Power Supplies

In short, regulated power supplies using a voltage regulator circuit, convert the AC main voltage into a clean, stable DC voltage without noise and variations. In contrast, an unregulated power supply provides a rippled-DC voltage by only rectifying the AC.

Key Differentials

Output Voltage

• Regulated – There is no variation in the output voltage of regulated power supplies irrespective of the current drawn by the load. The voltage is independent of the load current.
• Unregulated – The output voltage of unregulated power supplies changes with the output current, mainly because of the high internal resistance of the power supply.

Applications

• Regulated – Application areas of a regulated power supply include TVs, computers, and other highly sensitive electronic devices where voltage variations can be fatal.
• Unregulated – Electrical devices like LED lamps and DC motors that are not so sensitive to voltage fluctuations can use unregulated power supplies.

Cost

• Regulated – Regulated power supplies are comparatively expensive as they integrate voltage regulation circuits.
• Unregulated – Unregulated power supplies are low-priced due to fewer component requirements."

[matthias]

11 hours ago, Nenon said:

I personally don't feel like putting an AC SMPS in front of the Taiko ATX would be a step in the right direction. I think we get the best of both worlds the way this is.

 

The more I think about it the more I agree, it really seems to be the best of both worlds.

 

Matt

 

 

[Gavin1977]

11 hours ago, RickyV said:

Ah, okey.

I thought about using a Hypex SMPS to feed a HDPLEX DC-DC converter, however there is also an additional hitch in that these SMPS are untested in this application and are only used to power amplifiers which obviously only have high transient power peaks.  Providing a continuous 60w+ from a Hypex SMPS might cause it to fail pretty quickly.

[MarcelNL]

@Exocer, what you posted was pretty much what I have/had in mind, the more steady the current draw is the better 'regulation' you'll see using an unregulated PSU design. It is doable to get rid of ripple current the diagram shows by using good enough filtering (CLC for example) a regulated PSU does not do much for ripple filtering anyhow.

 

Not using upsampling and FIR etc in my system the load might be steady enough for the CPU as smaller power peaks can easily be handled by fast enough capacitors of which enough are needed anyway to keep the PSU hum down due to the high-ish current (not so sure about the rest of the board). 

 

All developments sound interesting yet committing to any which one without even seeing a more detailed description or specs (I'm not a great believer in specs) I'll wait a bit for now.

 

[RickyV]

56 minutes ago, MarcelNL said:

@Exocer, what you posted was pretty much what I have/had in mind, the more steady the current draw is the better 'regulation' you'll see using an unregulated PSU design. It is doable to get rid of ripple current the diagram shows by using good enough filtering (CLC for example) a regulated PSU does not do much for ripple filtering anyhow.

 

Not using upsampling and FIR etc in my system the load might be steady enough for the CPU as smaller power peaks can easily be handled by fast enough capacitors of which enough are needed anyway to keep the PSU hum down due to the high-ish current (not so sure about the rest of the board). 

 

All developments sound interesting yet committing to any which one without even seeing a more detailed description or specs (I'm not a great believer in specs) I'll wait a bit for now.

 

unregulated is only possible with this Taiko DC-atx module. Why?, because this module is designed for that. If you would try to feed an hdplex dc-atx module unregulated dc it will fail.

[Exocer]

Very excited about the unregulated approach. When funds allow, I will likely end up building the unregulated supply to feed the Taiko DC-DC converter. It will be interesting to compare SQ with the PHSR7T vs the unregulated supply. The SR7T is no slouch, but, I would guess the unregulated supply would provide a "less restricted" means of keeping up with sudden power demands. Plus, we will get @Nenon's voicing if we follow the exact recipe. Very interested in how that would translate to my system...

[JonD]

On 1/8/2021 at 1:00 AM, Ianderson said:

In the grand scream of things a higher priced case of a) better quality b) more room is universally advantageous. If one is shooting for great SQ, the price is secondary. Besides, the case is chump change compared to the rest of the BoM. (particularly the LPS’s)

 

As an aside I heard a rumour that Innuos was going to release a audio grade Network Switch for around $3k. Not sure how valid that might be.

 

I have heard they are looking at their ethernet boards, yes. Waiting to see what they have planned. Ethernet version of the Phoenix or an update to the Zenith etc.

[MarcelNL]

does anyone have a link to CPU desig specs? They must exist, f.e. for MB designers, so far I haven't found any.

[MarcelNL]

BTW; a (what I consider a solid with good information for folks new to PSU designs) response from Emile (Taiko) on a question around PSU over at WBF on page 276 and 277 somehow I cannot paste the link: 

[AME]

17 minutes ago, MarcelNL said:

BTW; a (what I consider a solid with good information for folks new to PSU designs) response from Emile (Taiko) on a question around PSU over at WBF on page 276 and 277 somehow I cannot paste the link: 

 

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-276

 

 

[matthias]

19 hours ago, Nenon said:

The last comment I want to make is that no matter how advanced the SMPS technology becomes, to me a well designed LPS is still the best option for powering low current devices. I do power my USB card on my server with a Sean Jacobs DC4 LPS and that makes a hell of difference. I find it best to use the right technology where it matters. For a dual CPU server that can easily draw 300W, this SMPS technology makes sense and works great. But for low current applications, nothing can beat a good old linear power supply. That's my 2 cents. I imagine that we can draw a line somewhere and say for everything under this line LPS is the way to go; for everything over that line SMPS is the way to go. I can easily see powerful amplifiers going over this line. Sorry about the generalizing paragraph :). 

 

Nenon,

I think it is really interesting that Emile in his "white paper" post on WBF mentioned that even in the range from 500mA up (which is typical for an USB card) DC-DC SMPS can have advantages in comparison to LPS.

 

Matt

[guiltyboxswapper]

19 hours ago, Nenon said:

The last comment I want to make is that no matter how advanced the SMPS technology becomes, to me a well designed LPS is still the best option for powering low current devices. I do power my USB card on my server with a Sean Jacobs DC4 LPS and that makes a hell of difference.

To be honest what very high quality SMPS options do we have at low voltages?  The likes of Benchmark Audio have been using SilentSwitchers to achieve this with SMPS, but they're often not implemented in any high (1a+) current form nor do they get much attention on the whole.

 

So yes certainly for now LPS is the way to go.

[Nenon]

14 minutes ago, matthias said:

 

Nenon,

I think it is really interesting that Emile in his "white paper" post on WBF mentioned that even in the range from 500mA up (which is typical for an USB card) DC-DC SMPS can have benefits in comparison to LPS.

 

Matt

 

If I have to draw a line, it would be like this:

3A or less - Well designed LPS rules. 1A or less has huge advantages here. 

3A to 5A - Depends on the design of the power supply.

More than 5A constant draw - Well designed SMPS rules. 

 

An AMD 7 3700x running Euphony is closer to the 3A limit. Maybe even less. But when I built LPS for that CPU I had it overprovisioned to be capable of 20A-40A.  That helped. 

The dual Xeon Cpus are more like 7A. You can see why I am going with the Taiko ATX for the dual Xeon. But also, as I previously said, it's a drop in replacement for the Hdplex and sounds much better. 

[matthias]

10 minutes ago, Nenon said:

 

If I have to draw a line, it would be like this:

3A or less - Well designed LPS rules. 1A or less has huge advantages here. 

3A to 5A - Depends on the design of the power supply.

More than 5A constant draw - Well designed SMPS rules. 

 

An AMD 7 3700x running Euphony is closer to the 3A limit. Maybe even less.

The dual Xeon Cpus are more like 7A. You can see why I am going with the Taiko ATX for the dual Xeon. 

 

I agree, this is the status of the current tech, maybe Emile is one step further with his DC-DC

SMPS.

But to have an unregulated LPS which outputs to an DC-DC SMPS is brilliant.

 

Matt