New HDPLEX 200W LPS

[austinpop]

13 hours ago, rickca said:

Who says the new 200W LPS is midfi?  It isn't a Paul Hynex SR4, but that costs about $425 for a single adjustable rail.  I bought my SR4 with a DC3FSXLR silver cable that cost $115.  And that isn't even high end compared to the SR7. But I'm not going to want an SR4 or an LPS-1.2 for every device.

 

I really haven't done enough listening yet.  My time for audio has been pretty limited recently.  I've spent quite some time recabling the Windows 10 i7-6700K server I built 3 years ago to try the HDPLEX 400W DC-ATX.  Mostly I've just been figuring out the best way to use the HDPLEX 200W LPS in my system, so I've done some testing on both my server and my NUC7PJYH/AudioLinux endpoint.  For the endpoint, I've settled on using my SR4.  

 

 

 

It would be nice to know several things, when you have the time to try them:

• what is the SQ of an LT30451 rail of the HDPlex 200 vs. the SR-4 vs. any other "audiophile" PSUs you may have (LPS-1.2?)
• what is the SQ difference between the fixed (non LT30451) 12V rail and the LT30451 rail set at 12V? 

The above 2 comparisons will give us a sense of where new HDPlex 200 stands in the pecking order, both its regular rail vs. its LT30451 rail.

 

Finally, it will be good to know how the HDPlex 19V rail driving the 400W DC-ATX adapter compares to your current ATX power supply.

 

That concludes your homework assignment.  Anything extra from the above will be awarded extra credit.

[austinpop]

Just now, bobfa said:

I will also take that homework assignment!  My HDPLEX 200 should be here tomorrow.  IT will be powering my NAS and my Roon Server.    

 

Love it!

[austinpop]

6 hours ago, rickca said:

I bought a $115 silver cable from Paul Hynes with my SR4.  So it isn't fair to compare it to the HDPLEX unless I get some cables of comparable quality for it.

 

Since the HDPlex also uses an XLR connector for output, could you not try Paul's cable on the HDPlex, and compare it to the stock cable, to assess the magnitude of change the cable makes?

[austinpop]

3 hours ago, rickca said:

The DC3FSXLR has a 4 pin XLR connector at the power supply end to mate with the chassis connector on the SR4.  The XLR connector on the HDPLEX 200W is 3 pin.

 

Ugh I forgot about that. 

[austinpop]

Could someone please summarize how to apply audiophile best practices (for example, power CPU separately from mobo?) to best utilize something like this (the HDPlex 400W PSU):

From what I can tell, the standard ATX set is 1x24pin, 2x8pin, and 1x4pin connectors. For a headless, fanless, server, which will only have an NVMe Optane SSD and a good network adapter (say the JCAT Net, powered by an external 5V PSU), which of the Molex'es will I need for the server? 24-pin to mobo, 4-pin to CPU, 8-pin for PCIe?

 

I am fairly clueless when it comes to ATX, so feel free to educate me.

[austinpop]

5 minutes ago, LTG2010 said:

This is a diagram of the 24 pin atx and 4 pin atx on your motherboad.

You will need a 24 pin atx extension cable and a 4 pin one.

Cut off the male ends, leaving the ends that will connect to your motherboard.

On the 24 Pin:

Strip the wire ends approx 1 - 2cm.

Connect the yellow 12V together.

Connect the orange 3.3V together.

Connect the red 5V together.

Connect the black ground together.

Connect the purple 5V SB to the red 5V

Connect the grey PowerOK to the red 5V.

Ignore / Cut off the Green PS ON, white -5V and blue -12V.

You now have 3 connections, 3.3V, 5V, 12V for your HDplex rails, the black is common on the board so all 3 power supply rails ground can be joined together or daisy chained, with the motherboard black ground cables.

Similar principle for the 4 pin 12V this will connect to the 4th HDplex (12V rail) seperately powering CPU.

 

Interesting. That is starting to make sense!

 

It would be nice if Larry @ HDPlex could supply the 200W with these connectors, so you don't need adapters, like he does the 400W, no?

[austinpop]

13 minutes ago, rickca said:

All you need is a 24-pin for the motherboard and a 4-pin or 8-pin for the CPU (the 8-pin is usually called EPS12V).  The PCIe is only needed for some graphics cards with either a 6-pin or 8-pin connector because they can't get enough power from the PCIe slot alone.

 

Thanks for that clarification. 

 

13 minutes ago, rickca said:

The connectors are on the 400W DC-ATX unit.  The 200W LPS 19V is designed to be used with this 400W DC-ATX unit.

If you get the 400W ATX LPS you don't need a DC-ATX unit.  Make sense?

 

Yes, understood. It appears that there is a tradeoff to be made between the 200W model (with DC-ATX 400W add-on) and the 400W model. Is the reasoning below valid?

 

With the 200W:

• Advantage
  • 2 rails with LT30451 (btw - is the LT30451 equal/better/worse than the celebrated LT3045 regulator?)
• Disadvantage
  • Need a DC-ATX converter (SQ hit due to inherent quality loss with DC-DC converter?)
  • All ATX voltages derived from one (19V) rail

With the 400W:

• Advantage
  • direct linear ATX supply (no ATX converter needed)
  • Not sure if ATX voltages (12, 5, 3.3) derived from same rail, or separate rails?
• Disadvantage
  • No rails with LT30451.

Did I get that right?

[austinpop]

7 minutes ago, LTG2010 said:

I"m guessing for your Dell server, but generally, on a high powered set up, you might have an 8 pin plus 4 pin, for CPU plus 8 pin for PCIE graphics high power cards. I would just connect the 24 pin plus CPU. The CPU will benefit from its own rail, the NVME can be powered from the 24pin.

 

Thanks! No, I'm toying with the idea of a custom fanless Roon Server, running Audiolinux in RAM.

[austinpop]

6 minutes ago, rickca said:

See my post

https://audiophilestyle.com/forums/topic/46748-new-hdplex-200w-lps/?do=findComment&comment=896924

 

 

Thanks a lot for the pointer!

 

Now I'm a bit confused again, because the slide on VIOC says this:

• LDO intelligently controls upstream switching regulator to regulate the LDO’s differential voltage (VIN – VOUT) to either a constant voltage or an adaptive voltage as a function of ILOAD

I thought high-quality linear PSUs did not use switching regulators, even if they are upstream of a linear regulator like the LT3045-1. Which suggests that the HDPlex uses switching regulators on these rails? I know - a question best asked to Larry. Just surprised to see this.

[austinpop]

14 minutes ago, octaviars said:

 

The big span of output voltages on the new HD Plex (3.3V/5V/7.5V/9V/12V/15V) from the LT3045 board would be impossible when the input to the board is 16V and needs a pre regulator as the voltage drop is so big. A switching DC/DC converter can do big voltage drops without producing much heat.

 

 

 

8 minutes ago, rickca said:

Yes.  See this post from Sbooster earlier in this thread.  

https://audiophilestyle.com/forums/topic/46748-new-hdplex-200w-lps/?do=findComment&comment=896553

 

 

Thanks for the clarification. I'll shut up now and go back and read the thread from the beginning.

[austinpop]

1 hour ago, rickca said:

OK I need to get off AS and do some actual listening now.  

 

You mean get off AS and sit on ASS? 😛

[austinpop]

46 minutes ago, tboooe said:

someone smarter in electronics

 

≠ me!

 

 

[austinpop]

1 hour ago, LTG2010 said:

There aren't any negative voltages on your board, or most boards for that matter that's why I left them out.

I also left out pin 16 PS on - since you cannot switch a linear power supply on or off from your computer / server unless it's specially designed to do so.

 

I’m no expert, but I was also told by a PSU designer that the -ve DC voltages are kegacy, and not used in modern motherboards. The last device he remembered requiring it was an ISA card. That takes you back doesn’t it!

[austinpop]

4 minutes ago, rickca said:

You mean legacy ... kegacy is something Supreme Court Justice Kavanaugh knows about.  Anyway, I think you're right.

 

Apologies about the in joke to our non-American users.

 

 

 

“I liked beer. I still like beer.”

[austinpop]

1 hour ago, bobfa said:

 

I ran the power supply for 24hours on the bench just loaded down with a small NAS and a NUC.  I then put the supply on my Sonic Transporter Server.  After another day I started listening on the small system in my office and found some interesting changes.  This afternoon I was able to "borrow" some time to sit down and listen on the Living Room system.  WELLLLLLLL,.....

 

The power supply is worth it to me.  It increases the server cost from $2000 to $2500.  The whole experience is better.  A couple of the BT tracks I use for reference show more detail and are more "open/relaxed".  The sound track to The Lord of the Rings; The Return of The King feels more "alive".  The choir vocals stand out better. The Billy Boyd vocal sounds more like you are in the hall listening to him sing.  The music is more moving.  Annie Lennox feels more present singing "Into the West" . 

 

It was hard to go back to work.  I really like this kind of homework.

 

 

 

Bob,

 

With what default PSUs are you comparing the HDPlex to?

 

 

[austinpop]

5 hours ago, rickca said:

This is an introduction to an upcoming post I have in mind with implementation experiences.

 

The NUC7i7DNBE running AudioLinux is a popular choice as a server.  At 15W TDP, it can be powered by a Paul Hynes SR4 (either 12V or 19V model).  But that 15W TDP assumes you're running at base clock speeds (no turbo boost).  The NUC7i7DNBE has base/boost clocks of 1.9 GHz/4.2 GHz, so disabling turbo boost significantly constrains its performance potential.

 

What if you need a really powerful server with base clocks at 4 GHz like an i7-6700K for HQPlayer's most CPU intensive filters/modulators?  Now you have a CPU with 91W TDP.  If you also want to exploit CUDA offload with an Nvidia card like a GTX 1080 (180W) or GTX 2080 (215W), how do you power this machine?

 

Once you move from a NUC to a mini-ITX or ATX motherboard, you need an ATX power supply.  These motherboards typically have a 24-pin connector and an 8-pin EPS12V connector for the CPU.  Can we significantly improve sound quality by using an LPS in such a configuration?  

 

The HDPLEX 200W LPS ($485) can deliver up to 10A on its 19V fixed rail.  Together with an HDPLEX 400W DC-ATX ($95), we have a solution that's a good fit.  With 4 separately grounded rails including two adjustable voltage 2A rails using 4 x LT3045-1 each, we can also power some accessories like a JCAT USB card, an ISO REGEN, a network switch, etc.  It's a really versatile and cost effective solution.

 

I have an i7-6700K system running Windows 10 with music files on an internal hard drive.  I use it as a server for my NUC7PJYH/AL/SR4 endpoint connected via wifi.  I also use it as a standalone system with an iFi micro iDSD and Focal monitors, so it's got to sound good in its own right.  The server and endpoint are in different rooms.

 

Does the sound quality of the server affect the sound quality of the endpoint?  Many users who have experimented have emphatically concluded that it does.  Will this still be true with an EtherREGEN in front of a wired endpoint?  We don't know yet.

 

I'd been powering my Windows 10/i7-6700K system with a Seasonic X650 Gold.  It's a very good 650W PSU that got 9.1/10 from power supply review site JonnyGURU. 

 
With the HDPLEX 200W LPS/400W DC-ATX, my server sounds so much better that I'm planning to replace Windows 10 with either AudioLinux or Windows Server 2019/AO 3.  I'm confident a more streamlined/lower latency OS will benefit even more from the HDPLEX solution.

 

Thanks for that great summary!

 

1 hour ago, tboooe said:

I wonder if this combination sounds better than powering the server directly with the HDPlex 400w AC to DC lps atx which I have?  I have the 200w lps on order and I have the first generation DC-atx so I can do a.comparison.  No exactly a perfect test because I don't the current version of the DC-atx but an interesting test nonetheless.  Based on specs alone, the 400w lps is better than the DC-atx in terms of ripple.

 

Yes, this is a burning question in my mind too. What is the penalty in SQ of the DC-ATX adapter fed from a single 19V rail, vs. using 4 independent rails directly driving the ATX and EPS voltages.

[austinpop]

1 minute ago, rickca said:

@austinpop and @tboooe you may want to read this post from @lmitche again.  He has already made some progress in answering this question with his configuration #3.  I don't have the cable necessary to try this myself.

https://audiophilestyle.com/forums/topic/55247-audiolinux-server-configurations-software-hardware-and-listening-impressions/?do=findComment&comment=924226

 

 

Yes, Rick, Larry's result is very promising, but is limited to the use of separate 12V rails for ATX and EPS.

 

One wonders now whether separating out the 3.3, 5, and 12V still remaining on the ATX to their own rails makes a further difference?

[austinpop]

And maybe someone can explain how you get a "400W Hi-Fi DC-ATX Power Supply" while fed from a single 19V rail of a 200W (or 160W?) LPS?

[austinpop]

4 minutes ago, soares said:

Just to remind that if you prefer you can order sboosters with up to 3 rails:

https://www.sbooster.com/OEM/

 

 

Very interesting! Thanks for the pointer. Has anyone had any experience with these units?

[austinpop]

Looking for some advice from the experienced hands here, who have used the HDPlex DC-ATX converter.

 

I want to use it with a Paul Hynes SR-7 19V/6A rail, but in a quick test we tried, the computer didn't power up. It also seemed to trip the over-current protection, as we had to replace the fuse on the PSU before it powered on again.

 

This is too expensive a PSU to experiment with, so I asked Larry for ideas. He suggested I try it with a laptop adapter. The most powerful unit I had lying around is a 90W 19.5V/4.62A supply, which conveniently has a 7.4x5.0mm tip, which fits perfectly in the HDPlex-supplied 7.4x5.0 to 6-pin Molex adapter. Here's the test:

1. connect power supply like this:
   • 19V 90W laptop adapter -> 7.4x5.5 - to 6-pin adapter -> 6-pin input on the DC-ATX converter
2. Nothing else attached to the converter

Result: the amber light on the converter comes on for a few seconds, then goes off. The laptop supply seems to go into overcurrent protection, as I have to disconnect and reconnect for it to come on again (LED indicator).

 

Before you tell me this converter is rated for 400W load, and requires a hefty input supply - I know this. My plan is to only drive ATX (not EPS or any other output) for a very low power use case, where the expected power draw should be in the 20-30W range. 

 

According to Larry, the DC-ATX Converter requires a significant inrush current to "charge the capacitors." but he does not know what the requirement is. Has anyone experienced this, and do you have any experimental data that suggests what the inrush requires from the input power supply? This seems to be the limiting factor to using this converter at this point.

 

I am currently looking for a more powerful SMPS (sadly I threw away all me old adapters in my last house cleaning ) to test this on.

 

 

 

[austinpop]

Thanks for the suggestions. Please note, the issue I'm seeing is technically during self-test of the HDPlex DC-ATX converter. Yes, I'm aware of the need for the jumper if not connected to the motherboard.

 

I was able to borrow a more powerful adapter from a friend, with some interesting results. So to recap - even with no output connectors connected, the HDPlex DC-ATX adapter has an amber LED that will light up and stay lit if it successfully passes self-test.

 

The failure case I was seeing was that the amber LED would come on for a few seconds, and then go off. Here are some new results with 3 power supplies:

• 90W laptop adapter - 19.5V/4.62A. This did not work
• PowerAdd Pilot batterY - 20V/4.5A. This did not work
• 120W laptop adapter - 18.5V/6.5A. This did work.

Of course, the next step is to actually connect it to a mobo, and continue testing further, but this was a test to see if the DC-ATX converter was defective or not. 

 

The finding here - at least for now - is that the DC-ATX converter is not defective, BUT it does require a significant amount of inrush current (even without any load). At 19V, this inrush current IC is:

• 4.62A < IC < 6.5A.

 

[austinpop]

Just now, rickca said:

Since your 19V/6A SR4 rail didn't work, haven't you proven that 6A is inadequate for the inrush current?

 

Not necessarily, The SR7 19V/6A rail should have a hefty amount of headroom, so this data will be sent to Paul for his thoughts.

[austinpop]

4 minutes ago, rickca said:

This is actually quite helpful to me.  I was considering using the 12V rail of my HDPLEX 200W LPS just for EPS12V on my motherboard.  My plan was to get a separate 19V LPS to use with a 400W DC-ATX for the 24-pin.  This could be another HDPLEX 200W, but I was considering something like a 19V/3A Farad.  Now it seems like the Farad wouldn't work with the 400W DC-ATX.

 

Yes, I think this is a key takeaway. When it comes to powering a mobo, either through ATX or EPS, even if the expected steady state power consumption is expected to be low, there are going to be current transients, especially during power on. 

 

The case I ran into was simply within the HDPlex DC-ATX converter, but I am sure once you connect this to the motherboard, there are further transients associated with all the other regulator modules in the system.

 

How this aggregates to determine the current rating for a suitable power supply is unclear. I do know someone who was able to successfully power the ATX input on his mobo, through the DC-ATX converter, using a 19V/5A rail on his SR7. I tried this experiment with another AS'er's SR7 with a 19V/6A rail and ran into an issue, but at the time I was unaware of this DC-ATX transient demand. I'm not sure what we did wrong, which is why I'm backtracking to figure it out.

[austinpop]

1 hour ago, Superdad said:

 

I am not not understanding the need/desire to use a 19V DC-DC switching converter board when all the mobo rails call for 12V and below.  

 

I am trying to replicate this:

 

On 5/24/2019 at 2:43 PM, romaz said:

but as I have figured out a way to independently power both the 8700K CPU and ATX motherboard (using a DC-ATX converter) with SR7 rails

 

Roy used an HDPlex converter driven by an SR7 rail for the ATX connection. I wanted to replicate this to see how it sounds. Even with the supposedly inferior SQ of the HDPlex DC-ATX converter, the so-called "hoarder of SR7 rails" found this a major SQ step up from the HDPlex 400 ATX.

 

There I go again - being  a messenger and proxy.