I recommend it to your attention
Analog power supply designed for high-quality audio PCs, with extremely low electrical noise, stable and efficient power management,
4 separate output circuits. (CPU, USB, HDD, ATX motherboard)
https://luxoraudio.hu/index.php/aruhaz/tapegysegek/luxpc01-18-detail
I don't know if it also ships abroad. It's worth a question.
The price is approx. 340 EUR, 370 USD
Analog power supply designed for high-quality audio PCs, with extremely low electrical noise, stable and efficient power management,
4 separate output circuits. (CPU, USB, HDD, ATX motherboard)
https://luxoraudio.hu/index.php/aruhaz/tapegysegek/luxpc01-18-detail
I don't know if it also ships abroad. It's worth a question.
The price is approx. 340 EUR, 370 USD
Last edited:
Almost same setup like Analog_sa, Intel motherboard with i3 (the 2100T version with 15W TDP) running for the last 10 years, with the exception that I use a picopsu before motherboard regs. Two LT1083 set to 12VDC, one for picopsu and the other for CPU.
Yes it makes difference and on the top of that, it never fails, it would still run after my death
Yes it makes difference and on the top of that, it never fails, it would still run after my death
And one more thing, the linear / transformer based power supply is 20% more efficient than the normal switching supply (450 Watt no name one) and just 2-3% than a Meanwell LRS 100W 12VDC with picopsu (tested idle / running web and videos)
Switching power supplies for low consumption PCs are not that efficient. Mine runs at 27 watts idle.
Switching power supplies for low consumption PCs are not that efficient. Mine runs at 27 watts idle.
I am running a crosshair vii with ryzen 8core 5600x..no graphics card...35W in total...
I am Starting to measure the current per rail, first results:
...rest to be measured soon...but feasible to use some nice little Lt3045 boards.
May I repeat my question from earlier:
Can I use a normal switch to switch all regs simply on ? Or do they have to be switched on by the board...like Peter did it ? I dont want to break this expensive motherboard...
I am Starting to measure the current per rail, first results:
- 8pin 12 CPU 12V connector: starts with 1.3A, goes during boot up to 2.5A, falls than during normal operation to 0.7A
- 3.3V on the 24Pin connector: 0.7A
...rest to be measured soon...but feasible to use some nice little Lt3045 boards.
May I repeat my question from earlier:
Can I use a normal switch to switch all regs simply on ? Or do they have to be switched on by the board...like Peter did it ? I dont want to break this expensive motherboard...
Its in a screened box with filtering. The noise comes from pcb tracks square waves radiating many megahertz.
I am grateful for this schematic.
But I assume there is another rail for the CPU - The LT1117 could not supply that, could it?
Are you doing something similar for the CPU as you did for the 3.3 and 5 volts rails?
No other rails, it's an older 2012/2013 i3, so no 1.8v rail. I think i used LT1084s throughout.
Don't think that higher current presents any real issues. Just use a 3 terminal reg with a transistor current booster and pile up parallel transistors until it meets the current requirements.
Don't think that higher current presents any real issues. Just use a 3 terminal reg with a transistor current booster and pile up parallel transistors until it meets the current requirements.
Thank you very much. Checked out the merchandise and I like it. László says he does ship to Germany, so I'm all set. I'm a DIYer and prefer this approach because I know what I'm getting since I'm the one putting it all together.
Brewed a Sean Jacobs recipe a couple of years back and it turned out great.
True, the Luxor will cost a few bucks more than one of the Chinese brands but the parts quality in some of the ChiFi offerings makes me feel a bit queasy...and considering you're talking $800 to a grand by the time you add shipping? NOPE! Have soldering iron...will burn baby, burn.
Absolutely NOT true. With a highly dynamic load like a PC most linear PSU's produce much more noise in practice. In static behaviour they might measure better but it is very very hard to make a linear PSU that outperforms a good switcher on noise levels on actual load. Most noise comes from earthing problems which comes from safety regulations. That's the area where you can gain a lot, not by going to highly inefficient and slow linear psu's
I TOTALLY disagree
Look at what tomshardware think is the best PC PSU, the Corsair CX450: the results are about -50 to -60dB re. nominal voltages.
I'm in the process of upgrading my test equipment's with the COSMOS ADCiso and Scaler. The Scaler has some sensitivities to the PSU, so I'm looking into suitable 5V USB power supplies. I tested every 5V PSU I have, and here's what gives:
| Brand | Model | Vout | Load | Noise 1MHz (mV) | Noise 22KHz (mV) | Noise LPF 1st (mV) |
| Excellway | HGJI | 5.11 | 500K | 3.20 | 2.00 | 0.60 |
4.89 | 18R | 7.80 | 3.60 | 1.40 | ||
| BPX-001 | 4.95 | 500K | 84.00 | 15.50 | 20.00 | |
4.55 | 18R | 81.00 | 14.00 | 15.00 | ||
| Charger | Router | 5.50 | 500K | 35.00 | 31.00 | 25.00 |
5.22 | 18R | 64.00 | 32.00 | 13.00 | ||
| Charger | Old | 5.18 | 500K | 27.00 | 10.00 | 2.00 |
4.73 | 18R | 10.00 | 45.00 | 2.50 | ||
| Charger | New | 4.97 | 500K | 6.00 | 4.50 | 3.00 |
5.06 | 18R | 72.00 | 11.00 | 1.70 | ||
| ZK-4KX | W/HGJI | 4.97 | 500K | 2.25 | 0.25 | 0.65 |
4.82 | 18R | 1.75 | 1.70 | 1.50 | ||
| ZK-4KX | W/BPX-001 | 4.97 | 500K | 87.00 | 14.50 | 19.50 |
4.82 | 18R | 76.00 | 16.00 | 18.00 | ||
| Battery | Blk | 5.08 | 500K | 1.20 | 1.10 | 0.35 |
4.94 | 18R | 0.70 | 1.30 | 0.50 | ||
| Battery | Blk charging | 5.08 | 500K | 80.00 | 38.00 | 10.00 |
4.94 | 18R | 83.00 | 33.00 | 17.00 | ||
| Battery | Blu | 5.09 | 500K | 2.10 | 1.35 | 0.50 |
4.98 | 18R | 36.00 | 1.60 | 0.60 | ||
| Battery | Blu charging | 4.70 | 500K | 120.00 | 48.00 | 26.00 |
4.65 | 18R | 110.00 | 45.00 | 3.00 | ||
| LT1085-R | 2x5V | 5.01 | 500K | 1.30 | 0.20 | 0.10 |
5.00 | 18R | 0.60 | 0.10 | 0.05 | ||
| LT1085-L | 2x5V | 5.00 | 500K | 1.15 | 0.10 | 0.10 |
4.97 | 18R | 0.46 | 0.10 | 0.10 |
"Noise 1MHz" is straight into my HP3400 RMS needle meter, 1mV full scale, 1MHz BW
"Noise 22KHz" is with a 1st order 14.3KHz lowpass filter for an equivalent noise BW of 22.5KHz
"Noise LPF" is with an additional .1R 100uF lowpass; it shows if the output cap is big enough
"Charger" is USB charger
"Battery" is power bank
"LT1085" is the only analog PSU, simulated noise is 20uV, measured is 100uV, OK-ish because it's done with croc clips on a much polluted bench
Overall, the results are in line with my expectations and tomsharware's findings. For me, analog is the way to go, with power bank a distant second.
Of course, only the rich can afford to power a whole PC with analog PSU's.
Last edited:
Hi YashN, didn't mean to ghost you, but life pulled me under for a bit.
Anyhoo...the Sean Jacobs setup was strictly a single rail job (adjustable voltage) for my Intel NUC. Simple, but not cheap. I spent just over five bills putting the whole thing together.
In the last few days I've built a server using a full size ATX ASUS board and the Luxor LPC250-RED LPS.
The Luxor comes with its own toroidal transformer, which weighs a ton for a 250VA transformer. In the pic it is housed in that Ali Express red finned case. JUST FOR COOL
. I may have retired from my job...but NOT from childhood. NEVER!Price and shipping for the Luxor unit from Hungary to Germany set me back just under $470 USD. You can see full pics and specs on it at the link provided above by Deenoo, to whom I am eternally grateful.
I would have bought one of the HDPlex LPS units, but they never have any. Tired of waiting.
The new build shows its heels to the NUC setup for several reasons. Even though the NUC was attached to a Singxer DDC the new ASUS build employs a Pink Faun I2S bridge, fed its 5 volts by my Sean Jacobs homebrew LPS. Both use APACER Industrial RAM, but the ASUS mobo makes full use of ECC.
Both also use Audirvana Origin, Windows 10, Fidelizer Pro, and Audiophile Optimizer.
Before Audirvana I was using Euphony Stylus. Audirvana is the hands down winner. Euphony makes a total mess of music file organization. Audirvana has turned it into a work of art. The great sounding Linux-based Euphony is lightweight and easy on the resources. Audirvana is anything but. But, WOW!
In the next few days I'll try the new system with a Streacom ZF240 switchmode power supply just to see if I can detect any differences.
Very interested in what your conclusions are. I am about to try to install some lps in my pc.
Theoretically, I thought ECC can only introduce more (instead of remove) noise and jitter because it's fundamentally more steps and checks.
What is that "abomination" (just kidding) ???
Better sound cards like the XONAR DX/STRIX derive 12V off a external connector , bypassing the PCI-E power bus.
Then they go overboard with overbuilt 5 and 3.3V regulators on the card.
I've used these ... they do sound noticeably than a "regular" (PCI-E powered) audio card.
I don't mess around , I use a separate USB audio interface/DAC with a separate 12V supply. For mobile , I will self-power off the USB
, but that is usually a laptop.
My main PC is a super quiet E5-2680 V4 Xeon , the PS is a Corsair RM550 (<10mV all rails).
What? , all the CPU and memory VRM's run off of 12V (below). Xeon sucks up 8A @12V to supply 53A .9V Vcore. ATX 5V and 3.3V aren't used
much anymore. CPU Vcore , PCI-E and memory all run off the 12V. Video card also can use 8-10A 12V....
3.3 and 5V are used mostly for drives (SATA).
I notice the Corsair PS 12V drops to 11.9V with all them low voltage amps flowing.
My external 12V (for the audio interface) is a small SMPS , no analog ?? under 1mV.
OS
Attachments
IMHO the key improvement the external power connector brings could be better ground connection which avoids the motherboard ground currents (GND directly to the PSU instead), lowering the risk of ground loop issues. I have not tested that though...
Load induced interactions on the shared PCI-E bus are the main issue. A power- hog gaming GPU still uses 25-50W through the bus (in parallel with
the audio card.
A separate PSU connection establishes a connection that has just the native ripple of the main PSU.
I made my own connector (when I had a internal audio card) that had an extra low HF ESR (panasonic) cap right at the sound card.
PS - yes , you also have the PC PSU main star ground right to the audio card , instead of the motherboard ground plane (PCI-E bus).
OS
I saw that after my edit time , me bad ...sorry...
. BTW ... I've never seen a use for (-)12V - [pin 14] .... Any sound card op-amp will use a separate (-) buck convertor.
Xonar uses +/- 5V for it's op-amp buffers.
As far as linear supplies , some newer "audiophile" SMPS's - the 5 and 12V supplies use advanced sine shaping/PFC and precision
feedback. Gives microvolt ripple with barely detectable >50Khz artifacts. I use one on analog synth modules - just a cap and small inductor
in series with each module. NO 50-60Hz .... the HF (50Khz) is so small I can barely measure , even trying to amplify the errata,.
I like my linear class AB amplifiers (Wolverine) , but even here , the builder's report better performance using instrument grade SMPS's
I suppose a linear supply has a "nostalgia" value ??? I use a toroid/capacitor bank on my main amp (had to use it somewhere).
OS