LPSU Choices?

[bibo01]

5 hours ago, LTG2010 said:

I think they are about to upgrade their supplies with LT3045's it may be worth investigating that. 

I would not aspect their more powerful ATX-PS (400W?) to be upgraded to LT3045 as this regulator only supports low amperage.

It was possible for their less powerful PS model. 

[Tomslin]

6 hours ago, LTG2010 said:

I think they are about to upgrade their supplies with LT3045's it may be worth investigating that. 

 

Regarding this discuss about different linear regulators, now I will come with maybe a remarkable statement: No regulator at all will be the all best. How is it possible, you may wonder? It is fully possible, only the voltage levels are quite close. As mentioned before, the motherboard is already full of different regulators. And it's definitely nothing about any nuclear physics here. And I promise it will give significally higer sound quality. Atleast it has been during my trials with this.

[the_doc735]

On 7/17/2018 at 10:58 AM, Tomslin said:

Hi,

Maybe you can get some tips from my discussion here:

http://jplay.eu/forum/index.php?/topic/1979-rebuild-and-improved-power-supply-for-a-singlepc/?hl=rebuild

@Tomslin 

 

Just got this from Larry at HDPlex:

 

"Hi

I looked at his example and don't understand how PC power management chip on his motherboard can communicate with his PSU and function as it should during power on process. Without the control circuit from PSU to tell power management chip on the motherbord that PS ON is successful, the PC will just think it as a failed start and shut down again.

The PS-ON line needs to be at 5V when PC is off and will be lowered to 0V when Power ON is pushed.

All 12v 5V 3.3V needs to rise within 500ms of time.

So I think it is not possible without additional control circuit. In fact, the reality is more complex than time sequence and control.

You also need MSM chip to help the timing for certain motherboards..." 

 

I was thinking of the HDPlex 200W with 4 outputs, but Larry said it is not possible to do what you have done with only the 3.3/5/12v?

 

So I'm hoping you can throw some light on his email response (above) please?

 

Cheers!

 

[Tomslin]

On ‎7‎/‎17‎/‎2018 at 11:58 AM, Tomslin said:

Hi,

 

Maybe you can get some tips from my discussion here:

 

http://jplay.eu/forum/index.php?/topic/1979-rebuild-and-improved-power-supply-for-a-singlepc/?hl=rebuild

 

But it was a long time ago and much has happened since, but still quite relevant.

 

Here is one thread I read at that time when I sought inspiration, thanks Nige

 

http://www.tirnahifi.org/forum/viewtopic.php?f=15&t=2832

 

 

 

 

 

The answer is simple: I don’t use control circuits, because it’s not needed. That explanation is also found in the links I published earlier in my post above.

We don’t really talk about the same things here. To only power a motherboard it will be enough with these three power lines and CPU power. However, if you want to build a PSU that’s fully compatible with the ATX standard, you need control circuits as he stated.

Less complicated is better, provided that it works, of course. That rule I always try to follow.

[the_doc735]

7 hours ago, Tomslin said:

 

The answer is simple: I don’t use control circuits, because it’s not needed. That explanation is also found in the links I published earlier in my post above.

We don’t really talk about the same things here. To only power a motherboard it will be enough with these three power lines and CPU power. However, if you want to build a PSU that’s fully compatible with the ATX standard, you need control circuits as he stated.

Less complicated is better, provided that it works, of course. That rule I always try to follow.

thank you!

[Solid-State]

A suggestion for YOU IS ,
Purchase a wide voltage PICO PSU and an Optima car battery to supply it with,
so you/we don't have to worry more
and its sounds good

[Solid-State]

On 9/30/2018 at 11:00 PM, the_doc735 said:

@Tomslin 

 

Just got this from Larry at HDPlex:

 

"Hi

I looked at his example and don't understand how PC power management chip on his motherboard can communicate with his PSU and function as it should during power on process. Without the control circuit from PSU to tell power management chip on the motherbord that PS ON is successful, the PC will just think it as a failed start and shut down again.

The PS-ON line needs to be at 5V when PC is off and will be lowered to 0V when Power ON is pushed.

All 12v 5V 3.3V needs to rise within 500ms of time.

So I think it is not possible without additional control circuit. In fact, the reality is more complex than time sequence and control.

You also need MSM chip to help the timing for certain motherboards..." 

 

I was thinking of the HDPlex 200W with 4 outputs, but Larry said it is not possible to do what you have done with only the 3.3/5/12v?

 

So I'm hoping you can throw some light on his email response (above) please?

 

Cheers!

 

 

 

and I think Larry should read this thread !!!!

 

http://www.tirnahifi.org/forum/viewtopic.php?f=15&t=2832&sid=c2e2c4fc08e90c101547bcabba8a5e95

[Tomslin]

7 hours ago, TubeMan said:

 

 

and I think Larry should read this thread !!!!

 

http://www.tirnahifi.org/forum/viewtopic.php?f=15&t=2832&sid=c2e2c4fc08e90c101547bcabba8a5e95

 

Indeed. It’s a very good DIY thread in this subject. I guess much inhold are a bit too basic for him, but anyway it provides many useful angles of view. And apparently he didn't know that it’s possible to start and power motherboards without control voltages.

[the_doc735]

13 hours ago, TubeMan said:

A suggestion for YOU IS ,
Purchase a wide voltage PICO PSU and an Optima car battery to supply it with,
so you/we don't have to worry more
and its sounds good

aren't pico's noisy SMPS? The hdplex version for instance has 70mV.

[the_doc735]

On 10/2/2018 at 3:47 PM, Tomslin said:

 

Indeed. It’s a very good DIY thread in this subject. I guess much inhold are a bit too basic for him, but anyway it provides many useful angles of view. And apparently he didn't know that it’s possible to start and power motherboards without control voltages.

May be this is what Larry was talking about?

"The Power Good Signal

In addition to supplying electrical power to run the system, the power supply ensures that the system does not run unless the voltages supplied are sufficient to operate the system properly. In other words, the power supply actually prevents the computer from starting up or operating until all the power supply voltages are within the proper ranges.

The power supply completes internal checks and tests before allowing the system to start. If the tests are successful, the power supply sends a special signal to the motherboard called Power_Good. This signal must be continuously present for the system to run. Therefore, when the AC voltage dips and the power supply can’t maintain outputs within regulation tolerance, the Power_Good signal is withdrawn (goes low) and forces the system to reset. The system does not restart until the Power_Good signal returns.

The Power_Good signal (sometimes called Power_OK or PWR_OK) is a +5 V (nominal) active high signal (with a variation from +2.4 V through +6.0 V generally being considered acceptable) that is supplied to the motherboard when the power supply has passed its internal self-tests and the output voltages have stabilized. This typically takes place anywhere from 100 ms to 500 ms (0.1–0.5 seconds) after you turn on the power supply switch. The power supply then sends the Power_Good signal to the motherboard, where the processor timer chip that controls the reset line to the processor receives it.

In the absence of Power_Good, the timer chip holds the reset line on the processor, which prevents the system from running under bad or unstable power conditions. When the timer chip receives the Power_Good signal, it releases the reset and the processor begins executing whatever code is at address FFFF0h (occupied by the motherboard ROM).

If the power supply can’t maintain proper outputs (such as when a brownout occurs), the Power_Good signal is withdrawn and the processor is automatically reset. When the power output returns to its proper levels, the power supply regenerates the Power_Good signal and the system again begins operation (as if you had just powered on). By withdrawing Power_Good before the output voltages fall out of regulation, the system never sees the bad power because it is stopped quickly (reset) rather than being allowed to operate using unstable or improper power levels, which can cause memory parity errors and other problems.

On pre-ATX systems, the Power_Good connection is made via connector P8-1 (P8 pin 1) from the power supply to the motherboard. ATX, BTX, and later systems use pin 8 of the 20/24-pin main power connector, which is usually a gray wire.

A properly designed power supply delays the arrival of the Power_Good signal until all the voltages stabilize upon turning on the system. Poorly designed power supplies, which are found in many low-cost systems, often do not delay the Power_Good signal properly and enable the processor to start too soon. (The normal Power_Good delay is 0.1–0.5 seconds.) Improper Power_Good timing also causes CMOS memory corruption in some systems.

Note: If you find that a system consistently fails to boot up properly the first time you turn on the switch, but that it subsequently boots up if you press the reset or Ctrl+Alt+Delete warm boot command, you likely have a problem with the Power_Good timing. You should install a new, higher-quality power supply and see whether that solves the problem.

Some cheaper power supplies do not have proper Power_Good circuitry and might just tie any +5 V line to that signal. Some motherboards are more sensitive to an improperly designed or improperly functioning Power_Good signal than others. Intermittent startup problems are often the result of improper Power_Good signal timing. A common example is when you replace a motherboard in a system and then find that the system intermittently fails to start properly when you turn on the power. This can be difficult to diagnose, especially for the inexperienced technician, because the problem appears to be caused by the new motherboard. Although it seems as though the new motherboard is defective, it usually turns out that the power supply is poorly designed. It either can’t produce stable enough power to properly operate the new board, or it has an improperly wired or timed Power_Good signal (which is more likely). In these situations, replacing the supply with a higher-quality unit, in addition to the new motherboard, is the proper solution."

 

Don't I need to worry about that in the 3.3/5/12v ONLY scenario? Doesn't it apply? Thanks!

[LTG2010]

4 hours ago, the_doc735 said:

Don't I need to worry about that in the 3.3/5/12v ONLY scenario? Doesn't it apply? Thanks!

You would connect the power good line to the 5V. This tricks the motherboard to thinkings its connected to an ATX supply, it should boot. Usually there is a short delay in the power good signal 200 - 500ms. If the motherboard does not boot the you can install a cheap delay timer off ebay or there are other methods. In most cases though it should boot.

You need 3 rails 3.3V, 5V and 12V plus your CPU rail 12V. The High current rail will be the CPU rail. The rest can be low current 1.5amp to 5 amp depends on your motherboard. Also you would run your SSD separately with its own 5V rail so ideally a supply with 5 rails.

[the_doc735]

7 minutes ago, LTG2010 said:

You would connect the power good line to the 5V. This tricks the motherboard to thinkings its connected to an ATX supply, it should boot. Usually there is a short delay in the power good signal 200 - 500ms. If the motherboard does not boot the you can install a cheap delay timer off ebay or there are other methods. In most cases though it should boot.

You need 3 rails 3.3V, 5V and 12V plus your CPU rail 12V. The High current rail will be the CPU rail. The rest can be low current 1.5amp to 5 amp depends on your motherboard. Also you would run your SSD separately with its own 5V rail so ideally a supply with 5 rails.

thanks. I need to address this further with you later on if that's ok? take care.

would this be ok: this ?

[LTG2010]

2 hours ago, the_doc735 said:

thanks. I need to address this further with you later on if that's ok? take care.

would this be ok: this ? 

I think I mentioned this supply somewhere earlier. It has 2 high quality LT3045 rails @ 2amps each. You can use 1 to power your SSD the second if you have a quality PCIE USB card or similar that can be powered externally. The 19V 10 amp rail can be used to power a HDplex DC to DC converter ( a high quality PICO ) for the 24 pin motherboard ATX connecter. The 12V - 10 amp rail to power the CPU seperately via the second ATX connector ( 4/6/8) pin.

[the_doc735]

13 hours ago, LTG2010 said:

I think I mentioned this supply somewhere earlier. It has 2 high quality LT3045 rails @ 2amps each. You can use 1 to power your SSD the second if you have a quality PCIE USB card or similar that can be powered externally. The 19V 10 amp rail can be used to power a HDplex DC to DC converter ( a high quality PICO ) for the 24 pin motherboard ATX connecter. The 12V - 10 amp rail to power the CPU seperately via the second ATX connector ( 4/6/8) pin.

My SSD is one of those M.2. samsung EVO things i.e. no typical SATA.

I have a PF i2s PCIe card that can be powered externally.

Isn't pico, SMPS though? That particular one comes in at 70mV ripple.

Yes the other 12v 10amp can be used for the 8 pin CPU mobo connector.

my PC has a maximum wattage of 63 W at boot up, i.e. that is the peak! ...then it settles back to around 33W when idle (no apps on!).

The HDPlex LPSU is rated at 200W so that should be fine, i.e. plenty of headroom overall.

However, in a certain TOMS HARDWARE quote it states that the usual requirement is: 2.0–4.0 amps  on the +5 V, rail.

So, when I adjust the variable  rail to 5v the requirement could be 2-4A !! (on that rail). The others look fine to me! However the HDPlex has a max. of 2 amps on that specific rail and may not be sufficient? Larry at HDPlex has offered me 12v + 12v at 10 amps each and 2 variable rails for the 3.3 & 5v respectively at a maximum of 2 amps each! He doesn't think it will work based on the TOMS HARDWARE quoted 'usual' requirements! 

That's:

ATX/24

3.3v

5v - 4amps! quoted by toms hardware (the HDPlex 200 only has 2 amps @ this rail).

12v

CPU/8 pin

12v

 

....the other 3 rails seem to be within tolerance.

 

MY HTPC:

MSI Z370i GAMING PRO CARBON AC mini-ITX + Intel® HD Integrated/embedded Graphics 630.

Intel Core i7 Coffee Lake 8700 SE Gen.8 2.9GHz LGA1151 TDP 35W CPU.

Pink Faun i2s Bridge PCI-e (expansion/riser) Card.

Samsung (only 250GB) 960 M.2 (2280) Evo PCIe 3.0 (x4) NVMe 3D V-NAND SSD MZ-V6E250BW (x1).

Patriot Viper 4 16GB Dual Ch. DDR4 3000MHz PC4-24000 DIMM PV416G300C6K.

Streacom st-fc9b-opt-alpha PC Fanless Chassis. 

 

I'm OK with GPU as I use the intel 630 embedded device because this PC is for audio/music only so I don't need the discreet GPU's. 

Peak wattage at boot up is 63W, then falls back to idle at 33W in total.

Currently using Seasonic Prime Ultra ’80+’ Titanium 650 Watt ATX M-PSU. But I want to try LPSU (s) without pico.

Like this: see pics, 

[LTG2010]

A few users here are using the HDplex 'Pico' on high end systems, I wouldn't worry too much about the 70mv ripple, the voltages are broken down further by the motherboards poor quality and 'very noisy' regulators. using the 19v rail with the Pico is a way to utilize the HDPlex you mentioned, you will need approximately 5amp for the 5v on your board so it will not work with the LT3045 rails.

A good quality supply starts from the Toroid or Rcore in the case of the HDPlex. A low impedance and it ability to respond quickly to load demands, transients, dynamics, throughout the frequency band. Then choice of quality components, whilst they may measure the same does not mean resistors, capacitors,  will sound the same. The lowest noise regulators may not sound as good as one with a slightly higher measerement, its about listening as well as measurements. Concentrate your best components  (LT3045 etc) for the outputs eg. your Pink Faun card, Hard discs, etc.

If you want to go the full ATX route then you would be looking at a custom build which can be expensive or a collection of smaller Linear supplies.

Moving onto the ATX pin diagram remember this is looking onto the motherboard or the underside of the connector (the pins) that pop onto that.

An easy way to connect up and also for testing is to use an ATX extension cable and cut off the male end.

We don't need the 2 negative voltages pins 14 & 20. We don't need pin 16 (PS On) you cant switch the supply on or off from the motherboard. Connect the orange, 3.3V together. Similarly the red, 5V and yellow, 12V. The black, Ground is common on the motherboard.

Connect the grey pin 8 (power OK) and purple pin 9 (5V standby) with the 5V (red wires)

Similar principle to the CPU ATX connector, bunch the yellow 12V wires together and same to Ground.

It should then boot up.

[the_doc735]

1 hour ago, LTG2010 said:

A few users here are using the HDplex 'Pico' on high end systems, I wouldn't worry too much about the 70mv ripple, the voltages are broken down...

I will answer later in more detail; off to bed!

[the_doc735]

Also see this:

 

Thanks!

[LTG2010]

OK, leaving out the 'PICO' the HDPlex would be more useful with 2 x 12V instead of 19V + 12V, I dont know if it could be supplied that way. That would cover the 2 x12V rails, leaving the 2x LT3045 rails. You could use one for the 3.3V the other preferably for your I2S PCIE card. That would leave the 5V rail which you could power temporarily from your current ATX supply.

[the_doc735]

20 hours ago, LTG2010 said:

A few users here are using the HDplex 'Pico'.....

 

A few users here are using the HDplex 'Pico' on high-end systems,

 

WOW!

 

I wouldn't worry too much about the 70mv ripple, because the voltages are broken down further by the motherboards poor quality and 'very noisy' regulators.

 

OK.

 

using the 19v rail with the Pico is a way to utilize the HDPlex you mentioned,

 

OK

 

you will need approximately 5amp for the 5v on your mobo so it will not work with the LT3045 rails on the HDPlex 200.

 

OK

 

A good quality supply starts from the Toroid or Rcore in the case of the HDPlex.

 

Yes, I understand this.

 

A low impedance and it ability to respond quickly to load demands, transients, dynamics, throughout the frequency band.

 

Understood.

 

Then choice of quality components, whilst they may measure the same does not mean resistors, capacitors,  will sound the same.

 

Yes I know as I max’d out the caps on my Cary 100t DAC, quite a good upgrade.

 

The lowest noise regulators may not sound as good as one with a slightly higher measurement,

 

WOW!

 

it’s about listening as well as measurements. Concentrate your best components  (LT3045 etc.) for the outputs e.g. your Pink Faun card, Hard discs, etc.

 

Will/would do! - My SSD is M.2. Samsung Evo i.e. no typical SATA.

 

If you want to go the full ATX route then you would be looking at a custom build which can be expensive or a collection of smaller Linear supplies.

 

True. Larry has offered a custom version of the ‘200’ with: 3.3/5/12+12 (2amps on the variable!).

 

Moving onto the ATX pin diagram remember this is looking onto the motherboard or the underside of the connector (the pins) that pop onto that.

 

Yes I know, it was me who turned it upside down/inside out to make the pic. match the 24 pin diagram pin out! LOL!! But the locking/catch gives it away, wink wink!

 

An easy way to connect up and also for testing is to use an ATX extension cable and cut off the male end.

 

Great idea! ~ saves me buying: wire, pins, housings etc. Still need to solder on the panel barrel sockets though for LPSU DC cables! 5.5/2.1mm.

We don't need the 2 negative voltages pins 14 & 20. We don't need pin 16 (PS On) you can’t switch the supply on or off from the motherboard.

 

Understood.

 

Connect the orange, 3.3V together. Similarly the red, 5V and yellow, 12V. The black, Ground is common on the motherboard.

 

Understood.

 

Connect the grey pin 8 (power OK) and purple pin 9 (5V standby) with the 5V (red wires).

 

Understood.

 

Similar principle to the 8 pin CPU ATX connector, bunch the yellow 12V wires together and same to Ground.

 

Understood.

 

It should then boot up.

 

Well, let’s hope so ay!

[the_doc735]

20 hours ago, LTG2010 said:

I wouldn't worry too much about the 70mv ripple, because the voltages are broken down further by the motherboards poor quality and 'very noisy' regulators.

Is there an upgrade for those regulators? e.g. paul hynes?

Apparently, supermicro can bypass these on their mobo's somehow? i.e. zenith SE.

[the_doc735]

15 hours ago, LTG2010 said:

OK, leaving out the 'PICO'....

 

OK, leaving out the 'PICO' the HDPlex would be more useful with 2 x 12V instead of 19V + 12V, I dont know if it could be supplied that way?

 

Yes, it can!

 

That would cover the 2 x12V rails, leaving the 2x LT3045 rails. You could use one for the 3.3V the other preferably for your I2S PCIE card.

 

Jord Groen @ Pink Faun said I had to use a 12 volt supply for my version of the i2s PCIe card! (set variable to 12, not 5v!). It uses the standard 4 pin molex plug, (yellow/red/black).

 

That would leave the 5V rail which you could power temporarily from your current ATX supply. 

 

Understood.

[the_doc735]

"An easy way to connect up and also for testing is to use an ATX extension cable and cut off the male end."

 

Great idea! ~ saves me buying: wire, pins, housings etc. Still need to solder on the panel barrel sockets though for LPSU DC cables! 5.5/2.1mm. 

 

 

[LTG2010]

1 hour ago, the_doc735 said:

Is there an upgrade for those regulators? e.g. paul hynes?

Apparently, supermicro can bypass these on their mobo's somehow? i.e. zenith SE.

Not really, but feeding them with a good quality supply helps to lower the noise floor.

The zenith SE uses a standard Pico,(noisier than the HDPLex) but this is fed from a quality supply by Sean Jacobs. But they also feed the CPU rail direct with a seperate 12V rail and the SSD with a 3rd rail, that's where the benefit is.

[the_doc735]

1 hour ago, LTG2010 said:

Not really, but feeding them with a good quality supply helps to lower the noise floor.

The zenith SE uses a standard Pico,(noisier than the HDPLex) but this is fed from a quality supply by Sean Jacobs. But they also feed the CPU rail direct with a seperate 12V rail and the SSD with a 3rd rail, that's where the benefit is.

got ya! - cheers!

[mourip]

5 hours ago, the_doc735 said:

"An easy way to connect up and also for testing is to use an ATX extension cable and cut off the male end."

 

Great idea! ~ saves me buying: wire, pins, housings etc. Still need to solder on the panel barrel sockets though for LPSU DC cables! 5.5/2.1mm. 

 

 

 

Consider using these locking mini-XLR connectors instead. Panel jack with locking cable connector. You will need to re-terminate your cable but it is MUCH more secure. Available easily on eBay.