Hi,
Most "universal PSU's incorporate PFC inputs, so the only thing your approach adds is "external".
My own PC runs on 12V DC anyway, so it already has the precise situation you suggest and the external SMPS does have PFC on the input.
It still needs massive filtering and safely breaking the earth connection...
Server Farms? They mostly use very large "whole rack" UPS systems, these are double conversion and always on-line, but they output AC. The losses in all these switched mode systems are quite low, if everything is sized/designed correctly.
You may be thinking of the so-called "Blade Centers", these are best thought of as an analog to the classic Mini-Computer, but with whole servers as plug-in modules.
Ciao T
Most "universal PSU's incorporate PFC inputs, so the only thing your approach adds is "external".
My own PC runs on 12V DC anyway, so it already has the precise situation you suggest and the external SMPS does have PFC on the input.
It still needs massive filtering and safely breaking the earth connection...
At least this will cut down on rectifier hash and SMPSUs will probably do better with nice clean DC seeing that's what they try to create anyway before chopping it up. I believe this approach is also used in server farms to save energy, so perhaps it has 'green' credentials too?
Server Farms? They mostly use very large "whole rack" UPS systems, these are double conversion and always on-line, but they output AC. The losses in all these switched mode systems are quite low, if everything is sized/designed correctly.
You may be thinking of the so-called "Blade Centers", these are best thought of as an analog to the classic Mini-Computer, but with whole servers as plug-in modules.
Ciao T
No, you missed 'heavily filtered'
I suspect that the earth might not actually need breaking (at LF) rather being given a high (and lossy) impedance at RF. But I'm speculating, haven't actually tried this on a real PC.
They do mostly now, but I'm thinking ahead here to when energy prices get back to economically sustainable levels.
http://hightech.lbl.gov/documents/data_centers/CEC-TB-40.pdf
I haven´t read the whole thread, but has anyone acyualy measured the performance of their onboard/outboard soundcard?
And what were its specs?
How did it compare to an outboard card?
What are the differences when using all these suggested power supply things?
Just curious. And just kick my butt to some links.
And what were its specs?
How did it compare to an outboard card?
What are the differences when using all these suggested power supply things?
Just curious. And just kick my butt to some links.
Last edited:
Do you have schematics for it? Or failing that, details of components used inside, preferably with photos of layout and internal wiring? I'm not familiar with it.
Hi,
Probably not.
The main noise components from PC's into the mains and earth relate to the SMPS Systems. These tend to run the gammut from around 30KHz to a few 100KHz. Most power filters are not very efficient across this range.
Simply enquire about the noise suppression at 30KHz, 100KHz, 300KHz and 1MHz in dB, as Monster does not seem to provide any relevant information anywhere.
For reference, a quite beefy "filter in a can" from NEC-Tokin (VC-215F) that I have just bought to fit to a Poweramplifier I am modifying filters common mode noise by 60dB in 1MHz-10MHz region but has less than 30dB filter effect at 100KHz (and 40MHz), for differential mode noise the situation is worse, barely 20dB at 100KHz...
Ciao T
Probably not.
The main noise components from PC's into the mains and earth relate to the SMPS Systems. These tend to run the gammut from around 30KHz to a few 100KHz. Most power filters are not very efficient across this range.
Simply enquire about the noise suppression at 30KHz, 100KHz, 300KHz and 1MHz in dB, as Monster does not seem to provide any relevant information anywhere.
For reference, a quite beefy "filter in a can" from NEC-Tokin (VC-215F) that I have just bought to fit to a Poweramplifier I am modifying filters common mode noise by 60dB in 1MHz-10MHz region but has less than 30dB filter effect at 100KHz (and 40MHz), for differential mode noise the situation is worse, barely 20dB at 100KHz...
Ciao T
Hi,
As an alternative, using a modified commercial PSU is also an option... ;-)
It depends on a number of factors.
I use PC's extensively for testing (be they attached to and AP2 or an EMU1616m or a Presonos Inspire for QC use in the factory) and I found that if I do not break the earth completely I can measure Jack with any reliability.
Of course, measurements are not actual subjective performance, but when the "grass" grows in your FFT up to the -80dB line or more you have a problem. Using common mode filters, big chokes with low DCR and all that to break the grounds did surprisingly little.
I think lower powerconsumtion overal does that better. If I consume only 100W instead of 300W for the CPU and Motherboard, my losses if they are 10% also go from 30W to 10W, but overall I have gone from 300VA consumption at an impossible 100% efficiency to 110W at 90% efficiency.
The big gains are elsewhere, not in AC Powered SMPS...
Ciao T
No, you missed 'heavily filtered'
As an alternative, using a modified commercial PSU is also an option... ;-)
It depends on a number of factors.
I use PC's extensively for testing (be they attached to and AP2 or an EMU1616m or a Presonos Inspire for QC use in the factory) and I found that if I do not break the earth completely I can measure Jack with any reliability.
Of course, measurements are not actual subjective performance, but when the "grass" grows in your FFT up to the -80dB line or more you have a problem. Using common mode filters, big chokes with low DCR and all that to break the grounds did surprisingly little.
I think lower powerconsumtion overal does that better. If I consume only 100W instead of 300W for the CPU and Motherboard, my losses if they are 10% also go from 30W to 10W, but overall I have gone from 300VA consumption at an impossible 100% efficiency to 110W at 90% efficiency.
The big gains are elsewhere, not in AC Powered SMPS...
Ciao T
If your FFT is 64k points and your grass is -80dB within those bins, then you've a serious problem. I suspect the common mode filters and big chokes you used had significant self-capacitance, They look effectively like shorts above, say, 5MHz. I have yet to see a spec for a commercial CM choke that does anything much above that frequency.
Not in dispute
Its merely one reason why PCs in their present form have not much longer to live...Its not either/or, its both and. But until ARM-style computing becomes mainstream for servers (perhaps with the A15) then let's be grateful for small mercies eh?
I reckon (though can't be sure) that such good figures for HF attenuation are only available when measured over a groundplane (i.e. low inductance) earth. That type of filter contains Y caps (mains to ground) which is why it has an earth leakage of up to 1.5mA. I'd be interested to know how much HF rejection is achieved in practice with an earth provided via a normal mains cable.
I wish there were ARM/MIPS boards without any SMPSs. Even the tiny Raspberry Pi | An ARM Linux box for $25. Take a byte! has one. In fact I know about only one tweakable board where I did not see any SMPS - the old Asus WL500b router. My little project to fit a PCI sound card to miniPCI slot on this board has been stalled at soldering remaining 40 wires of ATA80 cable to PCI socket for over a year :-(
Not only will there be the main SMPS, but also point of source SMPS's for the numerous core and IO voltages required, also another major source of noise is "simultaneous switching noise".
The problem with PC's and a lot of other gear and SMPS's is power consumption, a lot of gear has to have a max power consuption figure, this puts more constraints on the circuit design, so they arn't going to go away, but they can be made quieter it just requires more effort and more cost. A firm I worked for bought some battery chargers (12-24V SMPS based), the test units measured fine with no problems, when the container full arrived, we coudn't get them through even basic EMC tests, yet they ALL had a CE sticker on them. Result we had to design our own and bin 4000 chargers.
Also communication over power lines is adding more EMC mush to the equation!!!
The problem with PC's and a lot of other gear and SMPS's is power consumption, a lot of gear has to have a max power consuption figure, this puts more constraints on the circuit design, so they arn't going to go away, but they can be made quieter it just requires more effort and more cost. A firm I worked for bought some battery chargers (12-24V SMPS based), the test units measured fine with no problems, when the container full arrived, we coudn't get them through even basic EMC tests, yet they ALL had a CE sticker on them. Result we had to design our own and bin 4000 chargers.
Also communication over power lines is adding more EMC mush to the equation!!!
Well, I could not find any SMPS on the WL500b board. I will look carefully tonight.
It can't be so difficult to tweak it out - they say the consumption of the whole board is about 1W at full load, so an external linear regulator would be feasible. Do you know if the SMPSU just generates the core voltage for the ARM11?
They may not be on every board, but quite often with a processor DDR memory, etc you will find a lot of boards have voltages such as (from a real example) 1V0, 1V6_CORE, 1V25_VTTDDR, 1V8, 2V5, 3V3, 3V3_IO, 3v3_SW, 5V0 and that is just a selection, they do not all have there own supply, but the major ones will have a small SMPS, some are pi filter isolated and some have linear LDO's (the very low current).
A lot of point of source switches now have the inductor on silicaon, to help reduce the loop area of the switching nodes, eg Nationals Simple Switcher. It all depends on the devices used etc.
A lot of point of source switches now have the inductor on silicaon, to help reduce the loop area of the switching nodes, eg Nationals Simple Switcher. It all depends on the devices used etc.
Phofman, the multiple voltages are mainly found on PC motherboards, embedded computing etc, which is what I was refering to, to meet the power requirement budgets engineers tend to use SMPS as some of the switching currents are high, and often these SMPS's are cascaded, ie 5V will feed the 3V3 and 2V5, the 2V5 may then supply the 1Vx supplies, all this fed from the main SMPS supply. Noise on noise
That's clear, I am not looking for multiple input power lines, but for the actual SMPSs on board.
These are little SMPS's on the board, near to the devices in question, not line in from the main PSU. It will depend on the current required and the overall power budget for the design, whether an LDO or an SMPS is used. Then you have the requirements of the cards added to the board, the ethernet transiever and other chips used. Also as I stated on all complex digital boards where you have lots of IO switching (memory interface is included in this) you get simultaneous switching noise, which can cause more problems than the SMPS's.
Add to this that a lot of the DDR interface chips have been designed to allow motherboards to be designed with the minimum amout of layers (sometimes only 6) to keep costs down, you dont have a lot of ground planes to interspace between the signal layers to help return current coupling, so are also creating some noise due to the less than optimum return path. The power plane layers are often multiple voltages so signal layers next to these layers will couple causing the return currents to cross splits in the power layers which is as bad as splits and slots on ground planes.
These are just points regarding digital designs and layouts where I am pointing out sources of noise, basicly the whole of the digital board will contribute to the noise generated, digital signals are very tollerant of noise and depending on how much effort and cost in a product will quite often dictate how much design has been put into combating the noise, from just works so its OK, up to millitary, aerospace, medical where the noise and any circuit glitches can be life threatening so more expense is spent on solving the problem.
Add to this that a lot of the DDR interface chips have been designed to allow motherboards to be designed with the minimum amout of layers (sometimes only 6) to keep costs down, you dont have a lot of ground planes to interspace between the signal layers to help return current coupling, so are also creating some noise due to the less than optimum return path. The power plane layers are often multiple voltages so signal layers next to these layers will couple causing the return currents to cross splits in the power layers which is as bad as splits and slots on ground planes.
These are just points regarding digital designs and layouts where I am pointing out sources of noise, basicly the whole of the digital board will contribute to the noise generated, digital signals are very tollerant of noise and depending on how much effort and cost in a product will quite often dictate how much design has been put into combating the noise, from just works so its OK, up to millitary, aerospace, medical where the noise and any circuit glitches can be life threatening so more expense is spent on solving the problem.
Hi,
Let's not either forget the latest fad in audio chips (DAC's, Line and Headphone Drivers), build in charge pumps (that is gobbledeegeek for switched mode supply) to provide the negative rail from just one PSU. Handilt\y integrated directly on chip of your 32 Bit DAC and/or Line...
I wonder what the switching pulses from these do the ground bounce...
And no, these kind of on-chip switchers are not limited to audio either.
Ciao T
Let's not either forget the latest fad in audio chips (DAC's, Line and Headphone Drivers), build in charge pumps (that is gobbledeegeek for switched mode supply) to provide the negative rail from just one PSU. Handilt\y integrated directly on chip of your 32 Bit DAC and/or Line...
I wonder what the switching pulses from these do the ground bounce...
And no, these kind of on-chip switchers are not limited to audio either.
Ciao T
Abraxalito said:
I partially agree with Abraxalito, in that if you don't hear the difference between on-motherboard audio and external DAC audio, stick with lossy-encoded files. Since you may not have easy access to an external DAC, then that point is moot but, you may want to try some critical listening between two MP3 versions of a song, one at 192 kbps and another at 320. If you can't hear the difference, then either your repro chain as a whole is unable to resolve sufficient subtlety to convey a difference or, (sorry, no insult intended) your ear/brain combination isn't yet up to the task. Or, a combination of both.
Critical listening is a learned skill. The ability to perceive subtle differences is not something we're born with, whether we're talking aural or visual minutiae. It helps to have experienced live, unreinforced (unplugged) performances of your favorite genres, it also helps to have a skilled listener to act as a guide while you're improving your chops but, you can learn critical listening on your own.
So, bottom line: if you're happy with your current rig, enjoy it. If you'd like to wring more fidelity out of a computer-based system, you must move to an external DAC. Note that an internal AES/EBU interface produces the same result; allowing for external analog conversion. Once you have outboard conversion, you can gain additional improvements by moving to dedicated playback software.
I partially agree with Abraxalito, in that if you don't hear the difference between on-motherboard audio and external DAC audio, stick with lossy-encoded files. Since you may not have easy access to an external DAC, then that point is moot but, you may want to try some critical listening between two MP3 versions of a song, one at 192 kbps and another at 320. If you can't hear the difference, then either your repro chain as a whole is unable to resolve sufficient subtlety to convey a difference or, (sorry, no insult intended) your ear/brain combination isn't yet up to the task. Or, a combination of both.
Critical listening is a learned skill. The ability to perceive subtle differences is not something we're born with, whether we're talking aural or visual minutiae. It helps to have experienced live, unreinforced (unplugged) performances of your favorite genres, it also helps to have a skilled listener to act as a guide while you're improving your chops but, you can learn critical listening on your own.
So, bottom line: if you're happy with your current rig, enjoy it. If you'd like to wring more fidelity out of a computer-based system, you must move to an external DAC. Note that an internal AES/EBU interface produces the same result; allowing for external analog conversion. Once you have outboard conversion, you can gain additional improvements by moving to dedicated playback software.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.