|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
41.183.0.21
In Reply to: RE: cMP - the open source high-end Memory Player posted by cics on December 30, 2007 at 05:42:01
CPU (P4) power supply has seen improvements through the use of a dedicated 12V linear PS or the addition of smoothing capacitors. Most found this optimisation to yield excellent resuts.
This new and simpler approach is not available on the cMP² site. More testing is needed from others. On reviewing Gigabyte's (GA-G31M-S2L) mobo manual, it describes its P4 implementation as follows:
"... The 12V power connector mainly supplies power to the CPU. If the 12V power connector is not connected, the computer will not start."
If it "mainly supplies" power to the CPU, what else then? After testing with another mobo, it seems this is limited to fans (esp CPU fan). This is great news as we can treat the CPU as another "dirty" component (for example the HDD).
Idea:
Replace Grannite Digital (GD) PS' with one "dirty" ATX PSU. That is cMP² now has 2 ATX PSUs. This "dirty" PSU powers all "dirty" peripherals as before (as was done by the GD PS') and the CPU.
Procedure:
- Source a 2nd ATX PSU - preferably a good low noise one (Antec Earthwatts remains a good choice). Make sure you have power a switch on the PSU.
- Shutdown & disconnect all power.
- Remove GD PS' and wiring,
- "Dirty" ATX PSU sits either atop or aside your cMP². All DC power cables are run through a spare rear slot (remove screw and plate).
- Remove current "clean" P4 connector and replace with "dirty" P4.
- Connect all other "dirty" components (alternate USB ports, touch screen, ROM drive, HDDs/SSDs, ...).
- Check that "clean" ATX power ONLY supplies P24 (mobo).
- On the "dirty" P24 connector, hard wire it to permanently stay on. I used the end of a metal paper clip (which makes a nice "U" pin). Pins connect green and black wires of P20/P24 connector and is thereafter insulation taped.
- Connect AC power to "dirty" ATX ("clean" ATX remains off). Switch on.
You should see all your "dirty" peripherals come on. If there's no power, the ATX PSU is most likely not seeing enough load to turn on. I had to add a ROM drive to overcome this issue (ROM drive is not connected to mobo). Bad connections could be another cause for no power.- Connect power to "clean" ATX and turn on. Test playback, etc..
- If you intend to recycle power, shutdown as normal, then turn off "dirty" ATX. For startup, ALWAYS power up "dirty" ATX first.
Results:
After a few hours of playback, there was a tangible amount of improvement - more air and richer mids. Musicality factor of cPlay 2.0b32 improves even further...
Follow Ups:
I just did the P4 bypass tweak with a spare Antec PSU I had, and it's a nice SQ improvement! Whereas before, the DAC was slightly better than the 1212m analog, now it's a very close race, with the 1212m continuing to have a very slightly more solid bottom end, but the midrange is basically equal now (DAC won before). What's really interesting is the Mogami (1212m) vs. AZ Silver Ref (DAC) sound the same, but with completely different DACs. I must have a different weak link now, probably speakers, but maybe the amp equally so.
Are all the Antec Earthwatts PSU's ultra quiet running? My spare Antec 450w PSU is not silent but it's fairly quiet.
Anyone using fanless ATX PSUs like Silverstone makes?
Dave
Somewhere on the forum I saw a powerconnector extension enabling to switch on 2 PSU's at the same time, if I remember it was with an extension of the green and black wire? Can somebody point me to the correct post?
Can I use the same idea with a Pico too, so simply using the green/black to switch-on another PSU?
I had an Antec EA430 before replacing it with a linearly powered Pico. In the cMP^2 scenario, it is perfectly safe to run the PSU "fanless" (simply unplug the fan) as long as you're not concerned about voiding the warranty. My old configuration (G31M-S2L+E7200+EA430) could run 24/7 for weeks with no stability or heat issues.
"We should no more let numbers define audio quality than we would let chemical analysis be the arbiter of fine wines." N.P.
Are you using two Picos for the motherboard and P4?
I haven't tried a Pico yet.
I just removed the fan from the spare PSU, no problemo.
How did you get the physical P4 connector working with the linear 12v, cannibalize another ATX PSU?
I just disconnected the fan from my Corsair HX620, and it's a VERY nice improvement. Noticeably more ambiance, nice increase in bass extension. I'm back to going USB to the DAC with the Wireworld cable - that seems to sound most right to me.
When I went with a separate P4 a few days ago, I noticed horns sounding much more like horns. Before that they were recessed, with no bite.
Next step is a linear 12v for P4.
Love these cheap power tweaks!
All you do is cut off one end to directly solder to your linear supply and the ther into your p4 slot on the mobo.
What theob said. I recommend rewiring the P4 with higher quality cables though. The improvement over "stock" wires is quite noticeable.
"We should no more let numbers define audio quality than we would let chemical analysis be the arbiter of fine wines." N.P.
Only one Pico, I use a plain 12v linear for the P4. Pico + 12v, while not a complete linear setup, is still a significant setup from the Antec EA430. Well worth the investment imo.
"We should no more let numbers define audio quality than we would let chemical analysis be the arbiter of fine wines." N.P.
Idea: Replace Granite Digital (GD) PS' with one "dirty" ATX PSU. That is cMP² now has 2 ATX PSUs. This "dirty" PSU powers all "dirty" peripherals as before (as was done by the GD PS') and the CPU.cics’s post confirm yet again that PCs dedicated to music replay respond well even to modest improvements in the supply and distribution of power to their components.
The topic has been discussed at length on AA with suggestions ranging from the absurdly simple (smoothing the P4 line) to the intimidatingly complex (an all-battery system with power-on sequencing and elaborate soundcard mods).
However, all but the most rudimentary of a hierarchy of designs (see www.audioasylum.com/forums/pcaudio/messages/53957.html ) call for some competence in electronics and access to specialised tools. Many would-be users have neither.
It is a merit of the core cMP^2 design that it uses stock components and can be built by those with few constructional skills. However, although the proposed upgrade using two ATX PSUs maintains that low-skill approach (and gets switchers and mains power out of the case), it doesn’t follow that it is the best solution even for those without access to core skills.
In my opinion and with all due respect, even before plugging in a stray CD-ROM drive (or a few power resistors?) to keep switchers switching, it’s entry-level at best. In fact, it’s a bit of a kludge.
I’ll try to explain why I think that but would first like to clarify a couple of basic points.
** A 4-pin, 12-volt connector (the P4 or Pentium 4 line) was added to the ATX spec in February 2000 to power the CPU. If manufacturers now use it to power other components, they are arguably wandering off spec. Despite what it says in the manuals, I doubt that P4 powers anything on Gigabyte motherboards other than the processor and perhaps its cooling fan. As a cMP^2 system typically has no fans, it’s reasonable to assume that it powers only the CPU.
** ATX spec revisions in 2003 inter alia acknowledged that nowadays most power is required on the 12 V rails: power on 3.3 V and 5 V rails was reduced and -5 V was scrapped. As www.silentpcreview.com explained in 2005:
The high reliance of current systems on the 12V is dramatic compared to even just a couple of years ago and the evolution of the ATX12V spec reflects this change. Almost any system assembled from current components will draw the vast majority of current from 12V, in some cases, as much as 90% at load. (http://www.silentpcreview.com/article28-page2.html )In short, a modern motherboard is essentially a 12-volt device even if it needs power-sequencing circuitry and a couple of ancilliary lines.(For anyone sad enough, like me, to find these arcane topics mildly interesting, Intel specs are the official source but material on Wiki and elsewhere is more readable.)
++++
1. I have difficulty with the concept of the CPU as a “dirty” device that can be powered with impunity from a noisy source. First, sound quality is widely recognised as being improved at least as much by smoothing P4 as P24, arguably more. Second, there is no evidence to hand that a polluted P4 line will not pollute the supply to other, perhaps more noise-sensitive, devices. Separate SMPSs may to an extent mask that but it does not follow that they are the optimal solution.
2. The power demand of an under-clocked, small-format motherboard with a 45nm CPU is modest – typically under 25 watts. Allowing for accidental reversion to default BIOS settings, a 3-amp, 12-volt linear supply is likely to be adequate and a 5-amp one certainly is.
There may well be SMPS designs with better noise specs than off-the-shelf linear units but not ATX ones. The experience of several builders of cMP^2 and other systems is that a half-way decent linear supply outperforms even the best ATX designs in this context.
3. A Pico-PSU designed for use with a 12-volt regulated supply (i.e. not for car batteries) uses switching only for ancilliary lines. Even without additional smoothing (à la Greg), they are reported as bettering conventional ATX units. ( See: htttp://db.audioasylum.com/cgi/m.mpl?forum=pcaudio&n=47156 .)
A 12-volt “laboratory”-style PSU driving the P4 line directly and some P24 lines via a Pico PSU requires virtually no constructional skills to implement, costs much the same as two decent SMPSs and will comfortably outperform them . (Just about any 5-volt linear supply will perform as well as a Granite Digital switcher, perhaps better, for the “dirty” kit.)
Of course, one ideal would be to get a PCB laid out for John Swenson’s low RFI, low ripple design. Offers?
Just a few thoughts - sorry to go on for so long . . .
Dave
Edits: 11/16/09 11/16/09
If CPU power supply is isolated to just P4, using the approach suggested should be fine. So far my tests confirm this. CPU signaling to Northbridge is very important but this we control via software!
I'm suggesting an additional ATX over 1 or more GDs. It seems we don't need an expensive high powered ATX either but for an elegant in-box solution, one could try this:
Sparkle FSP300-601U would fit internally and may require less load to turn on. Note that it offers a P20 instead of a P24 connector.
I was very pleased to learn that GB mobos don't use P24 for the CPU.
Of course, using linear PSUs is the best way forward.
I'm using one of these to power all my HDDS and other devices, fits in a spare drive bay and is integrated with the power switch etc. I noticed a SQ improvement with my USB audio setup.
http://www.ocia.net/reviews/powerpartner/page1.shtml
cics,I'm not sure if I understand your recommendation correctly...
Your quote of the MoBo manual clearly states that P4 mainly powers the CPU. Maybe a few other components as well, but it seems as if we currently don't know which and to what extent. Nevertheless, the primary consumer of the current supplied through P4 surely is the CPU. Now if we consider this "dirty" and bundle it up on one line together with HDD, USB etc., we heavily "pollute" power supplied to the CPU, don't we? That cannot really be the target for improving power supply to our cMP system IMHO...
You surely make a very valid point that we should separate PSUs for P24 and P4, so if inmates only want to use ATX supplies, a good idea indeed can be to add a second ATX PSU for P4 while the first one only powers P24 - but why not keep the Granite(s)? Greg/GStew btw has already shown that even improving on these "dirty" supplies has a positive effect on overall system performance.
Many of us already e.g. use separate linear supplies or batteries for P4 and P24 (e.g. powering a picoPSU or supplying directly to P24) and a "dirty" supply for the truly "dirty" consumers like HDD and USB etc. But I find it hard to understand that taking away more or less clean power from P4 and making it one of the "dirty" consumers could be beneficial.
Have you compared 1 x ATX PSU for P4 plus 1 x ATX PSU for P24 plus 1 (or several) x Granite(s) for "dirty" consumer(s) versus 1 x ATX PSU for "dirties" and P4 plus 1 x ATX PSU for P24 and found the latter to be superior?
Thank you,
Robert
Edits: 11/15/09 11/15/09
Good points.
I'm looking to change our "base" design to 2 ATXs instead of 1 ATX and a few GDs. This way we reduce wiring needed and don't need any "Y" splitters. Also, we don't suffer from PS limitations (compared to GD where you must be under 2A). However, if you already use a dedicated PS for P4, this change will not be useful and should be ignored.
What triggered this design was reports on how sound improved when powering P4 separately. Are we getting improvement from better PS to CPU or is it that we have largely removed its "PS modulation" & ground effects that negatively impacts the critical audio chain (via P24)? My sense is the latter. This question grew larger with hfavandepas' 12V linear+pico implementation. Given that the pico also suffers from ATX like ripple noise, a normal ATX should do. Pico 150XT which passes linear 12V input untouched will likely be better.
My understanding of how P4 is used in the GB mobo suggests its limited to CPU and fans. There may be other components and if they were critical, sound quality would be worse. It's not.
It may seem strange that a "dirty" P4 works so well given that the CPU is tightly coupled to the audio chain. The reason is largely to do with soundcards interacting indirectly with the CPU. For audio streaming, it only sees RAM via the southbridge chipset. Another key factor is how well the CPU / Northbridge interface is significantly de-stressed. This seems to make the twin ATX package work very well.
cics,for the recommended "base" setup this is certainly good to start with, especially given your explanation on how little influence the CPU has during audio streaming (I wasn't aware of this and that the focus here is mainly on the southbridge chipset - would be great if we could improve its working environment, e.g. by adding caps to its voltage pins etc., but that's a bit tricky I suppose...).
Over the course of this year I have improved my system also hardware-wise step by step to a point where I use only batteries plus caps during playback, separate ones for P4 as well as P24 (no picoPSU or similar switching PSUs but battery power separately for every voltage rail) as well as for Juli@ and Buffalo DAC. Optimizing on P4 and P24 always brought little but pleasant improvements step by step - but I have to state very clearly that IMHO almost dramatic improvements can be achieved when further focusing on Juli@ (as we know already from the mods Alfred/sonics and Greg/Gstew made).
Powering Juli@ externally is a HUGE step as we know already for some time. There are two Juli@ versions, one only needs additional 5V for sample rates 176 and 192, the other at all SRs for clock timing. As Greg/Gstew has already pointed out cutting the PCI traces for all voltage input ensures that only the clean external voltages are present on Juli@, separating it even further from detrimental external influence.
In the past couple of days I had to learn how crucial even smallest variations are when you power Juli@ externally - it's a bit like open heart surgery, there's no more regulators and smoothing caps and different circuits having their influences, but you directly without any filters inject current in Juli@'s circuits, and parameters like voltage levels (3.25V instead of the nominal 3.33V as well as 4.85V instead of 5.00V seem to work best), output impedance of the current source, purity of power supplied etc. IMHO have a MASSIVE influence on sonics and sound quality - in my setup even small changes to these parameters had a much more noticeable and revealing influence on overall sonics than any other improvements mentioned above. I'm still in the process of testing and identifying and pinning down, as Greg/Gstew had mentioned before shunt regulators probably will be an excellent choice for powering Juli@ directly, that's next, will report as soon as I get there and have reliable and tested results :-)
Sorry for rambling, just wanted to put the relevance and effect of the various mods into the overall picture, at least as I see it ;-)
Best,
Robert
Edits: 11/15/09
Hi Bertel,
Since you PRACTICE (!) a lot of power supply improvements to the cMP2 setup, you might be able to shed some light on to the following question.
I would like to know the degree in (potential) sound quality improvement between: supply chain 2 and supply chain 3:
- Supply chain 1: original chain.
source: 5 volt ‘high quality power’ -> P24 5Volt input connector on MoBo -> MoBo to PCI -> circuit board of soundcard -> ‘sink’ components on soundcard (who need the supplied 5V)
- Supply chain 2: MoBo is bypassed.
source: 5 volt ‘high quality power’ -> circuit board of soundcard -> ‘sink’ components on soundcard (who need the supplied 5V).
- Supply chain 3: MoBo + (parts of) circuit board of soundcard are bypassed.
source: 5 volt ‘high quality power’ -> ‘sink’ components on soundcard (who need the supplied 5V)
I realise that it’s very hard to communicate about sound quality improvements in words, but I still would like too get an impression by description, because it’s better than no communication at all. May be it’s helpful to rank sound quality improvements as I (subjectively) experience them in my setup. I would scale/index them very ROUGHLY like this:
- linear PSU on P4: +90
- Low resistance wire (6AWG or similar) from linear PSU to both P4 and picoPSU/P24: +90 (*)
- Linear PSU on 12 volt rail on P24 through (specific) picoPSU model: +30
- smoothing caps on a switching ATX PSU: +20
- combined optimizations: XP, Kernel and BIOS: +15 <-> + 20
- various releases of cPlay - 2 <-> +4
Is it possible to say where in the above ranking roughly would be:
- by passing the MoBo (chain 2) with ‘high quality power’ from a linear PSU?
I don’t like soldering to my Lynx AES16 or Lynx L22 sound card. But I think I can manage to bypass the MoBo (with help of a PCI riser-card or something. don’t know yet)
(*)
Untill now no other inmates tried out or gave feedback on changing 18AWG wire between a linear PSU and the P4 / P24 for a wire type with much less resistance (6AWG or similar).
Probably many inmates consider me being an honoured member of ‘the ministry of silly tweaks’. :-) But I’m far from that. If I hadn’t tried it myself, I would be very much in favour of Ryelands comments on it. Http://db.audioasylum.com/cgi/m.mpl?forum=pcaudio&n=54333&highlight=sums+Ryelands&r=
But because you suggested I decided to give it a try. So thank you for the suggestion, because it made (very much too my surprise !! ) a very big improvement.
fully AOB optimized cMP2 PC -> Lynx AES16 -> XLR AES/EBU -> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
H,
I thought I had replied about your 6g wire tweak, but cant find the post.
Are you sure it is not the braiding that helps in sound more than the wire gauge??
Thank you for responding.
I really don’t have a clou of why it does sound much better. Most likely I corrected – without knowing- some mismatch or sub-optimal situation in my setup. (re-connected wires the right way, re-plugged all the gear in the correct way, ect)
I strongly favour Reylands post on this. http://db.audioasylum.com/cgi/m.mpl?forum=pcaudio&n=54333&highlight=sums+Ryelands&r=
But I still would like to know if I accidentally corrected something in my setup for the better without knowing. That’s why I decided to ask other inmates on the AA-forum if they can repeat this and report their findings back to the forum (too try this, one doesn’t need much soldering skills).
I was hoping that a few inmates would try this in their setup. If they would report that it didn’t made any difference in Sound Quality for the better in their setup, than I know for sure that I must have accidentally corrected something in my setup for the better without knowing.
But until now I haven’t got any feedback from inmates who tried it.
fully AOB optimized cMP2 PC -> Lynx AES16 -> XLR AES/EBU -> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Hi Mark,
thanks for asking, I'll try to produce something meaningful on this issue - but I'm afraid the answer is not so straightforward... ;-)
Those of us that use the ESI Juli@ soundcard are in a very comfortable position when it comes to judging which solution is best - we unsolder the on-board regulator (5-> 3.3V), cut the respective traces on the PCI connector that carry in the voltages from the MoBo, feed Juli@ with clean 3.3V and 5v lines, and we're there. Juli@ has dedicated input/output points (kind of little "drill holes") that are wonderfully suited to take this external voltage supply.
I currently use a Juli@ too but have used a Lynx AES16 before. I have just taken my now obsolete Lynx in my hands and am looking at it - and especially when comparing it with Juli@, it's so §$%§& complex, and sadly this is a complexity that for the most part we don't need... It is an excellent piece of engineering, but I have a hard time identifying a good injection point for clean external power. Ok, there are SMD devices that look like input rail smoothing caps to me, labelled C50 and C51 (far left front side down at the PCI connector) and C52 (far right side), from what I see those could potentially do the job. You also had to take care of 3.3V and +/-12V. While poking around with my multimeter I saw that the PCI traces for these voltages are connected on the board but couldn't find chips that would share that connection. This of course needs more work and thorough analysis. But overall I would say that I do not see why it shouldn't be possible to power at least the whole card with clean external voltages. I personally would prefer cutting traces and soldering regs directly to the board. One next step would be to further identify regulators further "downstream" that do additional regulation, and eventually inject here. But this is open heart surgery ;-)
So since you say you could imagine using a riser card or something similar, i think that goes without bloodshed ;-) and should be easily doable - give it a try!
Do I understand you correctly that solution 3 would be to inject external power directly to the chips or parts of the soundcard? This is complex and challenging and would take a lot of analysis, including soldering then of course, but indeed would be very elegant, since then you maybe could shut of whole parts of the card that are not needed by cMP2 which would be great! But to call that "advanced" would be a massive understatement... ;-)
So apart from my estimation that solution 2 (at least for a start) is the only feasible and viable, here's my judgement od the effect of such an operation: when I did this to my Juli@, i.e. cut the PCI traces, unsoldered the on-board reg, connected clean 3.3V and 5v lines (battery plus cap/resistor plus reg) to the respective onboard connectors, the effect was not subtle at all but in an order of magnitudes. See, by doing that we (i) isolate the soundcard from all the powerline "pollution" on the MoBo, and (ii) feed it with dedicated clean power, thus massively increasing the stability and purity of its working environment. Ok, a card like the Lynx creates its own mess and pollution through its complexity and all that stuff that's onboard and working, so maybe the effect is a bit less, but you could think of adding caps and replacing regs (see Greg/Gstew mods to Juli@) and further improve.
So again: I would expect a top-of-the-ranking improvement. The more regs and caps you can replace or upgrade, the more you can add to this improvement. And as you already suggest: the first step is simple, get yourself a riser card, identify which voltages are really needed, and replace them with external ones (that's how Alfred/sonics had proceeded when developing the Juli@ PSU mod). in a next step you then could follw the way the current takes and improve upon its lines. Give it a try, IMHO you can't loose! :-)
Good luck!
Robert
For those who wish to learn please listen to Bertel. As Cics is the king of the audiophile software, Bertel is the king of audiophile hardware. He is not saying but he has singlehandedly solved the metallic/static noise issue of a juli@/cmp^2 based system. He has helped me tremendously with tweaking my hardware to optimum results. It is not easy to implement a a total battery powered system and I don't recommend it for the feint of heart. I approached it but was limited by my soldering skills. But for those that want to achieve this, it is possible.
A fine young man not too disimilar from Cics in his knowledge of these pc audiophile matters and willingness to help others.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: