|
Home
/ FAQ
/ News Classifieds / Events |
Audio Asylum Thread Printer |
Get a view of an entire thread on one page |
84.138.116.83
In Reply to: RE: cMP - the open source high-end Memory Player posted by cics on December 30, 2007 at 05:42:01
I'm in the process of going "all battery" for P24 too (P4 done), so in order to right-size the respective power lines I considered it a good idea to do a few measurements (just as Dave and Mark have done on P4). Thought I'd share if anyone else is interested - brought a few interesting insights, at least to me.Current setup: Almost "standard" recommended cMP2 setup, i.e. Gigabyte GA-G31M-S2L, E7200, 1GB Mushkin EM2-6400 [superior to ValueRam and HyperX], Lynx AES-16 [will go Juli@ w/ external 3.3V PS next week], cPlay 2.0b23 SSSE3, all optimizations incl. Minlogon done - did I forget anything...?). Measurements done with Antec Earthwatts 430 for convenience, picoPSU showed same results.
All BIOS optimizations done. CPU host frequency at 150 MHz, SPD at 2.00, Mushkin RAM timed manually at 3-3-3-7, CPU GTLREF at 0.566V, CPU Voltage at 0.76875V (it's a shame but I can't get any lower, no way system would boot then...). CPU-Z reports Core Voltage to be at 0.752V. Core Speed is 900MHz, Bus Speed is 150MHz, FSB is 600MHz.
-------------
"Special" power rails:
* 5V StandBy:
- When PS turned on before pressing Power Switch: 170mA
-> During operation (idle / playback): 50 mA* Power_Good:
- No current could be measured with my rather "rough" multimeter -> seems that since this is a signal line, it is sufficient that just the voltage is provided with some very low current, certainly below 50mA, according to my measurements below 10-20mA, but 5V provided* Power_On:
- Switches from 5V to 0V at startup, so not relevant here* -12V:
- As little as ~10mA during operationNOTE: Most important (and for me quite surprising) :
All of the above can be TURNED OFF COMPLETELY during operation, they are NOT NEEDED during playback (at least in my [recommended] setup)! I ran all of the aboce through a switch, and I could turn them off at any time even during music playback without any adverse / negative effects. Didn't expect that... ;-)So the "relevant" lines to take care of come down to:
-------------
"Regular" power lines:
* 12V:
- Starting at 180mA at startup, quickly coming down to 140mA and staying at that level during idle and playback* 3.3V:
- I had supposed that being the "main line", but it isn't - starting at ~300mA and staying constant at that level during idle and playback* 5V:
- Now that's the real thing... Starting right off at 3 - 3.2A (!), going up to 3.6 - 4.2A during boot up - and breaking off as soon as cMP is loaded... This 5V line is so volatile, changing current so quickly and swiftly that it seems to not tolerate my rather "slow" multimeter - couldn't get the system up and running with the multimeter phased in - will retry with caps in place :-)
Nevertheless, the learning is: tis is the rail with the BIG draw at ~ 3.5 to 4.5A (more precise measurements to follow when caps are in place)-------------
Hope you find this as useful as I do - did that the best I could, please notify me of any flaws or errors I made, comments highly appreciated.
Cheers,
Robert
Edits: 05/04/09 05/04/09
Robert,Very useful information, sir! Thanks for doing & posting this.
This will be very useful for creating a fully-linear computer power supply... it's clear that the 12v P4 and the 5v ATX-24 lines are two of the key 'workhorse' lines.
One thought that may be useful for a fully linear CPS is to power each of the 5 5v lines with a separate power supply... regulator or battery. Powering them separately and measuring them separately may provide other information on how they are used and how we can best optimize each.
Another useful technique might be to use very high-quality supplies (again, regulated or battery) for the 3.3v and 12v, since they aren't pulling a lot of amps.
And of course, while it's useful to put additional filtering caps across each of these three voltage rails, based on your data I suspect that separate filtering caps across each of the 5 5v lines may provide additional benefits even if separate supplies are not implemented.
Its not surprising to me that the 'special' rails can be turned off after the computer is on (In fact, I bet that you could boot the PC with the -12v and the 5v standby turned off). One thing to note, tho, is that the -12v is used by the Juli@'s analog section (although in my setup before separately powering it, I was only measuring -9v at that point on the Juli@ board.
Two more things... give us more details on your battery-power setups, both for the P4 and the ATX-24. Then also, give us more details on the Mushkin memory... what other ones did you compare against, how is it superior to the others, and did you try other timings settings?
Thanks for the most provocative and useful post I've seen here in a long time (of course, besides the almost 2x weekly releases of new versions of cPLAY!).
Greg in Mississippi
Edits: 05/04/09
Greg,thank you. Yes, I'm completely with you that powering every power rail separately is a good idea - it might be a bit of overkill, but separation and elimination of interference and cross-balancing issues (quite a concern in "standard" ATX power supplies) will help bring the various components even more "at rest" and let them play as easy and effortless as possible. When I have swapped my Lynx card with Juli@ and powered it separately, P24 "only" has to power MoBo and memory (still and awfully complicated task), that'll help. I'm planning on using completely separated supplies (i.e. battery packs) for every voltage line anyway, including thorough regulating plus filtering caps - more to follow as I progress ;-)
Regarding booting with -12V and 5V standby turned off: You'd loose your bet ;-) PS for P4 and P24 has been my special field of interest for the past weeks, and while P4 is easy, P24 needs quite some attention... As you might know, ATX standard requires a proper and quite stict power up scheme (more details can e.g. be found here: www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf). 5V has to be there in the first place for the computer's power switch to have the MoBo's power control circuit (PCC) start the sequence, hold low the formerly held high 5V "Power_On" line (the green line on P24) to have the power supply ramp up the three main voltage lines and after successfully doing this issue the Power_Good signal (5V, the grey line on P24). This all has to be done in a corridor in a 100ms range with specific ramp rates, and I found the Gigabyte board be rather strict at that - I tried for weeks to emulate that timing and ramping manually (with switches and caps etc.) but didn't succeed, you need some kind of power management component to take care of this. I'm currently looking for someone to build such a power sequencing model from readily available components like Actel's Fusion or Linear Technologies' LTC2928 (cost currently quoted at 10-15k and above for developing that...), but in the meantime as a workaround just power up with my regular ATX PS (Antec) and then switch off the special rails not needed and switch over the main rails to battery power. Quite rough a method, I know, but works well for 12V and 3.3V for the time being, but still stability issues with the "workhorse" 5V rail. Will keep you updated once I have gathered more experience and details. Nevertheless, 5V standby and -12V line is a must to be present at startup, otherwise MoBo's PCC will not power up.
Regarding my P4 battery PS: No special timing etc. issues have to be taken care of here, powering P4 is pretty easy and straightforward as many have shown us here before (many thanks for their respective guidance and information!). In my case I power this with a pack of 3 boards with 5 high-quality 1.2V/1.8Ah NiMH cells each, i.e. 18V/1.8Ah (recharged by a transformator fed into the central control unit, all of these modules are manufactured by a small local electronics company). Since the current draw is low at ~350mA during playback, this is good for a 4 hour listening session (I usually don't have that much time in one go anyway...). Power then is fed into a voltage regulator which brings it down to precisely 11.94V as measured to be optimal by Bernd and confirmed by Theo recently. Since the regulator is designed to be quite "slow" and "weak" on purpose and thus is very forgiving, large caps can follow - I run the power into 5 x 10kuF Mundorf caps followed by a 47uF cap as quick buffer and a 1uF one as bypass (I am extremely positive about the Mundorfs, can't praise them enough). this setup works like a charm without any hassle for me. While the sonic effects were rather subtle from the start, I feel it overall added transparence and clarity and ease of playing, especially in midrange and upper bass - to my ears it rather sounded a bit as I could here the processor working now without the limits of a rigid, tight and polluted power supply as before, but that's what I wanted to hear for sure ;-) That's more or less a similar setup I'll try to follow with P24 too, will let you know once I get there (will require mighty PB cells instead of the NiMH for the 5v rail though).
Mushkin memory has been a recommendation by Alfred/sonics, so he's the one to both take the credits and comment on that in more detail. Haven't done much comparison myself - I almost exclusively play 16/44.1 WAV files upsampled to 24/192, had ValueRam 256MB before, swapped that with an "original" HyperX UL stick w/ 512MB which brought significant improvements to me , but still swapping that to the 1GB Mushkin stick made things even more "right", beautiful and in place - for me quite a substantial improvement. I started with SPD in BIOS set to "Auto", CPU-Z showed it had clocked itself in at 4-3-3-7, so the only thing I did was lower the first value (cycles) to 3 which had an unexpectedly clear effect in bringing back lateral spacing and soundstage.
I think that's more than enough for one post now - thanks for your patience... ;-)
Robert
Edits: 05/04/09
Post a Followup:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2009, Audio Asylum, all rights reserved ]
[ Go Green ]
[ Take5 ]