|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
82.156.128.190
In Reply to: RE: cMP - the open source high-end Memory Player posted by cics on December 30, 2007 at 05:42:01
Hi all,Since going all linear on the P24 is beyond my construction capabilities, I kept searching for some other way too improve the power supply too the P24 beyond the hybrid PSU solution I currently use (linear PSU + Pico on the P24).
* What is the next best option too a full linear cMP setup?
I noticed that amongst inmates it is communis opinio too consider PSU’s designs which are best for analog audio are also considered to be best for digital audio systems. But when reading about power supply noise suppression & decoupling for digital equipment, it appears that also much comes down too carefull and professional PCB design which addresses these issues on the PCB. Actually much of these issues can only be addressed locally on the MoBo. But how much of this is already done with sound quality in mind on a MoBo costing 45 euro’s? Clearly not much because when power supply to the MoBo is improved SQ greatly improves. But does this mean that analog audio and a cMP MoBo have same demands on the quality of power supplied?
Since the MoBo itself also generates lot of noise, delivering exceptional high power quality to the MoBo also might look like an attempt too make sewage water drinkable through mixing it with bottles best spring mineral water available. May be also less extreme PS improvements can still bring some further SQ improvements.* Why did I try ferrites + caps?
Since constructing an all linear PSU is beyond my DIY skills, the only option left I could think of was some sort of filtering on the power that is supplied by the Pico or the Antec ATX. Articles on power supply noise suppression & decoupling all use some sort of LC filter constructions on the PCB. Mostly through of a combination of ferrite beads and caps. So why not try and see if this also can be applied on the output of the Pico or Antec?
Since I only can solder some kind of filter for use on the P24 line, the filter region can only be in the lower HF area’s. Filtering VHF and UHF can only be realized on the MoBo itself close too the IC’s through special design and using highly specialized parts.I’m also not capable of doing any calculations on the use of wired coils/chokes (much too difficult) in filters. Especially the calculations on how too prevent ringing of coils/chokes are very complicated. So I was extremely happy that I stumbled upon articles in which ferrites where used in LC filters for noise suppression. Of course I already knew that clamp-on ferrites are being used on all sorts of signal cables and PS cables, but it never sprang too mind trying them on the P24 combined with caps too create an LC filter.
And lastly: ferrites are relatively cheap and it’s very easy too experiment with them on the P24 line.
* What I learned so far from 2 months messing around with ferrites.
1. Just 2 or 3 will have no audible effect.
When using ferrites (without the caps added) a ‘brute force’ approach is needed. At least some 5 to 7 ferrites in series on the P24 are needed too let you hear any subtle SQ improvement.
Although in a completely different situation, this ‘threshold effect’ is described in several articles and power points presentations on the http://www.audiosystemsgroup.com/publish.htm webpages.
The author suggests too clamp on at least 500 to 1000 ohms too get past this ‘threshold effect’.
‘Everything you always wanted too know about clamping ferrites on a wire’ is on these audiosystemsgroup webpages. Also much on proper grounding of your gear. VERY RECOMMENDED reading !
Also other authors report: ‘More is better’.
So put on the maximum amount of ferrits that will possibly fit in some way on your P24. Be creative in how too clamp a maximum amount. The longer the string of ferrites, the better. The impedance is proportional to length. This means that for a certain volume (and weight) of ferrite, best performance will be obtained if the inside diameter fits the cable sheath snugly, and if the sleeve is made as long as possible. A string of sleeves is perfectly acceptable and will increase the impedance pro rata.
See: http://www.elmac.co.uk/pdfs/ferrite.pdf2. The effect on SQ is no less than ‘jaw dropping’ when creating a filter through adding caps.
The effect of caps used together with ferrites is shown in figure 5-13.
(see also page 38-39 http://www.murata.com/products/catalog/pdf/c39e.pdf )
Van on filtering, demping, ect
3. Use good quality low ESR caps.
The use of good quality, low ESR caps is important. See graph.
Van on filtering, demping, ect I used Oscons
4. Filter all (!) lines/wires on the P24.
Noise might also creep onto the MoBo through the ‘STD-BY’-line, the ‘PWR-OK’-line and the -12V-line. So just to be sure I also put filter-caps onto these lines too.
However these lines don’t need bulk-capacitors which serve as a ‘current reservoir’. Extra bulk capacitors are only needed on the 3.3, 5 and 12 volt lines too suppress voltage fluctuation through extra current supply.5. What values too use.
* Ferrites
I first used clamp-on ferrites with the highest possible resistance www.conrad.nl would sell. 172 ohms at 100 mHz. I managed too clamp on 15 ferrites on the P24. But this way I also ended up with ‘a huge pile off charcoal’ on the P24 connector. Leaving no space for caps after the ferrites too create an π-section filter.
So I ordered flat ferrite clamps for ribbon cables (132 ohms at 100 mHz) and an extra P24 extension cable. See picture.
Van on filtering, demping, ect Carefully check if all off the combined P24 extension cable wires will fit into the flat ferrite clamp. I had too buy an P24 extension cable with extra thin 22 AWG wires too completely fit all P24 wires into the flat ferrite clamp. A P24 extension cable with 18 AWG wires (which are thicker than 22AWG wires) would not allow for fitting all wires into the flat ferrite clamp!
So carefully select the inner dimensions of the flat ferrite clamp for ribbon cable so it can accommodate all of the P24 wires.Avoid DC bias / saturation
At saturation the attenuation effect of the ferrite core disappears.
To avoid this make sure that both go- and return- currents are of equal strength and both flow through the same core. Used this way the ferrite is not subjected to a net magnetic flux and will not saturate.
See page 109-110 http://www.steward.com/pdfs/emi/technical/Use%20of%20Ferrites%20in%20EMI.pdf* Caps.
Since I’m not able too make any complicated calculations on what caps value’s too use in the π-section filter I had too find these values by looking at what values I saw being used in articles and papers ect.
The Murata paper recommends using value’s in a 10:1 rating but that still doesn’t provide any absolute value’s. By looking at what values also are being used in other articles and papers and looking at what Oscon value’s I could buy at Conrad I choose: 330 uF + 47uf Oscon’s (I liked 470:47 better, but couldn’t buy 470 uF Oscons at Conrad)* L-section filter
Since this is project is still 'work in progress' and I’m still experimenting, I did not yet solder any of the caps onto the P24 wires yet. I just pushed the legs of the cap into the inside of the plastic P24 connectors. This way the caps legs can connect too the metal cable connecters inside the plastic P24 connector.
Too construct an L section filter I simply put a triplet of a Panasonic FC 4700 uF + Oscon 47 uF+ 0,47 uF polymer cap on the 3.3V, 5V and 12V section between the Antec and the ferrites. See Picture.
Van on filtering, demping, ect
The effect is no less than amazing !
Hearing is believing.
Much more micro details, imaging, soundstage, etc, etc.
So the usual SQ improvements one gets when improving the power supply but only this time in some nice amounts ! :-)
* π-section filter.
In the past I hardly noticed any SQ improvement when I placed (good quality) caps on the P24 lines.
Putting on the above mentioned triplets didn’t do much, if anything at all when placed them on the P24.
So I was really amazed that putting little 47uF oscon after the ferrites (thus creating π-section filter) now resulted in a clearly audible SQ improvement. See picture.
Van on filtering, demping, ect
Too put in perspective somehow: the 47uF Oscon caps after the ferrites improved SQ again by 15 % of the improvement that the tripplet caps placed before the ferrites (L-section filter) already had brought.I wrote this interim report too encourage other inmates too experiment with π-section filters on the P24. In my setup an π-section filter made of ferrites + caps brought the usual SQ improvements but this time in GENEROUS AMOUNTS ! Highly recommended.
This is it for now. Do try this at home.
As always: any positive suggestions, comments, ideas, ect, which will improve the above suggested supply noise suppression on the P24 are very much welcomed
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Edits: 07/07/11 07/07/11Follow Ups:
Is the pi-filter connected after the linear psu to the Julia & P4??
So that said, the connection should be something like this :-
12v Linear psu => 4700uf cap +47uf+0.47uf => two ferrites=> 47uf=> P4
Is that correct???
Hi Jolida,Yes. If you want too filter the output too the Juli@ and/or the P4 than filter has too be between the PSU and the Juli@ / P4.
Following the Muratta Paper on the use of ferrites or the Altera AN 583 paper, I just copied the values they where using in their experiments.
The values I copied from these papers too create a pi filter are: 470 uF before and 47uF behind
If these are appropriate values for what I’m trying too do, I don’t know. I just copied.
Never the less, the effects on SQ in my setup are remarkable.=> P4
* 12v linear psu => 470uF => 2 ferrites with each 3 or 4 turns (!) => 47uF => P4=> P24
* Pico => P24 extension cable => 470 uF => 7 flat bead ferrites for flat ribon cable => 47 uF => P24
Too avoid a DC bias, all P24 wire must go through the ferrite. Also on all wires must be caps (except for the ON/OFF wire).
Van on filtering, de-coupling, demping, ect => SSD
* 12v power line => 470uF => 1 ferrite with 5 tot 6 turns (!) => 47uF => SSD
* 5v power line => 470uF => 1 ferrite with 5 tot 6 turns (!) => 47uF => SSD
Van on filtering, de-coupling, demping, ect Note:
- use as much ferrites as you can.
The intention is to create an impedance of 500 - 1000 ohm at the targeted frequencies.
This means you need 5 to 7 ferrites in series.
Or you need to make use of turns around the ferrites too reach this.- As the Muratta paper stresses it is important to use caps with good HF qualities.
Van on filtering, de-coupling, demping, ect Right now I’m also experimenting with other filter methods.
Like filtering with ferrite based chokes/coils.
My first impression is: these work equally well.
More on this too come.Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Edits: 08/26/11
Hey Mark I got another p24 extension, put some oscons on it and inserted it in series with my other cap/ferrite infested p24 extension and wow...another step forward. While I loved what Steppe's Cplay 42 stuff does this is way better. Depth, inner detail to die for, dynamics to snap your head off...I love it. Other inmates should consider this simple but effective tweak.
Hi Theo,
:-) Love too hear this.
If done right, ferrites + caps will indeed give a very nice sound quality improvement.
As the Murrata paper stressed: caps with good HF properties/qualities should be used.
Since you now use two P24 extension cables in series, you can ad extra ferrites.
More ferrites is better.
Did you also ad extra ferrites?
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Did not haev any extra ferrites to add but I will this weekend.
Ok to form a pi filter you need a cap ---ferrite----cap. If you can
count the caps in the atx power supply all you need are the ferrites
over the 5 v and ground lines etc and then the cap in the p24
connector on the mobo. Or you can use a p24 extension to push caps in
the extension connector, ferrites on the extension line and caps
pushed into the connector into the mobo. Which way did you do it?
Before I blew a mobo I pushed a .01 mfd cap into the 5 volt portion of
the p24 connector on the mobo. I thought the sonics were
great...better dynamics, very fast. Now on my new mobo I have caps
into the p24 connector for the 5 and 12 volt lines, ferrites on the
wires...very little improvement. Maybe my memory of what my old mobo
sounded like is not right or maybe I don't have a real pi filter,
maybe its only an L. Anyway I would appreciate your comments.More info: right now I have 47mf caps pushed into the 5v and 12 lines on the p24 connector into the mobo. Also I have an E2sl mobo (no spread spectrum bios setting) vs my old s2l mobo.
Edits: 08/08/11
Hi Theo,Good too hear you found a new mobo and have it up and running again.
I’m not sure if I correctly understand your questions (and thoughts behind the questions) too provide a useful answer. Some things puzzle me.
> If you can count the caps in the atx power supply all you need are the ferrites over the 5 v and ground lines etc and then the cap in the p24 connector on the mobo. <
* I don’t know if you can assume that the caps inside the ATX PSU will act as being a part of the 3 order filter on the P24 line. May be it works this way, may be it doesn’t. I just don’t know.
So just too be sure, I don’t consider the caps inside the ATX too be a part of the 3 order filter. So I use xxtra caps with the right values too create a 3 order filter. Since we are targeting (V)HF, caps with low ESR and excellent VHF characteristics are important.> all you need are the ferrites over the 5 v and ground lines etc <
Why do you think this? I see 2 problems with this.* (V)HF AC pollution travels from the ATX onto the MoBo on all DC plus voltage P24 lines.
You may prevent (V)HF AC pollution travelling on the +5 voltage onto the MoBo, but (V)HF polution than will simply travel onto the MoBo through the other positive DC voltage lines (+ 12, -12, +3,3, 5V stand-by, Power-OK) and return through the combined ground too the ATX.
A clean and steady ground is equally important. ‘Zero’ and ‘Ones’ are recognized as a voltage difference compared too ground. So both supply voltage and ground have too be ripple free. Both are of equal importance. Also the clocks on MoBo and Soundcard are very sensitive too (V)HF polution on supply and ground.
It is even important too prevent (V)HF traveling through the SSD ontoo the MoBo.
I also use a 3 order filter on the power lines too the SSD. And also this give a little SQ improvement.
See photo.
Van on filtering, demping, ect * How do you avoid DC bias in the ferrites (saturation of the ferrites)? The magnetic flux produced by DC plus and DC minus currents through the ferrite have too be equal too avoid saturation of the ferrites through a DC bias.
It’s not clear too me how you know that the magnetic flux produced by the % volt DC + and DC - minus return currents on the black wires will cancel out each other in the ferrite. The ferrites will saturate and loose there effect if no attention is paid too DC bias. Currents on the P24 are high. See my P24 measurements. Almost 1 amp for the combined +5 volt lines. Most ferrites saturate when there is more than 0.4 amps applied on them. I think one can easily go wrong when just grabbing the 5 volt lines and a some black ones.> Or you can use a p24 extension to push caps in the extension connector, ferrites on the extension line and caps pushed into the connector into the mobo. Which way did you do it? <
Back in may 2011 I just pushed caps into the connectors of the MoBo and/or connectors on the P24 extension cable. I did that too quickly verify if a 2nd or 3thrd order with ferrites + caps would have some effect.
By now I have soldered the caps onto the P24 extension cable connector. I cut away the plastic on the side of the P24 connector on each side, this way it's easily solder the caps onto the small metal parts inside the plastic connector. See photo.
Van on filtering, demping, ect > Now on my new mobo I have caps into the p24 connector for the 5 and 12 volt lines, ferrites on the wires...very little improvement <
With only a few ferrites clamped on the P24 In my situation there also was very, very little improvement. As I posted before at least 6 or 7 ferrites on ALL P24 wires are needed to make SQ improvement that is worth the trouble. But one only gets a real nice significant (!) sound quality improvement when creating a 3 order filter through adding caps in front and after the ferrites. I used 7 ferrites which clamp around ALL P24 wires (thus no DC bias) See picture.
Van on filtering, demping, ect
I tried this cap-ferrite-cap filter on the Antec Earthwatts ATX and on the 2 Pico’s I use. In all 3 situations I got the same sound quality improvement. Attention: same amount means: the pico 160-xt with P24 filter still sounds better with the filter than the Antec with the filter.So my experience with using ferrites so far is:
* ALL power lines coming from a switching PSU need filtering
probably all lines coming from a switching PSU are polluted with (V)HF noise. All these lines need filtering to prevent (V)HF travelling ontoo the MoBo and back through the ground too the switchting PSU.* As the author on the www.audiosystemsgroup.com pointed in his white papers on the use of ferrites : just a few ferrites will have no effect. You need at least 7 too 8 too create enough impedance too get past the threshold effect (see the white papers)
* A very nice sound quality improvement is realised when also caps are used too create a 2nd or 3thrd order filter
But a 3thrd order filter with ferrites is only effective from 1mHz and up. I also want too filter the 300 kHz switching noise (and it’s harmonics.) So I also need something too filter these frequencies.
So right now I’m experimenting with 10mH ferrite chokes + caps on the P4 lines and the 12V power line too the Pico 160-XT. And whow (!) this also makes a very nice SQ improvement. I simply use the ferrite chokes which are inside these little noise filters. See picture. They are very cheap 3,50 euro’s. And I can easily solder P4 connectors and caps too them.
Van on filtering, demping, ect It sounds really good. Thankx too audiodan I got this idea. More too come on these extra filters. I need more time for experiments and listening.
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Edits: 08/10/11
Mark I am interested is seeing a better picture(if its not too much trouble) of where you cut some plastic off the P24 extension connector to solder caps. I've gotten my extension cables and I need to solder some caps onto it.I've got some caps stuffed into it and so now I have a true pi filter which does do a better job of filtering but the stuffing is getting crowded.
Sonics so far: less bright and a lot less edgy on some music files...better dynamics but a bit of a lower mid upper bass suckout (which I easily fixed with my crossover points/volume settings). Net/net better sonics after all the adjustments.
Edits: 08/13/11 08/13/11
Hi Theo,
Here’s a picture.
Van on filtering, demping, ect
Although since I’m using a very small SilverStone SST-ML03B HTPC-case for already 4 months now, I still have still plenty of room inside the case for the P24 with 7 ferrites for ribbon cables clamped onto the P24 extension cable.
I use another P24 extension cable (with 3 extra black square ferrites on it. More is always better :-) ) too reach the P24 connector on the MoBo again. This way the P24 with flat-ferrites can be laid on the bottom of the case in parallel with the MoBo. Using this extra P24 extension cable allows me too efficiently use the space that is in front of the mobo. See picture.
Van on filtering, demping, ect
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
It looks like you got a real clean cut on the plastic which I have not been able to do. Hate to keep bugging you but what kind of tool did you use?
Hi Theo,I used such a knife. I think in the US they are also called ‘Stanley knife’
Van on filtering, demping, ect The trick is: first cut in at all 4 corners.
Van on filtering, demping, ect You than make a cut on the inside along the sides like this.
This way you separate the plastic outside from the inner plastic deviding sections.
Van on filtering, demping, ect
After that, you can cut off the plastic strip on both sides. Take it easy. You may need a few strokes too finally cut all the way through the plastic on the side. See picture.
Van on filtering, demping, ect Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Edits: 08/13/11
Ok Stanley did the trick. I had bought 2 p24 extensions one was in cmp^2 with caps stuffed here and there and ferrites galore. As reported it sounded good. So in the process of operating on the other p24 cable to have access to the metal .. for soldering I shut the system off and pulled a few of the caps off the p24 in the sys (to solder them on the other p24). When I turned the system back on it sounded dull and grainy and closed down. I missed the caps...even though it was only temporary. Anyway once I fitted the spare p24 with soldered caps and ferrites and put it back in the system I was not prepared for the sq pop. Wow very, very nice! I finally see what you were saying Mark. Sound space bigger, deeper highs quick but integrated with the mids (hard to accomplish imo). And the bass was so much richer,deeper and tuneful. Its one of those things where everything sounds great. I played all my favorites and they never sounded better.A smashing success...thanks again Mark!!!
Edits: 08/14/11
Hi Theo,
Good too hear you finally got results!
Did you also put a filter on the P4?
Cleaning the power too the P4 will also result in better sound quality.
Cleaning the power of the P4 with this type of filter is very easy.
- Take a P4 extension cable and 1 or 2 ferrites.
- make 4 or 5 turns around the ferrites (as many turns as the length of the P4 extension cable allows)
- put caps before and after the ferrites.
And again you will have a very nice sound quality improvement!!
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Yep I did it already...it is a nice improvement. Thanks again.
I'm starting to think that my implementation of the stuffed capacitors into p24 connectors was seriously handicapped from sounding right. There is so much better sound from the soldering that I believe my pushing of the caps into the p24 slots was not making proper contact or intermittent contact at best. Anyway for those who want to try this my recommendation is don't stuff, solder right from the beginning.
Again a big shout out to Mark for this excellent tweak!!
perfect...thanks again!
thanks
Hi Mark and welcome back from holiday.
'.....(V)HF AC pollution travels from the ATX onto the MoBo on all DC plus voltage P24 lines.
You may prevent (V)HF AC pollution travelling on the +5 voltage onto the MoBo, but (V)HF polution than will simply travel onto the MoBo through the other positive DC voltage lines (+ 12, -12, +3,3, 5V stand-by, Power-OK) and return through the combined ground too the ATX.
A clean and steady ground is equally important. ‘Zero' and ‘Ones' are recognized as a voltage difference compared too ground. So both supply voltage and ground have too be ripple free. Both are of equal importance. Also the clocks on MoBo and Soundcard are very sensitive too (V)HF polution on supply and ground.
It is even important too prevent (V)HF traveling through the SSD ontoo the MoBo.
I also use a 3 order filter on the power lines too the SSD. And also this give a little SQ improvement.
See photo...'
I realize I have to do all the voltages I was just surprised I did not get a bigger sq improvement on the 5v line, thats all.
'...How do you avoid DC bias in the ferrites (saturation of the ferrites)? The magnetic flux produced by DC plus and DC minus currents through the ferrite have too be equal too avoid saturation of the ferrites through a DC bias.
It's not clear too me how you know that the magnetic flux produced by the % volt DC + and DC - minus return currents on the black wires will cancel out each other in the ferrite. The ferrites will saturate and loose there effect if no attention is paid too DC bias. Currents on the P24 are high. See my P24 measurements. Almost 1 amp for the combined +5 volt lines. Most ferrites saturate when there is more than 0.4 amps applied on them. I think one can easily go wrong when just grabbing the 5 volt lines and a some black ones...'
I don't know if I am avoiding dc bias...as I posted before I use brute force (I now have approximately 10 ferrites on my p24 line...trying to include a black with another color line within a ferrite).
'...By now I have soldered the caps onto the P24 extension cable connector. I cut away the plastic on the side of the P24 connector on each side, this way it's easily solder the caps onto the small metal parts inside the plastic connector. See photo...'
I plan to solder also once I believe I have the right combo of ferrites/caps in the right places. Do you still use 440 uf before the ferrites and 47 uf after (roughly a 10:1 ratio)?
'...So right now I'm experimenting with 10mH ferrite chokes + caps on the P4 lines and the 12V power line too the Pico 160-XT. And whow (!) this also makes a very nice SQ improvement. I simply use the ferrite chokes which are inside these little noise filters. See picture. They are very cheap 3,50 euro's. And I can easily solder P4 connectors and caps too them...'
I would like to see a link to these devices to understand them better. What really baffles me is that with my old mobo I got a significant sq pop but not on my new mobo. Maybe it needs more breakin time, maybe its the difference between an ES2L and an S2L Ga-G31m mobo.
Please hang with me on this. I believe you are getting good sq pops but I alas am not...yet.
Hi Theo,
Yes I still use this 10:1 ratio. But there is no real calculation or knowledge behind this ratio. I just copied values that I red in 2 documents on decoupling (the Murata paper and the Altera paper).
> I would like to see a link to these devices to understand them better. <
Nothing special about these filters. They all have the same schematics and they are widely used. You will find them everywhere. Since they are cheap, it’s an easy way too combine 2 chokes in a handy enclosure onto one can easily solder the PSU wires. I don’t connect them too a real earth. It’s just the chokes I’m interested in. It’s easy too create a 3 order filter with them. I place caps before and after chokes so I can experiment with much higher capacitance values.
Ontstoringsfilter 0,6 A (Conrad Electronic)
http://www.produktinfo.conrad.com/datenblaetter/525000-549999/534625-da-01-en-Printfilter_0_6_A_5500_2000.pdf
On the 12 V DC power line which powers the Pico, I use one which can handle 1,6 amps. On the P4 12 V DC power line I use one which can handle 0,6 amps. I chose hight mH values. The Voltage drop across these filters is about 0,3 – 0,4 Volts. But this is easily corrected through raising the output voltage of the linear PSU’s a little. I’m really amazed that these filters give a profoud positive SQ effect on the 12 V DC power line which powers the Pico. Not in my wildest dreams I would have expected this. It was audiodan who suggested trying this.
7 or 8 ferriets on the 12 V DC power line to the Pico have no effect. I expected this. But I didn’t expect 2 x 6mH chokes + caps placed BEFORE the Pico too have such an effect.
So the last stop, before the top (an all linear PSU) is no longer a hybrid PSU solution (P24: Linear PSU + Pico and a Linear on the P4)
The last stop before the top is: a hybrid PSU solution with heavy filtering. What filters and what way is the most easiest DIY solution, still has too crystallize from experiments. So far I found:
- caps + (much)ferrites + caps give a nice sound quality improvement and is easy to do. But It only effective in the mega Hertz regions.
- The Pico uses the LM2642 chip from National Semiconducter which has switching frequency of 300 kHz. This frequency (and harmonics) should also be filtered. So chokes have too be brought in.
Using ferrite based coupled chokes which can be found in these ready made line filters is an easy way the experiment with chokes in the P4 12 V dc power line.
Too my (big) surprise these filters also had big effect on SQ when placed on the 12 DC power lines that power the Pico on the P24. So cheap linear bench PSU’s (50 – 90 euro’s) may give a nice SQ pop, but the output of cheap linears appears to be very noisy somehow. I see no other explanation for the very significant SQ improvement when ferrites, caps and/or chokes are used on the outputs of a linear.
My overall conclusion is:
A hybrid PSU solution is no longer the last stop before the top (an all linear PSU).
The last stop before the top is: a hybrid PSU with heavy filtering.
I hope other inmates with more skills and knowledge will join these filter experiments.
When done right: ferrits + caps give a real significant SQ improvement, but will not deal with the 300 Khz swithing frequency from the Pico.
> What really baffles me is that with my old mobo I got a significant sq pop but not on my new mobo. Maybe it needs more breakin time, maybe its the difference between an ES2L and an S2L Ga-G31m mobo. <
I’m not a member of the ‘breakin school’. I don’t ‘believe’ in it the way audiophiles handle the subject.
I also don’t think it’s the difference between mobo’s types. I used several brands and types and the difference are very little too none.
I really have absolutely no clou on what is wrong and what might be the cause, as all details about your setup and connections matter very much.
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Hey Mark ....thanks for the info.
I have several P24 extension cables on order to experiment with. Also I thought about the caps in the antec and there is no reason why they can't be part of 3 order pi filter. But putting in a cap ferrite cap filter on the extension 1st connector/extension wires and extension connector to mobo will be another pi filter in series with the caps (in Antec), ferrites, caps in extension connector. So this combo is probably very effectiver in rfi rejection/dissapation, Anyway I'll continue to work on this.
But in my system its very difficult to do (i.e., limited space). But where there is a will there is a way. So I added a cap (440 mfd) across the 12v and ground pins of Juli@. I already have a cap across the 5v pins on juli@ so I figured the 12v feed line was the 2nd highest current draw after the 5v feed. Anyway....yes yes you were right very nice sq pop along with the ferrites. I still only have L filters I can't yet figure how to do pi filters. So thanks again Mark.But now some interesting stuff. You know I have some ferrites on my spdif feed out to my dac. I know many experts believe I might be throwing away some baby with the bathwater, but I tried anyway. I thought it cleaned up the sound a bit. Then I added a .0015 mfd cap too and it made the sound darn near perfect. Remember your curves for caps and ferrites? Ferrites basically handle suppresion above 10 mhz. I use 44 sample rate which IIRC is under 5Mhz so no problem. And my listening confirms this. So if a very small cap is good why not put a large cap in...440 mfd. I did and it suppressed the whole signal. This confirms the reference data which shows a significant increase in suppression for a cap this size especially under 10mhz. I tried this cap in several different places and got the same result. This whole thing got me to thinking that maybe the best remedy to emi suppresion is through these caps. The ferrites help but the caps are the optimum solution in suppresion. And I believe I am right at the razors edge for throwing away bits on the spdif line. But my listening tells me I'm throwing away bathwater and not the baby. When I added a large cap whoa I deepsixed the baby too. Why did I try these values? It was not via calculations it was because I had them at hand and just decided to try. Looking at your refenced docs seem to confirm this and my experimentation seems to support.
Yes thanks again Mark. Others are really encouraged to try this especially the caps. Now if I can figure out how to do pi filters I'll try that next.
Edits: 07/09/11
Hi Theo,
Thank you for feedback.
* About the caps on the Juli@.
I don’t know what you mean and how you do that: “across the 12v and ground pins of Juli@” and an L-filter (ferrites + caps) on the Juli@ ?
My Juli@ digital part only has some litte pins on the top of it. See picture.
Van on filtering, demping, ect
How do you manage too fit an L-section filter on there? Is there a thread or picture where I can read/see what you mean and how you are doing that?
* On electrical S/Pdif and noise traveling on these lines.
I agree you are definitely at risk of throwing the baby away with the bathwater.
From your experiments I would conclude there is unwanted noise traveling along with the S/Pdif square wave signals onto your DAC. And your DAC doesn’t like it.
Also some nice clean thinking approach on intercepting the highest noise frequency’s by cutting it off through use of little caps, without cutting (too much) into the S/Pdif frequencies.
But one thing is clear: if placing ferrites and/or caps on the S/Pdif-line is filtering unwanted noise, than there is noise on these S/Pdif lines traveling from your computer to your DAC.
What too do about it?
0. Just let it travel there. Who cares.
1. Use another DAC that has better ‘cleaning capabilities’ or is more immune.
2. Try sending the noise on a D-tour too the ground, before it reaches the DAC (filtering)
3. prevent the noise from traveling across to your DAC (galvanic isolation)
4. kill the source which produces the noise.
May be also other electronic parts on the mobo will like this and might function better as a result.
5. anything else I didn’t think of to stop the noise from traveling too the DAC.
Whether ‘caps only’ (adding capacitance) or ferrites (Resistance/inductance) or a ‘combination of caps + ferrites’ (L-filter, л–filter, ect) are best, depends on which HF noise frequencies you are targeting and how steep you want your filter slopes.
In my situation, placing only caps on the P24 did almost nothing for sound quality. Also placing only a few ferrites improved SQ only somewhat. But putting enough ferrites on the P24 (I tried too realize a 1000 ohms) combined with caps before and after the ferrites, brought generous amounts of sound quality improvement.
One last question. I have received 2 reply’s, yours and JBPZ’s. Both reply’s are about the Juli@ card on adjusting/filtering the power regulation and signaling. But I posted about filtering the power on the P24 and P4 lines with L-section and/or л-filters based on combining ferrites and caps. How does these two relate too one and another?
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
All I meant was I already have ferrites all over p24 and caps across Juli@ pins for 5 volt feed and now I have added caps on the 12 v feed. I know that eliminating noise on the power lines are the optimum solution for killing emi but I don't have any room on p24 to be able to get at installing caps so I take what I can get. Thats all. I am not willing to add more stuff to P24, already no room.
Hi Mark
I've read your thread with great interest.
Effect of a pgreek filter is reasonable but your realization is just a bit complex for me, my PC is plentiful of cables yet so I've some difficult to put in an extension P24 cable with a lot of Ferrite rings plus capacitors.
So I ask you this question:
have you tried a pgreek filter only on the couple of cable that feeds the pico psu? Following the "garbage in/garbage out" philosophy it could be a simpler solution? Can you try this one against your solution and report the results?
I'm very curious about it, maybe could be not the best solution at all but I'm sure that could be a simpler solution, if effective.
Thanks a lot for your interesting work
Daniele
Hi Daniele,
What is a ‘pgreek filter’? I never heard of it.
Also google-ing came up with no results.
What is it?
* About the Pico and the switching noise it (probably) produces.
I’ve red that a Pico is also a switching PSU and thus ‘chops’ 12V DC into 5V DC and 3.3V DC.
So being also a switching PSU, the Pico probably also produces unwanted switching noise.
( I’m not completely sure about this. But I guess it must be)
I can understand your view and philosophy on the "garbage in/garbage out" principle.
I agree with: what you clean before the Pico, you don’t have too clean after the Pico.
But since the Pico probably is adding a lot of noisy garbage by itself, filtering AFTER the Pico can’t be left out.
Most likely this is why replacing the Antec Earthwatt ATX -> P24 with a Linear PSU -> Pico -> P24 in my setup, only gave very little and subtle sound quality improvement.
(Again: I’m not sure about this. But I guess this must be the cause)
Constructing and building a high quality linear PSU myself too power the P24 is far beyond my DIY skills. But constructing a L-section filter or an π-section filter through use of ferrites + caps is within reach of my DIY skills. Actually it is rather easy. Anyone can do it. No special skills needed.
And (very!) much to my surprise, the results are nothing less than truly spectacular.
Especially when compared too all previous PSU improvements I’ve already done in my setup.
Studying the picture of the inside of your Zahlman cMP setup I’m very curious what the schematic is of the DC filters I see being used before the Pico and also on the right side of the case.
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
π filter=pgreek filter . It's my trivial translaction, I cannot find the right symbol :-).
About picoPSU is it a switching or not? It could be possible, or better probably, but in my system I've noticed a great difference in SQ changing from Seasonic to Peaktech to my diy linear PSU, being last one the better.
So i think it could be possible that the disturbance of picoPSU could be amplified by a "dirty" 12V current so if you ameliorate the current quality before pico could ameliorate his 5V and 3.3V currents.This was true in my system.
Beyond the regular capacitor in each PSU of my linear PSU proposal, I've saved only the Silmic run (6000 mF) in the P4 line of my previous setting.
My "noble" Belleson superregulator based PSU units (2 for the soundboard, 1 for pico and last one for P4) are assembled on a single board inside an old ATX PSU case.
The others board you can see on the picture are the 12V and 5V PSU feeding SSD, HDD,Fans, USB and CD-ROM.
No filters on picoPSU line at all.
Hi Daniele,
Sorry I mistakenly studied the picture posted by nagual192 as being your setup. On nagual192’s picture I see a big wired choke on the 12V DC power line that powers the Pico (mounted on a yellow board which is resting on the cpu cooler). I guess that choke is there too filter power too the Pico. So I was curious which frequencies nagual192 targets and how he does the math on preventing the choke from ringing.
This time I studied your pictures.
Do you use your Asus Xonar card as a DAC or as a digital interface too an external DAC?
I guess optimizing power for when the Xonar is used as a DAC, will be extremely challenging and difficult. Slightest micro levels of ripple & noise in power and/or in ground (0 Volt) produced by the MoBo will impact in the DA conversion process. Optimizing this situation is much too difficult for my level of knowledge and skills.
Until some years ago I used a Lynx L22 soundcard but I finally settled on an external DAC.
And only recently I decided too follow Cic’s advice too 100% isolate the external DAC from the computer through use of an (good quality) optical connection.
This way my power supply modifications now only impact the digital processing on the mobo and on the digital inferface. I don’t have too worry any longer on how these modifications will also impact the digital too analog conversion process.
In both situations (Xonar used as a DAC or used as a digital interface) it is important too know if a Pico produces noise on the 12, 5 and 3.3 output. If the Pico is producing noise, than the clean, high quality power coming from the exquise Bellesons, is spoiled by the Pico. Too direct filter actions where they are most effective, it’s important too know if the Pico spoils your carefully generated clean power.
On the other hand it is quite simple too try out your suggestion.
Here’s what springs too mind on what I can do.
* For the experiment I will use the pico PW-200-M.
It is said that this pico regulates all 3 DC lines. (12, 5 and 3.3)
* Use of ferrites
I will put ferrites beads on the 12 DC power line (coming from a linear) too the pico pw-200-M
I will try too realize at least 1000 ohms
I think I can do this through putting 3 or 4 ferrites in series with each a decreasing numbers of turns.
See http://www.audiosystemsgroup.com/publish.htm why.
* Use of Caps
I will put Oscon caps in front and after the ferrites to create a л filter. 47 uf after and some 300 uF before the ferrites.
* Results.
Give some 2 weeks too come up with first impressions, as I don’t know how much time I will want to spend on critical listening sessions indoor.
Besides a very nice spring, until now we had lousy summer weather in The Netherlands. But when weather changes, I will be out with friends in the park in the evening. Enjoying BBQ and drinks. When living in the northern part of Europe, one has too grab all chances too enjoy the summer when weather is good. ;-)
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Hi Mark
first I agree with you about BBQ in summer time. I live in Sardinia, I'm bored by sun and summer, and just now I were dreaming some cold: here we have 40°C! So good BBQ in the north country!
I use only the analog output of the Essence and after my last upgrade ( sandy bridge CPU+MoBo and Belleson PSU) I think that picoPsu (150 and 160XT, I've on my 2 cMP2) or wears only simple regulators for 5V and 3,3V or is really the best Switching PSU in the world, IMHO!
SQ of my system is now astounding, background silence is incredible, each instrument rises from nothing, with music cue ( and decay too) so natural and dynamic that seems impossible to me they could come from a digital source. Sandy Bridge is a very effective upgrade in the war against electric rumor and picoPSU is a fantastic ally.
All this is too difficult to acheive from a switching PSU, and I must repeat that, in my experience, linear PSU are'nt all the same: my peaktech sounds worse than the Belleson, surely when feeds the PicoPSU. So Is it the pico sensible to the current quaity? My response is yes.
I'm starting to use ferrite rings. So I'll wait to yours next "cold" tests, in the meantime I carry on with mine.
Buon appetito e salute! Enjoy
Daniele
Hi Daniele,
Last 2 weeks the lousy summer weather even got worse in The Netherlands. In some places it rained for 20 hours on a row. According our national weather forecast service (KNMI) this hasn’t happened since 1954. : -(
Despite this weather, I did not spend much time on the project on filtering the power supply (filtering the 12V DC power input) too the Pico PW-200-M.
Although I made some progress. I did choose and order the parts needed for a variety of filter setups. I not only want to do some experiments on filtering the input side (as you requested), but I also want too do filter experiments on the output side of the Pico. As it is cleaning the output of the Pico, that I’m most interested in.
Last week I did some research on the WWW too see if I could discover what switching frequency is used by the Pico. I discovered that both Pico’s in my setup (the PW-200-M and the 160-XT) use the LM2642 Two-Phase Synchronous Step-Down Switching Controller (with a magnifying glass one can see the part number on the controller). This controller is manufactured by National Semiconducter.
See: http://www.national.com/ds/LM/LM2642.pdf
This LM2642 chip consists of two current mode synchronous buck regulator controllers with a switching frequency of 300kHz. This 300Khz switching frequency will probably result in widely scattered band of noise on the Pico outputs. As I don’t have a scope, I don’t know how well this noise is filtered, but putting an π-section filter made of ferrites + caps on the pico output, resulted in a very nice sound quality improvement. So I think the output of the Pico is not filtered much. Probably only so much filtering is done as is needed too comply with EMI regulations. Since an π-section filter from ferrites + caps only ‘bites off’ noise effectively in the upper regions (1 mHz - 300 mHz) of the widely scattered hash of 300 kHz switching noise, I’m also looking for ways too filter the HF noise from 100 kHz too 1 mHz.
Since ferrites are not effective in this region, I will need too use filters made of inductors + caps. So I probably will end up with some sort of 2-stage filter. First stage for filtering the 100 kHz too 1 mHz region with a combination of inductors + caps. The second stage (1 mHz - 300 mHz) with a combination of ferrites + caps.
So Right now I’m searching the web for idea’s on how to filter SMPS noise with a combination of chokes/coils + caps and what value’s too use for L and C.
I also want too experiment with such a filter design on the input and output of the Pico.
See: http://www.hamill.co.uk/pdfs/azrtatad.pdf
Some inmate on the diyaudio forum pointed too this.
* on the P4, a standard 230 V AC power line filter, combined with a π-section filter from ferrites + caps gave a very nice pop sound quality!
Just out of curiosity, I threw a standard EMC- 230 V AC Power Line Filter for 1-Phase Systems (2 x 10mH + 2x 0.015 uF + 1x 2.2 nF) into the 12 V DC power line too the P4. See: Ontstoringsfilter 1 A (Conrad Electronic) .
I already use an π-section filter (2x 100uF + 2 ferrites (with 4 turns) + 2x 47uF) on the P4 12 V DC power line coming from the Velleman Lineair PSU.
I put the standard 230V AC power line filter before the π-section filter.
It made a real nice SQ improvement !
Every time I’m still surprised on what a little noise –cleaning can do for sound quality in a cMP setup. So I can’t wait too start experimenting with various combinations L + C filters. I expect the ordered parts too arrive at the end of this week.
So coming weeks/months I will be busy experimenting with various concepts and various combinations of L + C filters on the input and output of the Pico’s in my setup.
I will start on the input side of the pico.
As soon I have have results I let you know, but this surely will take another few weeks.
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
Hi Mark
Great job and very good questions to ask.
I'll wait for your experiences, you are a prodigy of learning!
Daniele
Are you updated regulators?
Hi JPBZ,
I'm sorry but your question is too cryptic for me. What regulators do you mean?
I only experimented with L-section filters and π-section filters on the P24 line from the Earthwatts ATX created through a combination of (many) ferrites and conductive polymer capacitors (Sanyo os-con).
As suggested by the Murata paper (and other papers as well) this combination of ferrites and capacitors VERY EFFECTIVELY filters power supply noise on the P24.
Very much too my surprise the sound quality ‘quantum leaped’ through use of π-section filters on ALL (!) wires of the P24 coming from the Antec Earthwatts ATX.
No kidding! I wasn’t expecting such spectacular sound quality improvement at all. I only expected some subtle improvements over the PSU combination I normally use on the P24: linear PSU -> Pico 160 XT -> P24 connector.
In no part of this amazing PSU noise filter project I touched or changed any regulators.
What regulators do you mean?
Mark
fully optimized cMP2 PC -> ESI Juli@ -> Van den Hul Optocoupler MkII-> Lavry Black DA10 -> XLR Mogami Gold -> Klein & Hummel O300
The Juli@ card regulator, as some guys has used (Gstew and others). Thankx
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: