This is sick. I managed to find out the 7805 working condition. 12V goes from molex to the 3*3.3ohm resistors and directly to the 7805 input pin. There's no other filtering. The first caps that seems the 7805 are the 2 FG caps. The first gets got, second father less. The 12V given to the 7805 is practically unfiltered and there is only protection with resistors. I'm moving 100nF filtering directly to the 7805 pins. Also I noticed that the 7805 and 7812 case is used as a ground. Moving the ouput filtering cap from the FG cap to the regulator ping only sligthly changed the sound in a good manner. I'll add input filtering cap. I may do the same with the 7812. I may post pictures.
The current Xonar 5V voltage system for analogue stage. I wonder if the 1000uF cap may require to add diode to the 7805 between input and output, but it's still working. There's a place for another cap after the coil. The 5V regultion is now pretty good compared to other places. I may say it's now the best. The DAC 5V may need fixing. The -/+12V seems to be far behind the rest while it was the best at the begining when I took care of it. While improving 5V with addition 1000uF in its current state I noticed good bass improvement and the imaging is much better. The bass control is much better. Without proper caps it's not so good making also 3D positioning much worse. Before upgrade it was like the lower sound harmonics were somewhere else than higher. Now they are more centered 
I removed half caps on the card. The 5V uses 3x1000uF and it's a must. For the +/-12V it's easier and plays different role. I put attention into 7812. Using Nichicon KZ was naturally not enough (not to mention Nichicon FG). Oscons were better but KZ + MLCC was better. Oscon + MLCC in place of existing FG? Bad. Oscon + MLCC? Better. Adding MLCC o the regulator pins was a must. This 1cm makes a difference. One the distance was minimized the quality improved much. Highs are so smooth, choirs are smooth.
I noticed the difference between my 3 favourite opamps: the internal power decoupling and power noise resistance. As long as there is a difference, the power is not good enough. By improving power the difference shortens. LM6172 seems to be equal in mid and high freqs. OPA627 to be good at high and bad in mid. AD797 seems to be good but too soft on mids while highs are little worse.
I noticed the difference between my 3 favourite opamps: the internal power decoupling and power noise resistance. As long as there is a difference, the power is not good enough. By improving power the difference shortens. LM6172 seems to be equal in mid and high freqs. OPA627 to be good at high and bad in mid. AD797 seems to be good but too soft on mids while highs are little worse.
It's not so easy. I noticed I did all I could with Muse KZ it wasn't natural and details were lost. I switched back to Silmic II to get great details and smoothness boost. Even small MLCC interfere with the Elna caps. First time I noticed in one of my reference recording turned on neon/light bulbs.
The DAC filter caps. The also play role in sound and can be improved here.
I want to add 7912 for check but I find the pins for it are probably wrong. The board gnd for it has the -12V potential. The input or output pin has the GND, the other is not connected. Is it the faulty connection so the regulator wasn't inserted?
The DAC filter caps. The also play role in sound and can be improved here.
I want to add 7912 for check but I find the pins for it are probably wrong. The board gnd for it has the -12V potential. The input or output pin has the GND, the other is not connected. Is it the faulty connection so the regulator wasn't inserted?
With those cuts (output leg of 7812 and the diode beside L20) you may isolate the power rails (+/-) of the I/V stage, final opamps and some other analogue stages on SPDIF in/out).
The caps on right hand side of the regulator (my picture) may be ignored. What is on the left hand side of that regulator is important. There are the rails I named above.
There is no other better way to get symmetrical +/- rails, than using an external PSU for this purpose. The design of this card is wrong in this power stage. It works how it works, but is not right do it. More than this 12v is not just right tension level for the opamps used in I/V and finals. There may be 15v. Even the original opamps it may have +/-15v to work at the full parameters. Those are the reasons for this mod: an external, better PSU for this stage. With an transformer/toroid giving 2x18v, one may easy heave +/-15v regulated. On the picture are indicated where the 15v power rails it may be inserted.
The ferrite beads may be there where are planted, as it are used for filtering reasons. The tension drop on those ferrite beads are very low, and not important in this case.
Asymmetrical tensions on original power rails, come from bad design of the board.
The caps on right hand side of the regulator (my picture) may be ignored. What is on the left hand side of that regulator is important. There are the rails I named above.
There is no other better way to get symmetrical +/- rails, than using an external PSU for this purpose. The design of this card is wrong in this power stage. It works how it works, but is not right do it. More than this 12v is not just right tension level for the opamps used in I/V and finals. There may be 15v. Even the original opamps it may have +/-15v to work at the full parameters. Those are the reasons for this mod: an external, better PSU for this stage. With an transformer/toroid giving 2x18v, one may easy heave +/-15v regulated. On the picture are indicated where the 15v power rails it may be inserted.
The ferrite beads may be there where are planted, as it are used for filtering reasons. The tension drop on those ferrite beads are very low, and not important in this case.
Asymmetrical tensions on original power rails, come from bad design of the board.
The Essence ST headphone amp is not the best one. The stock card requires very clean power supply. Full potential of this Essence or your HD600? I made all my DIY headphones very dependant on the amp quality due to ease of drive. Stock headphones use a lot of components which make them sound in the maker's idea way. Stock Essence is not good enough to drive stock headphones all the way. There's still some sonic trace left from the headphones. I mainly plugged in 32ohm drivers including the DIY drivers with 32ohm voice coil. The highest impedance I tried was 150ohm.
If the Xonar Essence can take the full potential of headphones then these headphones are poor ;-) If you know what I mean...
Anyway going back to the caps: I noticed oscons bypassed with ceramics can be good. I will try to capture caps impedance plot with Agilent impedance meter and do the 20Hz-110MHz plot. From what I heard with ear I can tell:
- MLCC - very low ESR, very low ESL, low capacitance
- oscon - very low ESR, medium ESL
- elna silmic II RFS - medium ESR, low ESL
- nichicon muse kz - low ESR, medium ESL
- nichicon muze fg - medium ESR, higher ESL
- nichicon HV - high ESR, low ESL
That's why I find good to use multiple Elna Silmic II or use oscons bypassed with MLCC. It makes the impedance more flat when trying to get rid of power supply noise. Silmics II have quite flat impedance but not so low. Just putting more make the sound better. Oscons however leave some unpleasant harshness in the reverberation. Adding MLCC in the 100nF-2uF range helps deal with it. Single MLCC or single oscon in my case? Not. Why? Maybe because I've got no larger MLCC than 1uF yet ;-)
Decoupling digital is a pain. Multiple oscons and MLCC when there's no room for it. The lower the overall impedance, the better.
If the Xonar Essence can take the full potential of headphones then these headphones are poor ;-) If you know what I mean...
Anyway going back to the caps: I noticed oscons bypassed with ceramics can be good. I will try to capture caps impedance plot with Agilent impedance meter and do the 20Hz-110MHz plot. From what I heard with ear I can tell:
- MLCC - very low ESR, very low ESL, low capacitance
- oscon - very low ESR, medium ESL
- elna silmic II RFS - medium ESR, low ESL
- nichicon muse kz - low ESR, medium ESL
- nichicon muze fg - medium ESR, higher ESL
- nichicon HV - high ESR, low ESL
That's why I find good to use multiple Elna Silmic II or use oscons bypassed with MLCC. It makes the impedance more flat when trying to get rid of power supply noise. Silmics II have quite flat impedance but not so low. Just putting more make the sound better. Oscons however leave some unpleasant harshness in the reverberation. Adding MLCC in the 100nF-2uF range helps deal with it. Single MLCC or single oscon in my case? Not. Why? Maybe because I've got no larger MLCC than 1uF yet ;-)
Decoupling digital is a pain. Multiple oscons and MLCC when there's no room for it. The lower the overall impedance, the better.
Essence can very well drive 300 ohm headphones. Even lower and higher impedance works well. The headphone chip on this card is a very good one. But this chip alone can not do to much... There are stages before this chip which it may be improved to have a really audiophile headphone output from this card.
But anyway, the headphone output is better than RCA as it is. One may chose correctly in the software driver, the right impedance one may use for the headphones.
I ordered Vanguard TCXO. Is it sensitive to the given voltage? Built-in 7805 is enough or building additional one may help? Low noise voltage regulators are to hard to find. Decouple voltage? I've got no ferrite bead. I see you used ferrite bead. The 10uF tantal is however between Vdd? and some place on the board.
It have to be only the right voltage for it`s Vdd - 5v. 7805 is not the right regulator for an oscillator.
I can not agree with your affirmation about hard to find low noise regulators... If you have access to Farnell or another electronics vendor, you may for sure find such. You may also use an shunt regulator for 5v, f. ex. T431. You may connect the regulator to the oscillator after you are sure about that regulated tension is right on it`s output.
But anyway, it may be a quite important improvement with this oscillator, than the original one.
I can not agree with your affirmation about hard to find low noise regulators... If you have access to Farnell or another electronics vendor, you may for sure find such. You may also use an shunt regulator for 5v, f. ex. T431. You may connect the regulator to the oscillator after you are sure about that regulated tension is right on it`s output.
But anyway, it may be a quite important improvement with this oscillator, than the original one.
I tried various xonar caps bypassing in the most important places in my system. While through-hole MLCC as electrolytic bypass gave clear high frequency, the resonance gave a little unpleasant sound. I start do a little trick the card. There are a lot of ferrite beads. The example is stock 7805 || cap 47uF || ferrite bead || cap 47uF. Or 100uF. When putting 1uF MLCC on the 7805 the best I could do is to remove first electrolytic cap and then replace second with a good sounding 1000uF.
Decoupling digital stage is a pain. I went away from the idea of oscon || MLCC. Last setup I used was 2xoscon || tantal || 2xMLCC. Maybe try (large cap + series SMD R0.27-1.00) || MLCC. Or a stack of different capacitance MLCC.
What's funny is that I noticed some in my opinion low-end opamps now are pretty good. The difference became so small and everything came mostly to the different internal power decoupling if there was any.
Decoupling digital stage is a pain. I went away from the idea of oscon || MLCC. Last setup I used was 2xoscon || tantal || 2xMLCC. Maybe try (large cap + series SMD R0.27-1.00) || MLCC. Or a stack of different capacitance MLCC.
What's funny is that I noticed some in my opinion low-end opamps now are pretty good. The difference became so small and everything came mostly to the different internal power decoupling if there was any.
I'm still waiting for my vanguard to arrive. I bypassed ferrite bead on the 5V with 4,7ohm. Is it noise or just faster response? I don't know. I took my lying unused Panasonic FC caps (they've got higher ESR than FM) and I noticed they are very good in the digital stage. I may have to measure them with good impedance meter. The card gets to its limit in my opinion.
I had to change the RCA IC to my headphones to the of the Sommer cable and decided to still try to use Amphenol plugs. I had to move from one of the Klotz cables and try lower capacitance. The sound became unpleasant but I decided to tune my Xonar for it. I had to lower opamp gains to maximize SNR, the noise floor was limiting the dynamic range of the music in the quitest parts. The added 4,7ohm resistor was really introducing noise from the regulator. Then I tried to make sound more coherent and clear. I fully removed 5V DAC decoupling caps I noticed.... almost no difference. I tried different caps and sizes and stayed with Panasonic FC as the main 5V at the filter output. Then I will go for proper decoupling. I noticed there can be a trick applied for better imaging on headphones to simulate some tube-like soundstage: make caps work more like a dual power system, less global caps, more channel decoupling. This goes away from my other the idea of using really small deceoupling caps. They seem to be slightly better at higher frequencies, probably due to little lower ESL.
Comming back to the 7912 - the potential at the 7912 ground doesn't seem right. Seems like it's not connected to anywhere. The input and ouput pins are shorted on the stock card with ferrite bead. Ok. However the ground pin is not connected to any of the -12V caps legs: should be on the positive pin.
Comming back to the 7912 - the potential at the 7912 ground doesn't seem right. Seems like it's not connected to anywhere. The input and ouput pins are shorted on the stock card with ferrite bead. Ok. However the ground pin is not connected to any of the -12V caps legs: should be on the positive pin.
Some pictures of my xonar from last 2 weeks. Comments are already in this thread. Right now I'm removing almost all +/-12V caps and building from scratch. There are 2 caps in the line in which are connected to the +/-12V and ground. Also I didn't notice the 7912 has switched input and ground legs. Now it's ok. The Silmic II caps after break in are almost non existent when filtering high freqs. I could remove them from the decoupling opamps and I noticed very, very small difference. Stock FG can go higher. I could install 7912 but I've got no smd resistor to increase voltage on the input before regulator. DAC Vcom caps are very important. I noticed Coris replaced them later. I also added larger Vishay 1845 for filtering 5V after 7805. There are already 1839 for filtering PSU 5V. I find them very good at filtering noises.
Coris, do you know the inductance of the CLC filter in the 5V stage? I almost completely removed 5V stage on the card. I remodeled CLC to filter to be more a C(R+L). Also used the metal cover. The sound become fresh much more detailed. More than ever on this card. Sound effects are fresh. The previous noise is non-existent. Knowing the L value would help me properly model the rest.
I stripped the 5V CLC filter to the C. I now the ripple is bigger and it should be noise but... it's great! It's totally different from what I have ever heard on this card. The background sounds are very, very clean and pleasant like the dynamic range went up by 10dB and there is no added noise to it. Voices are thin and the imaging is more varied. Shouldn't it be vice-versa?