I forgot to answer you to a previous question about the Xonar driver I use.
I have tried Unixonar. I found out that it had some issues (for me) in the way it were put together, and the sound it were not specially improved because the mods done on the original driver.
I found out my self that to modify the original driver, update some part of it and so on, is not very difficult. So, I installed the last original one, I`ve done my self the necessary mods (ASIO update, and few other) and I run the original one.
I noticed that is quite important for the sound and sound stage out of the STX/ST how is made and function the motherboard and/or the processor, and maybe the clock system in the computer. I can not say exactly what it may be in this case, but I`ve only noticed differences in sound quality (especially changes in sound stage) when used one or another motherboard/processor with the same sound card.
I recommend to install and control the windows timer resolution (by the application with the same name). This have in my opinion an quite big impact in the way the data is streamed in the system, and in this sound board case too.
I have tried Unixonar. I found out that it had some issues (for me) in the way it were put together, and the sound it were not specially improved because the mods done on the original driver.
I found out my self that to modify the original driver, update some part of it and so on, is not very difficult. So, I installed the last original one, I`ve done my self the necessary mods (ASIO update, and few other) and I run the original one.
I noticed that is quite important for the sound and sound stage out of the STX/ST how is made and function the motherboard and/or the processor, and maybe the clock system in the computer. I can not say exactly what it may be in this case, but I`ve only noticed differences in sound quality (especially changes in sound stage) when used one or another motherboard/processor with the same sound card.
I recommend to install and control the windows timer resolution (by the application with the same name). This have in my opinion an quite big impact in the way the data is streamed in the system, and in this sound board case too.
guys i was wondering about all these mods for external psu for the card.what about if somebody makes two serial psu one with 5v and one with 12v volt and feed them to the molex connector (i read somewhere that 5v must be high current maybe 2A)without cuting any traces to the soundcard.wouldn`t this be a much easier improvement?do you see any problem to this?
The point with those external PSUs is to power the board components/stages in a right way. Specially the analogue stage of the board which need +/- power. Minus part for the analogue is supplied on original board by an oscillator which use the +12 rail to do this. More than this , the final opamps need an +/-15v to give all of its. The original regulator on the board is 7812, which it will regulate what is coming from 12v rail... What kind of regulation is that? Not enough, the power which is to be measured on the analogue stage is +12v (more or less) and -11,5 (more or less)... They who designed this board did not care about output offset. They used AC coupling to prevent offset issues...
The board have many design faults and that because one may need to do it right to get the most of that this card really can... A part of this "doing right" is using good/clean analogue PSUs for the opamps...
This is the story of mods for this sound card...
The board have many design faults and that because one may need to do it right to get the most of that this card really can... A part of this "doing right" is using good/clean analogue PSUs for the opamps...
This is the story of mods for this sound card...
you are right i never thought iy`s +-12 so i feed very clean power at the molex the oscilator will somehow spoil it.i guess it`s better to power the opamps directly.do you know what the 5v is powering also?
i see at your photos that 3.3v is supplied from pcie so what is molex 5v for?
Last edited:
5v it feed some control circuits, control of the relays, and so... There is a circuitry on the board which give informations to the main processor that the board is powered on as it should, and this is managed by the driver software. That because the Molex it have to be intact and connected in place. Else errors messages pop up on the screen and no sound... Eventual modifications it have to be made on the power traces on the board, taking care that only the targeted chips have to be powered from an eventual external PSU. The external PSU is in fact dedicated to the +/- rails for the output analogue stage (I/V and final opamps), but part of this power is used too for the digital in stage of the board. The rest of the power system it have to be in function as it were designed. I have removed completely the headphone stage from my board, as I never needed such. So that chip did not take any power from the rest of the system. Is important that even though some relays may be not used, their connections to the control power have to be intact (they have to be clicking to give the messages to the main processor that all it still be OK in the system).
Is now quite long time I`ve been involved in this project. To give very precise answers now, it mean I have to go back to may papers (find those...) and do researches... But the most of the informations are wrote down on the thread.
So, please read carefully my posts from the beginning of this thread to get more details.
I have used from Molex only 12v rail, to make 5v for the analogue stage of the DAC and clock, and 3,3v for the rest, using cascaded regulators. 5v for the relays it still come from the Molex/computer rail. Details are in the previous posts.
Last edited:
you need to cut traces to isolate the original regulators on the board... But you can of course cut the output terminal of the old regulators. You may chose what is more convenient for you.
Is better let the regulators on the board, than desoldering it. The board is very destroyable when desoldering components from it.
Care about where is to solder the GND of the new PSU. Chose very short connection in the the near area of the opamps, no mater the side of the board.
Is better let the regulators on the board, than desoldering it. The board is very destroyable when desoldering components from it.
Care about where is to solder the GND of the new PSU. Chose very short connection in the the near area of the opamps, no mater the side of the board.
I played more with the card. To the existing caps I added monolithic ceramic to the digital stage, headphone amp and 5V decoupling. There were still some inconsistencies in higher tones, but better than the time I was using only opamp sockets.
Then I tried the 5V filtering. Coris mentioned there are 2 of them. I don't know the connection, maybe they are in parallel. Due to existing caps I could only access one of them. There are 2x100uF. I added one 470uF Nichicon HV which is quite slow and sounded breathless in the opamp stage. This filtering changed almost every aspect of the sound in every stage and is quite global like 12V power decoupling.
There may be an error for the blue text: 12v power filtering.
Then I tried the 5V filtering. Coris mentioned there are 2 of them. I don't know the connection, maybe they are in parallel. Due to existing caps I could only access one of them. There are 2x100uF. I added one 470uF Nichicon HV which is quite slow and sounded breathless in the opamp stage. This filtering changed almost every aspect of the sound in every stage and is quite global like 12V power decoupling.
There may be an error for the blue text: 12v power filtering.
You may not expect that only decoupling caps or filtering is enough to improve the whole sound out of this card.
I will suggest that you do what do you consider to be done about those filtering/decoupling caps (you may even not notice a clear/dramatic improvement in the sound), and then just go further to another mods, which it may be necessary to get more improvements.
Those caps are only a part of the mods to be done for to notice a dramatic improvement of the sound... You can go back to this stage maybe after the rest of the mods are in place. Then you may notice better the differences when one or another caps are in the right place.
I will suggest that you do what do you consider to be done about those filtering/decoupling caps (you may even not notice a clear/dramatic improvement in the sound), and then just go further to another mods, which it may be necessary to get more improvements.
Those caps are only a part of the mods to be done for to notice a dramatic improvement of the sound... You can go back to this stage maybe after the rest of the mods are in place. Then you may notice better the differences when one or another caps are in the right place.
What are the other mods? There is mainly quartz to change and what else? Play with the feedbacks.
What I read in the this thread is that the 5V after 7805 is used to make +12 and -12V and then regulators like 7812 or similar. If so, then why the card uses molex 12V and 5V to power the card? The power drain from both lines are similar as mentioned in the thread. So maybe the +12V goes to the +12V regulator, +5V to the +5V, and then -12V is made from one of them? I'm just curious to know which filtering changes what.
What I read in the this thread is that the 5V after 7805 is used to make +12 and -12V and then regulators like 7812 or similar. If so, then why the card uses molex 12V and 5V to power the card? The power drain from both lines are similar as mentioned in the thread. So maybe the +12V goes to the +12V regulator, +5V to the +5V, and then -12V is made from one of them? I'm just curious to know which filtering changes what.
I find the digital decoupling important for high frequency time resolution. With improper decoupling they are harsh, not precise, bit too loud. Now I use oscons with cermic. I may try smd ceramics and smd tantals once I buy them.
Clearing 5V path let me use oscons at the TPA +/- 12V decoupling. Normally they are artificialy harsh. Now they are clear. Using little slow but big caps at filtering stage and faster, smaller caps at decoupling works best for me. With bad filtering, the decoupling caps can't be too fast.
What about 1000uF smd tantals for 6V in one piece? 1500uF/4V are available but I'm not keen on using them with such a stress.
I plan to buy the XO Coris used. I may connect it to the 5V available on the xonar board.
Clearing 5V path let me use oscons at the TPA +/- 12V decoupling. Normally they are artificialy harsh. Now they are clear. Using little slow but big caps at filtering stage and faster, smaller caps at decoupling works best for me. With bad filtering, the decoupling caps can't be too fast.
What about 1000uF smd tantals for 6V in one piece? 1500uF/4V are available but I'm not keen on using them with such a stress.
I plan to buy the XO Coris used. I may connect it to the 5V available on the xonar board.
Do not forget to parallel the big decoupling capacities with 10n -100n good quality caps. I prefer SMD film as those have not long terminals/legs as the normal caps, which it may increase ESI at high frequencies. With SMD film caps one may be very carefully when solder it. To much warm, or to long time used to solder it, it may destroy those type caps.
About something else one may remember when is to decouple with large capacities.
ESR and ESI of two capacities/caps decrease (1/2) when they are paralleled. The clue here is to make an need it capacity, soldering together smaller capacities caps. Do you need 1000µf, then solder you together 10x100µf. or 2x500µ, or another combination (the many paralleled, the lower resulting ESR/ESI). Using small SMD caps this is very convenient.
Ceramic caps have not any ESI and extreme low ESR. Tantal caps have enough ESI/ESR as usually. Special tantal caps (quite expensive) have quite low ESR/ESI. _Use the described above strategy to reduce it.
The best when use decoupling caps is to not use an capacity in one piece, but many smaller soldered together.
Be carefully with the parameters of the regulators, and care to use regulators which are over current protected, when use large decoupling capacities. Do not forget that an regulator or PSU see a very large capacity at its output, when one use big capacity caps to decoupling spread in many places (many chips) on a board.
There is not only the type of dielectric or how many µF one use when decoupling, but very important are ESI and ESR of the caps. So read the datasheet of those components when you chose what you want/need of caps. Chose the lowest ESR/ESI as possible for that convenient price...
About something else one may remember when is to decouple with large capacities.
ESR and ESI of two capacities/caps decrease (1/2) when they are paralleled. The clue here is to make an need it capacity, soldering together smaller capacities caps. Do you need 1000µf, then solder you together 10x100µf. or 2x500µ, or another combination (the many paralleled, the lower resulting ESR/ESI). Using small SMD caps this is very convenient.
Ceramic caps have not any ESI and extreme low ESR. Tantal caps have enough ESI/ESR as usually. Special tantal caps (quite expensive) have quite low ESR/ESI. _Use the described above strategy to reduce it.
The best when use decoupling caps is to not use an capacity in one piece, but many smaller soldered together.
Be carefully with the parameters of the regulators, and care to use regulators which are over current protected, when use large decoupling capacities. Do not forget that an regulator or PSU see a very large capacity at its output, when one use big capacity caps to decoupling spread in many places (many chips) on a board.
There is not only the type of dielectric or how many µF one use when decoupling, but very important are ESI and ESR of the caps. So read the datasheet of those components when you chose what you want/need of caps. Chose the lowest ESR/ESI as possible for that convenient price...
Last edited:
You may notice that larger caps have lesser ESR. For 1000uF it can be 15m to 50mohm while smaller can have up to 1 ohm.
The caps that was a little hot is now cold. Now the 5V filtering caps keeps getting little hot. I replaced mine with a different one and larger. The same thing happens. I wonder what it supports: the 5V analog DAC lines (there is already decoupling) or another chip? Somewhere may be the lack of power and probably I moved the buffer closer to it: from the 12V filtering to the 5V filtering.
The caps that was a little hot is now cold. Now the 5V filtering caps keeps getting little hot. I replaced mine with a different one and larger. The same thing happens. I wonder what it supports: the 5V analog DAC lines (there is already decoupling) or another chip? Somewhere may be the lack of power and probably I moved the buffer closer to it: from the 12V filtering to the 5V filtering.
If you make 1000µ from 10x100µ SMD ceramics, then you have much lower ESR (µohm), and no ESI...
1000µ may have quite low ESR comparing with lower capacities, but large ESI. And ESI is the worst when about decoupling... Big components (caps) and theirs long legs introduce parasitic impedances, and this is only bad...
But anyway, you may chose as you consider.
The filtering/decoupling caps do not have to get hot. No warm either. It is wrong functioning somewhere. There is about ripple currents which may be quite high. You may review the power of your PSU, and its delivered current.
1000µ may have quite low ESR comparing with lower capacities, but large ESI. And ESI is the worst when about decoupling... Big components (caps) and theirs long legs introduce parasitic impedances, and this is only bad...
But anyway, you may chose as you consider.
The filtering/decoupling caps do not have to get hot. No warm either. It is wrong functioning somewhere. There is about ripple currents which may be quite high. You may review the power of your PSU, and its delivered current.
I've got the feeling that the 4 FG caps to the most right when looking at the card front are little hot. I added the 1000uF cap and it's also little hot. I know the lower two are for the 5V filtering but for what are the top two? I noticed you didn't change them.
I switched to the OPA627 and then to LM6172. There's a difference. They become much clearer than before. The LM6172 in my config works at much lower gain: 1kohm||100ohm in feedback loop and is slightly to soft. When digitaly decreasing volume the sound lacks resolution and speed. Increasing this opamp gain helps this.
I switched to the OPA627 and then to LM6172. There's a difference. They become much clearer than before. The LM6172 in my config works at much lower gain: 1kohm||100ohm in feedback loop and is slightly to soft. When digitaly decreasing volume the sound lacks resolution and speed. Increasing this opamp gain helps this.
Can you mark that caps on the picture? The Oscon are filtering for 12v rail coming from computer.
To appreciate more objective the sound quality in connection with different opamps, I suggest you to find a better oscillator and do this mod first. The oscillator it may have its own regulator. It works very well an 3,3v oscillator (own regulator). You will be very surprising by the resulting sound with a quality oscillator in place. Then you can prove some different opamps...
I will also suggest to cut the 3,3v coming from motherboard, and create this tension from f. ex. 5v rail on the board. Isolate this 3,3v rail from the computer power system it improve much too, the sound.
To appreciate more objective the sound quality in connection with different opamps, I suggest you to find a better oscillator and do this mod first. The oscillator it may have its own regulator. It works very well an 3,3v oscillator (own regulator). You will be very surprising by the resulting sound with a quality oscillator in place. Then you can prove some different opamps...
I will also suggest to cut the 3,3v coming from motherboard, and create this tension from f. ex. 5v rail on the board. Isolate this 3,3v rail from the computer power system it improve much too, the sound.
The lower oscon was hot first. Now these 4 FG caps are little hot with the one on the back side no matter the size or brand. I'm not sure if all 4 were hot. It may be that the stress went from the 12V filter caps to the 5V filtering caps by added larger caps to the 5V. I wonder where try to touch the next element. Is it the 5V analog DAC decoupling which I already try to improve or maybe some next regulator? I wonder if by proper filtering I can decrease caps stress and improve quality in the same time.
I marked these caps on the image below.
I marked these caps on the image below.
I replaced most 100uF silmic II with freshly bought 220uF. There was an improvement on AD797. Then switching to LM6172 was easier. I started replacing all elna 12V caps with a Nichion Muse KZ 100uF I just bought. They are a bit expensive. I used 47uF before and now 100uF. With LM6172 it's much clearer. There are 2x100uF KZ for TPA and 2x100uF for LM6172 plus 2x47uF KZ on rail-rail on opamp. Clear. Some instruments sounds 3D in the terms the sound comming from it is not a flat sound played in the space but some are more round. It's quite fast and has a lot more presence. Too bad I couldn't play the 5V line. I fought I could find SMD ceramics and SMD tantals in local stores as I bought SMD resistors. No way, I have to order them online too.
The uper caps in the picture are filtering for +/-12v rails. The lower ones are 5v filtering for DAC analogue stage, clock and others stages on the board.
You still use the board`s power system, and I`m personally quite sceptic for the results you may have... It is possible you have some high frequency oscillations, because the added caps, or something, and this may cause the heating of that caps. Else it may be few other causes to got this faults in functioning.
I will want to remove the original caps completely (as I did my self) and replacing with some better quality ones.
You may also try to see what happen when powering the board from another 12v/5v PSUs plugged in to the Molex. Find some 12v/5v PSUs you trust and power the board this way on your bench (out of the computer). Watch if any changing in heating of that caps. If is not happen, then it may be something wrong with the power in your computer (not about the DC tensions, but some pulses or some HF AC component). If is happen again with another PSUs, then it may be some inaudible oscillations in that area of the board. Is good to use an oscilloscope to find out more. Or just do some measurements anyway, with an multimeter, on few important points on power rails on that area.
I suspect that by adding some more caps on the back side of the board, you have created some loops and positive feedback somewhere. Do not forget that the switching PSU for the -12v rail is still working on your board...
Anyway, is something which goes wrong when you get hot caps... I will even do not think at all to appreciate the output sound in such conditions...
Else, yes, Crystek is the best solution. Vanguard is much better than that cheap original oscillator, but Crystek is much better than that Vanguard...
I use right now this last edition of Crystek oscillators, and the results are very good. There is not only about how many/less ppm is about, but how low is the jitter, phase noise and so on. Crystek is one of the best in quality for theirs oscillators. But it may be quite expensive... Anyway, I think is worth.
Last edited: