The shipping cost to Poland from DigiKey is $60. That's too much. The price of the oscillator is $26 and I can accept it. There are no Crystek on ebay. It will be hard to find it.
I had problems removing these caps. I removed digital decoupling cap. One leg stands out, one came out parially like on some of your pictures. I can't solder anything there even with soldering flux. That's why I'm trying to find most fine solution and remove these FG when I'm pretty sure what will be put there.
This oscon was hot before I added there 2x1000uF. It's still not much hot, it's just not cold.
I may try to put my external PSU (half-wave, +11.5V/GND/-11.5V on LM317 and LM337) on the +12/GND molex line and leave 5V/GND to the old PSU. Similar to the idea you came to. Which power planes on the Xonar will it change?
I had problems removing these caps. I removed digital decoupling cap. One leg stands out, one came out parially like on some of your pictures. I can't solder anything there even with soldering flux. That's why I'm trying to find most fine solution and remove these FG when I'm pretty sure what will be put there.
This oscon was hot before I added there 2x1000uF. It's still not much hot, it's just not cold.
I may try to put my external PSU (half-wave, +11.5V/GND/-11.5V on LM317 and LM337) on the +12/GND molex line and leave 5V/GND to the old PSU. Similar to the idea you came to. Which power planes on the Xonar will it change?
OK! Hot is something, a little bit warm is a little bit different...
To remove the caps: just move it back and forth (is not very big movement in the beginning). It will take a little time, but this movement will broke their legs at last. If you can not solder right in that point (broken leg), then you can scratch a little bit the isolated surface around that rest of the cap terminal. So it will be fine at last...
When use another external PSUs (board on bench) for measurements or another tests, you have to start in the same time both PSUs (12v/5v). Is quite easy to fix this...
Is right that is quite expensive the shipment at Digikey... Maybe you have Farnell in Poland... Maybe they have the same...
I did not get it your question about the changing of Xonar planes...
I removed FG caps for the TPA using slightly diffent technique. It's somewhat clearer, higher tones are louder, some little distortions are being heard in the sound. For my ear the frequency response in higher tones is now more flat in the 10k-20kHz range.
Current setup:
12V filtering: 2x1000uF Elna RJH
5V filtering: 1000uF Elna Silmic II
DAC 5V: 2x47uF Os-con SC or 4x220uF Elna Silmic II
DAC internal filtering: no change
3.3V: (removed FG cap) 2x47uF Os-con SC + 1uF ceramic
some caps near the 12V regulator: don't know what are they for, no change
I/V opamps LM6172: 2x100uF Nichicon Muse KZ, rail-rail 2x47uF Nichicon Muse KZ (one on each opamp solfer to the opamp pins)
near RCA opamp socket: not used, added 4x47uF Nichicon Muse KZ
headphone amp TPA: (removed FG caps) 2x100uF Nichicon Muse KZ
I find the LM6172 to be very, very detailed opamp. If I improve sound for the LM6172, the sound will improve on most of other opamps. What about suggested input resistors for these opamps? I heard some tweaks may be needed for them to fuction properly lub fixing input bias.
Current setup:
12V filtering: 2x1000uF Elna RJH
5V filtering: 1000uF Elna Silmic II
DAC 5V: 2x47uF Os-con SC or 4x220uF Elna Silmic II
DAC internal filtering: no change
3.3V: (removed FG cap) 2x47uF Os-con SC + 1uF ceramic
some caps near the 12V regulator: don't know what are they for, no change
I/V opamps LM6172: 2x100uF Nichicon Muse KZ, rail-rail 2x47uF Nichicon Muse KZ (one on each opamp solfer to the opamp pins)
near RCA opamp socket: not used, added 4x47uF Nichicon Muse KZ
headphone amp TPA: (removed FG caps) 2x100uF Nichicon Muse KZ
I find the LM6172 to be very, very detailed opamp. If I improve sound for the LM6172, the sound will improve on most of other opamps. What about suggested input resistors for these opamps? I heard some tweaks may be needed for them to fuction properly lub fixing input bias.
+12v on Molex power a quite big area of the entire board. The clue here is to let the power coming in to Molex as it as. One may cut only the power input in to the final amps, and feed those from an dedicated/clean +/- PSU. It may also cut it the -12v for the final analogue stage, which is made it on the board (little switching PSU on Xonar).
The rest of the +12v coming from Molex it may be also used to create (cascaded regulators) an local 3,3v PSU for the digital stage on the board. The same 12v from Molex have to be used to synchronize the board Power ON sequence with the Power ON of the computer. One may not remove this 12v rail from Molex and change it with an different PSU. This is quite important to prevent bad functioning, errors, and maybe damages on the board/computer.
If you modify the 12v on Molex input (anyway the Power ON sequence of this rails it have to be the same as for the rest of the computer), then it will be affected the entire board functioning. This is not a very good idea. If you isolate the power rails of the final amps from the rest of the power system on the sound board, then you can modify as you wish this tensions values (accordingly to the opams data sheet), to find the best results. The GND of the entire power system it have to be common, and with very short connections.
To be honest, I have to say that I`m enough critic to how it looks your board with all that quite big caps soldered everywhere. Is not about the aesthetically point of view, but about the functioning consequences...
How is done your mods in this area, it may introduce oscillations, high ESI values, and the caps quite big areas can capture/induce all sort of electrical parasitics from the rest of the computer.
One may not forget that digital circuitry are very noisy. Spikes and high frequencies noises are everywhere around those circuits. The digital system is not affected to much of such, because strong filtering of all sorts it is happen in the digital signals paths. But eventual sensitive analogue circuits placed near to digital stages it may be quite strong affected of such intense electrical activities. The noises is very hard to be removed from analogue stages without strong perturbations on the useful signals. That because one may care much to shield accordingly such analogue signal processing stages.
The original shielding of the STX/ST board it may be in its place. That is a not very well shielding (aluminium), but anyway is better than nothing. You may find a solution to have the better caps in right places, and in the same time care for an good shielding, specially for the analogue stages on the board.
Yes, I/V stage is quite important to get the right sound. Here is enough to work to find the best. On both I/V converter resistor and the feedback caps. If you read carefully the LM6172 data sheet, you may find that setup which can lead you to the best sound. When you experiment with this I/V stage, is better for appreciations to let in place that opams which are there for the moment. After you are satisfied with the result of tweaking I/V stage, then you may go further end experiment with another final opamps. Then back to I/V stage for another tweaks details ... and so on, back and forth until you find the best setup. It is quite much time consuming this kind of mod, but it brings the best results... After you set up the analogue stages, you may want to put in place a better oscillator. That will be the final improvement.
Last edited:
Your schematics is quite right...
I`m not very sure now (is quite long time ago I have been in this project), if the -11v or -12v is created from +12v or +5v. I think it is used +12v. You may go back in this thread to find those informations. I`m sure that I have wrote about this subject in the beginning of the thread...
There is an regulator, 7912 which it were designed to regulate the -11v from inverter... Just stupid, but is like this on the board... You have to cut the output of that regulator to isolate the - rail for opams. The +/- rails goes to headphone chip too. You may read the description of this power supply mod in the beginning of the thread or so. There are som important informations about this in earlier posts (mine and others).
The 3,3v rail feed too the main processor and the chips for SPDIF in/out. There is an quite important current on this 3,3v rail for digital stage on the board (few hundred mA).
I`m not very sure now (is quite long time ago I have been in this project), if the -11v or -12v is created from +12v or +5v. I think it is used +12v. You may go back in this thread to find those informations. I`m sure that I have wrote about this subject in the beginning of the thread...
There is an regulator, 7912 which it were designed to regulate the -11v from inverter... Just stupid, but is like this on the board... You have to cut the output of that regulator to isolate the - rail for opams. The +/- rails goes to headphone chip too. You may read the description of this power supply mod in the beginning of the thread or so. There are som important informations about this in earlier posts (mine and others).
The 3,3v rail feed too the main processor and the chips for SPDIF in/out. There is an quite important current on this 3,3v rail for digital stage on the board (few hundred mA).
I often return to the beginning of the thread. I searched for the 7912 on the card but didn't it find it. I'll try tomorrow.
I returned back to the only 1 oscon on the 3.3V decoupling. Sound became soft. It's not transparent, but soft and pleasant. Everything sounds fine, reminds me of tube amp. However I prefer more transparent and clear sound.
Will the regulator work fine with the output pin cut? I mean can the leg be in space not connected to anything while the other leg is connected?
I returned back to the only 1 oscon on the 3.3V decoupling. Sound became soft. It's not transparent, but soft and pleasant. Everything sounds fine, reminds me of tube amp. However I prefer more transparent and clear sound.
Will the regulator work fine with the output pin cut? I mean can the leg be in space not connected to anything while the other leg is connected?
Sorry, I have mixed up about negative regulator. There is place for component on the board, but is not planted.
Anyway that situation when 12v from Molex is regulated with an 12v regulator it still exist, but for +12v rail (as you can see on the picture here - 7812).
The negative rail it have no regulation more than what is done in the inverter.
I have marked up on the picture the points where is to be cut or lift up the component`s pin.
There is no problem when the regulator have the output in the air. You can of course cut both or all its terminals, but is more easy to cut one... If you cut the input only, then when you put your external PSU on that it`s output pin, is not just what the regulator likes... BTW, if you cut out it`s output pin, that regulator do not work any more... for the purpose it is there. You have to use another PSU in that point...
This picture here is an old one. There is not my last modification here, but one of the many...
Anyway that situation when 12v from Molex is regulated with an 12v regulator it still exist, but for +12v rail (as you can see on the picture here - 7812).
The negative rail it have no regulation more than what is done in the inverter.
I have marked up on the picture the points where is to be cut or lift up the component`s pin.
There is no problem when the regulator have the output in the air. You can of course cut both or all its terminals, but is more easy to cut one... If you cut the input only, then when you put your external PSU on that it`s output pin, is not just what the regulator likes... BTW, if you cut out it`s output pin, that regulator do not work any more... for the purpose it is there. You have to use another PSU in that point...
This picture here is an old one. There is not my last modification here, but one of the many...
Last edited:
That's a little strange. The reason was probably to support regulators with higher voltage to make them work properly. So some 5V filtering are still before being converted to +/-12V and not being used in the 5V DAC analogue stage anymore. More, some 12V
filtering is only being used for 5V DAC analouge stage and not the opamps.
I'm trying different settings for the digital stage. Ceramics are the must. The 47uF oscon is too soft while bypassed with 100nF MLCC is little clearer. Still 2x47uF + 1uF MLCC was better but sounded unnatural. Moving to the SMD is probably unavoidable. Stack of a few >10uF MLCC may sound good
I'll take a look at the path near the 7805. If the 2 capacitors near it are a little hot, then maybe one of them is at the input and the other is at the output. One closer to the output pin and any placed in the parallel are mediocre hot. One closer to the input pin is little hot. 12V filtering oscon near the 7805 is also little hot.
Still I try to figure out proper digital decoupling. I leaved only 1uF MLCC, the card wasn't recognized and relays didn't switch. Then I added 22uF tantal. Still the same. So I placed 470uF electrolitic with probably quite high ESR and 1uF MLCC. Worked. Each configuration produces very different sound. According to the appliction note the 10uF should be sufficent but even 22uF wasn't.
Are the intern filter caps for the VcomL and VcomR DAC input? There's also the Vcc1 input in the datasheet.
Still I try to figure out proper digital decoupling. I leaved only 1uF MLCC, the card wasn't recognized and relays didn't switch. Then I added 22uF tantal. Still the same. So I placed 470uF electrolitic with probably quite high ESR and 1uF MLCC. Worked. Each configuration produces very different sound. According to the appliction note the 10uF should be sufficent but even 22uF wasn't.
Are the intern filter caps for the VcomL and VcomR DAC input? There's also the Vcc1 input in the datasheet.
I leaved the digital at 470uF Nichicon HV + 100nF MLCC as the compromise. Now I'll try to figure out from which the hot comes. It seems it's the high frequency. The further from the source, the low ESR/ESL caps only gets hot like oscon yes but elna RJH not. Closer the source and all caps starts to get hot. The original track seems like this: 330uF oscon, 330uF oscon, 7805, 47uF (or 100uF) FG, 47uF (or 100uF) FG. It's most hot at the caps closest to the 7805. If both 5V filtering caps gets the 5V (I was adding 1000uF/6.3V even before), then there's probably no filtering at the 7805 input which is sugested for longer tracks. Let's say it is short ;-)
So I added 100nF MLCC to the hotter 330uF cap. There are also 1000uF RJH caps but they are cold, they are slow also ;-) Sound slightly become natural but just slighty. This reminded me my first whole card decoupling changes:
- decoupling +/-12V made the sound clearer and removed harshness
- decoupling +5V made the sound more natural and very pleasant to hear
- decoupling +3.3V made the reverberation more present.
So I added 100nF MLCC to the hotter 330uF cap. There are also 1000uF RJH caps but they are cold, they are slow also ;-) Sound slightly become natural but just slighty. This reminded me my first whole card decoupling changes:
- decoupling +/-12V made the sound clearer and removed harshness
- decoupling +5V made the sound more natural and very pleasant to hear
- decoupling +3.3V made the reverberation more present.
There's a possibility that the 2 input filtering oscon caps are for 5V and 12V. That would make the sense. Also It's possible that the 7805 is oscillating. Some people made it stable by adding 100nF and the input/gnd and 100nF at the ouput/gnd. If only I could measure voltages easily on the working board or see what happens at the oscilloscope it would be easier.
I tried to figure out connecting between filtering caps but I failed. However I removed 12V filtering 2x1000uF caps and didn't notice difference now except slightly quieter bass. I added another 100nF MLCC. The soundstage got wider. Distant sounds are clearer, with beter granularity/resulution: it's more pleasant to listen to them. Close sounds are however too soft with low resolution. The digital stage is at fault. Still this distant sounds resolutions is sometimes amazing: lot of high frequencies may play and still snares has their own tone and are clear.
I was listening to the music a lot now. So much difference. In the end I added 4 100nF MLCC caps near the 12V -> 5V regulation. Then I went to the 7812 caps. I added also 100nF for each.of them. The whole 1k-10kHz got great speed up. The 1k-3kHz was slightyly muddy before. Now it's clear. The dropdown is that the 6k-8kHz got louder. The sound has now lightning speed and some sounds are like the natural one. There is no mud at all. What I miss the most is making the sound much more in front of listener. It's very hard for me to achieve.