Xonar ST/STX mods...

I didn't connect the clock directly to the pin. Playing with these legs with my tools is very hard. I connected it to the second smd element, the one near the old quartz positive pad. When changing sampling frequency from 44,1kHz to higher the sound gets slower, but playing 48kHz or 192kHz has the same speed. Bought the Vanguard from the only HK seller on ebay. I've got no access to the proper measurement devices. 24,574MHz may require oscilloscope with higher than 10MHz or 25MHz range. I already helped myself with the datasheet of the CMI8788, CS2000 and the eeprom with the xonar pin placement. The source linux drivers were also helpful.

I'll try to find better ground. The STX card is easier if they don't have the CS2000. I heard the newer STX may have the CS2000. On the STX the clock gives signal to the CMI8788. Then the CMI8788 multiplies it and feeds to the PCM1792.

The ST is more advances. The CMI8788 is not doing the multiplication. The CS2000 does the multiplication. It also lowers the jitter. That's why the clock change on the ST is not so much improvement like on the STX. When I speed the music for normal playback it's not much different in the tonal balance. The sound gets little smoother, there's no sharpening, it's also more precise. There are more sounds in the center which is more detailed than before reclocking. I expected much more, but well... I've got to fix this clock. On other forum there was a thread where after reclocking the sound was slowed down by 30%. Also I found some images on the net with wrong clock connection like the one using the eeprom on your image. They suggested to use the ground near the 7805 from which the power is taken.

I'm happy the card can still run. I fought it was dead like broken CMI8788 legs.
 
Fixed timings! It's a small xonar driver bug or internal CMI8788 bug (sometimes called the feature!). The fault was disconnecting eeprom during the CMI8788 operation. The card may report as xonar, but when disconnected, is switches to the default CMI8788 routines and clock selector. This may require faster clock. After restart, the card was reported as c-media. Fixed eeprom connection and now it's working. I don't hear any large improvement in the sound. Bit more sharp with slightly better imaging, more air between instruments. Most of this was already fixed with better cables and faster power management.
 
The noise comparison. After recklocking seems more noisy with more and larger spikes. And it's just noise measure. Maybe the master clock is being used when recording and the noise and spikes come from better quality clock or... worse quality, spikes are doubled. The clock may interfere with the 5V too.
 
I removed last 7805 filtering and soldered the clock output pin between the R and C of the original circuit. It looked strange. The card XTALO and XTAL1 were probably shorted by a large R and then there was a serial C on each line. Then the old quartz. I'm not very sure about this. The suggested connections always look different like in the attached images. Now I tried to solder skipping the first smd element on the card near the old quartz pad.

So, I'm one element closer to the XTAL input pin and removed more filtering. The higher tones are not gray, it's more pleasant to hear, the mids are however gray ;-)
 
By dealing with the proper power supply and short connection, I managed to eliminate most cons of the new oscillator. Still, it's not easy. I wonder why the MLCC decoupling near the regulator make the oscillator output very unpleasant... clock ;). I might add the regulator for the clock. It would be better to supply it with 8V than 12V. I expected big gap after replacing the clock, but at first I got something worse. At least it can be improved.
 
Member
Joined 2009
Paid Member
You may remove the 18pF caps and 100k resistor, and then solder the oscillator output to the right pin. The oscillator do not behave as an quartz resonator... and that because it may not have that caps and resistor network on the clock pins of the processor. Clock out pin it may be unconnected, when an oscillator is connected to the clock in pin.
I personally lift it up the clock in pin of the processor, and then solder it (very carefully - the procedure is described in the thread beginning) directly to the oscillator output. That`s all and worked from the very beginning!

Changing the original quartz with the new oscillator it not have to induce any problem. It just work from the very beginning, and the improvement is obviously. This mod is quite old now, and are many informations about how have to be done. If you should follow that informations, you should have a very working audio card...
I suggest to read from the beginning this thread. There are some links to this mod made by others, pictures, and so. Or you can google about.
 
Last edited:
not sure whether I understand completely (Im new to all this), atm I only want to change the clock of my STX and exchange it for that Vanguard 24.576MHz TCXO 0.3ppm clock.

already saw that picture http://www.diyaudio.com/forums/atta...9d1361467492-xonar-st-stx-mods-photo260bb.jpg which showed where to solder the clock to the chip.
but wasnt there also something else, the clock does need power connection as well, doesnt it? so which picture and description of what to connect to where was there again? does this picture http://www.diyaudio.com/forums/atta...32d1318534933-xonar-st-stx-mods-photo261c.jpg show how it has to be done? could you please elaborate a bit further on this what I exactly I have to do here? and is that thick white/glossy stuff actually a cable or cable protection/isolation? where to buy such stuff?
 
In comparison the sound is much sharper. The last setting on the card with the old quartz was awful with the new clock. However I notice the different filtering and decoupling now sounds totally different like:
- the setting was okay, now it's sharp and upleasant
- the setting was too soft with lack of sound dynamics, now it's ok and clear
- the sound was too sharp, now it's... I didn't test.

I went back to electrolytics and notice that after break-in they sound different than before. Seems they are better now. When using old quartz, the fresh RFS were ok, after long break-in they were almost useless and didn't help, too soft, not precise. KZ were so-so, after break-in hard to say.

Now the RFS after long break-in are pretty good. Now I test KZ and Y filtering, not only the X filtering. First they are not good, maybe due to small resonance between X (2x10uF) and Y (47uF) line. Then after break-in (the break in time of KZ is shorter than RFS) the reverberation seemed to improve.

From this I may say the old quartz was soft. The softness may come from the jitter which causes the sound spectrum to be smoothed. That's why some sharpening may had a good result and transparent setup was a bit too soft.
 
In some 5V settings the clock seems to be very bad. The music starts playing after after a few seconds seems to be out of buffer. Pop and clicks comes out, after a few seconds, the sound is not available. On all drivers. In worst case when trying to turn off device in Windows and start again, the yellow sign pops up. Slowly I start to figure out what can be the fault. First, the solid separation between DAC 5V and clock power has to be made. Putting large caps doesn't help. I added second MKP cap to the clock and it made everything worse.

I've got the idea of builing regulator for the vanguard on a small board. Get the voltage from the 7812 on xonar card, regulate and get 5V. Filter on input and output. For the oscillator, in the different thread, the CRC filter on input was proposed with the R large as 50ohm.

I may also try the audio-gd clocks with integrated PSU. However due to increasing cost I may not buy it.
 
I gave the card more tries. I was inspired by the provided impedance and ESR curve of the KZ and the RFS caps. Now I'm removing all distortion present in the sound. It's not measurable. When playing 1kHz and measuring THD if there are harmonic distortion present, they are in the 6-16kHz range, all even harmonics: 6k, 8k, 10kHz and so on. Even when they are at -100dB or -110dB, they change sound.

After all, when minimizing distiortion, the background sound becomes clearer, the main sound goes away to the back. Sometimes it becomes soft, sometimes it's sharp. I start to utilize Nichicon HV caps, some call them miniaturized HD. I don't know their ESR/Z/phase/impedance/capacitance curves. They still produce soft sound, but it seems the distortion level is much lower.