Slow playback after replacing crystal on soundcard

[terranigma]

I tweaked with my cmi8788 based soundcard by replacing traditional crystal oscillator with Rakon brand oscillator chip at same 24.576mhz.

I thought that if i could supply oscillator chip via clean power source then it would be an improvement.

I removed two little capacitors and crystal and applied oscillator chip like schematic below.

It worked but playback is too slow. Rebecca Pidgeon sings like Hulk Hogan.

I tried 10 nf cap, 100r resistor at output and 3.3v supply but same result.

I used oscillator chip from an old soundcard. Maybe it's a faulty or non accurate chip.. have you any idea?

 

[kevinkr]

Have you measured the oscillator to make sure it is actually producing a 24.576MHz output? You might also try a new oscillator module from one of several sources like Tent Labs, etc.

 

[Eric_R]

Why don't you remove the 1M resistor also, it leaks signal from XO to XI.

 

[wa2ise]

Why don't you remove the 1M resistor also, it leaks signal from XO to XI.

I agree, that resistor may be what's goofing it up.

 

[weissi]

I would also put an series resistor of ~100Ohm between Oscillator Output and CMI8788 Input. Otherwise you might get some ugly overshoot.
But I'have modded my CMI8788 in a similar way, so it should work.

 

[terranigma]

Have you measured the oscillator to make sure it is actually producing a 24.576MHz output? You might also try a new oscillator module from one of several sources like Tent Labs, etc.

I have no any facilities for measuring. I'm trying to find another one at the moment. If i could find Crystek or Petermann Technik brand low jitter oscillators, it would be good enough instead of expensive ones.

Why don't you remove the 1M resistor also, it leaks signal from XO to XI.

First, i tried without it, when this problem occurs, then i resolder it after checking for similar modifications. Then, i will try without it again...

I would also put an series resistor of ~100Ohm between Oscillator Output and CMI8788 Input. Otherwise you might get some ugly overshoot.
But I'have modded my CMI8788 in a similar way, so it should work.

My soundcard is Club3D Theatron DTS (Bluegears B-enspirer). I tried with 100ohm resistor as i mentioned in first message. Which parts did you remove? Did you remove 1M resistor? Can you give a little explanation please?

 

[weissi]

I also have the Theatron. Basically I have removed the 1Meg Resistor plus the to ceramic caps and connected the output of the oscillator via a 100Ohm resistor to XI of CMI8788. But I used an 3.3V oscillator. Normally this shouldn't effect the CMI8788 but I don't know if its Ports are 5V tolerant....

 

[terranigma]

But I used an 3.3V oscillator. Normally this shouldn't effect the CMI8788 but I don't know if its Ports are 5V tolerant....

Maybe this is the trick. That seems it wouldn't be clear unless getting a new 3.3 oscillator.

Thank you all.

 

[weissi]

I have used this one, although not low jitter, but I guess it will work much better than the crystal oscillator that's onboard:

http://at.rs-online.com/web/search/searchBrowseAction.html?method=getProduct&R=4785216

 

[terranigma]

I ordered Crystek C3391 's from Mouser. I have to find a low noise regulator solution. ON Semi NCP700 looks up for this job with 15uV rms noise level.

 

[weissi]

This is how my implementation of the new clock looks like:

 

[terranigma]

weissi, you have done quite smooth job. which kind of a regulator did you use for oscillator?

I'm looking for a ldo regulator with max 20uV rms noise level.

These ones look like up for this job quite enough.. have anyone other suggestions please?

Linear Tech LT1761 (20uV)
http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1778,C1764,P1750

National LP5900 (6.5uV) !!
http://www.national.com/pf/LP/LP5900.html

On Semi NP700 (15uV)
http://www.onsemi.com/pub_link/Collateral/NCP700-D.PDF

 

[weissi]

regulator is a plain LP2981 (!), don't have the exact part number in my head. But it follows a LM317 and an LM7805, so it' should be pretty low-noise. What you see in the background is the regulator part. Every AK4396 got it's own analog +5V. see picture...

An externally hosted image should be here but it was not working when we last tested it.

 

[weissi]

Ah, th regulator part was done under a microscope I had in my last job.... made me fell in love with smd parts 😉
if you take a closer look you will see that it's pretty easy to acces every DAC's analog supply stage. Just lift those small ferrite beads: FB19-FB22

 

[SwedishWings]

Maybe this is the trick. That seems it wouldn't be clear unless getting a new 3.3 oscillator.

Thank you all.

The spec says that absolute maximum rating is 3.6V on all I/O ports. I would not run it for a second like that, as the voltage swing from your 5V oscillator moslt likely exceeds the absolute maximum rating. You might fry it.

/Mike

 

[terranigma]

Ah, th regulator part was done under a microscope I had in my last job.... made me fell in love with smd parts 😉
if you take a closer look you will see that it's pretty easy to acces every DAC's analog supply stage. Just lift those small ferrite beads: FB19-FB22

If you don't use ultra low noise type regulator for oscillator, you can't get any audible improvement related by jitter. I strongly recommend you to find one of the regulators that i mentioned in my previous message.

Also, why didn't you use 5v regulator's (LD1117 mine) output pins for supplying DACs ?

The spec says that absolute maximum rating is 3.6V on all I/O ports. I would not run it for a second like that, as the voltage swing from your 5V oscillator moslt likely exceeds the absolute maximum rating. You might fry it.

/Mike

This is the last thing i want to hear in this topic I resoldered smd parts and checked it again... no more problems at the moment

 

[weissi]

f you don't use ultra low noise type regulator for oscillator, you can't get any audible improvement related by jitter. I strongly recommend you to find one of the regulators that i mentioned in my previous message.

I cannot agree on this. As the original oscillator circuit is inside the CMI8788 chip, there are quite a lot of noise sources located extremely near to the main clock. Pulling the main clock generator outisde this digital mess must result in an improvement, not as big as with low-noise type regulators, but to some degree significant. This was a ~0 budget attempt, since I only had to buy the oscillator (total 5Euros) The rest i've "stolen" from company (except the FR4 board).
Another thing is the operatin frequency.
@25MHz noise rejection is mostly a task for the decoupling capacitors connected after the regulator. Since I used 10n X7R paralled with 4u7 tantalum this is quite a good compromise. Of course this can be done much better, but I've done it as a side project. My EMU1212m is modded much more sophisticated. I've got my Theatron primary as a fun-to-mod soundcard.

I've an EMU0404 too, and I wanted to compare the AK4395 in my EMU against the AK4396 in the Theatron....
well, the EMU provides a much more pleasant listening. I have to say that I didn't changed anything in the analog section of the Theatron except the capacitors, while the EMU has AD8022 opamps in the Anti-Alias filters....

LD1117? remember that this is a low-drop type, output noise is 100uV compared to 40uV with the 78L05 (typical values).
PSRR is a little bit better with 1117 types, but since it's connected to a LM317 before @8V, this is much superiour to a single 1117 runnning from 12V alone.

When you want to get the most out of the Theatron, why don't you take the I2S signals via LVDS connection out of the PC to external PCM1794 D/A's with discrete I/V conversion. That would be my approach to get serious quality audio... expensive though...

Anyway, maybe this post sounds a little bit harsh, it's just my approach to these things, nothing more....

cheers, Markus

 

[terranigma]

Theatron's most sophisticated part is AK4396 DACs. If you are using analog part of souncard, maybe that's the why of you couldn't satisfy as well as EMUs.

Please take a look on this topic:

http://www.diyaudio.com/forums/showthread.php?postid=1140979#post1140979

I cancelled all of the analog stage which consists of crappy parts by getting the signal directly from DAC pins. Since AK4396 has benefits like "direct-out" very high quality balanced outputs, then i decided to go with these pins. Then i realized, why these DACs called as "Miracle DAC" without another effort.

Also, I'm not about to keep using 1117's on the board, I'm about to supply all of DACs directly from 1117's single output point by isolating its legs completely.

 

[weissi]

That's why I'm wondering... AK4395 had 110dB stopband rejection, while AK4396 only has 75dB.... This difference put's more stress to the post-filter, so you have to have high speed opamps in there.
All modern D/A's have more than 100dB... I think AK4396 isn't an upgrade to the '4395, just cheaper to produce....
Since no one knows how the output stage of those AK439x D/A'S lokks like I'll not say more on this.
EMU engineers put simple voltage followers opamps right after the AK4395 outputs.
I'm sure they wouldn't throw 2 opamps for nothing in there. Remember that everything inside these D/A's is CMOS stuff, including the analog part. CMOS has no good reputation for beeing the ideal audio devices.

Those AKM D/A's have only 2 separate supply voltages, whereas all other manufacturers use minimum 3 different power supply input.
Take a look at the AK4397 datasheet and count the different supply in's. Wonder why they call this D/A "high performance premium"?
TI (BurrBrown) is also using 4 separate supply pins, like CirrrusLogic.
AnalogDevices managed to handle the task also with only 2 separate supply pins, but ADI parts are one of the most expensive and sophisticated parts on this earth (they'll take some money for this)

But I think it's mostly a matter of implemetation.
I have build a combo consiting of PCM2707(USB to I2S)==>PCM1730==>AD8022 I/V and this was by a great margin the best sounding D/A unit I've heard. (16bit MP3@320kbit)
Unfortunately I've covered the board whith epoxy.... one channel broke and I'm not able to repair it

My opinion is that one will never get serious high-resolution audio out of the Theatron... It simply wasn't designed to be in the pro league. Get yourself an EMU card and you'll hear/know why.

cheers, Markus

 

[terranigma]

My opinion is that one will never get serious high-resolution audio out of the Theatron... It simply wasn't designed to be in the pro league. Get yourself an EMU card and you'll hear/know why.

Due to there is no classification between AK4396's (all of them same) and I'm not using external analog stages fullfilled with various components like opamps, I believe it's not necessary to pay too much for those EMUs.

My approach in diy aims keep as close as possible source and amplification. I believe in adding something between not making something best. Maybe adding something makes it "hi-end" which regarding to the cost completely.