[bichi]

Quote:

Originally Posted by holland /img/forum/go_quote.gif Bichi, How did you get your noise floor on the XtremeMusic down to 110+ dB?

- if you are referring to post #1410, RMAA Summary, it's for a SB0550, X-FI Elite version. (Germanium's repaired card)
- best I got with a SB0460, X-FI Fatality, (same core board as Music) is RMAA Noise Level: -101.8db @ 48-24.

Main difference between SB0460 and SB0550 are the DAC's and ADC's (from datasheets):

SB0460:
- Cirrus Logic CS4382 DAC (SNR: 114db)
- Wolfson WM8775 ADC (SNR: 102db)

SB0550:
- Cirrus Logic CS4398 DAC (SNR: 120db)
- Asahi Kasei AK5394 ADC (SNR: 123db A-weighted)

 

[ter1]

bichi, I measured the voltage of line-out, for left channel: 0.25mvdc, but for right channel to ground: 2.9mvdc. Is that okay?? the RMAA shows +0.00, +0.02 and I can't hear something abnormal using the HD650 earphone.

 

[ter1]

In Vista, in "Speakers Properties" --> "Advanced" --> "Default Format", I can select the sample rate and bit depth for "Shared Mode",

my question is: for ASIO, these settings will not affect the sound, right? How about kernelstream as in foobar2k?

I looked through MS KB, the ASIO and OpenAL will go directly to soundcard driver, not go to kmxer like DirectSound does. So I don't think the "shared mode" will affect ASIO. Am I right? But how about kernelstream? Can someone says a little about "shared mode" and "Exclusive mode"?

Thanks!

 

[Maniac]

Quote:

Originally Posted by ter1 /img/forum/go_quote.gif bichi, I measured the voltage of line-out, for left channel: 0.25mvdc, but for right channel to ground: 2.9mvdc. Is that okay?? the RMAA shows +0.00, +0.02 and I can't hear something abnormal using the HD650 earphone.

The reason you can't hear anything abnormal with your HD650 is a very long one... but I can shorten it down to a few words...


that is because everything is normal and working.

 

[bichi]

Quote:

Originally Posted by ter1 /img/forum/go_quote.gif bichi, I measured the voltage of line-out, for left channel: 0.25mvdc, but for right channel to ground: 2.9mvdc. Is that okay?? the RMAA shows +0.00, +0.02 and I can't hear something abnormal using the HD650 earphone.

I don't run the SB0550 with Line-Out coupling capacitors shorted. Would be better to ask Germanium for current data. Here are some tips:

a) When you say "ground," are you using PC "chassis" ground reference or cable "sleeve" as ground reference? At very low voltage measurement levels, depending on sensitivity/quality of your DVM, ground-loop noise can be misleading.

b) Look for small right/left channel deviations in RMAA graphs.

c) Root cause could be deviations of internal DC BIAS network of CS4398, opamp and passive component leakage, 2-pole filter design, etc....

d) Guessing 2.65mvdc delta shouldn't impact sound quality with low-impedance headphones (30~320ohm), but might when driving high-impedance (10k~50kohm) input to high-definition mixer/pre-amps/amps. Potential problem if driving high-gain, direct-coupled amps.

 

[bichi]

Had a bit of time and did a quick trace of SB0550 (Elite Pro) Line-Out section, specifically the 2-pole filter, after the Cirrus Logic CS4398 DAC. Did not trace further than "mute" transistors, relays or possible 3rd pole.
- wanted to get filter circuit to model with Texas Instruments new TINA v7.0 SPICE simulator release.
- interesting to see effects of changing "coupling-cap" values, shorting coupling caps, etc.
(virtual tools, ie., DVM, O-scope, signal-analyzer, etc)
- TI's "FilterPro" also gives interesting modeling of specific active filter types, compared to Creative's design vs. Cirrus Logic recommended design: (see AN048 app note below)
- any corrections/additions welcome.

Close-up Picture of SB0550 Line-OUT section:
- green "Q2" via --> "R63" via
- green "Q10" via --> "R41" via
http://www.esnips.com/doc/667641c9-6.../X-FI-MOD-092b

TI TINA circuit of SB0550 Line-OUT section:
- stock TI OPA134 opamp selected (simulator is specific to TI parts)
- note: DC BIAS section is internal to CS4398 DAC
http://www.esnips.com/doc/1fc4ecde-7.../X-FI-MOD-093b

TI TINA AC and Noise simulator results:
http://www.esnips.com/doc/22981264-d.../X-FI-MOD-094b

AudioTester, German company offers an alternative to RMAA, with advanced tools:

Whitenoise response of SB0550, loop-back mode, 96/16:
http://www.esnips.com/doc/3987f290-5.../X-FI-MOD-095b

Dual tone 60hz + 1khz:
http://www.esnips.com/doc/46215e8b-3.../X-FI-MOD-096b


REF:
Texas Instruments TINA v7.0:
http://focus.ti.com/docs/toolsw/fold...t/tina-ti.html
Texas Instruments Active Filter Design:
http://focus.ti.com/docs/toolsw/fold...filterpro.html
Cirrus Logic: Design, 2-Pole Filter, Differential Input:
http://www.cirrus.com/en/pubs/appNote/AN048Rev2.pdf
AudioTester: (shareware)
http://www.audiotester.de/

 

[Maniac]

Nice work bichi,

regarding the DC offset, 2.**mV is really not going to affect anything at all, even the highest powered systems. If you do measure those commercially available audio systems, a lot of them will have higher offsets than that.

Since you have already setup your card to take anything in a socket, I'd really recommend you to try using a copper wire to replace the coupling caps. The DC will be gone by the time it had passed through the OP-Amp.



As a side note, I've found my pre-amp (a very old Proton AP400 Pro for some reason was never opened until recently), actually have a small DC bias at its imput, about 0.4mV. I'm guessing it is from the length of the trace onboard. Since the meter will read something even higher when it is not connected to anything. Before the input cap, it have about 10~20mV of DC, and I'm already thinking about getting rid of the coupling caps.

 

[bichi]

Quote:

Originally Posted by Maniac /img/forum/go_quote.gif Nice work bichi, "...Since you have already setup your card to take anything in a socket, I'd really recommend you to try using a copper wire to replace the coupling caps..."

Been there, done that, with both SB0460 and SB0550 models (see post #1413 link below, for most recent "short" test)
- objective is to compare "shorted" vs. "cap-coupled," started months ago, where Germanium "shorts" and I chose "cap-coupled" route.
- if you read past posts, you can see Germanium mention increased IM and Noise values when shorting.
- Creative's 2-pole differential filter design involves Cirrus Logic's DAC inherent DC BIAS and has an effect on performance.
- if you install TI's TINA, load the SB0550 circuit and model it, you can see the effect, albeit, subtle.
- also, there is 1.55ma current across each Line-OUT coupling cap, which the CS4398 DAC has to sink, if shorted. (also mentioned in past posts)

Quote:

Originally Posted by Maniac /img/forum/go_quote.gif "...The DC will be gone by the time it had passed through the OP-Amp..."

Not true, at least on a SB0550, from post #1404:
*BIAS: 0.023uvdc, measured across sleeve-ring and sleeve-tip, 47uf coupled
*BIAS: Coupling-caps shorted (100ohm): +3.1 mvdc Right, -2.2mvdc Left, measured across sleeve-ring and sleeve-tip (AD8599)
*BIAS: 2.5vdc @ 1.55ma, measured across coupling capacitor, post initialization (supplied after relays set, un-mute)

Bottom line, "shorting" is a personal decision, relative to playback equipment, sound quality judgement, safety, etc.
- all I am doing is making the observations and reporting technical results, such that persons can make an informed decision.

Stay tuned, comparison/tests of "verboen" tantalum and MLC capacitors in audio-coupling....

REF:
Post #1404:
http://www.head-fi.org/forums/showpo...postcount=1404

Post #1413:
http://www.head-fi.org/forums/showpo...postcount=1413

 

[Maniac]

Personally I would not worry much about single digit level DC bias in the output, what I would really be concerned is how would there be a 1.55ma current through the capacitor. Especially when DC bias is an extremely low 2.5VDC for the caps to block. Plus I'm also curious about what equipment is used to measure 0.023uV, that's like 23 nano volts.

If you got the time, give your home stereo equipment a measure, I'll be willing to bet that most of them would measure as much if not more than what you got out of a direct coupled X-Fi.

 

[ter1]

Hi bichi, I want also to mod and improve the +/-12v and +5v power parts with better film caps (like RIFA or WIMA) instead of the stock ones, could you please indicate which ones are the stock ones for power filtering that can be modded/replaced? and the values of them?

The 2nd concern is for the analog part beyond the DAC that is the part including the circuit surround the opamps. Are the stock ones (caps and resistors and other components) good enough to make good sound? Are there any possibilities to mod them? I think the analog circuit there is most vulnerable to interference and the quality of the components are more contribute to good sound than power supply circuit on board.

Could you please give some advice on the above 2 questions (better with pictures )? There are lots of such questions in our local forums...

 

[stormcloud]

This thread has been most informative, went through all 144 pages and I can tell you sometimes it makes my head spin. That argument on shorting the 22uF coupling caps vs bypassing them especially.

I've got a question for bichu/germanium or anyone else that can shed some light into it :

Based on general opinion, electrolytic caps are bad for coupling in the direct signal path. Assuming there is a DC offset and one rules in favour against shorting/forming a wire bridge, would a straight forward replacement of those 22uF electrolytic caps with a polyester or polypropylene (values between 0.47uF - 1.5uF) work?

 

[gameguytw]

I Fix my Auzen X-fi Like this:

Before Fix:
http://media.twango.com/m1/original/...33d5d3f0d5.jpg

After Fix
http://media.twango.com/m1/original/...da4a8f66f9.jpg

Change some CAP to BG N CAP on sound card,and get out oscon SVP CAP。

 

[LawnGnome]

Quote:

Originally Posted by stormcloud /img/forum/go_quote.gif This thread has been most informative, went through all 144 pages and I can tell you sometimes it makes my head spin. That argument on shorting the 22uF coupling caps vs bypassing them especially. I've got a question for bichu/germanium or anyone else that can shed some light into it : Based on general opinion, electrolytic caps are bad for coupling in the direct signal path. Assuming there is a DC offset and one rules in favour against shorting/forming a wire bridge, would a straight forward replacement of those 22uF electrolytic caps with a polyester or polypropylene (values between 0.47uF - 1.5uF) work?

Your low frequency response will lose a lot this way.

 

[stormcloud]

Quote:

Originally Posted by LawnGnome /img/forum/go_quote.gif Your low frequency response will lose a lot this way.

Thanks for that confirmation. It's been a long time from the EE I took at university, I'm just slowly getting back into the scene. Re-learning all this back is fairly exciting. It's perverse, I used to hate analog/power electronics

So it looks like I'm pretty much stuck to 22uF (or higher till 220uF) value for a straight forward replacement. Loosely basing the freq response on the simple high pass filter below :

f = 1 / 2 * 3.1415 * R * C

f = low freq cut-off (at 3dB)
R = sum of impedance
C = coupling capacitance

- am I correct to assume that bypassing generally decreases impedance, increases the low feq cut-off, but gives you a (freq dependent) better mid-range and high-end?

- any good suggestions for decent polyester/polypropylene caps 22uF - 220uF?

- if using electrolytic cap, it's obvious that super low ESR caps like the OSCON offers better benefit than the standard cap

--------------------

updated :

Went back and trawled to post #1312 from bichi. Looks like 33uF is a good value.
Looks like most (if not all) poly caps with the above sort of value are either uneconomical or simply too big a monster to mount on the X-Fi or both. I guess Blackgate N/NX bipolar is the next best bet ... oh well ..... gosh ... did I just answer my own questions? lol

 

[LawnGnome]

Quote:

Originally Posted by stormcloud /img/forum/go_quote.gif Thanks for that confirmation. It's been a long time from the EE I took at university, I'm just slowly getting back into the scene. Re-learning all this back is fairly exciting. It's perverse, I used to hate analog/power electronics So it looks like I'm pretty much stuck to 22uF (or higher till 220uF) value for a straight forward replacement. Loosely basing the freq response on the simple high pass filter below : f = 1 / 2 * 3.1415 * R * C f = low freq cut-off (at 3dB) R = sum of impedance C = coupling capacitance - am I correct to assume that bypassing generally decreases impedance, increases the low feq cut-off, but gives you a (freq dependent) better mid-range and high-end? - any good suggestions for decent polyester/polypropylene caps 22uF - 220uF? - if using electrolytic cap, it's obvious that super low ESR caps like the OSCON offers better benefit than the standard cap -------------------- updated : Went back and trawled to post #1312 from bichi. Looks like 33uF is a good value. Looks like most (if not all) poly caps with the above sort of value are either uneconomical or simply too big a monster to mount on the X-Fi or both. I guess Blackgate N/NX bipolar is the next best bet ... oh well ..... gosh ... did I just answer my own questions? lol

You can get Solen poly caps in that range for surprisingly not much more than the BG NX's. If you have space above your card, you can fit them on the back of the pcb.