or Connect
Head-Fi.org › Forums › Equipment Forums › Computer Audio › Hotrodding the X-Fi: A Layman's Guide (No 56k)
New Posts  All Forums:Forum Nav:

Hotrodding the X-Fi: A Layman's Guide (No 56k) - Page 75

post #1111 of 2194
X-FI SB0730 (compact) ExtremeGamer:
- replaced G-Luxon 220uf @ 25vdc with Panasonic FK, 1500uf @ 16vdc (C164)
- will not be able to change NJR4556 (SSOP) with LME4920; 49860 or AD8599 (SOIC)
- SSOP vs SOIC package size difference: http://www.esnips.com/doc/9da5669f-0...I-MOD-019b-XFG)
- M33078 is not Line-In, but MIC-IN.

- short "subjective" listening: bass definition less "muddy" and improved highs, but still "tinny.."

RMAA v6.0.5 test results:
- cap change only:
44/16: http://www.esnips.com/doc/7e3d3920-5...I-MOD-020b-XFG
96/24: http://www.esnips.com/doc/9d981375-3...I-MOD-021b-XFG

Quote:
Originally Posted by germanium View Post
Audio creation is what I meant, sorry. I was just trying the different modes but audio creation did in fact work. Have a good dinner ok.
- ahhh, right - had me worried there for a sec.....
post #1112 of 2194
Quote:
Originally Posted by bichi View Post
X-FI SB0703 (compact) ExtremeGamer:
- replaced G-Luxon 220uf @ 25vdc with Panasonic FK, 1500uf @ 16vdc (C164)
- will not be able to change NJR4556 (SSOP) with LME4920; 49860 or AD8599 (SOIC)
- SSOP vs SOIC package size difference: http://www.esnips.com/doc/9da5669f-0...I-MOD-019b-XFG)
- might replace Line-In M33078 with LME49860, just for comparison sake, later date...

- short "subjective" listening: bass definition less "muddy" and improved highs, but still "tinny.."

RMAA v6.0.5 test results:
- cap change only:
44/16: http://www.esnips.com/doc/7e3d3920-5...I-MOD-020b-XFG
96/24: http://www.esnips.com/doc/9d981375-3...I-MOD-021b-XFG



- ahhh, right - had me worried there for a sec.....
Sound with coupling caps shorted & power supply caps bypassed with metalyzed films is vibrant & lively. Cymbles have natural shimmer & body, not weak, harsh & tinny sounding at all like stock. High frequency sounds are brought out better without sounding harsh or out of place. Some sounds near the top of the hearing range may be heard for the first time as they now have sufficient energy. This in spite of no appearant change in actual frequency response according to RMAA. Bass is rock solid but with ambiance not heard before. This can still be improved apon with bigger bypasses on the +/- 12 volt caps. Very open sounding mids with lots of ambiance extraction from recording which seemed lacking in stock card. This gives the modded card a huge advantage in the creation of a huge believable soundstage with lots of depth both projected forward & back up to 50 feet behind the speakers.

You will not likely see huge differences or even small ones with your card in distortion or noise specs on the any card except the Elite Pro as the limiting factor on the lower cards is the ADC. Specs on everything except on the noise will not change much even on the Elite Pro except the spectrum analysis of the vaious tests for noise & distortion looks much better on the modded Elite Pro. Noise specs improve by about 3db or more on modded Elite Pro. The lower cards ADC seems to be limited to 16 to 17 bits resolution even though they do put out 24 bit word length data. The Elite Pro on the other hand has ADCs that truly have near 24 bit resolution & specs that at least suggest 20 bit resolution. Only in the spectum analysis can you see the true 24 bit like noise level on the Elite Pro especially through the critical midband where noise hovers around the -150db mark when modded, before it hovered in the -145db range but with a fair amout of spikes going up into the -130db range as can be seen in creative tests. The modded card has absolutely no spikes at all in the critical midrange or top end until you get past 20 KHz & then only one or 2 spikes that goes up to about -138db. The rest is clean all the way up to 44KHz.

This goes to show that specs can be very very misleading considering the very small changes in specs but the total revamping of the audible results.
post #1113 of 2194
does anyone know if I hardwire a DIP8 socket on to where the SMT opamp should be, will that affect the sound quality (due to the length of the wire)?
post #1114 of 2194
Quote:
Originally Posted by hyyam85 View Post
does anyone know if I hardwire a DIP8 socket on to where the SMT opamp should be, will that affect the sound quality (due to the length of the wire)?
Not recommended as it would be too easy to damage the solder pads when changing the opamps out & sound quality issues not limited to lenth of wire as soldering direct to the board makes a much better connection.
post #1115 of 2194
Quote:
Originally Posted by bichi View Post
- damn bastid!



- so you let me be the "sacrificial" rat-boi, eh?
- still think yer a chicken-poo, LOL! (its much easier with solder-paste with SOIC sized parts, than with wire solder...)
No one forced you to change the opamps & I for one have never advocated it. My feeling is that there are other much more revealing mods than the opamp mods alone.
post #1116 of 2194
Quote:
Originally Posted by germanium View Post
Sound with coupling caps shorted & power supply caps bypassed with metalyzed films is vibrant & lively.
Sorry, just catching up on this thread after a couple of months, so if this question is redundant forgive me. What do you mean by bypassing the power supply caps?

When you say bypassing, do you mean running the 3.3uf in parallel to the 220uf power cap, for instance?
post #1117 of 2194
Quote:
Originally Posted by holland View Post
Sorry, just catching up on this thread after a couple of months, so if this question is redundant forgive me. What do you mean by bypassing the power supply caps?

When you say bypassing, do you mean running the 3.3uf in parallel to the 220uf power cap, for instance?
Yes that is exactly what I mean, I found though that it is better to bypass the caps that power the opamps than the one by the DSP chip.
post #1118 of 2194
That's interesting. Which cap did you use? I did a quick search and haven't turned up anything with a small enough lead spacing. Would SMD tantalum caps be good?

So, it's an increase in total capacitance to 103.3uf but I'm not sure how the different materials of the 3.3uf cap act with 100uf electrolytic.

Do you have an understanding of what's happening? If so, would you mind explaining it to me?

Thanks.
post #1119 of 2194
Quote:
Originally Posted by holland View Post
That's interesting. Which cap did you use? I did a quick search and haven't turned up anything with a small enough lead spacing. Would SMD tantalum caps be good?

So, it's an increase in total capacitance to 103.3uf but I'm not sure how the different materials of the 3.3uf cap act with 100uf electrolytic.

Do you have an understanding of what's happening? If so, would you mind explaining it to me?

Thanks.
I used polyester metalyzed film. They had long leads from the factory so I was able to cut them down to what I needed & bend the leads to get the spacing.

Metalyzed film caps have lower ESR & much lower dielectric absorbsion than electrolytics do but they are huge for the amount of capacitance you get out of them. Used in combination with the electrolytic cap can significantly lower the impedance in the power supply & provide a better return path for the signal ground. Most people don't think of the powersupply being in the signal path but it IS in a vast majority of transistor amps. All signal must return to the power supply so the lower the impedance the better & the less the dielectric absorbsion the better. There are other factor that may contribute too as metalyzed films I use are noninductive & have frequency capabilities well into the MHz. Electrolytics are generall inductively wound internally & hence are far from a perfect capacitor.
post #1120 of 2194
Going run down a different path, not going to accept the "classic" power cap, "paralleling" with smaller cap just yet...
- the math just does not work out... (also ref Rod Elliott's article, page 14, Figure 11)
- to be clear, the only reason I use RMAA is to detect any radical changes from it's "spec'd" behavior, not it's "sonic" quality.
- achieved improved lows and cleaner highs, at least for my taste, with no coupling cap changes or shunting power caps, but with faster slewing op amp and increasing power filter cap values. (C16, C47 and C177)

Took the time to read Cirrus Logic CS4382 datasheet, Eval board description and Filter App Note:
- datasheet, page 4, I can see I am still well within the DAC's performance tolerances.
- datasheet, page 38, figure 42, shows the "recommended" output "filter/buffer" and can better "guess" at Creative's design.
- datasheet, page 13, figure 6, note VD, VA, VLS, VLC DC inputs and how they "typically" treat power bypass.
- traced VD, VA on a SB0703 and appears 10uf @ 16vdc polar electrolytics are installed. (C91, C119)
- of interest to me is FLIT+ (C107, 47uf) and VQ (C108, 10uf) functions and the capacitor "network" they show, relative to ground. (will leave those alone for now)
- cannot tell how VLS is wired and related cap(s) (could be C43, C16?)

As an experiment, will replace C91 and C119 with 6.8uf @ 25vdc tantalums (parts on hand, otherwise, 4.7uf)
- will post back results in a few hours.
(modified card is sounding wonderful at this point, not sure why any more modifications are needed, except for curiosity's sake....)

REFS and Remedial reading:
Capacitor Characteristics: (Audio Applications, Rod Elliott)
http://sound.westhost.com/articles/capacitors.htm

Cirrus Logic CS4382 datasheet, Eval board description and Filter Design App Note:
http://www.cirrus.com/en/products/pro/detail/P223.html

AVX TAP Technical Summary & Application Guidelines:
http://www.avxcorp.com/docs/Catalogs/techsuml.pdf

Electronic Power Supplies: (page 141)
Irving M Gottlieb - ISBN0-8306-8540-5

Art of Electronics: (Chapter 7)
Paul Horowitz; Windfield Hill - ISBN 0-521-37095-7


ps: Gemanium
- got yer note and will PM back info
post #1121 of 2194
You will notice in the spec sheet that the power caps are already bypassed with .1uf. All I'm doing is increasing that by a huge margin so that it can benefit the whole audio spectrum. Also you will notice that the DAC is rated fo 114db signal to noise & dynamic range. The ADC is limiting the ability to read that spec & so dynamic range is way out of spec due to the ADC.
post #1122 of 2194
Current Changed List (SB0460):
- LME49860 (Line In and Out)
- C177: Panasonic FK, 1500uf @ 16vdc from Jamicon WL 220uf @ 16vdc
- C16, C46: Panasonic FC, 220uf @ 25vdc from Jamicon SS 100uf @ 16vdc
- C91, C119: 6.8uf @ 25vdc tant from Wincap 10uf @ 16vdc

RMAA v6.0.5 results vs., Creative RMAA v5.5, after Cirrus Logic DAC, VA, VC +DC bypass cap change to tantalum:

24b/96k
Frequency response (from 40 Hz to 15 kHz), dB +0.01, -0.07 (Creative: +0.02,-0.17)
Noise level, dB (A) -100.8 (Creative: -102.2)
Dynamic range, dB (A) 100.8 (Creative: 102.0) (Cirrus DAC: 111 typ)
THD, % 0.0016 (Creative: 0.0008)
THD + Noise, dB (A) -91.2 (Creative: n/a) (Cirrus DAC: -100, 0db)
IMD + Noise, % 0.0030 (Creative: 0.0025)
Stereo crosstalk, dB -97.1 (Creative: -98.6)
IMD at 10 kHz, % 0.0057 (Creative: 0.0052)
General performance: Excellent (for both)

16b/44k
Frequency response (from 40 Hz to 15 kHz), dB +0.01, -0.08 (Creative: +0.02, -0.08)
Noise level, dB (A) -93.9 (Creative: -94.1)
Dynamic range, dB (A) 93.8 (Creative: 94.0) (Cirrus DAC: 94 typ)
THD, % 0.0016 (Creative: 0.0009)
THD + Noise, dB (A) -86.8 (Creative: n/a) (Cirrus DAC: -94, 0db)
IMD + Noise, % 0.0059 (Creative: 0.0057)
Stereo crosstalk, dB -94.8 (Creative: -97.0)
IMD at 10 kHz, % 0.0080 (Creative: n/a)
General performance: Excellent (Creative: n/a)

Subjective Listen Tests:
- much tighter bass definition, less "muddy/boom."
- clearer, well defined highs with better definition, without "shattering, tinny-ness."

Mods with greatest "sonic" improvement:
- C177 cap value increase
- LME49860 change

Mods not as obvious, sonically:
- C16, C46 changes (op amp power source)
- C91, C119 tantalum changes (CS4382 DAC power input, VA and VD)

Probed around with oscope, looking for any unexpected ringing, between 20khz to 10mhz
- nothing unusual found

Appears LME49860 (or LM4562, LME49720) opens frequency response range and is technically measurable.
C177 cap value increase appears to "clean-up" lows and highs. Not too sure about DAC VA/VD (C91, C119) cap changes to tantalum yet, need to listen for a few days...

Gotta get some dinner now, before it's all gone....

PICS:
Cap Change: http://www.esnips.com/doc/4e51b1cf-3.../X-FI-MOD-022b
44k/16b: http://www.esnips.com/doc/94f9b2d7-9.../X-FI-MOD-023b
96k/24b: http://www.esnips.com/doc/fe9e0802-2.../X-FI-MOD-024b
post #1123 of 2194
Quote:
Originally Posted by bichi View Post
Current Changed List (SB0460):
- LME49860 (Line In and Out)
- C177: Panasonic FK, 1500uf @ 16vdc from Jamicon WL 220uf @ 16vdc
- C16, C46: Panasonic FC, 220uf @ 25vdc from Jamicon SS 100uf @ 16vdc
- C91, C119: 6.8uf @ 25vdc tant from Wincap 10uf @ 16vdc

RMAA v6.0.5 results vs., Creative RMAA v5.5, after Cirrus Logic DAC, VA, VC +DC bypass cap change to tantalum:

24b/96k
Frequency response (from 40 Hz to 15 kHz), dB +0.01, -0.07 (Creative: +0.02,-0.17)
Noise level, dB (A) -100.8 (Creative: -102.2)
Dynamic range, dB (A) 100.8 (Creative: 102.0)
THD, % 0.0016 (Creative: 0.0008)
THD + Noise, dB (A) -91.2 (Creative: n/a)
IMD + Noise, % 0.0030 (Creative: 0.0025)
Stereo crosstalk, dB -97.1 (Creative: -98.6)
IMD at 10 kHz, % 0.0057 (Creative: 0.0052)
General performance: Excellent (for both)

16b/44k
Frequency response (from 40 Hz to 15 kHz), dB +0.01, -0.08 (Creative: +0.02, -0.08)
Noise level, dB (A) -93.9 (Creative: -94.1)
Dynamic range, dB (A) 93.8 (Creative: 94.0)
THD, % 0.0016 (Creative: 0.0009)
THD + Noise, dB (A) -86.8 (Creative: n/a)
IMD + Noise, % 0.0059 (Creative: 0.0057)
Stereo crosstalk, dB -94.8 (Creative: -97.0)
IMD at 10 kHz, % 0.0080 (Creative: n/a)
General performance: Excellent (Creative: n/a)

Subjective Listen Tests:
- much tighter bass definition, less "muddy/boom."
- clearer, well defined highs with better definition, without "shattering, tinny-ness."

Mods with greatest "sonic" improvement:
- C177 cap value increase
- LME49860 change

Mods not as obvious, sonically:
- C16, C46 changes
- C91, C119 tant changes

Probed around with oscope, looking for any unexpected ringing, between 20khz to 10mhz
- nothing unusual found

Appears LME49860 (or LM4562, LME49720) opens frequency response range and is technically measurable.
C177 cap value increase appears to "clean-up" lows and highs. Not too sure about DAC VA/VD (C91, C119) cap changes to tantalum yet, need to listen for a few days...

Gotta get some dinner now, before it's all gone....

PICS:
Cap Change: http://www.esnips.com/doc/4e51b1cf-3.../X-FI-MOD-022b
44k/16b: http://www.esnips.com/doc/94f9b2d7-9.../X-FI-MOD-023b
96k/24b: http://www.esnips.com/doc/fe9e0802-2.../X-FI-MOD-024b
All your tests are within a margin of error & may improve very slightly with more time the computer is booted so of no particular concern. How does the spectrum analysis look though or did you compare those?
post #1124 of 2194
Quote:
Originally Posted by germanium View Post
You will notice in the spec sheet that the power caps are already bypassed with .1uf. All I'm doing is increasing that by a huge margin so that it can benefit the whole audio spectrum. Also you will notice that the DAC is rated fo 114db signal to noise & dynamic range. The ADC is limiting the ability to read that spec & so dynamic range is way out of spec due to the ADC.
Yes, well aware of what Cirrus Logic suggests and tracing SB0460, Creative implemented the same "parallel-bypass-cap" technique.
- experiment was to "improve" supply decoupling and see if there were any "sonic" changes.
- to test Elliott/Bateman's "stacked-bypass," conjecture, removal of the .1uf SMT caps is required. (will do this soon, if time permits)

CAUTION for casual capacitor bypassing: (SB0460)
- watch out for C107 47uF and C108 10uf, where shorting may cause unexpected results.
- they form capacitive coupling to ground, for pins FILT+ and VQ: (see page 24, Cirrus CS4383 datasheet)

FILT+ = Positive Voltage Reference (Output) - Positive reference voltage for the internal sampling circuits.
Requires the capacitive decoupling to analog ground, as shown in the Typical Connection Diagram.

VQ = Quiescent Voltage (Output) - Filter connection for internal quiescent voltage. VQ must be capacitively coupled to analog ground, as shown in the Typical Connection Diagram. The nominal voltage level is
specified in the Analog Characteristics and Specifications section. VQ presents an appreciable source
impedance and any current drawn from this pin will alter device performance. However, VQ can be
used to bias the analog circuitry assuming there is no AC signal component and the DC current is less
than the maximum specified in the Analog Characteristics and Specifications section.

Quote:
Originally Posted by germanium View Post
All your tests are within a margin of error & may improve very slightly with more time the computer is booted so of no particular concern. How does the spectrum analysis look though or did you compare those?
Yep, again, well aware of limits of DAC and it's influence on op amp performance, relative to RMAA results.

Spectrum analysis? You mean the graphs?
- link to 24b/96k RMAA Result HTML and GIFs (28kBytes):
http://www.esnips.com/doc/9346e7bf-f.../X-FI-MOD-025b

We having PHUN yet?
post #1125 of 2194
Your graphs look excellent but it appears that the test was for 96KHz but the card sample rate was set for 48KHz. The noise floor looks very clean. Any distortion in the dynamic range test is completely buried by noise from the less than ideal adc chip.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Computer Audio
Head-Fi.org › Forums › Equipment Forums › Computer Audio › Hotrodding the X-Fi: A Layman's Guide (No 56k)