X-Fi sound bugs? I think I have a solution.

[Deleted member 24505]

I would really like a xonar D2X.Droool

Are the DX's TI dacs burr brown ones ketxxx?

 

[imperialreign]

CPU utilization is defi moot in terms of dual and quad-core setups; but if we compare to onboard, that's a whole 'nother realm . . .

My biggest thing is the audio latency, how long it takes for a file to be pulled from the SYS, processed, and played back. The more voices you tack on, the longer it takes to continue playback, and you can start getting into clipping and buffering issues if the card is moving faster than the system . . .

actually, the initial X-Fi releases (specifically the Fatal1ty, Xtreme Music and Elite Pro) were overly prone to audio clipping because of this; the system PCI BUS couldn't keep up with the cards and how often the device was flagging it's IRQ wanting access. Second revision cards were being released where the card's "BIOS" were effectivelly slowing the APU down to cut back on the audio clipping. Even still, the X-FI APUs latency times trump everything else currently out there.

The move towards PCIEx1 should allow these APUs to go back to full boogie, which is just sick . . .

To put it into comparison - the CA201K is a full-blown audio processing unit, not a DSP or chipset like the C-Media units are; the X-Fi APU is capable of handling up to 128 hardware voices (literally up to 128 seperate audio sounds at a given time); the Xonar series can also support 128 hardware voices . . . but when we go into what the cards are capable of with software voices, the X-Fi APUs lead the pack with capability of up to 65,535 voices simultaneously . . . no C-Media chipset on the market can even come close to this capability, most topping out only within the 10,000-15,000 range (IIRC).

For the absolute best positional audio and playback possible, you need higher software voice support - big reason why CMSS-3D is still considered much better than Xonar's offering (Xear3D, IIRC); number of software voices also plays a big role in how many different times the same sound can be played at the same time, as well as software occlusion effects, and countless other aspects of the final playback.







I'm by no means saying the Xonar cards aren't worth their cost - far from it! They are very respectable cards, IMO - if you're looking for premier audio quality for a range of uses, they're more than capable. They offer great audio quality, without having to pay an arm and a leg like you would for an Auzentech or high-end HT Omega.

I understand there are a lot of people that have had many issues with Creative's cards . . . but likewise with ASUS and Auzen and HT and all the other manufacturers (especially in terms of Vista issues). Creative's issues just dwarf everyone elses due to the sheer volume of their cards on the market compared to others.

But, relatively speaking, they do have more driver issues than hardware issues - compared with everyone else.

 

[Mussels]

CPU usage: well on a quad....

EAX: well lol, EAX isnt shared to anyone other than creative, so ofc they're on top. Oh except alchemy universal lets even my old AC97 soundcard have EAX 5.0, hehe.

Crystalliser: to each their own. i dont actually like it, as it just makes my music strange (possibly because i have winnar speakers)

 

[Ketxxx]

CPU utilization is defi moot in terms of dual and quad-core setups; but if we compare to onboard, that's a whole 'nother realm . . .

My biggest thing is the audio latency, how long it takes for a file to be pulled from the SYS, processed, and played back. The more voices you tack on, the longer it takes to continue playback, and you can start getting into clipping and buffering issues if the card is moving faster than the system . . .

actually, the initial X-Fi releases (specifically the Fatal1ty, Xtreme Music and Elite Pro) were overly prone to audio clipping because of this; the system PCI BUS couldn't keep up with the cards and how often the device was flagging it's IRQ wanting access. Second revision cards were being released where the card's "BIOS" were effectivelly slowing the APU down to cut back on the audio clipping. Even still, the X-FI APUs latency times trump everything else currently out there.

The move towards PCIEx1 should allow these APUs to go back to full boogie, which is just sick . . .

To put it into comparison - the CA201K is a full-blown audio processing unit, not a DSP or chipset like the C-Media units are; the X-Fi APU is capable of handling up to 128 hardware voices (literally up to 128 seperate audio sounds at a given time); the Xonar series can also support 128 hardware voices . . . but when we go into what the cards are capable of with software voices, the X-Fi APUs lead the pack with capability of up to 65,535 voices simultaneously . . . no C-Media chipset on the market can even come close to this capability, most topping out only within the 10,000-15,000 range (IIRC).

For the absolute best positional audio and playback possible, you need higher software voice support - big reason why CMSS-3D is still considered much better than Xonar's offering (Xear3D, IIRC); number of software voices also plays a big role in how many different times the same sound can be played at the same time, as well as software occlusion effects, and countless other aspects of the final playback.







I'm by no means saying the Xonar cards aren't worth their cost - far from it! They are very respectable cards, IMO - if you're looking for premier audio quality for a range of uses, they're more than capable. They offer great audio quality, without having to pay an arm and a leg like you would for an Auzentech or high-end HT Omega.

I understand there are a lot of people that have had many issues with Creative's cards . . . but likewise with ASUS and Auzen and HT and all the other manufacturers (especially in terms of Vista issues). Creative's issues just dwarf everyone elses due to the sheer volume of their cards on the market compared to others.

But, relatively speaking, they do have more driver issues than hardware issues - compared with everyone else.

Ahem.. I'll summarise the guru3d review for you 128 vioces, just like an X-FI, build quality thats lightyears ahead of the Creative POS, EAX 2.0 (lets face it.. other versions of EAX don't actually do that much to make things more "immersive") Comparable framerates in games, miles better audio quality for music and DVDs.. I think thats enough

Asus even bundle Rightmark 3D Analyser to prove how good their Xonar is, don't see Creative doing that for their precious X-FI, do ya?

And to hammer the nail home about CPU utilisation, the ADI1988B CODEC on the Crosshair frequently had spikes of where task manager would report 30% CPU utilisation on my 3500+. Did it affect FPS in games? Pfft. Like hell it did everything flew by and sounded amazing to boot.

 

[Ketxxx]

I would really like a xonar D2X.Droool

Are the DX's TI dacs burr brown ones ketxxx?

I has absolutely no idea. guru3d is probably your best place to go to find that out, they have the most detailed review for the Xonar I've seen.

 

[imperialreign]

Crystalliser: to each their own. i dont actually like it, as it just makes my music strange (possibly because i have winnar speakers)

TBH - the Crystallizer feature is junk - except for with compressed audio and average bit rates below 190kbps. It can improve how the file sounds, but it's still dependant on the actual quality of the compression itself. Personally, I only recommend it's usage for gaming, or when listening to shite mp3s. ALthough, I still recommend that users rip mp3s at at least 200 kbps; and I push using lossless codecs as well. TBH, I hate mp3s, no matter what their bit rate.

In terms of uncompressed or lossless audio, though, Crystallizer actually hurts the output quality quite a bit . . . you can't improve on something if there's nothing there to improve on y'know? Big reason why I recommend lossless or high-quality variable bit rate.

Ahem.. I'll summarise the guru3d review for you 128 vioces, just like an X-FI, build quality thats lightyears ahead of the Creative POS, EAX 2.0 (lets face it.. other versions of EAX don't actually do that much to make things more "immersive") Comparable framerates in games, miles better audio quality for music and DVDs.. I think thats enough

Asus even bundle Rightmark 3D Analyser to prove how good their Xonar is, don't see Creative doing that for their precious X-FI, do ya?

And to hammer the nail home about CPU utilisation, the ADI1988B CODEC on the Crosshair frequently had spikes of where task manager would report 30% CPU utilisation on my 3500+. Did it affect FPS in games? Pfft. Like hell it did everything flew by and sounded amazing to boot.

I've read all that stuff before - yes the Xonar series can support 128 hardware voices, just like the whole X-Fi line-up, all of Auzentech's cards, HT Omega's, Razer's, and every other reputable audio card on the market.

But none of them support more than 15,000 software voices - the C-Media chipsets cannot handle that kind of driver workload.


As to RMAA; I had no idea ASUS actually provided this free utility with their cards . . . but do they provide the testing methodology .pdf as well?

Even more so, how many users actually understand how to read and interpret RMAA results, or are ASUS kind enough to also provide basic documentation for that as well?

Again, back on with CPU utilization - I never mentioned anywhere that high CPU use (as with onboard audio) affects in game FPS; that's a bollocks point nowadays, as even a nutburst Pentium 4 isn't noticeably affected by CPU use. At the most, onboard audio might cause you 1-2FPS, and that's under the worst hardware circumstances. Use of a standalone card might net you 1-2FPS at the most.

Is it an amount we'd actually notice in game? Hardly.


Again, I reiterate, CPU utilization is a moot point with audio hardware. But what is a point to stress, though, is system audio latency. That equates to many of the audio issues people complain about, and was the biggest sole contributor to the X-Fi snap/crackle/pop issue. The cards were working to fast for the system to keep up with.

When your audio hardware processes too fast, you end up with audio clipping as it has to continue to buffer the audio while waiting on the system; when it moves to slow, you end up with clipping as well, as the audio isn't done being processed before the DSPs and system start loading up more work.

It's a matter of fine tuning that balance - ASUS, Auzen, HT, and everyone else running C-Media chipsets don't have that hard of a time, as the chipset can only work but so fast; they work great with the standard PCI BUS. It's very easy to design within those parameters, as PCI has been an industry standard for over 10 years now. Even still, with the newer PCIE cards, they still must use a bridge chip, which for the most part slows the BUS stream down for the audio chipset.

Remember back when audio cards finally started moving from ISA to PCI? How many issues cropped up with the new PCI hardware? The cards were working too slow for the faster BUS, and the easiest fix many times was to shorten the PCI latency of the installed slot. 4-5 years later and it wasn't a concern anymore, as the PCB components on audio cards were capable of working faster.

But, in 2005 with the release of the X-Fi line-up, we were running into the opposite issue - the APU works too fast for the PCI BUS, which lead to a lot of audio issues. The typical fix here was to increase the PCI latency to allow the card more time to the BUS; but with so many mass-produced systems, not everyone had the ability (or the know-how) to do so; so the cards had to be revised with a new BIOS that forced the APU to work slower, which eliminated a lot of issues in most systems.

It's not about the CPU or BUS utilization, but about the card latencies - and when it boils down to a competitive edge for gaming, it's hard to beat out 65,000+ software voices being processed at any given time. That can spell the difference whether you hear your opponent ages before they've snuck up on you, or after you've handed them the frag.



Again, I'm not saying the Xonars are bad cards - again, far from it! They are very reputable cards for their price segment. They beat out the X-Fi's in audio quality definitely, but when it comes to processing performance, the X-Fi's hand the Xonar's their ass.

For the most part, every company on the market offers better audio quality than the X-Fi's (except for the Elite Pro and Titanium), but none of them can truly compete on a hardware level.



As to the X-Meridians, Xonars, Preludes, Barracudas, Infernos and all the other high-end audio cards . . . yes, there are countless threads on the internet of people modding these as well looking for the best output quality possible.

The biggest mods, though, typically involve swapping out the capacitors. Sure, the Xonar and Prelude use better quailty capacitors, but all solid states are not ideal for audio cards; especially for filtering audio output channels. In this sense, all the high-end cards have followed the same stoopid route that Creative have of slapping the same type and brand of capacitor for every circuit on the board. In that sense, none of the other manufacturers are better than Creative as they're making the same mistake.

Sure, they use better DACs, ADCs, OPAMPs, VRMs and other chip components compared t the X-Fi lineup . . . but the sources these chips influence and change will always be seriously hampered by the capacitors found on the circuits.

 

[trodas]

Ketxxx -

nobody should spend £200+ on a soundcard then have to recap the entire PCB to be absolutely sure no problems with it will arise

Well, I did

Xonar D2X, I've just seen some reasonably detailed pics, and things like solid state capacitors are never a bad thing

It is, when come to audio. Polymers are horrible for audio. No caps or polyruetane roll ones is the best choice.


imperialreign -

Xonar D2 - LM4562 for front channels; TI R4580I for all others
Xonar D2X - LM4562 for front channels; TI R4580I for all others
Xonar DX - TI R4580I for front channels; TI 5532 for all others

Mediocre opamps. LM4562 give a lot of details, but are too sharp and too RFI picky... These 5532 are solid performer, but still nowhere near AD8599 quality. I did not want that card

 

[Ketxxx]

Cant agree on that. Doesn't matter what part is used weather its one from a range of opamps or electrolyite caps or something else, each has its strengths and weaknesses. The key is finding a decent balance, and Asus have done that. The Xonar is a quality audio card with that balance between producing great results and longetivity thats provided with the solid state caps.

 

[Mussels]

not to mention all the auzen cards are socketed, so you can replace the opamps without soldering (and without trashing the warranty, i believe)

 

[imperialreign]

not to mention all the auzen cards are socketed, so you can replace the opamps without soldering (and without trashing the warranty, i believe)

100% correct; although, the best OPAMPs don't come in a DIP fashion; but, there are still some superb choices out there . . .

and for those Auzen owners who aren't the most keen on actual PCB components, Auzentech even goes the extra length by providing quite a bit of information on OPAMPs in general, as well as even giving a list of OPAMPs you can purchase directly from them (so you know they'll work . . . or at least fit the socket ). http://www.auzentech.com/site/products/opamp.php


Auzentech is still hands down the cream of the crop for premier audio cards, no matter what audio processing setup they've used - either the C-Media chipsets or an X-Fi APU.

If you're looking for the best audio hardware, coupled with the best sound quality, and outstanding customer and technical support and service, they are definitely the way to go. Hell, you even have your choice if you want to purchase a fully functional X-Fi (no Creative drivers! ) that even includes Dolby encoding capabilities (unlike Creative's line-up).

 

[Mussels]

auzen just have more reliable drivers. that alone makes them better.

with creative (audigy 4 in lan rig) i've ran into:
*features cut back/removed
*crackling audio (still happens to this day)
*BSOD's (less common, but mostly related to the fact A4 owners dont get any software in vista)
*useless mic port (its too quiet, and turning up the volume/using boost results in nasty static)

with my auzen... nothing. i enable dolby encoding, plug the speakers in and dont touch it. its never crashed or bugged out on me, ever (and i have an older model too)

 

[imperialreign]

which I give Auzen a superb, well-deserved for

hell, even their X-Fi equiped models run flawlessly; even more amazing, Creative supply them the APU drivers!

 

[trodas]

X-Fi output DC offset measuring with LM4562 opamps
---------------------------------------------------------------
L: -190mV DC, 10mV AC
R: -174mV DC, 10.2mV AC
RL: -210.2mV DC, 10.1mV AC
RR: -204.7mV DC, 10mV AC
CENTER: -209.5mV DC, 10mV AC
SW: -231.9mV DC, 10.2mV AC


X-Fi output DC offset measuring with AD8599 opamps
---------------------------------------------------------------
L: -219mV DC, 9.5mV AC
R: -205mV DC, 9.8mV AC
RL: -209mV DC, 9.7mV AC
RR: -205mV DC, 9.3mV AC
CENTER: -210mV DC, 9.6mV AC
SW: -232mV DC, 9.7mV AC


Conclusion - X-Fi opamps stage design does produce the -200mV DC offset. That probably can be cured by just increasing the positive opamps supply voltage from 5V to 5.2 or 5.4V ...


Ketxxx -

Cant agree on that.

Not to worry, me neither. Polymer caps is great for high-frequency voltage filtering (just check their frequency response curves, for christ sake!) while using them for anything getting close to audio is UNFORGIVABLE mistake that completely erase any filtering in low-frequency, hence the strong bass line kill the quality effectively.
You can't measure that on the RMAA, to show this problem you need the DAC/opamps output a loud bass line and then the polymers crumble and kill the voltage filtering = the card output get much less that satisfactory.

Just because todays there are some kind of "hype" about polymers does not mean that they are usefull for everything. They have a very strict usability around 200 - 500kHz, and that it is.

The key to understand this problem is thing called ripple rating coeficient. It basicaly says what multiplier has to be applied on the rated ripple for different frequencies. Nothing show this better that this table:

Samxon X-con URL polymers: 120Hz = 0.05; 1kHz = 0.30; 10kHz = 0.70; 100kHz = 1.00
Samxon elyctrolyte GA caps: 120Hz = 0.50; 1kHz = 0.80; 10kHz = 0.90; 100kHz = 1.00

The specs are very much the same as with all other polymers - even spoken - the better polymers, the worser they run at low audio frequency.

Hence polymers are, once again, an unforgivable fault and you can dance around this all the way you want, it is just an unforgivable serious mistake made for the marketing crap and stupid people around here...

Doesn't matter what part is used...

Ehm, sorry? Then what are we talking about there, lol?


Mussels -

not to mention all the auzen cards are socketed, so you can replace the opamps without soldering

True, but the best suited opamp (gotta take the input sensitivity into account too!) is made with the SOIC NARROW version only, so this in kinda useless then... sadly. Yet you still can exchange the opamps for something better and less general-use-like like the 5533 ones are.
They are steady performers, don't get me wrong, but they just aren't real audio grade opamps. And the LM4562 are too much picky and hi-frequency and sharpness like... So, not good.

*crackling audio (still happens to this day)

Normal thing with bad caps.

 

[Mussels]

no the crackling audio is software related. for example, supreme commander works in 2.0 or 5.1 on the auzentech just fine, but the creative card crackles nasty like when in 5.1 mode. in 2.0 the problem doesnt happen (and its over ALL channels, so its not just the rear outputs doing it)

 

[lokonsky]

Conclusion - X-Fi opamps stage design does produce the -200mV DC offset. That probably can be cured by just increasing the positive opamps supply voltage from 5V to 5.2 or 5.4V ...

Hi trodas ...i'm so interest with your Great work...
How you increase opamp supply voltage?

 

[trodas]

Mussels - none of what you say contradict my statement that this IS about caps. This is exactly how bad caps HW behave... and only in multichannel mode? Exactly why this is caused by bad caps. Multichannel = way more load on the caps... Don't you get it?

lokonsky - Easy. Check out how the L7805 regulate the voltage. The IO manage the 5V difference between the output and common ground, usually connected to the ground. So, to bump the output voltage, you just have to add a resistor between the common ground and the ground That one, that has like 0.2V lost on it with the given current Easy.
But that is lame.
Correct way is use TWO resistors and create a voltage divider on the output of the L7805 chip that feed the required 0.2V to the common ground and you got your 5.2V on the opamps witch might (or might not) cure the -200mV offset.

 

[imperialreign]

final testing with the AD8599 OPAMPs with copper heatsinks installed over them to partially act as an EMI shield;



AD8599 with no HS:

and with a HS:

negligible difference, unlike running similar tests with the LM4562s. TBH, there's no difference even noteworthy enough to bring to attention.

Although, the AD8599 without using any form of shielding still produce better results than LM4562 with shielding.

 

[lokonsky]

Correct way is use TWO resistors and create a voltage divider on the output of the L7805 chip that feed the required 0.2V to the common ground and you got your 5.2V on the opamps witch might (or might not) cure the -200mV offset.

So if 5.2V not cure the DC offset, a coupling capacitor must be installed then...

 

[trodas]

Who says so? DC offset can be neutralized in the input amplifier caps - and there is usually way more space, so a proper polypropylen caps can be used there, to the great benefit for the sound!



X-Fi output DC offset measuring with original UK X-Fi, no heatsink and L+R channel NJM4556 opamp, rest NJM4558 opamps
---------------------------------------------------------------
L: 29.9mV DC, 9.8mV AC
R: 29.6mV DC, 9.8mV AC
RL: 0mV DC, 9.8mV AC
RR: 0mV DC, 9.8mV AC
CENTER: 0.2mV DC, 9.6mV AC
SW: 0.2mV DC, 9.7mV AC


X-Fi output DC offset measuring original UK X-Fi, shorted DAC - opamps caps
---------------------------------------------------------------
L: -179.4mV DC, 9.8mV AC
R: -204.5mV DC, 9.8mV AC
RL: -222.9mV DC, 9.8mV AC
RR: -232.2mV DC, 9.8mV AC
CENTER: -225.2mV DC, 9.6mV AC
SW: -238.9mV DC, 9.7mV AC


X-Fi output DC offset measuring with LM4562 opamps
---------------------------------------------------------------
L: -190mV DC, 10mV AC
R: -174mV DC, 10.2mV AC
RL: -210.2mV DC, 10.1mV AC
RR: -204.7mV DC, 10mV AC
CENTER: -209.5mV DC, 10mV AC
SW: -231.9mV DC, 10.2mV AC


X-Fi output DC offset measuring with AD8599 opamps
---------------------------------------------------------------
L: -219mV DC, 9.5mV AC
R: -205mV DC, 9.8mV AC
RL: -209mV DC, 9.7mV AC
RR: -205mV DC, 9.3mV AC
CENTER: -210mV DC, 9.6mV AC
SW: -232mV DC, 9.7mV AC


A word about the DC offset.

maybe it's not safe to remove the coupling capacitors

Maybe it is not safe to live on Earth
Got my point?
Nothing is safe, ever. However the X-Fi did not have by default ANY cap between the opamps output and the jacks. So...

And the main point is, that the DC offset is neutralized on the amplifier input coupling caps. So nothing to worry about.

Unless you have custom amp with stripped input coupling caps, then you have absolutely nothing to worry about. It can only make your sound nicer and more detailed.

In some cases the input caps in the amplifier aren't even need to be there, depending on the amp input opamps and their gain versus voltage. If the gain is low and voltages high, the DC offset is just amplified like the rest of the signal and that it is. No clipping happen then. But it is wise to let one last coupling cap to have in the input of power amplifier... if you did not want end up with like 16V DC offset on the speakers :tongue:

 

[trodas]

My modification of X-Fi Fatal1ty, UK edition w/o heatsink on main chip and with caps wires bent to sides and then soldered from the bottom of the PCB for user Tez, Head-Fi forum:

...and his card is working just great!

 

[trodas]

If you remove all the eight audio caps, you won't be able to do a RMAA measuring using loopback cable, since the input jack will not work anymore. So, to make it work, you gotta add a C48 and C49 caps. The original are bipolar 4.7uF 50V Jamicons. I used audio quality polypropylene film caps to replace them, 4.7uF 63V MKT ones (Digi-key order number 495-1131-ND ):

And then I can make my first RMAA test: http://ax2.old-cans.com/X-Fi Fatality MKT caps 24_48.htm

It is much better that trying to pass the -200mV DC offset right back to the input: http://ax2.old-cans.com/X-Fi Fatality NOCAPS.htm

...