Unraveling X-Fi frequency response in Windows Vista

In our last couple of sound card reviews, we’ve noticed some odd behavior from products based on Creative’s X-Fi audio processor in RightMark Audio Analyzer’s frequency response tests. The X-Fi had no problems with the frequency response test when we initially reviewed the XtremeMusic more than two years ago. However, that was with RMAA 5.5 and Windows XP. RightMark Audio Analyzer is now up to version 6.06, and at least under Windows Vista, X-Fi frequency response curves are all jacked up. Here’s what I mean.

In the graph above, taken from RMAA 5.5 under Windows XP, the X-Fi XtremeMusic’s frequency response curve looks fine.

Fast-forward to Windows Vista with RMAA 6.06, and the frequency response curve is a mess. These results were taken from the exact same sound card and associated driver control panel configuration.

This behavior goes beyond just the XtremeMusic, as well. Again in Vista, the frequency response curves for the X-Fi Fatalty and Auzentech’s X-Fi Prelude are all over the map.

What’s particularly odd about these results is that they only appear in RMAA “loopback” tests that simultaneously play back and record the test signal on the sound card. When we use a separate sound card on a second test system to handle the recording, the response curves even back out.

The fact that X-Fi frequency response is unaffected with these straight playback tests is even more curious considering we actually used a second X-Fi as the recording card on our secondary system. So there’s nothing wrong with X-Fi frequency response on the playback end or on the recording endβ€”only when the two are used together in a loopback test on a single card.

We’ve contacted Creative about the issue and hope they’ll respond with a solution. However, unless you’re in the habit of playing back and recording audio simultaneously, you shouldn’t be affected by the X-Fi’s poor frequency respose in RMAA’s loopback tests.

Comments closed
    • Prototyped
    • 11 years ago

    /[<(never mind, I hadn't noticed the followup blog post.)<]/

    • yogibbear
    • 12 years ago

    My onboard Realtek doesn’t look too shabby for vista…

    • kitsura
    • 12 years ago

    Good luck on Creative responding to anything if it doesn’t earn them money.

    Btw, I’m still pissed that their latest driver update more or less ruined my X-Fi Elite Pro. My breakout box has just become a very expensive and ugly paper weight.

    • ssidbroadcast
    • 12 years ago

    In what real-world circumstance are you recording and playing back at the same time?

    • just brew it!
    • 12 years ago

    Y’know what that looks like to me? Constructive and destructive interference. If two copies of the test signal are mixed together with one of them greatly reduced in level and slightly delayed, you’d get a response curve something like that as the frequency varies. I wonder if some of the test signal that RM outputs is leaking back into the recording signal path through a mechanism /[

      • UberGerbil
      • 12 years ago

      Interesting. I’d buy that.

    • Krogoth
    • 12 years ago

    I suspect that the culprit maybe audio-level DRM at work with Vista.

      • just brew it!
      • 12 years ago

      If it was happening to /[

      • Mikael33
      • 12 years ago

      lol more fools thinking DRM is the cause for all thing bad in vista lolol

        • sigher
        • 12 years ago

        Well since the drivers are part of the DRM kill-bit nonsense it might just be that creative failed but just because of the stupid DRM requirements, or it’s even possible that creative succeeded in applying full DRM ‘protection’ where the others fail and that that’s the issue when other cards have less of it..
        The vista DRM stuff is not a myth you know (for those here that are simpletons and believe that) , it’s fully documented by MS.

          • ChronoReverse
          • 12 years ago

          Yup, the DRM that’s so restrictive that it doesn’t do anything at all. Come back when you can provide a real example instead of the conspiracy theories.

            • sigher
            • 12 years ago

            Yeah let’s listen to MS forum suckups rather than what countless experts including the freaking MS engineers reported.
            Let’s all blame gremlins and ferries whenever vista goes screwy instead of being a mad conspiracy theorist and blaming the software.

            But yeah, Vista is great, I’ve been using it 14 years now with no issues AT ALL.

            (Just checking what benefits exactly you get when you do the MS PR job)

    • MadManOriginal
    • 12 years ago

    Think Creative would ifx this in drivers? haha, good one πŸ™‚ Maybe auzentech will push them to fix it though, they’ve been good at getting driver updates and fixes.

    • Saber Cherry
    • 12 years ago

    I think “Unraveling” is a cute headline, in light of the tangled-looking response curve =)

    By the way, even when “Working correctly” the Realtek and Creative have a frequency-response plunge starting at ~16kHz. Can you guys find some young whippersnapper that can hear in that range and see if it sounds as bad as it looks? Because there’s really no excuse for not supporting sounds above 16kHz even in an audio stream with Nyquist frequency 22050Hz, let alone the Nyquist frequency of 48kHz at 96kHz samples. So I wonder if this is a readily audible difference in someone who can hear well in the range.

    Also, does the test software go above 20k? Because if it does, then it looks like Asus is also putting in an arbitrary cutoff (though ~4khz higher). And if it doesn’t… I have to wonder why not.

    P.S. If you do end up retesting anything… I would love to see the results of retesting the onboard sound /[

      • MadManOriginal
      • 12 years ago

      Don’t know how young they have to be, I can hear to slightly above 18kHz and I’m over 30. Then again that was coming out of a Prelude so maybe I CAN hear higher but the card cuts it out too much :p

      • liquidsquid
      • 12 years ago

      You reduce at 16kHz and up because to maintain a high S/N you want maximum attenuation before nyquest. When you are talking areas of 110dB, that is a LOT of attenuation. If you rose the “knee” up higher in frequency, your drop-off would need to be much sharper, which would produce phase issues and much more “bumps and lumps” in frequency response. In other words, they erred on the side of conservative caution.

      I used to be able to hear to 18kHz, now only 16.5kHz. I could never hear below 30Hz, but I sure could feel it in my speakers! I miss my 1.2kW of audio madness and home-brew folded horn refrigerator speakers. Those things would move furniture.

      -LS

        • just brew it!
        • 12 years ago

        The phase distortions introduced by a sharp cutoff analog filter are the reason that good CD players have used oversampling D/A converters since the 1980s. Basically they interpolate between the samples digitally, eliminating the need for a sharp cutoff analog filter. (Another way of looking at it is that most of the low-pass filtering is done digitally, where it can be done without introducing as much phase distortion.)

        There’s no reason a sound card couldn’t implement the same sort of thing. In fact, I think that may be what the cards that are flat out to 20kHz are doing.

        • Saber Cherry
        • 12 years ago

        Bear in mind that CD players sample at 44.1 kHz. Any decisions made based on that sampling rate would not be relevant to a 96 kHz sampling rate, with a nyquist frequency of 48 kHz.

        • MadManOriginal
        • 12 years ago

        Soo which curve is ‘better’? I’d think the flat curve is but with all the voodoo that can be involved in audio along with synergy of the whole playback system maybe it’s not always.

        Are the X-Fi based cards just doing things a cheaper way, in that it’s easier to attenuate earlier? -5 dB at 20kHz is a pretty strong dropoff, almost 1/4 the percevied volume, but I suppose that’s academic for all but the bat-eared among us. I’m reading the wikipedia page about (Nyquist )sampling and I think I get how the attenuation and 2x sample rate relate but I don’t know what the result of an insufficient attenuation would be.

          • Saber Cherry
          • 12 years ago

          For CD audio, supposedly, a sudden cutoff at 22.050khz sounds “harsh”, so a steep gradient from 20 – 22 is “better”. But 16khz-20khz should sound exactly like the input.

          For 96kHz sampling, there should be no cutoffs below 30kHz, period. Probably 40-44kHz would be a good place to start filtering.

            • MadManOriginal
            • 12 years ago

            So the X-Fis that start filtering at ~16kHz are far from ideal huh.

            • Saber Cherry
            • 12 years ago

            Far from ideal, but it might not be noticable to most people over 40. For people under 40 it really just depends on how sensitive they are and how well they protect their hearing. For people 20-30 who have never attended rock concerts or worked in construction/airports/etc, there should be an audible difference.

            With good speakers, I expect the difference should be blatantly obvious to most people under 20, using lossless source music with a lot of high treble (MP3, for example, strips out the upper frequencies). But I don’t know for sure, which is why I would like to read the perspective of someone in that age range who has access to both cards.

            It’s also hard to predict because the entire 10kHz – 20kHz frequency range is squished to represent less than 10% of the graph, and the x-axis is at -5 dB, so you can’t clearly see what’s going on.

            • just brew it!
            • 12 years ago

            q[

    • deruberhanyok
    • 12 years ago

    interesting… you guys are awesome for looking into this and posting about it.

    • Forge
    • 12 years ago

    Very interesting. Thanks very much for looking into this, Geoff. I look forward to Creative’s doubletalking excuses.

    • Dposcorp
    • 12 years ago

    Wow. I don’t have any of the cards, but I dig you guys taking the time to document the issue on a 2 year old piece of hardware. Kudos to you.

    Did the two systems both run Vista?

    I wonder if it is a “Vista” issue, either driver related (alchemy) or new vista audio components………on the other hand, maybe it is some kind of internal card sound loop, kinda like when you plug a switch into itself, or record a display that is displaying what is being recorded, etc…

    Is the card maybe swinging back and forth trying to compensate for itself raising or lowering playback and recording frequencies?

    • Madman
    • 12 years ago

    “We’ve contacted Creative about the issue and hope they’ll respond with a solution.” -> Creative? LOL

    If I got it right then you cannot record ‘what you hear this way’, right?

Pin It on Pinterest

Share This