Got my M-Audio Audiophile 2496

and I'm loving it a lot. There's a transparency there I didn't quite have with the old Turtle Beach Santa Cruz, and the soundstage is definitely much deeper and larger. I'm enjoying a more detailed and more enveloping sound. Thanks to all of you for the many recommendations of this card--right on. I got it from Guitar Center for 100.00, a good price and comparable to mail order. (They're pretty good about matching online prices, esp. from other online music retailers.)

I just finished my first needle drop at 24/96: Steely Dan Pretzel Logic, original ABC release cut by Alan Zaentz (I hope I got that name right). The LP isn't perfect--the low mids are a little on the warm side and there's a midrange suckout that dulls the sound a bit--but these are relatively minor flaws in a nicely cut LP that sounds much, much better than the most recent CD remaster. The Audiophile 2496 did a great job on the transfer, and the bit-conversion/downsampling on Sound Forge got the drop to redbook with very little loss, to my ears.

Question: when y'all record at 24/96 and get ready to convert to redbook, do you do the bit-conversion or the downsampling first?

 

I like to avoid all conversions by recording at 16/44.

 

Yes, it's very important to dither when converting to 16 bit. Is there a reason why are you not recording in 32bit? Are you recording in 96? My SF 6 won't support it.

 

If your end product is redbook, don't dance around with up/down. It's always nicer and truer to input at 16/44.

 

That assumes he's not going to do any processing in the meantime.

 

I don't think there are any cards that record in 32bit. Grant may come in and explain it better, but as I understand it, "32bit" is a file format that adds extra information to a 24bit recording giving it "floating point" accuracy - sort of like the difference between a vector-based CAD program and raster-based Microsoft Paint program.

 

IIRC the Audiophile 2496 records at "only" true 18 bit.

 

Unless the internal sample rate is switchable, most all of them are 48k by default. If this true with your card, your card is doing sample rate conversions anyway. In this case, the best plan of action would probably be to record at 48k unless you trust the hardware to do transparent SRC. Otherwise, do the conversion with the software, or dedicated hardware.

Also, if you are not doing anything to your files, 16-bit is the best if your are going to red-book CD. If you do anything to your files, it's better to record or increase your resolution to 24 or 32-bit, and dither back down for CD.

 

ALWAYS do the SRC FIRST! Then do the dithering second. Some software, will do both in one process. Dither and noise shaping should ALWAYS be the VERY LAST thing you ever do!

But, if you like, you can do one first, and then another later, in another program, perhaps. I will use either Sound Forge or Cool Edit/Adobe Audition for the SRC, but will ALWAYS use Audition for the Cool Edit/Audition for the dither and noise shaping. I found the dither in SF to be less than passable. Besides, Audition gives me TONS more options for dithering/noise shaping. I can get the exact results i'm looking for.

 

Are you sure about that???

 

My card is indeed switchable, by means of a knob on the front.

 

Basically, that's correct. 32-bit offers an infinite dynamic range. In some programs, 24-bit will automatically be converted to 32-bit, as most quality ones such as Cool Edit/Audition, Sould Forge, Waves, and tons of plugins use internal 32-bit floating-point processing.

Again, all this is necessary IF you process your files in any way, even gain changes. If you just record and don't do anything other tyhan a few simple edits, 16-bit is OK.

 

I’ve had a few different M-Audio cards. Currently I have a Delta DIO 2496. I use the analog RCA outs to a preamp and I also use the optical out via a glass fiber cable to a DAC, the same one I use from my CD Transport.
I also have the Sonic Theater which has the very same ‘audiophile-2496’ chip. I use it to record in from a tape monitor off the preamp via a micro to RCA adapter.
I had a Delta 410 card for some time and used its RCA inputs to record vinyl. There weren’t any better. The DIO 2496 has two discreet sections, a digital in/out (toslink and coaxial) and an analog out for high quality monitoring. Its optical out is superb! Shame it doesn’t have an analog input. The Sonic Theater USB device is really nice. I use it for games and DVD’s on the PC, the Windows integration is better then the Audiophile or Delta cards. I can also put the device on the rack with the other HiFi gear and patch cables in and out of it. Hell, I have hundreds of dollars of IC’s running to a $90 USB sound chip – hehehe, its sounds great.
MP3’s sound amazingly great to my tube DAC but the analog outs are damn good too.

I tried recording at 24/96 too and its just not necessary if your not doing any major editing or filtering and EQing. The native 48k thing I wasn’t aware of, that’s a good point, maybe 16/48 is best. I’ve been recording in 16/44. I usually record an entire album side, the cut and paste in Sound Forge, fade ins and outs and sometimes a tiny bit of gain control and EQ’ing.

I remember when you couldn’t touch that card for under $200!

 

I have the Audiophile 2496 card, and I master using Peak 4. I have compared every configuration available to me with this combo of hardware and software: recording at 16/44, 24/44, 24/88, and 24/96; dithering using POW-R 1, 2 and 3; downsampling first, and bit-conversion first. My results:

Best transfer option is 24/96, no question. When playing it back off the hard drive, transfer sounds very "analog": warm, transparent, with good depth and realism. BUT... problems begin with conversion to 16/44 redbook standard. Converting 24/96 to 16/44 preserves the most detail, but imparts a hard digital edge to the sound that is unpleasant to hear. Transferring at 24/88 and then converting to 16/44 misses some detail compared to 24/96, but preserves a more "analog" sound and tonal balance. I find this to be the best choice overall. In any case, downsample first, then convert the bit rate.

I find that the results of converting a 24/88 transfer to 16/44 are superior to a straight transfer at 16/44. However, every choice is a tradeoff. If it were possible to burn a straight 24/96 transfer to CD, then that's what I'd do. Everything else is unsatisfactory. That's why, at the end of all my tedious experimentation, I decided the best option was to just listen to the analog source and forget about transferring to CD-R! However, to make copies for archiving, listening to in the car, etc., I find the best compromise to be 24/88 --> 16/44, as I said.

 

Oh, and POW-R 3 dithering is better than 1 or 2.

 

The Audiophile 2496 sample rate is indeed switchable. That's one big reason I bought it. I believe it's the AC97 spec that ties most consumer cards to the 48k rate, with other rates involving SRC on the fly.

I'm not doing much to my files aside from simple edits that split the tracks for burning, and fades in and out for the sides. I will fix whopper clicks-and-pops by hand, one at a time. For annoying surface noise between tracks (only), I'll use Sonic Foundry Noise Reduction 2.0. I never use NR on the music.

Oh, and my SF 6 (I've got 7 waiting to be installed) does support 96khz recording.

 

My specs say true 24bit/96khz recording. My old Turtle Beach Santa Cruz was limited to 18 bit.

 

Thanks for that detailed information. I bet the downsampling from 88khz goes a lot faster, too. I have noticed how liquid and detailed the 24/96 transfer is. I don't notice the digital edge going down to redbook, but I'll try the 24/88 method you outline and see if I can hear a difference.

Question for you and Grant and any other dither/bit-conversion gurus: what options are you choosing, exactly, and what are the advantages and disadvantages to each? I'm using the option Sound Forge says is the best in most cases: highpass triangular dithering and high-pass contour noise-shaping.

 

Yeah, they all do these days. Audition supports 192k.

 

In Sound Forge 6, I notice a veiled sound that I can't seem to get past. SF doesn't allow you to customize the amount of dither you can apply, nor does it have many noise shaping curves. I read a few tests on the internet that poo-poos SF dithering/noise shaping, saying that the problem is that it doesn't supply sufficient enough dither for passable sound. If I can find these tests results, i'll post them here. I may have seen them on the isoTope/Ozone website, and someone on the old Syntrillium posted a link to another study that noted the same kinds of remarks about SF. One of Cool Edit's curves was somewhere in the middle or upper middle ranking.

But I have found SF SRC very good, a wee bit better than Audition, IMO.

My guess is that the reason SF dither and noise shaping curves are lacking is that they were, in fact, among the last companies to go 24-bit or 32-bit. From what i've observed, most of their rank and file users record and process at 16-bit/44.1, and they have Direct-X for those who wish to use superior SRC and dither. SF bills themselves as a "professional" software, and indeed, many pros use it. It should be known that the pro-recording world is still very much a 16-bit world! 16-bit is used quite frequently. I was told by a current, active remastering engineer that instead of tapes, most of the oldies he gets are on 16-bit/44.1 CD-R that were dubbed from the master tapes. Knowing this, I suppose that SF didn't see the need to move up to high-res so quickly.

 

So what dither settings do you find yourself using in Audition, Grant?

 

Even if you are doing a change in digital levels on the sound file, you should be in a higher resolution. If I could find it again, there's a good article online that describes the math behind working in 24-bit or higher as opposed to 16-bit if you're going to be manipulating the signal at all.

I do everything at 24/96 (or 24/88.2) now. My goal is that these are going to end up on DVD-Audio, and I can downsample for CD if I ever need to. If my sound card supported it (SoundForge does), I'd be doing everything at 24/192. Now I just need a DVD-Audio authoring program that's worth a crap and doesn't cost a small fortune...

 

After much experimentation, I found that .7 bits is a good dither amount that balances between distortion and too much noise and veiling of the sound. I also favor a C2 noise shaping curve because it retains the crispness of the highs. But, it does not sound as quiet as the 44.1 curve.

If you use noise shaping, you can get away with using lower bits for dithering.

You can't adjust dither bits in SF.

 

Oh, I agree completely. Sckott's advice to not futz around and record at 16-bit is only valid if the user does not intend to process the file.

 

Sad, but true. If I hadn't been using SF for years [and being comfortable with it] plus getting a "deal" [$100 for a used copy] on SF 6, I suppose the smart thing to do would be to use the likes of CE/Audition. A bit pricey new, however.