What We Heard

The "veil"

If you read the audiophile press, or listen to guys that like to think they’ve got really hot ears, you’ve probably heard expressions like, “Well, you know, when you go to 16-bit from 20-bit it’s not so much a difference, like, EQ or distortion or something, but it’s like, well, there’s this veil that’s been added. Yeah, that’s it, a veil.”

Well, they’re right. There’s a veil. When you bring up the level 40 dB (I’m not sure you wanna try this at home, with your speakers – if you DO decide to try it and end up blowing a hole in your living room wall with a suddenly liberated voice coil, well, don’t call me), it’s actually very obvious and audible. The 4-bit loss of resolution adds a kind of low-level murk to the recording.

The effects of truncation

Truncation is when you simply throw away extra bits, and don’t bother to re-dither the new Least Significant Bit (you might want to refer back to my earlier article if you’re confused by this).

What does it sound like? A variety of badnesses show up. Most noticeable and troubling is what happens with fades. Have you ever fooled around with noise gates? You know, set the threshold at some low level so the gate is always open and then slowly turn down the signal level? Well, what happens is that as the signal approaches the threshold of the gate it starts switching on and off. Well, that’s what happens with undithered signals as they approach the Least Significant Bit. They start cutting in and out. Most Annoying, as Winston Churchill used to say about bombs dropping on London.

Also, I noticed, on the string quartet recording, a definite change in timbral quality. One other listener commented that it sound like the players had changed their bowing, which I think is an apt description. I also heard a timbral change on the Bowie remaster when listening to the truncated, undithered 16-bit version, though it wasn’t so obvious there. Edgier, grittier, harder, less refined, less well-defined are all terms that apply here.

The various dithers

Well, it turns out you can just about always hear the dither noise! Make no mistake. It’s really there! In some of the instances we experienced, it was even moderately offensive. It also worked. It permitted us to hear audio signal a good 10 dB or more below the Least Significant Bit (and the level of the dither), and none of the problems arising from truncation that I described above showed up, except for a curious problem with Toby’s fade-to-digital-black example. In that case, the transition from dither to digital black yielded an ugly and chunky noise artifact just before the black signal arrived. I suspect that mastering engineers wisely maintain dither noise across digital black transitions between tunes, just to head this little cutie off.

White Noise Dither

What’s to say? You can hear it pretty easily (it sounds like white noise, duh!), it is generic, and it does the job.

Super Bit Mapping II

There were lots of unkind remarks about Sony’s Super Bit Mapping I (which we didn’t audition), but to my ears, Super Bit Mapping II, the latest version of SBM from Sony, worked very well. It was the least audible, to me, of all the dithers and it maintained the audio quality very nicely as we approached and went below the LSB.

Apogee’s UV-22

I personally didn’t hear any advantage over white noise with UV-22, which boosts level significantly above 15 kHz. The dither remained audible at all times, and I couldn’t hear any appreciable improvement in low-level signal behavior. On the plus side, it worked well.

Sonic Solutions’ Turbo-Bit

This one was a curiosity and, frankly, underwhelming. The noise was more audible, as a function of its shaping, than the other dithers, and it tended to mask the music more. There was general consensus that it was the least desirable of the dithers presented.

Putting All The Bits Together

I figured, after spending my time at 100X resolution (or, listening 40 dB down), that I’d really be able to hear this stuff at normal levels, now that I was grooved in to what to listen for. At 100X, it was all sooooo obvious. Actually, Toby had been concerned that most people wouldn’t be able to pick out stuff, but I’ve got to say, it really seemed clear to me.

Anyway, 40 dB is a heap ‘o level change. Try setting your mixer to unity and adjust the monitor level so that things sound nice and sort of “reference-level loud.” Then pull back the mixer faders to -40 and listen again. It’s a long, long way down. Yep, no doubt about it. It certainly is a whole lot softer.

So, when we went back to reference listening levels, we were, in effect, turning down the Least Significant Bit and the sound of dither by 40 dB. Guess what. It all disappeared. Man, I couldn’t hear the different dithers, the funny crunchy stuff approaching digital black, any of that stuff. I think I could hear a slight difference, still, between truncated 16-bit and 20-bit. That was about it. The veil? Well, maybe. Sort of. Yeah, there might’ve been a veil there. Yeah. Not sure, though. Maybe.

Brad Meyer, one of the other listeners, also pointed out that the 16-bit truncated signal is not representative of what CDs sound like today, but rather is a considerably worse signal than what is served up at your favorite record store.

What does it all mean? As Scott Hull pointed out in my mastering engineer interview with him and Toby last year, it is useful and informative to work at a greater resolution than will be present in the final product. It makes for a better final product. When the opportunity presents itself to work at 20 bits or more, avail yourself.

However, the final 16-bit product available over the counter today is really quite good, and the audible difference between it and a 20-bit recording is actually quite small. From my viewpoint, the difference pales into insignificance in comparison to the differences that occur when you move two feet in any direction while listening to your favorite monitors. Or move a mic. Or add a compressor. You know, creative things.

Happy dithering!