Thursday, February 14, 2013

Lousy Sound

I have noticed a trend -- lousy sound. I have been a music lover and audiophile (not audiophool, which is unscientific superstition, like anthropogenic climate change) since high school.

Listening to Pandora, I was struck by a new kind of distortion. It's an obnoxiousness in the upper midrange, almost like a harsh, noise modulation being kicked up by things like electric guitars and vocals. It isn't harmonic, and it isn't normal intermodulation. And no, it's not this.

What a long strange trip it's been, getting to the bottom of it. At first, I figured it was Pandora's standard AAC bitrate. So I paid for the subscription version so that I could get the high quality feeds. No joy. Besides, AAC actually sounds really good at reasonable bitrates.

So then I suspected my speakers. Maybe they were just obnoxious, so I started listening with headphones -- three different ones. No joy. Next was the sound card, so I tried a number of different ones on various machines I have at my disposal. They all exhibited the same obnoxiousness.

Heathkit AA-1214
Crossover distortion in my amplifier? I replaced the finals in it a while back, because they went into Vce breakdown and burned up. I used the original part, but with a higher breakdown voltage that wasn't available in 1972. Did I get the bias wrong? It turns out, one of the bias resistors had changed value. So I re-biased both channels. Distortion measurements are well within their original spec. Besides, this amp has never sounded bad to me in the 40 years since I first assembled it. The obnoxiousness persists.

My ears? Maybe... I do have tinnitus... but no, it doesn't happen on every song. I don't notice it with my old personal CD or vinyl collections (although the distortion does sound somewhat similar to vinyl damage from a worn out stylus). But I have been familiar with that phenomenon since the early '70s, when my ears were brilliant. Although I miss with great anguish the acuity of my youth, I do know how to listen. This ain't it.

To reiterate: it doesn't happen on every song. I have heard of the loudness wars that CD producers have been engaged in. It's the same idiotic loudness wars that broadcasters have been engaged in for many years, except it's even more idiotic when we're dealing with a medium that so doesn't need any additional processing to sound good. There's more than enough headroom on a CD for the dynamic range of any music known to humans. There's no reason to pack it all into the top 6 dB of the 96 dB available.

Not only do they compress the snot out of it, they clip it too. If this clipping occurs before the A/D converters, the higher harmonics that would cause aliasing get filtered out. But if they clip after the A/D converters, it will cause aliasing, and aliasing sounds nasty.

Aliasing has no correlation with the harmonic structure of the music (in fact, it's inverted). But it will kick up as noise whenever the digital clipping occurs. I can't imagine that recording engineers and record producers would be so stupid as to clip in the digital domain (without using the proper anti-aliasing filters), but maybe they do. If anybody knows, please drop me a line, or post a comment here.

I would not expect this obnoxiousness to occur with the classics, but with so many classics being remastered and reissued, I expect that the new releases of old material are also getting the same treatment. It certainly sounds like it is. Even our proud legacy is being mutilated.

Thursday, February 7, 2013

Realtek Sound Improving?

A favorite pastime among gamers and audiophiles is to bash Realtek audio codecs. This may have been justified at one time. A build I did several years ago using an Intel D945GCCR motherboard with the Realtek ALC883 audio codec certainly deserved the drubbing. Even an old Sound Blaster 16 PCI was a dramatic improvement.

But a recent build using an Intel DH67BL motherboard with the Realtek ALC892 audio codec was a very pleasant surprise. In A-B listening tests, I was unable to discern the difference between the Realtek and the Asus Xonar DG (a very well regarded audio board).

It's possible that the ALC892 sounds better than the ALC883 chipset. However, the published specs of the two chipsets are too close to quibble, and frankly the horribleness of the sound of the D945GCCR motherboard would have made any specification pointless. The thing that curled my eyelids was the harsh upper midrange distortion, which if anyone measured it, would have pegged the needle. It was that nasty.

You want to know what I think? I think both Realtek chipsets are just fine. I think Realtek is a victim of industry-wide poor motherboard design. I think the problem is with the analog portion of the D945GCCR motherboard, which was designed by Intel. Intel are digital logic designers, not analog audio designers. I think they botched it with the D945GCCR, but they finally got it right with the DH67BL. The former requires an add-in sound card. The latter really is a pleasure to listen to.

I think Realtek's problem is that it's their face on the audio subsystem -- their volume controls, equalizers, effects controllers, media players (if you use them). If Realtek were smart (or a bit smarter anyway), they would let the motherboard manufacturers put their logo on those apps. Then the blame (or praise) for the sound quality would go to the board manufacturer, which actually has more to do with the sound quality anyway.

The sound really depends on the board layout, the analog design and choice of components, whereas if you get the audio codecs right, they're right -- period, end of story. It really isn't any more expensive to make a good audio codec than a poor one, so might as well make a good one. If you get a justifiable reputation for making poor audio codecs, your bottom line will suffer terminally, regardless.

Now, having said all that, I should point out that the Realtek chipsets are not as quiet as the Xonar (with noise levels in the -90 dB range, the Realteks are comparable to 16-bit conversion), whereas the Xonar approaches -120 dB, which consistent with 20 ~ 24-bit conversion. They all claim 24-bit resolution, but only the Xonar approaches actual 24-bit performance.

Make no mistake though, unless you're using this for professional studio work, 16-bits is far more resolution than you'll be able to discern with your naked ear in anything but an anechoic recording studio. As a point of reference, the noise level of a good vinyl analog recording corresponds to about 8 ~ 10-bit resolution.