Fm Sonic Scoop: Why All We Know About Bit Depth is Wrong

Music as represented in 1's and 0's. Discuss anything pertaining to D hardware - CD, DVD, SACD, DAC, etc.

Fm Sonic Scoop: Why All We Know About Bit Depth is Wrong

Postby rtsyrtsy » Sun Jun 14, 2015 8:29 am

Well, granted that the headline is meant to get clicks, the content is actually quite intriguing. I'd love to hear the thoughts of our more experienced digital experts on WS.

http://www.sonicscoop.com/2013/08/29/why-almost-everything-you-thought-you-knew-about-bit-depth-is-probably-wrong/#sthash.ReUqugmW.EVsUabWf.dpbs
User avatar
rtsyrtsy
Moderator
 
Posts: 3908
Joined: Tue Jan 28, 2003 5:25 am

Re: Fm Sonic Scoop: Why All We Know About Bit Depth is Wrong

Postby JackD201 » Sun Jun 14, 2015 8:09 pm

Sadly the author left out one very important aspect. He assumes like the dude at Xiph that the voltage steps are the same. This is not the case. More sad is that when pressed on this issue the advocates clam up and divert the topic. The formulae for voltage step can be found on wiki! When people read that the difference in dynamic range is all you get when you go from bit to more bits the assumption is that you are simply piling on more data points. The fact is that the data points are increased in between as well. This is what gives higher bit rates more resolution. In other words one needs to use less noise to fill in the gaps (dither). This drops aliasing errors.

It is mentioned in the article that the type of dither and filters matters a lot. That is very true. My current DAC allows for three different filter types. Many software players like Amarra will allow you to choose types of dither as well. All this is audible. This is the chink in the argument. While digital is not a stair step, as is commonly believed, and that a simple sine wave can be almost perfectly replicated, in the case of complex compound sines errors happen. Largest of which occurs when the chip must "guess" the voltage values to be placed between the binary command points. I am convinced that the math of Nyqvist and Shannon is solid. The implementation so far is is not. So, while it is true that by the time you get to 16bits the difference between that and 24 is very small compared to say 8 and 12. It is definitely there however and is most apparent in low level signals like ambience and decays, the stuff that is often masked by noise with lower bit rates. It is not only about dynamic range. THAT is the myth that needs to be busted. Something that recording professionals worth their salt know.

The crazy thing is that the industry has been pushing sampling rate as the factor. 24/96 must be inferior to 24/192 for example. That or up sampling makes everything sound better. This is a case to case thing because sampling rate is not about extension limits per se but rather about what filters can be employed. Higher sampling rates allow filters with values high enough that fold back is outside the audible range. A 16/44 brick wall filter for example can ring at 8kHz. Sampling rate is a side issue unlike word length which is where resolution lives.

*this applies to multibit PCM, sampling rate for PWM (D-S/DSD/DXD) is another matter altogether
User avatar
JackD201
Immortal
 
Posts: 10024
Joined: Sat Oct 09, 2004 12:10 am
Location: Bozania

Re: Fm Sonic Scoop: Why All We Know About Bit Depth is Wrong

Postby rtsyrtsy » Wed Jun 17, 2015 10:10 pm

JackD! Well, I asked for an experienced opinion and boy did I get one.

Thanks for this. I may need to read this 10 more times to get the gist of what you are saying. :-) And I need some aspirin after the first reading. :devil:

Seriously, I have stretches when free Spotify is good enough and stretches when nothing less than lovingly ripped FLAC will do. I don't have DSD software and hardware so for now, ignorance is bliss!

I did try the Tidal test and if I remember correctly, got 4 out of 5 correct--and I was really on the fence on the one I got wrong. Net, it is audible (to me and in my system) but not quite night and day. I hear it in the higher frequencies and as you pointed out, during the quiet passages. It's more a nuance thing than an outright, in your face sort of difference.

Maybe my ears are just ageing and not gracefully!
User avatar
rtsyrtsy
Moderator
 
Posts: 3908
Joined: Tue Jan 28, 2003 5:25 am

Re: Fm Sonic Scoop: Why All We Know About Bit Depth is Wrong

Postby JackD201 » Thu Jun 18, 2015 3:03 am

Past around 320kbps or 16/44.1 the differences can actually be easily ignored (with enough booze) :D Seriously though (hindi bagay ah) it get's really hard to tell the difference. Differences might be small enough to be masked by say the aircon running. Bigger differences are more likely due to the production of the tracks themselves.

Going back to the voltage steps the easy way to think of it is that 16bits has over 65k values and 24 bit has closer to 17 MILLION possible values. When people see how dynamic range goes up as well, we assume that we just keep piling on the output volts and that with 24bit you'll blow your head off or liquefy if you use it all ergo it is useless. If I understand correctly these values are split between zero and the highest output on the analog end of the conversion. This is a poor analogy but I think it kind of works. The output voltage is the size of your screen and the values are the size of the pixels. The pixels aren't the same size. By going from a 42" plasma, adding pixels didn't grow your screen into an iMax screen, the pixels just got smaller and denser, you still have a 42" screen. If you max out level, max level will still be, say, the typical 2v output.

I agree with much of the article actually especially the part about the advantages of recording on 24 bit vs 16 bit having done so myself. The added headroom towards 0dBfs means I can set levels higher and not be afraid of clipping as early. The effect is that I don't need a safety net or parachute in the form of a compressor/limiter. This alone makes a recording sound less processed because it is less processed.
User avatar
JackD201
Immortal
 
Posts: 10024
Joined: Sat Oct 09, 2004 12:10 am
Location: Bozania

Re: Fm Sonic Scoop: Why All We Know About Bit Depth is Wrong

Postby makinao » Thu Jun 18, 2015 7:43 pm

JackD201 wrote:I agree with much of the article actually especially the part about the advantages of recording on 24 bit vs 16 bit having done so myself. The added headroom towards 0dBfs means I can set levels higher and not be afraid of clipping as early. The effect is that I don't need a safety net or parachute in the form of a compressor/limiter. This alone makes a recording sound less processed because it is less processed.


That kinda reminds me of the difference between 15ips and 30ips back in analog tape recording days. When budget permitted (because 2" tape was EXPENSIVE) I would always opt for 30ips. More headroom, lower noise, and easier to edit.
User avatar
makinao
Citizen
Citizen
 
Posts: 170
Joined: Thu Aug 23, 2007 12:13 pm


Return to Digital

Who is online

Users browsing this forum: No registered users and 2 guests

cron