Re: dynamic range and sample bit depth (James Johnston )


Subject: Re: dynamic range and sample bit depth
From:    James Johnston  <audioskeptic@xxxxxxxx>
Date:    Wed, 3 Dec 2014 23:34:27 -0800
List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>

--001a113f1c587e1d5b05095efd16 Content-Type: text/plain; charset=UTF-8 I think Dick has the truth of it here. For capture (i.e. recording) it's nice to have at least 20 bits, it gives you lots of editing room and processing room. For home delivery, let us consider: Dithered PCM is white noise. The white noise due to air being made of molecules (and in the case of argon, atoms) is somewhere in excess of 6dB SPL at the ear drum, probably circa 8dB SPL. Jont and David can argue over the exact number. For now, I'll err on the side of caution and say 6. A really good stereo can hit about 110dB SPL, unless it's a sound reinforcement system. Now, subtracting the two is not quite valid, I am raising the noise at the quiet end by some 4dB (assuming proper dither), but let's do that just for a good estimate. 110-6=104. 104dB/6.02 db/bit = 17.2 bits. So let's call that 18 bits. It is likely that for all but the most extraordinary circumstances, 18 bits is plenty. I'm not including the noise in a usual home listening room here, which will push the lower end way up. Now, for acquisition, if you have something captured at full intensity at 20 bits, but you need to drop it by 48dB in order to make the mix right, what's the problem, its noise floor is way, way under the 18 bit delivery described above, which is, frankly, gross, mass overkill. For reasons that I don't want to go into in this thread, involving signal processing issues, filtering issues, etc, I would prefer 64kHz at 20 bits for delivery, that's only to make sure I get the 18 bits to the DAC. But how many of us have a system with 104dB dynamic range? So... I think the original question was sincere, but somewhat confused. This, of course, assumes proper TPD on the reduction to 18 bits. Don't dither, and you've done it wrong. JJ --001a113f1c587e1d5b05095efd16 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable <div dir=3D"ltr"><div><div><div><div><div><div><div><div><div>I think Dick = has the truth of it here.<br><br></div>For capture (i.e. recording) it&#39;= s nice to have at least 20 bits, it gives you lots of editing room and proc= essing room.<br><br></div>For home delivery, let us consider:=C2=A0 Dithere= d PCM is white noise. The white noise due to air being made of molecules (a= nd in the case of argon, atoms) is somewhere in excess of 6dB SPL at the ea= r drum, probably circa 8dB SPL. Jont and David can argue over the exact num= ber.=C2=A0 For now, I&#39;ll err on the side of caution and say 6.<br><br><= /div>A really good stereo can hit about 110dB SPL, unless it&#39;s a sound = reinforcement system.<br><br></div>Now, subtracting the two is not quite va= lid, I am raising the noise at the quiet end by some 4dB (assuming proper d= ither), but let&#39;s do that just for a good estimate.=C2=A0 110-6=3D104. = 104dB/6.02 db/bit =3D 17.2 bits. So let&#39;s call that 18 bits.<br><br>It = is likely that for all but the most extraordinary circumstances, 18 bits is= plenty.=C2=A0 I&#39;m not including the noise in a usual home listening ro= om here, which will push the lower end way up.<br><br></div>Now, for acquis= ition, if you have something captured at full intensity at 20 bits, but you= need to drop it by 48dB in order to make the mix right, what&#39;s the pro= blem, its noise floor is way, way under the 18 bit delivery described above= , which is, frankly, gross, mass overkill.<br><br></div>For reasons that I = don&#39;t want to go into in this thread, involving signal processing issue= s, filtering issues, etc, I would prefer 64kHz at 20 bits for delivery, tha= t&#39;s only to make sure I get the 18 bits to the DAC.<br><br></div>But ho= w many of us have a system with 104dB dynamic range?<br><br>So... I think t= he original question was sincere, but somewhat confused.<br><br></div>This,= of course, assumes proper TPD on the reduction to 18 bits. Don&#39;t dithe= r, and you&#39;ve done it wrong.<br><br></div>JJ<br><div><div><div><div><di= v><div><div><div><div><div><div>=C2=A0</div></div></div></div></div></div><= /div></div></div></div></div></div> --001a113f1c587e1d5b05095efd16--


This message came from the mail archive
http://www.auditory.org/postings/2014/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University