• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Dither Unnecessary for Acoustic Recordings?

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
If you are recording acoustic music using microphones, then the mic, the mic preamp, etc. add a certain level of noise. This noise is typically higher/louder than -96 dB. So in effect, the bottom 1-3 bits of each sample are already randomized for you by noise from the mics & mic preamps. You might say the signal already has "analog dither". In this case it seems unnecessary to digitally dither the recording - it would do nothing more than randomize an LSB that is already randomized. It wouldn't hurt, but it wouldn't help.
@j_j or any other of our technical experts care to elaborate?
 

DVDdoug

Major Contributor
Joined
May 27, 2021
Messages
2,916
Likes
3,831
Right. Most analog recordings are "self dithered" (unless you're downsampling to 8-bits).

And, at 16-bits and normal listening conditions you can't hear dither or the effects of dither anyway...
 
OP
MRC01

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
Sure, analog recordings would definitely qualify due to tape hiss and other noise.
Yet it seems that even digital recordings of acoustic music would be "self-dithered" due to the noise from the mics and mic preamps!
 
Last edited:

Frgirard

Major Contributor
Joined
Apr 2, 2021
Messages
1,737
Likes
1,040
If you are recording acoustic music using microphones, then the mic, the mic preamp, etc. add a certain level of noise. This noise is typically higher/louder than -96 dB. So in effect, the bottom 1-3 bits of each sample are already randomized for you by noise from the mics & mic preamps. You might say the signal already has "analog dither". In this case it seems unnecessary to digitally dither the recording - it would do nothing more than randomize an LSB that is already randomized. It wouldn't hurt, but it wouldn't help.
@j_j or any other of our technical experts care to elaborate?
No one records in 16 since the beginning. Decca start in the 70's in 18bit.
 
OP
MRC01

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
If you're recording in more than 16-bit, the noise from the mics & preamps covers even more of the least significant bits. So dither would seem even less necessary.
For example suppose the total noise on the mic feed is -60 dB. At 8-bit you would have to dither because this is below the LSB. At 10 bit or more you wouldn't have to dither because noise at -60 dB is already randomizing the lower bits for you.
PS: it wouldn't hurt to dither, as you'd be twiddling bits below the noise floor. But it shouldn't be necessary, as the noise already twiddled the bits for you.
 

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
The topic makes no sense. Dither (in 16 bit resolution) is added in the final step, after processing and mastering which is done at much higher resolution because of math operations to avoid truncating. Dither is added when finally converted to 16 bits and it is a necessary step.
 

JustAnandaDourEyedDude

Addicted to Fun and Learning
Joined
Apr 29, 2020
Messages
518
Likes
819
Location
USA
Dither is added when finally converted to 16 bits and it is a necessary step.
Am I interpreting this correctly to imply even if random noise were present in the PCM signal from bits 24 to say 14, uniformly distributed wrt frequency, and the signal is ultimately truncated to 16 bits, that subsequent dither is necessary? That the truncation removing bits 17 to 24 somehow introduces tones into the 16th bit that could be audible through the random noise in bits 16 through 14? What are the conditions on the musical signal itself for this risk of audibility of truncation without dithering to arise?
 
OP
MRC01

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
My understanding is that if random noise were present up to bit 14 (say, a mic preamp with SNR of 84 dB), then the bottom 3 bits of 16-bit are already randomized (bits 14, 15 and 16). You should be able to simply truncate bits 17 and lower. There is no point to randomizing bits that are already randomized.
If that is incorrect, please correct & explain.

Now, to @pma 's point: suppose your peak levels were only -12 dB so in post-processing you shift it left 2 bits (amplify it). You have introduced 2 new LSBs and you need to dither them to avoid quantization distortion.
 

digitalfrost

Major Contributor
Joined
Jul 22, 2018
Messages
1,521
Likes
3,086
Location
Palatinate, Germany
From my understanding, what dither does is it allows you to better store values below 1bit, that would otherwise be truncated. From this perspective adding correct dither should always be done, because you loose nothing but you gain accuracy.
 

JustAnandaDourEyedDude

Addicted to Fun and Learning
Joined
Apr 29, 2020
Messages
518
Likes
819
Location
USA
My understanding is that if random noise were present up to bit 14 (say, a mic preamp with SNR of 84 dB), then the bottom 3 bits of 16-bit are already randomized (bits 14, 15 and 16). You should be able to simply truncate bits 17 and lower. There is no point to randomizing bits that are already randomized.
If that is incorrect, please correct & explain.

Now, to @pma 's point: suppose your peak levels were only -12 dB so in post-processing you shift it left 2 bits (amplify it). You have introduced 2 new LSBs and you need to dither them to avoid quantization distortion.
Your reasoning is plausible. I did not understand the reason for pma's statement, which is why I asked the question about whether the mere truncation itself somehow unrandomizes the 16th bit. Thanks for explaining about the shift which introduces zeros into the last few LSBs and requires dither.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
No one records in 16 since the beginning. Decca start in the 70's in 18bit.

Apparently this is a reference to an in-house built digital recorder 18bit 48kHz, built by Decca.


All Denon's Nippon Columbia) PCM digital recording release in the late 70s were 14bit.
The Mitsubish X-80 was 16bit
The Soundstream recorder was 16 bit
The 3M 32 track digital recorder was 16 bit.

12,13,14 and 16bit covered all all the commercially sold rotary head PCM digital recorders as of 1979.
12,13,14 and 16bit covered all the commercially sold stationary head digital PCM digital recorders as of 1979.
 
Top Bottom