First, I'm not interested in debating formats, equipment, etc.I'm not looking to create reference quality archival needle drops.I rip at 192/24 to WAV and then plan on converting to FLAC.
I always record in floating point, then convert to 16 bit after everything else is done.Another suggestion is to produce a number of output files, one for each major step of processing (preferably with a naming scheme that identifies what was done). I never keep anything but the final output -- once the LP is totally finished, so those intermediate file take up disk space only temporarily.
I reject the commonly spewed FUD about quantization error caused by successive edits.
Since people have tried to tip-toe around this, I thought I would make myself more clear. Any decent software will process at higher than 16-bit and then dither back.
First, high pass filter to remove low frequency turntable noise like rumble. Common filters roll off below 20 Hz. Few LPs of the ~ 800 I recorded and cleaned up had any content below 32 Hz. Often 40 Hz is a better choice.Declick next. Noise reduction first will make it significantly more difficult to find (especially for programs to automatically find) smaller clicks and often harder to remove more obvious ones cleanly. I frequently do four or five passes of automated declicking, each with different settings, then do manual declicking on what I can hear. I use three basic settings for manual fixes. Since the settings depend very much on the application program, I can't suggest settings for you but you must start out with setting for the largest clicks and progress to the smaller and smaller.There have been cases of recordings so noisy that too many smaller clicks were masked by the background noise but became more audible after noise reduction. Therefore, at least some noise reduction before final declicking is sometime useful.The Younglove process is quite often useful . I use four variations, depending on the recording. This should be the last step of automated declicking.Noise reduction is next. Some LPs are fine without it but it is more often than not beneficial in my experience. This will depend very much on the music and the condition of the LP. I have found, contrary to much advice, that the same process can successfully be done on the entire recording at once instead of doing each track, or piece of track, with its own settings.Again, I often do more than one pass; two most frequently. Then, on some material, especially with significant fade in or fade out sections, I may do one or more passes on just those sections.Also, on particularly damaged or noisy recordings, one or more passes of the Younglove processing can usefully be applied for noise reduction, used on just those selected parts that need it.I don't do EQ manipulations but if you are intent on that, this is the time. Normalize for the last step, with one exception. Without arguing ay thing about the audibility of the quantization noise at different bit depths when there is musical content, the data is definitely different after processing operations. With simple test tones the differences are very obvious in spectral analysis and can be heard at higher volume levels.
The largest undesired signals and noise should be removed first to minimize the artifacts that any nonliniear processing stepts might add.
Then comes tics and pops due to dirt. Finally, remove hiss, as is possible while retaining the best possible fidelity.
...Take two copies of a normalized 24-bit vinyl capture: reduce the effective resolution of one copy to 14 bits and leave the other copy unchanged. Subject each copy to the same set of edits and ABX for differences.
Quote from: greynol on 11 October, 2016, 05:45:16 PM...Take two copies of a normalized 24-bit vinyl capture: reduce the effective resolution of one copy to 14 bits and leave the other copy unchanged. Subject each copy to the same set of edits and ABX for differences.Why would anyone record anything by his/her hearing skills? Aren't the question here about something else.If there's a need for post-processing the vinyl recording wouldn't it be better to have it done at higher sample rate and bit-depth? My work flow is:For Reaper, I have created a 'base project' (24/96 recording) which has all the needed pre and post plug-ins (RIAA, FreeG (fine tune L/R balance and gain if needed), HP (<20Hz, rumble), LP (remove >20kHz), SPAN (metering) and Hum (50Hz/60Hz) remover) included but only
...But if you lowpass above 20 kHz then 24/48 would be well sufficient for your recording (as it is for 99,9 percent of other cases). ...
Quote from: greynol on 11 October, 2016, 05:45:16 PM...Take two copies of a normalized 24-bit vinyl capture: reduce the effective resolution of one copy to 14 bits and leave the other copy unchanged. Subject each copy to the same set of edits and ABX for differences.Why would anyone record anything by his/her hearing skills?
If there's a need for post-processing the vinyl recording wouldn't it be better to have it done at higher sample rate and bit-depth?
This is not to say that there is something inherently wrong with keeping stuff at >16-bit and then reducing depth at the end (or not reducing the depth). However, claims of better sound quality are indeed subject to TOS8.
I know 44.1 can be/is tight for filtering without aliasing but at 48 kHz 4 kHz filtering space is not enough for you ?
People might consider whether content that has been low-passed at 17kHz can be ABXed before worrying about not having enough room for a transition band.
Doesn't this depend on filter implementation ?
Quote from: greynol on 13 October, 2016, 01:58:26 PMPeople might consider whether content that has been low-passed at 17kHz can be ABXed before worrying about not having enough room for a transition band.Doesn't this depend on filter implementation ?
Aren't the limitations of the human ear also relevant?Of course they are. I am fascinated when people start pontificating about filter types and slopes, illustrated with nicely drawn FR curves from theoretical simulations, and completely and perhaps conveniently forget to do any DBTs, Above we are presented with two frequency response curves, The supporting text seems to take it as revealed truth that they are distinguishable in an ABX test. This might be so, but is the reason for the differences above or below 10 KHz?
Filter: ON IIR Order 2 Coefficients 0.8760666490683719 1.752133298136744 0.8760666490683719 1.466061529573591 1.504266596273488 0.533938470426409Filter: OFF IIR Order 2 Coefficients 0.8760666490683719 1.752133298136744 0.8760666490683719 2.691575899010475 1.504266596273488 -0.6915758990104754Filter: OFF IIR Order 2 Coefficients 0.8760666490683719 1.752133298136744 0.8760666490683719 2.023192926460408 1.504266596273488 -0.02319292646040805Filter: OFF IIR Order 2 Coefficients 0.8359412478697531 0.1948669448559074 -0.112155492262995 1.0 -0.03000145304555539 -0.05134584649177911
http://www.musicdsp.org/files/Audio-EQ-Cookbook.txtI don't mind to do DBT but, here are the coefficients for those various filters shown in plot if someone wants to run ABX test