Editing of digitized vinyl music - Pros and Cons

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Editing of digitized vinyl music - Pros and Cons

During the discussions in another thread, related to my floating tangential tonearm, it came out that the digital editing of the digitized vinyl music is a quite disputable practice. After some suggestions, I open a new thread, devoted to the digital editing Pros and Cons.

First I must clarify that I apply digital editing of ripped vinyl music many years before the construction of my turntable with floating tonearm. I used it intensively with recordings made with classical pivoted tonearms.

It is good to define first the purpose of editing. I would formulate it like this: to make the digital copy sound as close as possible to the original audio material (the "master") cut as a vinyl disk . If we name:

"O" - the original sound material
"V" - the sound extracted from vinyl, and
"D" - the edited digital copy of V

we want to make, formally talking, the difference D-O as small as possible. In any case much smaller than V-O.

D - O << V - O

First of all, we have to list shortly the main types of differences between O and V. Later I will discuss in details any type separately.

1. V has an added number of clicks and pops.
2. V has an added disk surface noise.
3. V has some added very low frequency components.
4. V has a different frequency distribution compared to O.
5. V has a different channel balance compared to O.
6. We can consider also the hiss of older records, although it is often a part of O.

These are the main editable / correctable types of added sounds. We can mention also some additions that can not be corrected digitally:

a. The wow caused by the eccentricity of the disk grooves with respect to the axis of rotation. This can be minimized by careful disk centering.
b. The harmonic distortion components added by the cartridge. Not editable digitally. The only cure is a higher quality cartridge.
c. The harmonic distortion added by incorrect tracking geometry (tonearm length, tracking angles). Not editable digitally, but minimizable through careful tonearm/cartridge adjustments.

After listing of all the above negative addons, I will discuss in more details the editable ones and the methods of editing. I think to do it post-by-post and with no hurry .. :)

Meanwhile, please feel free to add any note / opinion / comment, not necessarily related to what I have posted.

Thank you.
 
Before starting the basic analysis, I want to clarify that using high quality digitizing is very essential for obtaining good editing results. Today I use WAV 48 kHz 32 bit. Some years ago I used experimentally 96 kHz recordings, but found that the benefits are quite small with respect to the 48 kHz version.

I record directly to Adobe Audition 3. Most of the editing procedures are a part of Audition 3. They have typically a lot of adjustable parameters, about which I will comment "in due time". But one of the most important procedures - this of click removal - is made by an external software - the wonderful ClickRepair of Brian Davies, USA. So, after some initial Audition 3 procedures*, the file is saved, de-clicked with the Brian program and opened again in Audition 3 for further editing.

The final product is archived in OGG format - the highest quality variable bitrate version, with peak values of recording bitrate approaching 600 kbps. The WAV version is used for incorporation to the MP4 clip sent to YouTube (where the clips since some time are 'emitted' in WEBM format). In special demo cases I use the lossless FLAC format.

* Brian suggests to run his program before any noise reduction procedure.
 
Removing of clicks and pops

Clicks and pops are the more severe 'addons' to the original sound. In most of the cases they totally 'kill' the original material in their time region, so we have nothing to 'preserve'. Except an idea of what was killed, given by the waveform in the proximity of the click - just before and just after. As a result, removing the click by some interpolation , based on the unaffected proximity is the best possible action.

First of all I have to declare my strong belief in clicks and pops 'usefulness'. Therefore, I will not discuss this point with anybody declaring something different. I will only follow with (silent) sarcasm this type of declarations and discussions.

A. The origins of clicks and pops.

The most common origin is some permanent mechanical disturbance in the disk surface, caused by:

- Too quick stylus drop down
- Malicious stylus dragging over the disk, caused by accidental push on the cartridge
- Cutting by some sharp 'third side' object.

Not permanent sources of clicks are the dust particles of various sizes. A good brushing of the disk before playback will reduce considerably the quantity and the strength of this type of clicks. Using a permanent brushing device is doing a good job too, as long as the dust felt or electrostatically attracted to the disk surface during the - say - 20 min of playback of one LP side may be quite significant.

Non permanent sources are also the electric shocks in the AC supply, when not enough filtered and reached the signal path. These shocks are mostly "pops" = clicks of longer duration, with higher presence of medium and low frequencies in their spectra.

B. The problems of false detection.

As one looks carefully some waveform, it is easy to recognize 'optically' the strong clicks. They have a remarkably short duration and look like thin and long needles even in the background of some high frequency content. If something is easy recognizable by the human eye, it is (or should be) also recognizable by some clever mathematical algorithm. In fact, most of the severe clicks are recognized and removed by all the software i have used.

Nevertheless, there are some specific cases, when the click removal algorithms create false detection: they remove some signal peaks that are not clicks, but parts of the original sound signal. The result in most of the cases are not just 'some loss', but actually some adding of scratch-like artifacts, clearly audible and very unpleasant.

Some typical cases of sound with potential risk for false detection:

- The sound of some brass instruments, usually when played in a 'sharp' mode. Trombones, trumpets (especially with mutes).
- The strong sibilant components of some human voices, more often female ones.
- Some percussion instruments with very sharp and short duration sounds.

The excellent Brian Davis software uses different tricks to copy with the cases of false click detection. It has special "modes" for Brass and Percussion. It uses some tricks, e.g. declicking also the time-inverted waveform, where some effects of the sharp signal front are not present. Or uses the quick repetition of the brass 'click-like' sounds, which is not typical for real clicks.

But ... Given that Brian has issued a very detailed manual on his ClickRepair program, I have to stop talking and recommend a careful reading of this manual to anybody deeply interested in this specific issue:

CRManual.pdf - Google Drive
 
Last edited:
Member
Joined 2014
Paid Member
My view on this is that it's entirely up to the ripper of the vinyl how much they want to post process and I would not argue with anyone's preference on this. Prefessional restorers will actually edit the sonogram but most of us can't spend days on a single track.



I do agree that getting the initial rip as clean of artificts as possible, including correct centering of the record is vital to getting a good end result. There were some tests on here to see if you could correct for eccentricity in software and the results were mixed and generally not an improvement.
 
Prefessional restorers will actually edit the sonogram but most of us can't spend days on a single track.

Hi. The implementation of my standard editing cycle lasts about 1 hour. But this is today. 15 years ago it needed more than 10 hours and I simply did not do it.

There were some tests on here to see if you could correct for eccentricity in software and the results were mixed and generally not an improvement.
I agree with you, but still: do you have some link describing these tests and, maybe, some musical files demonstrating the results?

I forgot to mention in the list of the general preliminary steps the use of stabilizers. The internal one is widely used today, but adding the extremal one - the ring - is doing a better job in lowering the disk warps and hence - the resonant tonearm vibrations and the trackability.
 
Last edited:
In theory it should be possible to reduce the harmonic distortion by modelling the non-linear channel, and dynamically characterizing the behaviour and compensating for it.

From V you can infer C, the output from the cartridge, and then G, the groove displacement. using a non-linear cartridge/stylus model in reverse. Tweaking the parameters of your model you can evaluate the G signal for intermodulation product density, or some other indicator of distortion in order to hunt for the minimum distortion - perhaps you have be selective with excerpts of the recording to get tractable results, but in theory you only need some estimator of distortion and some knobs to twiddle on the model.

The devil is in the details, and I'm sure there are a lot of details. It would be much simpler to do this with a test-disk with known reference signals, and then assume the model is constant (though parameterized by track radius). With real music the loud sections are usually not simple in structure, yet they are the bits that provide the information you need.
 
Removing the very low frequencies

This operation is, actually, the first of my standard sequence. But I started with clicks removal because I consider it very important.

The frequencies lower than 20 Hz normally do not belong to the original music material. Well, if some other types of sounds have been recorded, e.g. volcanic eruptions of a Bugatti Veyron power start, they may be present. But our old good science, the human acoustics, learns us that these frequencies are not audible. That's why we are considering as a standard object of interest the region 20 - 20000 Hz, and not because we are lasy to go down to 10 Hz or up to 30 kHz. Yes, measuring in wider frequency regions gives us additional assurance about the quality in the audible region.

Under usual conditions, the frequencies below 20 Hz are caused mainly by:

a. The platter bearings (rumble)
b. The disk warps, which activate tonearm resonance vibrations
c. Some external shocks, activating the turntable suspension vibrations.

It is a bad practice to neglect the origins of the very low frequencies as listed above and to rely to their digital suppression. But even after applying the best possible mechanical measures, we have a residual parasite sound material which is good to remove.

The Fast Fourier Transform filters of the Audition 3 and later allow the application of very sharp high-pass curves, with actual slopes (measured with white noise samples) even higher than 60 dB per octave. The main question is the selection of the turnout frequency, that where the signal goes down by -3 dB. In most cases I use 30 Hz. When Grand royal is present, i go down to 25 Hz. In the case of big Church organ music I select 20 Hz.

The removal of not audible frequencies is not a 'nonsense' (if not audible, why to remove?). We all know that the loudspeakers are the most not perfect components of the whole recording/reproduction chain. Even the most perfect ones have considerable distortion rates in their low frequency end. So, a noise of 15 Hz may be not audible, but it's second, third and so on harmonics ARE. Why to add listenable noise without any logical reason??

----

BTW, I follow with interest the discussion between Mark and billshurv.
 
Member
Joined 2014
Paid Member
The frequencies below 20Hz are generally not the 3 you listed. The main ones are
1. Center hole offset (eccentricity) at 0.55Hz
2. Arm cartridge resonance at 7-12Hz
3. Bearing drag/belt issues causing speed changes.


Attached is an analysis of a linn LP12 rather in need of a tuneup. The circular graph shows how frequency changes over a revolution and the X-Y is a frequency analysis. As you can see the 0.55Hz eccentricity is the biggest contributor. Why does this matter? Well because that modulates the wanted signal and cannot easily be removed. High pass filters will remove the signal, but the damage has been done at that point. This could be argued to be part of the 'vinyl sound' of course, but at least it can be addressed and minimised by careful setup.
 

Attachments

  • LP12.JPG
    LP12.JPG
    325.1 KB · Views: 159
The frequencies below 20Hz are generally not the 3 you listed. The main ones are
1. Center hole offset (eccentricity) at 0.55Hz
2. Arm cartridge resonance at 7-12Hz
3. Bearing drag/belt issues causing speed changes.

I mentioned your 2. in my b. I call 'tonearm resonance' exactly what you call "arm cartridge resonance".

b. The disk warps, which activate tonearm resonance vibrations

I think that the effects of the 0.55 Hz eccentricity movements are quite low, from the signal level point of view. The cartridge response goes down very quickly for frequencies below the resonance one.

Your 3. - yes, this is another source, but again of very low principal frequencies.
 
Member
Joined 2014
Paid Member
Sorry yes, I misread your 'b'
ref eccentricty B&K did measure this back in the 60s. The graph I showed is from an actual recording showing relative levels!


The eccentricity causes a large FM component close in. How audible that is depends but it is very easily measured and IMO somthing to be worth taking the effort to minimise if you are digitizing a record.
 
ref eccentricty B&K did measure this back in the 60s. The graph I showed is from an actual recording showing relative levels!

I am very surprised to know that the results of Brüel & Kjær refer to the cartridge signal and not to the groove displacement! A normal cartridge with frequency response going down to 0.55 Hz?? I can't believe this! May be they used some special type of measuring cartridge, responding to displacement (e.g. piezo) and not to velocity??
 
The cartridge response has nothing to do with it, the effective groove velocity varies at .55 Hz and the frequency of everything on the disk varies with it.

Are you talking about the wow caused by some eccentricity ±dR? Yes, every playback frequency will vary within ±dR/R with a period of 0.55 Hz. But this has nothing to do with the level of the cartridge signal with frequency 0.55 Hz.
 
But this has nothing to do with the level of the cartridge signal with frequency 0.55 Hz.

Neither does the plot Bill showed, this effect is clearly audible on piano recordings in particular. Take an opera LP you hate and widen the hole and play it very off center. The only point is the frequency response of the cartridge has nothing to do with it.
 
b. The disk warps, which activate tonearm resonance vibrations

Here, I assume you rip the LP for listening only.

If the disk warp excites tonearm resonance frequency. You can't track the LP at all. It means you can't even make a digital copy of the LP. Once you are able to make a digital copy of the LP. The arm/cartridge resonance frequency is not a problem at all. In your digital copy of the LP, you cut all the information under 30 Hz out. In fact, you are not cutting out the excited arm/cartridge resonance frequency caused by warps.

Here is a video to show you what happens when the arm/cartridge resonance frequency was excited.

YouTube
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.