Restoring audio from vinyl for SQ, QS decoding.

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

See Why Audio

Member
Joined
Apr 25, 2007
Messages
28
Hi all, I 'm not sure if I already introduced myself to any of you on this forum, new computer, no saved favourites and stuff. But anyway here I am. Colin Young is the name, restoring audio from vinyl sources is the game.

I have been using the SQ script found elsewhere on this forum to decode a few SQ albums recently and I have to say I am amazed at the quality of the sound that is achievable. Compared to my distant memories of listening to SQ records 'live' through a decoder (although never a Tate) it's a whole new ball game. What fun!
My specialty is restoring audio from vinyl sources. I have opinions and some advice on restoration that you might choose to take, or not as you wish because there is no absolute right and wrong but I’ll give you what I feel is ‘right’. Forgive me if I go over any old ground or bore anyone! Please ask away if you have questions, I will get around to replying – promise.

I use Adobe Audition 1.5

Further to the question; should you do audio clean up before or after decoding? My personal view is emphatically do it before decoding but be careful.
Damaged vinyl? Hunt for a cleaner copy of the vinyl rather than fart around restoring severely damaged discs. Life’s too short.
Like computers: put rubbish in, get rubbish out. No-brainer: only recordings made on a well set-up, high quality deck/arm/cartridge will achieve good results from a SQ decode. The SQ system requires a very phase-accurate signal to work properly. A mis-tracking cartridge is losing phase info, as is a cartridge that is badly aligned. The things that make SQ work are the most tricky to replay from vinyl: Accurate phase and accurate tracking of transients. You need a good deck. For example if your system produces only a ‘hash’ kind of high frequency output (lots of content but no finesse, like a Stanton DJ cartridge on a SL 1210, say), you have lost the phase and the SQ system will not work properly on those frequencies. The result will be more bleed-through or crosstalk in the higher frequencies.
The flatter the frequency response of the cartridge, the more perceived accuracy in the final separation. An SQ decode I made with a moving magnet cartridge seemed nowhere near as well separated as a subsequent decode using the same record and a nicer moving coil cartridge although I can’t be sure that’s anything more than snobbery ;-)

Like others, I found the centre-channel extractor in AA 1.5 did weird things to bigger pops and clicks, especially those that were only on one channel, making them virtually impossible to fix later on. In most cases it made them 'longer' by introducing false overhang so it makes sense to tackle them before decoding. The only things that affect the decode are those actions that mess with the phase or channel separation, so your declicking and restoration should avoid any techniques that do either of those things unless you have chosen to compromise, as I will detail below:

Here’s my basic A to Z for restoration to SQ decode: I’m going to assume you all know your way around AA 1.5. If you don’t, umm, well...

(I always work nowadays at 96khz/32bit)
1) Record the waveform as a whole side of the record. Make sure the recording level is well short of peaking out. Save it. Only edit and save to separate tracks as a last step if at all.
2) Repeat for side 2, 3, 4 etc.
3) Using the FFT filter, Remove frequencies below 12 Hz. Create a preset for this starting at 12 Hz and rolling off downwards to 0Db at 8 Hz and below. This is controversial for some people but I find that it makes editing a lot easier by ‘flattening’ the waveform while only removing stuff that is, arguably, inaudible. It also makes for less glitches at track changes. Warning! The FFT filter doesn’t work properly for 192Khz recordings in AA 1.5 – I don’t know why or if later editions of AA fixed this problem. There is an alternative and that is to use the marquee tool to select (carefully) the region below 12 Hz and mix to mono because almost all the low frequency information is ‘flattened’ by the resulting phase cancellation. Save the file.
4) Fix large pops and clicks. (See the methods below). Yawn. Save the file.
5) Optional: Run a fine declicking script (see below) on areas you think need it. Save the file. (OK you get the message!)
6) Optional: Manually remove remaining clicks. (see more below for advice).
7) Choose final amplification*: Check if all ‘sides’ were mastered equal in amplitude – most albums are, some are not. Tip: avoid the ‘normalize’ function. Instead use the ‘amplify’ function and click the ‘find normalize’ button. It is way more accurate a process. Note the settings and repeat across an album once you have made sure it’s all recorded the same amplitude.
*I use MLP lossless packing software for preparing my final files before burning my DVD-A’s and I can down-mix the levels in that so I tend to save the master WAVs at 70% before decoding but of course you would probably normally save the file at 70% amplitude before most decodes anyway. Aaaaand, save the file.
8) Decode and burn DVD-a, etc.

Those are the basics, here, in more detail but in no particular order are the finer things:

For reasons of ergonomics and to make things a LOT easier and faster, I would recommend anyone who wants to do a lot of restoration to create a set of single-key keyboard shortcuts for the following: (this is for a normal lay-out qwerty keyboard - I'm right handed too !)
1) 'Fix single click now' (I chose ‘D’).
2) Toggle between waveform view and spectrographic (I chose ‘V’).
3) Three keys for declicking in increasing severity, a sort of ‘light’, ‘medium’ and ‘heavy’ choice if you get my drift. (I chose G, H, J) I don't have presets for you because all recording equipment is different, you need to find your own!
4) Zoom to selection. (I chose ‘Enter’ on the number pad and I wish I wish I wish there was a function on the mouse for it! There isn’t – I tried..)
Then, get used to, and learn to love the ‘marquee’ tool in spectrographic view. It allows you to select exactly how much of the frequency range (height) and across how much sample-time (width) to work on, allowing the single-click-fixer algorithm to assess and fix the click with the minimum of unwanted artefacts remaining.

For really BIG pops and clicks, you must compromise: decide whether you want a momentary loss of channel separation or an audible residue. IMHO it's better to suffer the momentary loss of separation, because it’s harder to detect while listening, but that's just my view. The trade off is this: the longer the sample-time you select, in order to make the click fixing more effective, the more the possible loss of separation in the final decode. This is because the algorithm is reconstructing the click from surrounding information, possibly at the expense of phase information (sometimes it guesses right, other times not – it’s unpredictable and depends on the nature of the waveform). On the other hand, the shorter the sample time you select, the less info the algorithm has to work with and the more thump-style residue you will have remaining. Too long a selection with fix-single-click tends to create a ‘drop-out’ so be careful here – you’ll soon get used to it.
The best compromise I have found is to use the spectrographic view and the ‘marquee’ tool to select a shorter sample time for the upper frequencies of the click (use the fix-single-click now) and repeat using a longer sample-time for the remaining lower frequencies (again use the fix-single-click now) – slightly overlapping the two steps in frequencies and biasing your lower frequency selection to the click and slightly after it to take account of the inevitable low frequency overhang as the stylus settled back from the click. (another reason for using a really good turntable which should produce less of this overhang).
Tip: mixing to mono hides a great deal of nastiness from clicks that are out of phase but you will destroy the SQ phase for that instant – you decide what’s best.
Tip: Employing the marquee tool when mixing to mono allows you to sneakily mix only the affected frequencies.

For clicks that exist only on one channel: you can sometimes ‘steal’ the ‘good’ channel and paste over the ‘bad’. Again phase might be lost but it’s surely a better thing than a bang! The marquee allows you to select only the affected frequencies for doing this.
I have created two keyboard shortcuts (comma = left channel and period = right channel, optional with shift) that first fix-single-click then mix to mono or if I use the shift key, do it in reverse order. Again the marquee tool allows you to do this on only the affected frequencies. Tip: such as what a declicker leaves, i.e. only a low frequency thump.

That’s it for really BIG bangs. Smaller clicks in practice should present no trouble to the SQ decode, provided you are careful in how you remove them.

For recordings that require the removal of very fine clicks – the kind that ‘dust’ the signal rather than ‘crackle’, I sometimes use a script that employs the noise reduction algorithm (whilst not actually removing any noise), alongside the automatic click removal, to remove only the tiniest clicks. What it does is to sample the noise, keep only the noise which you then declick and paste over the results inverted over the original waveform, keeping the noise but removing the very tiniest clicks. Magic. You do run the risk of removing very tiny natural clicks too but it’s much safer than just doing a normal declick. Things like sax, trumpet and other similar sounds with natural clicks in them, that would normally suffer at the hands of the auto-declickers, come out stunningly clean with their (larger) natural clicks intact. Slightly bigger clicks are better removed manually and you would do that after the declicking script finishes.
I’ve added this script below for those that are interested. Note: it takes not quite as long as a decode so it adds a lot of time to the project.

Then it’s just a case of going through the rest of the track listening out for stuff that shouldn’t be there, zooming in and fixing it. I do this in spectrographic view and, using the three keyboard shortcuts I created for declicking at three increasing levels of severity. What this allows me to do is; locate the area that needs declicking either by hearing it or/and seeing it, use the marquee tool to select only that area that visibly needs declicking and use one of the three keyboard shortcuts to ‘zap’ the clicks, starting with the mildest and only moving to more severe settings if the click(s) remains, or use the single click fix. What do I choose to fix? Me, I choose anything that annoys me. I work on the theory that if it doesn’t annoy you it should probably be left intact. Natural clicks should be left intact anyway. How to tell the difference? A lot easier than you might think but you need to get used to what they look like. They are easier to identify the closer you zoom.

Finally it is worth noting that if any tracks are ‘segued’ or continuous with no gap they should be treated as ‘whole tracks’ and processed as one file rather than split up before decoding. That applies to the declicking script as well as decoding SQ. The centre channel extractor has a little overhang and also takes time to ‘kick in’. If the music has natural silences between tracks, it’s safe to process each track individually (taking care to preserve the volume level across ‘sides’.) and use fades to get to and from silence.

I know I could have described the above in less words but I never was much good at précis!

I would love to provide more tips and tricks so please feel free to communicate with me!
And also feel free to hire me ;-)
Anyone care to do a little comparison of my SQ work against what they feel to be the best they got? I’m up for a challenge!

Laters all,

Colin AKA See Why Audio

PS N Wilkes, I lost your number!

Fine declicking script:
This script employs the Noise Reduction and Automatic declick functionality of Adobe Audition 1.5. It works by declicking only the saved noise of a file and pasting the results, inverted, on top of the original file, replacing the noise but phase cancelling the very fine clicks that were removed when declicking the saved noise. Simple!

Conditions:
File is stereo wav 96khz/24 or 32bit
File exists: C:\Program Files\Adobe\Audition 1.5\fft_files\younglove_decrackle_96.fft (you can edit line 11 of the script to change this directory or file name)
Script works on either whole file (as part of a batch process) or highlighted section on current open file.
Script works on highlighted area in spectrographic view and marquee tool selection.

Instructions:
Add script to your script collection or create a new collection.
Open audio file or add to batch process list.
Find area of file that has noise, but not including the very fine clicks, with no audio signal. Select it.
Open Noise Reduction Dialog box, get noise profile and save as C:\Program Files\Adobe\Audition 1.5\fft_files\younglove_decrackle_96.fft. (or your alternative as defined in line 11 of the script)
Close NR dialog box
Highlight area to declick
Run script

Copy all below------

Title: younglove_declick_96
Description:
Mode: 4
Undo: 1

Selected: 0 to 397766 scaled 397766 SR 96000
Freq: Off
cmd: Channel Both

Selected: 0 to 397766 scaled 397766 SR 96000
Freq: Off
Comment: Noise Reduction\Noise Reduction
cmd: {974C27BF-7D77-4136-B697-0941F91251D7}
1: 1
2: 0
3: C:\Program Files\Adobe\Audition 1.5\fft_files\younglove_decrackle_96.fft
4: 8192
5: 400
6: 19
7: 96000
8: 2
9: 32
10: 8
11: 4.48699e-007
12: 0.0035547
13: 2.77803e-005
14: 0
15: 400
16: 0
17: 0
18: 2
19: 2
20: 0
21: 1000
22: 16384
23: 1000
24: 1
25: 1
26: 1
27: 0
28: 0
29: 22
30: 65
31: 0

Freq: Off
cmd: Copy
1: 0

Freq: Off
Comment: Noise Reduction\Click/Pop Eliminator
cmd: {DC611066-C59B-4551-879D-C1218E942EDF}
1: 30
2: 32
3: 12
4: 9
5: 10
6: 0
7: 40
8: 1
9: 209
10: 1
11: 12
12: -12
13: -50
14: 8
15: 1
16: 18
17: 0
18: 15
19: 28
20: -24
21: 0
22: 0
23: 0
24: 0
25: 0
26: 0
27: 0
28: 0
29: 0
30: 0
31: 0
32: 0
33: 0
34: 0
35: 0
36: 0
37: 0
38: 0
39: 0
40: 0
41: 0
42: 0
43: 0
44: 0
45: 0
46: 0
47: 0
48: 0
49: 0
50: 0
51: 0
52: 0
53: 1
54: 60
55: 0
56: 0
57: 14
58: 20
59: 1

Freq: Off
cmd: Paste Special
1: -1
2: -1
3: 3
4: 1
5: 0
6: 0
7:
8: 0
9: 0

Freq: Off
cmd: Copy
1: 0

Freq: Off
cmd: Undo

Freq: Off
cmd: Undo

Freq: Off
cmd: Undo

Freq: Off
cmd: Paste Special
1: 1
2: 1
3: 3
4: 1
5: 0
6: 0
7:
8: 0
9: 0

Freq: Off
End:
 
Can't wait to give this a try. Thanks for all the info Colin!! And welcome to the forum.
 
Back
Top