IPB

Welcome Guest ( Log In | Register )

2 Pages V   1 2 >  
Reply to this topicStart new topic
Tape artifacts, still learning
Dogway
post Aug 15 2013, 15:59
Post #1





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



Hello, I am ripping some tapes and have some issues I don't know how to address. I post a chunk of the problematic audio.

RAW
half fixed (denoised for isolating purposes)
Different 'fixed' track showing the artifact.

It's like if the sound was broken, interrupted or distorted, can't tell, like gain drop peaks. The problem is it's not clipped. What is it and how can I fix it? There are also eventually channel fades...


I would also like to know how I can check if a track has DC offset or not, so I know when to apply.
I'm also in the beginning of some audio restoration solution and was aiming my eyes to the new RX3, I do video restoration and want to also add audio restoration to my arsenal. I would like to know if they have audio compressor and DC offset fixing too. I normally use audacity, but I would like to work without switching applications really.

Thank you!
Go to the top of the page
+Quote Post
DVDdoug
post Aug 15 2013, 18:13
Post #2





Group: Members
Posts: 2594
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



I can't listen to your files right now, because I'm at work. And I don't own Izotope RX....

RX has a good reputation and you probably can't beat it for general-purpose restoration restoration. I suggest you download the trial version.

You can get artifacts with noise reduction... Sometimes, "The cure can be worse than the disease". It's something you have to experiment with (maybe trying various settings) and then make a judgement if it's better with or without it.

If you have a very-small constant background-noise, noise reduction usually works pretty well. Or, if you have a single-pitch noise or hum, a notch filter can be very effective. Short-duration "clicks" and "pops" like vinyl record noise can often be repaired by deleting that part of the audio or by replacing the defect with the just-preceeding or just-folowing audio.

If the noise is bad (loud hiss or "static", a dog barking, or someone coughing/sneezing) there is often nothing you can do about it...

Professional recordings are still recorded in soundproof studios with high-quality low-noise equipment. There's no substitute for starting-out with a good recording. Even the best professional software can't fix a bad recording.

You can often make a good recording into a great recording. If you have a bad recording, you may be able to make some improvement, or there may be nothing you can do.

QUOTE
I would like to know if they have audio compressor
I don't think so. Compression is not for "repair". Izotope Ozone does have a compressor. You can find all kinds of VST compressors (free or professional-commercial) and they should run in Audacity. (Most VST plug-ins are not tested or specified to run on cheap or free audio editors or DAWs.)

QUOTE
I would also like to know how I can check if a track has DC offset or not, so I know when to apply....

DC offset is caused by a hardware problem. Usually a defect in your soundcard/recording interface.

... and DC offset fixing too.
I didn't see anything on the RX website about DC offset. But, there is a DC offset repair "effect" for Audacity, or you can simply use a high-pass filter (10 or 20 Hz), since DC is "zero Hz". A high-pass filter may be the best, most-foolproof, solution... I once used a DC-offset repair tool that made it worse!

DC offset repair will have no effect on the sound, other than removing the "tick" at the beginning/end (assuming it's working properly). Running it a 2nd time, or on a file that does not have any DC offset won't have any effect (sort-of like normalizing twice has no effect the 2nd time).

If you have DC Offset, silence will be a flat line in your waveform above or below zero or it may be obvious that the whole waveform is shifted up or down. If there is silence at the beginning or end of the file you'll hear a "tick" when the waveform suddenly jumps up or down from zero. A fade-in or fade-out will "hide" the offset and make it inaudible, since you will no longer have a "jump" in the waveform. So, it's possible to have DC offset that's hard to detect.

Some audio waveforms naturally have higher positive-peaks than negative-peaks, or vice-versa. That does NOT always indicate a DC offset.

QUOTE
I normally use audacity, but I would like to work without switceing applications really.
I underatand that, but sometimes it's best to use the RIGHT TOOL tool for the job. A screwdriver handle doesn't make a very good hammer! biggrin.gif I've got a lot of audio/video tools, many of them free. If I find a free tool that works better than what I normally use, I'm going to use it... If I have to pay, I may try to get-by with what I already have (and already know how to use). For example, I've been using GoldWave as my main audio editor for many years, but there are times Audacity can do things GoldWave can't.

This post has been edited by DVDdoug: Aug 15 2013, 18:41
Go to the top of the page
+Quote Post
Dogway
post Aug 15 2013, 18:46
Post #3





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



Yeah, I know compressors are not directly related to restoration, but they are necessary if you want to reduce sporadic high peaks that prevent you from normalizing the track (and hence improving the dynamic range).

About DC Offset I read that the old Sonic Studio could tell you if there was offset or not. It can be tricky as you say if there's not a silence part, offset is too small, or offset varies in the middle. I really don't want to be high-passing every track I run to. Either way it seems I won't be able to free myself from using 2 or more applications (?)

About the commented artifact, it's not broadband noise, I don't worry much about that since is so much discussed. I'm starting into audio restoration so first I need to know to identify the problems, it's like pops but it's not, I think they belong more to the "distortion" realm, but any insight on this is welcome when you have the time to listen. Thanks lot for the input!

BTW I made an analogy with video characteristics to help myself get into audio easier, is this true?:

CODE
Resolution:
ej: 1080p   48000Hz

Dynamic Range:
ej:Contrast   Gain

Depth:
ej:8bit   16bit

Detail:
ej: Sharpness   Cutoff


This post has been edited by Dogway: Aug 15 2013, 18:47
Go to the top of the page
+Quote Post
Dynamic
post Aug 15 2013, 19:04
Post #4





Group: Members
Posts: 821
Joined: 17-September 06
Member No.: 35307



I opened it in Audacity. That has a Normalize effect that can be used only to remove DC offset if any is present (uncheck the Normalize maximum amplitude to: ___dB). You can usually see if any offset is present using Waveform (dB) view having set the Dynamic Range in dB mode to about 96 dB.

A lot of software will analyze the signal to show DC offset and RMS levels and peak values, but I've not found this function in Audacity. (It was in CoolEdit 96 and CoolEdit 2000, which became Adobe Audition)

The RAW sample is slightly clipped (7 successive samples starting at 1s + 31654 samples for example). From the slope, I'd imagine it's not likely to be very audible, but it's a flattened peak, so it's a distortion.

There's ample dynamic range in 16-bit (way more than tape, which is usually nearer to about 9-bit equivalent) that you can turn down to leave 6 dB or 12 dB below full scale without worrying about digital noise or noise in your soundcard being anywhere near the level of tape noise present. You can then edit cleanly and adjust the levels later to make as loud as you prefer without hard clipping. The trouble is that some forms of editing (even filtering or reducing noise) can actually increase peak levels (e.g. the Gibbs phenomenon), so giving plenty of headroom will only help.

There are tools like the Tape Restore Live plugin for Winamp that you can use to reverse Dolby B NR (if it was used in the original - the crispness of the click at 18.4 seconds is hard to interpret compared to splashy cymbal sounds when Dolby is used but not reversed). It can also try to track and compensate for Azimuth, though this is best done by aligning your tape heads (assuming your deck allows you to). You can then save the output using Disk Writer output device or whatever it's called.

It sounds like the tape was dubbed from some source onto a blank tape as there's tape hiss from the tape material at the start then the sound turns on at 1.46s with a spike that's typical of a turn-on and the speed (i.e. pitch) seems to stabilise over the next second and the tape seems to get lined up and stretched over the recording head and tensioners. From then on it's fairly clean with the sort of stereo image stability and mild drop outs you tend to get from tapes. When I applied heavy noise reduction (too heavy - too much distortion creeping in) it seemed to reveal that the original tape had been recorded from another tape with an Automatic Gain Control riding the gain up and down, so the noise would periodically increase (and it was a markedly large and noticeable increase after I'd applied the noise profile heavily (default settings) having obtained the noise profile from the first 1.4 seconds).

On the fixed version I can hear further speed instability on the high 'string' or synth string sounds at around 15 seconds, resulting in brief pitch shifting (oddly I can't make this sound out in the raw version). It's the sort of thing that happens on some tape decks if someone pushes down on the play button harder and it increases the tape tension and stops it moving over the rollers at a consistent speed. I had an old Sony WM36 Walkman with Dolby B NR that sometimes struggled to play certain types of commercial tapes cleanly at full tension but I could hold the play button in slightly less than all the way using a small bench vice (vise in US English) and get consistent results when digitising with Dolby NR, low noise and good frequency response, and transferred some recordings to MiniDisc or PC pretty nicely, back in the day. I couldn't afford a HiFi tape deck at the time, but that was one of the best Walkman products (so long as you left the Graphic EQ at the 0 detents, otherwise that became very hissy).

I did also pick out a little stereo fluctuation in there, which is very tricky to improve greatly and might have come from the original tape-to-tape dubbing, not from your capture. I remember how fortunate I was that my own tape-to-tape dubbing was done on a Sharp Back-to-Back boombox that used a single motor to maintain speed stability, did not apply AGC to the dubbing signal and even had a fade-in envelope when starting dubbing, allowing clean transitions on mix tapes. The extra tape hiss was largely eradicated by Dolby B NR on my Walkman so I enjoyed pretty decent mix tapes.
Go to the top of the page
+Quote Post
Dynamic
post Aug 15 2013, 19:44
Post #5





Group: Members
Posts: 821
Joined: 17-September 06
Member No.: 35307



Oh, and by the way, listening on headphones (earbuds actually) at a little less than full volume, it sounded largely clean. When I listened on laptop speakers at first it sounded quite distorted, almost like heavy tape saturation, oddly.

Sample 2, I wasn't sure what it was supposed to sound like. In some ways it sounded a bit like AM radio that's not quite tuned in to the centre of the band so it lacking bass. It also sounded a bit rough, but maybe it was supposed to. Electronica is hard to second guess compared to acoustic instruments. Sample 2 was certainly far below clipping.

If you use a compressor on tape transfers, do that at the end of your processing. Any hiss reduction, EQ or drop-out repairs (and anything TapeRestore Live can do) should be applied to the cleanest possible signal. Personally, I've never wanted dynamic compression on my transfers, even in the case of a very quiet track. Its quietness was certainly artistically intended.

Another thing I found was able to disguise some of the flaws in some otherwise fairly good tape transfers (probably cleaner than yours in the first place) was a little stereo channel swapping reverb which adds a sort of 'produced' sheen to recordings without audible comb filtering. I'm planning to write a Nyquist plugin for Audacity when I get time to learn the language, and have something that does a little of what I'm aiming for, based on someone else's plugin. You can copy from the Codebox in this post on the Audacity Forums and save it as a plain text file called C:\Program Files\Audacity\Plug-Ins\Delayfli.ny. The defaults will be fairly good to give you an idea of the effect I'm aiming for, but I haven't optimised for memory use or speed or used the mult functions I'm aiming to use eventually to swirl the sound around. It's not great for incredibly sharp isolated impulses (they sound rather distinctly echoed), but you don't usually find those on tape, so it works OK for most music. Not wanting to veer too far off topic, I've also got this inkling that a variant of this plugin might be a fairly benign way to add a modicum of perceived loudness to files without robbing music of its short term dynamics, because it works similarly to Haas-delayed sound reinforcement at concert venues (which can give up to a 10 dB boost without greatly impacting the stereo localization of the precedent sound, apparently) and also Haas kickers in old studio control rooms. I should add that I'm in no way a fan of the Loudness War, however, before somebody theatens to kick my Haas for suggesting such a thing. wink.gif
Go to the top of the page
+Quote Post
Dogway
post Aug 15 2013, 20:01
Post #6





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



@Dynamic: I see, I was under the impression that recording at loudest as possible (without clipping) was a nice start, obviously, because sporadic peaks may arise eventually I on purpose opted to ignore these 7 samples here and there for the benefit of the whole. Actually it's not that one clipping the artifact I talk about, but on the flip side is good to know that I can actually go way lower on the gain side without losing dynamic range or honoring more device noise.

I really didn't want to discuss this sample in depth, I mean yes, but for learning purposes. It's only a radiocassette recording. I have many radio or even friend's sessions recorded onto tape that I would like to keep. Later on I will also start fixing VHS audio, so I think it's wiser by starting to learn what first intrigues me. What I think I refer in my OP is this "...the sort of stereo image stability and mild drop outs you tend to get from tapes". I want to fix those drops. Here for now (nothing extremely serious) I'm using a Sony WM-EX678 walkman for recording, it has Dolby B NR. I first cleaned slightly the head with some cotton buds and alcohol. I might have screwed the "restoration", that's why I included the RAW clip too, the "restored" ones are for easier listening of the artifacts I mentioned, otherwise it would be very hard to hear. It's like if even removing broadband noise, the voice or background synth are still very dirty... if you know what I mean...


PD: On you second post, the intention with compressor is basically to shave the 2 or 3 TOO HIGH track peaks down to logic levels, so the whole track can be pumped up. I wouldn't do this to clean sources, but this is a tape, and anything should be done, it's either a compressor or using a limiter (chopping the spikes off) producing clipping.

This post has been edited by Dogway: Aug 15 2013, 20:36
Go to the top of the page
+Quote Post
Dogway
post Aug 17 2013, 20:18
Post #7





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



Can somebody help?

What is this artifact called and how to fix it?

(it was previously hiss reduced and level equalized)
http://www.mediafire.com/listen/4z9csqv2zzd1kau/Artifact.wav
Go to the top of the page
+Quote Post
Specy
post Aug 17 2013, 21:08
Post #8





Group: Members
Posts: 22
Joined: 27-November 08
Member No.: 63320



It sounds like a very bad quality tape. I once got a tape from someone from the brand Yoko and it sounded a bit like this - as if the material on the tape is not spread evenly.

I must say that the audio quality is really really horrible, and the amount of hiss is really enormous. I don't know on what kind of recorder you recorded it, but I have restored my old tapes and they sound incredibly good, if there are no silent parts no-one would guess that they are listening to a more than 20-year old cassette.

I have checked your recording with Tape Restore Live and AZIMUTH is good, so it must be either the recording quality (if so, you're doomed....) or the walkman (!) that you used to record it.

If you seriously want to get decent quality, go to a second-hand store and get a good quality cassette deck. (Hint: Bring a tape with a beep tone, to test for jitter).

Btw: If you're going to do things like noise filtering, do all that BEFORE you remove any Dolby processing! Because Dolby removes hiss in quiet parts but not in loud parts, and noise reduction should be done on a signal with constant noise, not varying noise which you get after using Dolby.

This post has been edited by Specy: Aug 17 2013, 21:15
Go to the top of the page
+Quote Post
Dogway
post Aug 17 2013, 21:21
Post #9





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



This is not my standard tape quality, it's only a chunk to illustrate the problem. So you don't know how to call this artifact?

And you also recommend me to play my tapes with any dolby (dolby b for example) disabled? I had the impression that everything that could be done by hardware should be done first, as in (hardware denoising > software denoising).
That small clip had hiss/broadband noise removed, I don't think that's hiss, it doesn't even sound homogeneous.

This post has been edited by Dogway: Aug 17 2013, 21:54
Go to the top of the page
+Quote Post
Specy
post Aug 18 2013, 01:56
Post #10





Group: Members
Posts: 22
Joined: 27-November 08
Member No.: 63320



You are talking about the 'damaged tape'-like sound right? No idea what it would be called...

The reason why you need to fix the audio before applying Dolby B is this:
Dolby B acts differently depending on the amount of highs in a recording, and it needs to be calibrated to work correctly. If for example you lost a lot of highs (which could be the case with older tapes, and calibration may be off if you play back on a different cassette deck than you recorded on), then Dolby B will remove even more highs except on parts where the highs are very loud, leading to very annoying effects (too much highs reduction, but not always, so for example loud "S" sounds would stick out more than they should). Also if you have AZIMUTH issues (which doesn't seem to be the case in your sample), correcting it first restores highs, and yet again that helps Dolby to function correctly.

Hiss removal doesn't affect Dolby much (since the hiss level should be very low already, and of course Dolby was designed to handle hiss!), but it's still better to remove hiss before applying Dolby because the tape hiss is constant - but Dolby makes it non-constant. In louder parts (with more highs) of the recording, there will also be more hiss after applying Dolby; since you need to setup hiss removal for the lowest hiss levels to avoid too many artifacts, you won't remove most of the hiss when Dolby kicks in if you do it afterwards.

Note: If your original signal (before recording to tape) already contained hiss, THAT hiss should be removed AFTER applying Dolby! Because that hiss is non-constant on the tape, but constant after applying Dolby....

Hope this is all a bit clear smile.gif
Go to the top of the page
+Quote Post
AndyH-ha
post Aug 18 2013, 07:24
Post #11





Group: Members
Posts: 2207
Joined: 31-August 05
Member No.: 24222



I can't comment on your distortion mystery as I don't know what you are trying to get at, or how it differs from the original but I do have a few more general comments that I don't think have been covered already.

It has been pointed out that DC offset, which is a hardware problem, and best corrected there, can be taken care of in the recording by filtering out the lower frequencies. What hasn't been mentioned is that those lower frequencies should always be filtered out regardless, so you don't actually need to know where there is an offset to be corrected. I've never seen any sample recorded from cassette or LP that doesn't have a great deal of very low frequency noise, with not a bit of music in it.

I commonly use an FFT filter with a cutoff of 32 or 34Hz on my recordings. Music that low in frequency is rather rare; I sometimes use a significantly higher cutoff. The first file you provided, sample.flac does very nicely with a 100Hz cutoff.

You have a misconception about hard limiting. It does what you say you want compression for. It does not clip waveforms. I don't know just what the calculations are but it does a local negative amplification without significantly altering the waveform.

You can demonstrate this quite readily with a test tone such as a sine, triangular, or some other simple waveform. Start with it at 0dBfs, or any value you want. Hard limit at 3 or 6 dB lower and examine the resulting waveform. Zoom in so you can see the individual cycle. You will see just as perfect a waveform as the one you started with.

I downloaded two of your samples, the one I just mentioned and the third, sample2.flac. These took well over 10 minutes each so I felt disinclined to pursue them further. As I said, I don't know what your main question is directed towards. However, for the first file, it is easy to reduce the tape noise enough for it to be insignificant. If you aren't achieving that, you probably just don't yet know how to use the tools well enough.
Go to the top of the page
+Quote Post
slks
post Aug 18 2013, 08:52
Post #12





Group: Members
Posts: 400
Joined: 31-March 06
From: Houston, Texas
Member No.: 29046



I listened to the unprocessed recording and honestly I don't hear any major issues with it, besides what you'd normally expect from the format (a little bit of fluttering and stereo image issues). Certainly nothing that would impede my enjoyment of the music itself.

The only thing I did notice that wasn't typical to tapes was the high amount of noise in the signal - hiss and rumble. Like previously stated, this is probably because it was dubbed from another tape at some point. Thus doubling the noise.

Noise filters can try to reduce the hiss/rumble, and they might be successful at it - but that kind of thing always requires manual tweaking to get optimal results. As for the other tape issues, I don't know of any software out there to correct those.

What I would do, personally, is to see how much noise you can remove with noise reduction tools. (GoldWave has one built in, but other suggestions that were made here are probably better.) If you can't remove the noise without creating audible artifacts, then I would go for keeping the audio unprocessed. The thing to keep in mind is that, from a technical perspective, the audio you have is damaged when compared to the original recording. There's no clean way to reduce that damage - you can only attempt to mitigate it with filters after the fact.

If the filters make things worse, I'd say to leave it unedited. This recording is far from unlistenable and it's much better than some of the needle drops or bootlegs I've come across.

I'd stay very far away from compression or anything that's aimed at editing the characteristics of the audio, not just reducing noise and tape artifacts. If you have just a couple of peaks on the song that go full-scale, compression (or a brick wall filter) might be a useful tool if its going on a compilation where it has to match the volume level of other tracks. I haven't run into any tape transfers where freak peaks go above +3 or +6 of the ReplayGain-processed output track. So if this is poppish music, and the album is Rep"layGained at the end, you should be able to have the track at the proper volume without any kind of limiting or compression. Or very minimal compression on just a couple of peeks throughout the whole track.

I have done rips from cassette, vinyl, and 1/4" tape reel. Vinyl by far requires the most processing. Tape is usually much better performing for this task. Unlike vinyl, you don't get that random grain of dust that'll add a peak 12dB above the loudest musical content.

This post has been edited by slks: Aug 18 2013, 09:03


--------------------
http://www.last.fm/user/sls/
Go to the top of the page
+Quote Post
AndyH-ha
post Aug 18 2013, 10:48
Post #13





Group: Members
Posts: 2207
Joined: 31-August 05
Member No.: 24222



There are tools that can fix wow, flutter and miscellaneous speed variations in tape (and other) recordings but few people are likely to use them. I've read very good things about the first. The second company has a long standing and lofty reputation, although this is a fairly new product.
http://www.celemony.com/cms/index.php?id=capstan
http://www.cedaraudio.com/news/respeed_launch.html

There are good and poor noise reduction tools. The one developed by Syntrillium, in Cool Edit and Audition, is very good and the NR2 tools from Sonic Foundry (now owned by Sony and available as a plugin and in their editor) do a pretty good job too. Improperly used they can produce very audible artifacts but they can also do considerable reduction with no artifacts. Once you have the idea, application is usually very straight forward. Ozone has a good reputation but I've not used their products.

Removing rumble is no big task.

The discussion was about removing the occasional very out of range peak, not compression in general. The sound is rarely changed by any noticeable amount by chopping such peaks 3 to 6 dB with hard limiting.
Go to the top of the page
+Quote Post
AndyH-ha
post Aug 18 2013, 10:53
Post #14





Group: Members
Posts: 2207
Joined: 31-August 05
Member No.: 24222



Last time I looked at the Sony editor the NR2 tools were only included in certain editions. The basic NR tool, part of all versions, is just about useless.
Go to the top of the page
+Quote Post
Dogway
post Aug 18 2013, 13:47
Post #15





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



Wow, seems I missed the party.

@Specy: You talk about Dolby as something I can do afterwards (in software). So Dolby is kind of a psychoacoustic non-uniform "denoiser" and I should apply after my typical software denoiser, so the remaining temporal inconsistencies on noise levels can be equalized, is that right? So the problem is not Dolby, but denoising after Dolby, Dolby NR should be applied at the last in the chain. What I couldn't understand is your last sentence. You say I should apply dolby to the source of my recording, but I am recording from radio (for example), and also you say the source hiss is non-constant, why? why in this case (source denoising) I should apply Dolby and not typical software denoising?

@AndyH-ha: My question is simply that after I fixed everything I knew what to fix, I found that the track (or parts of it) still sounds very bad and I don't know why. Some people call it "mild drops", others "fluttering" or "rumble". The noise resembles that of wind hitting the mic, or plastic bag crumple if you know what I mean. It's not a constant noise so you can't apply a typical hiss denoiser there, the silence parts are okeyish, but then when audio enters in on occasions some "dirty" sound also comes in. Thanks a lot for the hard limiting tip, I thought compression was only affecting peaks.

So in general you don't recommend me to go for the new RX3 software? isn't it up to Audition or Sonic Foundry denoisers?

This post has been edited by Dogway: Aug 18 2013, 13:49
Go to the top of the page
+Quote Post
Specy
post Aug 18 2013, 14:48
Post #16





Group: Members
Posts: 22
Joined: 27-November 08
Member No.: 63320



@Dogway: Question #1: Was this recorded with Dolby? Because if not of course you should not use Dolby for playback either.

For TAPE noise, perform removal BEFORE Dolby. Dolby is basically a type of compressor (during recording) + expander (during playback), and the frequencies that it works on depend on the audio level, mainly of the high frequencies. Which means that when you add noise, the expander will also expand the noise. So if the hiss on tape is constant, Dolby will make it non-constant. And filtering it out afterwards is hence the wrong place.

This post has been edited by Specy: Aug 18 2013, 14:49
Go to the top of the page
+Quote Post
Dogway
post Aug 18 2013, 15:02
Post #17





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



Most likely not recorded with dolby, I recorded stuff with old radiocasette or boombox.
So Dolby (at playback) is just a real time EQ? As in an EQ based denoiser of some kind? On my walkman (WM-EX678) I have the remote control that says to be using Dolby B (I could disable it), but when I record I don't use the remote control, I opt for less cabling and plug the jack ends back to back with the PC line in.
Go to the top of the page
+Quote Post
Glenn Gundlach
post Aug 18 2013, 21:06
Post #18





Group: Members
Posts: 369
Joined: 19-April 08
From: LA
Member No.: 52914



QUOTE (Specy @ Aug 18 2013, 05:48) *
@Dogway: Question #1: Was this recorded with Dolby? Because if not of course you should not use Dolby for playback either.

For TAPE noise, perform removal BEFORE Dolby. Dolby is basically a type of compressor (during recording) + expander (during playback), and the frequencies that it works on depend on the audio level, mainly of the high frequencies. Which means that when you add noise, the expander will also expand the noise. So if the hiss on tape is constant, Dolby will make it non-constant. And filtering it out afterwards is hence the wrong place.


I think that will vary with the type of noise being removed. Dolby B is the top processor from a Dolby A unit. Anything that changes the response will alter the Dolby decoding. You put in a slight rolloff or boost in the band (approx 1000 Hz and up but the turnover slides between 500 and 2K) will become a _larger_ rolloff or boost after the decoding. Alterations below 500Hz will have virtually no effect.

The LF noise in 'sample fixed' may have been the tape getting near a strong magnet rather than from the recorder.

Depending on how much 'repair' you want it could get expensive and time consuming.

G
Go to the top of the page
+Quote Post
AndyH-ha
post Aug 19 2013, 00:46
Post #19





Group: Members
Posts: 2207
Joined: 31-August 05
Member No.: 24222



My discussion was meant to convey my ignorance, except for its reputation, which is good, of the Izotope product. Its noise reduction may be excellent, or not. I have no experience with it.

I take it, from your last comments, that you mean the low frequency thumps in sample2, such as at about 0.4 seconds and 2.25 seconds. These sound to me much like the result at the microphone of plosives in speech. If these are in your files before you began improvements, they are most likely a result of the original recording effort via microphone from the real world. Maybe they can also be produced by some over compression on the bass, post recording (but before you got it). You probably don't want to eliminate the sound but rather ameliorate it to a more pleasant state. I've done this successfully on guitar recordings, where it obviously isn't from voice, but the treatment is much the same as with ill recorded voice plosives.

The easier way is to use a high pass filter on the file to eliminates enough of the lower frequency that the thump becomes acceptable. Try a cutoff of 100Hz to start. Depending on the particular music, this may also eliminate certain quality aspects of your recording -- much of the bass.

The more surgical way is to only adjust the plosive itself, which may be as small as 25 to 30 milliseconds (or larger). You select the relevant bit and apply the filter. You can also probably use a more aggressive high pass filter, perhaps 200 to 300 Hz, maybe more. My experience is that for the majority of instances in any particular recording, the same filter settings work each time. You can't depend of that, however; only your ears can tell for sure. Obviously this is a slow and tedious process if there are very many instances.

I find it much easier to work in spectral view. The exact segment that contains the thump becomes fairly obvious with some experience. Also, sometimes it seems to work best to select the material, then adjust to zero boundaries (another tedious step) before applying the filter. Occasionally it is necessary to select a somewhat larger segment (i.e. beyond what seems to be the excess low frequency segment) to get the best results.

I'm not at all familiar with this applying Dolby B in software but I'm going to comment anyway. Last time I paid any attention, the rule was always to use Dolby on playback if it had been used when recording. Dolby processing is a complex set of actions that constantly adjusts parameters to the input signal. No software attempts to get around that were very successful, perhaps in part because the patent holder was aggressively jealous of any attempts. Maybe that has changed.

Regardless, if you want to do on-computer processing out of the normal sequence, I suggest trying a few samples both ways and carefully considering the results. Many years before the computer I often preferred to listen without Dolby engaged but I've always been more satisfied with it applied to music as intended when I've done cassette transfers.

I've done considerably more voice only recordings where I find using Dolby B on playback preferable even when it was not used to record. My interests are no doubt different than some people's. My aim with this kind of material is a file that sounds pleasant and is easy to understand, easy to listen to. Any slight change from the original tone, timber, or whatever, is totally irrelevant if it help meets the main requirements, and the reduction in hiss is a definite plus. I've always found that noise reduction, if it was still needed, was more successful post Dolby that as a replacement for Dolby, but then I never had any software Dolby mimic.
Go to the top of the page
+Quote Post
Specy
post Aug 19 2013, 01:59
Post #20





Group: Members
Posts: 22
Joined: 27-November 08
Member No.: 63320



QUOTE (AndyH-ha @ Aug 19 2013, 00:46) *
I take it, from your last comments, that you mean the low frequency thumps in sample2, such as at about 0.4 seconds and 2.25 seconds. These sound to me much like the result at the microphone of plosives in speech.

These sounds are what you get on some cheap cassette recorders when you record a new recoding over an old one, and press the stop button. If you want to get rid of them, just cut them out.

QUOTE
I'm not at all familiar with this applying Dolby B in software but I'm going to comment anyway. Last time I paid any attention, the rule was always to use Dolby on playback if it had been used when recording. Dolby processing is a complex set of actions that constantly adjusts parameters to the input signal.

Exactly that is why you should fix the tape first - any (temporal, or one channel) loss of highs will greatly affect the sound with Dolby enabled. That's why old tapes played back with Dolby often sound bad, some of the highs have been lost; you need to fix that before applying Dolby. Note: In many cassette decks, there's a potmeter that you can use to adjust 'Bias'. You could try that as well. But other things, such as AZIMUTH shifts, high frequency loss in one channel (actually it's often a volume loss, I think it's due to some dirt that got between the tape and the recording head, but after appyling Dolby the loss in highs is often much bigger), and in some cases there's a constant difference between the left and right channel which you cannot fix with a single Bias potmeter.

My experience - and I have digitized over 100 tapes - is that applying Dolby *after* fixing all the other errors gives much better results, even if the software doesn't exactly match the specification. (Having said that, I also have one newer cassette deck here in which the Dolby B filter somehow destroys soft sounds, making them sound like low quality MP3's, with older decks that's not the case, I have no idea what the problem is - but compared to that deck software does a much better job even without any restoration.)

This post has been edited by Specy: Aug 19 2013, 02:04
Go to the top of the page
+Quote Post
Dogway
post Aug 19 2013, 03:13
Post #21





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



I don't know what you guys are talking about, really. Sorry if it sounds harsh but that at 0.4s, 2.25s, 4.2s, 6.0s are break kicks.

I have been trying during these 4 days and 20 posts to put a name to a single artifact and you trained-guys can't tell me for Pete's sake the name of it.
Something as simple as;

"yes, that's 'blocking', use a 'deblocker'..."


I don't know, I feel very frustrated. I uploaded recently a better sample, and nobody mentioned about it.

@Specy: You are asking me to apply a (software) denoiser before Dolby, but since Dolby is a hardware thing, what you are basically saying is "drop Dolby altogether and use a software denoiser". Why don't you tell it all?
Go to the top of the page
+Quote Post
AndyH-ha
post Aug 19 2013, 04:53
Post #22





Group: Members
Posts: 2207
Joined: 31-August 05
Member No.: 24222



Has anyone recognized the existence of this artifact that bothers you. Obviously I haven't. In this condition there seems no possibility of naming it; recognition has to come first.

As for the item I identified, I did so because it sounds, quite actively, unpleasant to me and it can be modified as I described. I know it is supposed to be there, but is it supposed to sound quite the way it does?. Since it is your material, only you need be happy with it as is, however.

I'm sure Specy can explain this without help but since I'm here: the discussion is about a software alternate for Dolby B, something that can be applied to the on-computer recording if said recording is made without Dolby engaged on the playback deck.
Go to the top of the page
+Quote Post
Dogway
post Aug 19 2013, 15:04
Post #23





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



Well, I just rechecked and I was playing the tapes with Dolby disabled, I doubt they were recorded with Dolby, so everything is fine in that regard, can we leave aside the Dolby discussion? It's not related nor I am interested on it.
Go to the top of the page
+Quote Post
Specy
post Aug 20 2013, 21:30
Post #24





Group: Members
Posts: 22
Joined: 27-November 08
Member No.: 63320



I don't really know which artifact you mean. I basically hear 4 things:

1. A lot of noise.
2. Jitter.
3. Bad quality or damaged tape
4. Near the end: Distortion in the piano, which I guess was probably present in the original

About Dolby (which is now irrelevant, but if anyone else reads it): It is of couse best to feed the whole thing through an original hardware Dolby decoder. But you should still restore the audio first. So if you can somehow hook up your sound card output to a Dolby decoder, that would be perfect. I don't have the technical knowledge to do this, so I just used the software Dolby decoder mentioned earlier in this thread; it's not perfect but it gets very close, and it really sounds better than what I got with the Dolby on my cassette deck without fixing the recordings first. I'll see if I can find some recordings to post...
Go to the top of the page
+Quote Post
Dogway
post Aug 20 2013, 22:11
Post #25





Group: Members
Posts: 26
Joined: 15-August 13
Member No.: 109633



You mean my last sample? (Artifact.wav)

I only hear one type of artifact and I don't know how to call it.

"3. Bad quality or damaged tape"

What artifact is that called? can you post an example?
Go to the top of the page
+Quote Post

2 Pages V   1 2 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 20th September 2014 - 02:02