Digitising & Restoring Tapes & Records On Your PC
Technique : PC MusicianHere's a carefully tweaked RIAA de-emphasis curve that's provided by Waves in their collection of Q10 Pre/De-Emphasis presets. Notice the -12dB on the input faders to prevent digital overload.
A PC with CD burner and soundcard provides the ideal means to preserve old tape recordings or transfer your vinyl collection to CD.
Martin Walker
One of the reasons many PC users have for buying a soundcard is to transfer recordings from older analogue media such as vinyl or tape, and burn them to CD so that they can easily be listened to and archived. In this month's PC Musician I'm going to take you through the three stages of this process: getting the audio from your tape or record into the PC, carrying out any processing or restoration work that may be appropriate, and burning it to CD.
Preparing Your Material
The obvious thing to ensure first of all is that whatever analogue playback device you're using to get the audio into your computer, be it a reel-to-reel or cassette tape deck or a turntable, is set up to yield the best possible audio quality. If you're using a tape machine, for instance, make sure that the heads and mechanism are cleaned, and it's also worth cleaning vinyl records properly before you transfer them.
The output of tape decks can be directly connected to a suitable line-level input on your soundcard. Most cassette decks will have -10dBV consumer output levels, while more professional reel-to-reel tape machines will need your soundcard's +4dBU input sensitivity setting. If you're transferring records, however, you'll need to amplify and correctly equalise the low output level from your record deck's magnetic cartridge before you can connect it to a soundcard. Typical signal levels are around 5mV from an MM (moving magnet) pickup, and 0.5mV from an MC (moving coil) pickup, so you should keep the deck well away from both your PC and its CRT monitor.
With a suitable editor like Wavelab, you can keep an entire album's worth of tracks in one file using suitable markers, and process them on a global or individual basis.
Many musicians will still have hi-fis with a suitable phono input, and some mixers provide them, but otherwise you'll need a preamp with the RIAA (Recording Industries Association of America) equalisation that provides a flat overall frequency response. Phono preamps are available as stand-alone boxes from various specialist hi-fi retailers, and in DIY form from companies like Maplin (www.maplin.co.uk) but soundcard manufacturers have also spotted this niche. Terratec, for instance, offer a stand-alone phono preamp that draws its power from the 15-pin Gameport connector found on many consumer soundcards, while you can buy Steinberg's low-cost Clean! software (reviewed in SOS July 1999) bundled with an RIAA preamp for £79.99. Most phono preamps will probably have the lower -10dBV output level.
If you don't have a dedicated RIAA-compatible preamp, you could try using a mic preamp and then an EQ plug-in to boost the bass end and roll off the treble end to compensate for the response of the cutting lathe. I've detailed some suitable settings in the RIAA Equalisation Using An EQ Plug-in box.
When it comes to actually playing back your tapes, it's important to get the right playback settings on your tape recorder. Obviously, you should set up cassette recorders appropriately for ferrochrome, chrome or metal tapes, while reel-to-reel recorders may give you a choice between IEC or NAB playback EQ (although this can be corrected once in the PC using a suitable EQ plug-in such as Waves' Q10). For most recordings, you'll also need to decide whether or not you need noise reduction enabled, and if so of which type — Dolby A, B or C, Dbx and so on. Nearly all prerecorded cassettes use Dolby B, but if you don't know what type was used on a particular recording, it's often possible to guess. If you hear background noise levels 'pumping' with noise reduction switched out, the chances are that the recording used noise reduction, and most likely Dolby B. If you don't know, choose the type that gives the most natural-sounding result compared with other known recordings.
Recording Into A Soundcard
I can see little point in using a soundcard sample rate other than 44.1kHz if you're transferring audio to be put on CD, while 16-bit recording should provide more than enough dynamic range to capture any analogue tape. However, vinyl freaks with pristine record collections may want to instead choose 88.2kHz and 24-bit, to preserve the maximum amount of detail and high-frequency extension until the last possible moment. If you've got a consumer soundcard with loads of inputs like the Soundblaster Live! range, make sure you mute all but the one you're using, to minimise background noise levels.
Although many audio editing packages (such as Wavelab, shown here) provide automatic click removal functions, you can also edit out short clicks by hand, by simply recreating the likely waveform using a pencil tool.
It's best to experiment a little to find the optimum recording level, to make sure you get the maximum dynamic range onto the hard disk recording. Many PC audio recording applications and soundcard utilities provide an input meter with text readout of the highest recorded level, so start by running through a couple of tracks to see what this is, and then adjust the output level of your tape player or mixer accordingly, so that you still have one or two dBs of headroom left. For instance, if your peak meter reads -5dB, you can safely increase input level by at least 3dB without worrying about running into clipping. If you know the recordings well, you'll probably already know where the loudest parts are likely to be. You may well only have to do this once for a particular deck, and can then run through a whole batch of tapes or records with the same settings, so it's worth doing carefully.
Before you get stuck in, take a careful look at the recording facilities offered by your audio application, since there may be some real time-savers among them. For instance, Wavelab provides a selection of auto-start and auto-stop options that can automate the process of recording albums. I find its 'Auto-create markers at silence points' function particularly useful, since it places suitable CD track markers in gaps in the audio. Although you'll probably have to move their positions forward so that they coincide with the start of the next track rather than the fade out of the previous one, it's better than doing everything by hand.
Before you hit Record, check that you've got enough hard drive space. At 10Mb per stereo minute at 44.1kHz/16-bit, you'll need about 650Mb per CD, and if you envisage doing any PC audio editing you'll need double this amount of space to store the temporary files. During vinyl transfers you should keep your monitor levels quiet: otherwise, acoustic feedback could modify the frequency response of the recording. For the same reason, avoid moving about during the recording if your floor isn't solid.
RIAA Equalisation Using An EQ Plug-in
Even if you haven't got a properly equalised RIAA preamp, you can create the appropriate playback filter response using a suitable plug-in. For both 33rpm and 45rpm vinyl, you'll need a shelving bass boost with a turnover frequency of 500Hz and a slope of 6dB/octave, peaking with about 17dB boost at 50Hz, and then optionally rolling off below this using a low-pass filter. At the top end, you need another 6dB/octave shelving EQ, this time turning over at 2.1kHz, and rolling off to -13.6dB at 10kHz.
A similar approach can be used for playback of 78s, although there were many different curves used by different companies at different times, so you'll have to use your ears. Once again 6dB/octave slopes are needed, and a good start is a bass turnover at 250Hz with a boost up to about 12dB at 50Hz, and either a flat response at the top end for early pressings where there is little top end, or a gradual roll-off starting at between 3kHz and 6kHz, to perhaps -12dB at 10kHz.
Since there's a considerable LF boost, you'll need to be careful in avoiding digital clipping, and many plug-in EQs incorporate input and output level controls for this purpose. Waves' Q10 Paragraphic EQ already has a suitable preset, while Cool Edit Pro users can download a text file that adds a suitable preset to its FFT filter from www.a-reny.com/restauration/avance.html.
Splits, Stops & Starts
Once the recording is safely on your PC's hard drive, it's worth playing it back all the way through (at least the first time you do it) to check that no background tasks, screensavers and the like have caused momentary glitches or dropout. Making long recordings can root out problems with your PC setup that simply don't show up on short takes.
The next step is to decide whether to deal with songs individually, or leave them en masse. Some processes such as noise reduction can take a long time to compute, even on a fast computer, so you may prefer to split your file into separate song files and treat them individually, especially if you're using budget software. However, with a more advanced editor like Cool Edit Pro, Sound Forge or Wavelab, it's easy to add markers to delineate tracks, gaps between them, and unwanted areas. You can then easily select and process individual sections if required, as well as applying processes to the overall file.
Some people also trim out unwanted sections and deal with fades in and out at this point in the proceedings, but I prefer to perform global noise reduction first, and tackle these later on. Remember that for many noise-reduction algorithms to work their magic, you need a short section of noise by itself, so don't get rid of it yet! On vinyl the lead-in is often noisier than the lead-out, because it bears the brunt of pickup arm landings, so keep both sections.
Clicks & Crackles
Transfers of analogue recordings can exhibit a number of faults such as hiss, hum, and a decided lack of top end. Most can be corrected to some extent, but the first items to deal with should be ticks, clicks and pops, since their distinctive waveform shapes are easier to identify before you apply any other plug-ins, whether you're doing it by hand or using a plug-in with automatic detection.
Hum-removal processing needs to deal not only with the 50/60Hz fundamental frequency, but also with its harmonics, as in this example using Waves' Q10 Paragraphic EQ.
In the special case of mono vinyl records, you may first want to choose either the left or right channel of the stereo digital recording, or alternatively merge the two stereo channels together, before proceeding any further. Overall background noise might be lower when the two channels are merged, but if some surface clicks and pops are quieter on one channel than the other, you could copy this data to the other channel. Most editing packages let you do this, but of course the final CD will still be burned as a stereo digital file, and you can still add stereo enhancement if you wish later on using band-splitting or comb filtering. Many audiophiles frown on this practice, but if you want to try it, plug-ins such as PSP's PseudoStereo (part of their excellent StereoPack), TC Works' Native EQ Parametric and Waves' Q10 Paragraphic EQ have suitable presets.
If there are just a few short clicks, they can be dealt with by hand in most editing packages: locate the clicks, zoom in on them, and use the pencil tool to draw in an approximation to the missing section of waveforms. Many software applications including Cool Edit Pro, Sound Forge and Wavelab contain dedicated click-detection functions that make the search process a lot less tedious. They work by searching through the audio file for sudden changes in amplitude, and invariably need to be set up carefully to avoid picking up valid musical transients (too sensitive) or missing some clicks altogether (not sensitive enough).
Individual treatment of clicks does have the advantage of only processing your audio for the duration of each click, and leaving the remainder unchanged. However, some applications and plug-ins also have more intelligent algorithms that automate the redrawing process as well, and can process entire tracks automatically, detecting and removing clicks as they go; the simplest ones draw a straight line between the beginning and end of the click, while more sophisticated designs interpolate data based on what's on either side of it. Here it's even more important to make sure that only unwanted defects are being removed. If there's a function such as 'Residual', 'Audition', or 'Difference' that lets you hear just what's being removed, this makes the process a lot easier: good examples of this are Waves' X-Click (part of the Restoration bundle), the Virtos Declicker, and the Spectral Design DeClicker bundled with Wavelab 4.0.
When clicks have been dealt with, you can attempt to remove any remaining crackles. Badly treated vinyl can exhibit almost continuous low-level transients, often due to dust on the surface being gouged into the groove by the stylus, and some 78's can even suffer from tiny pockmarks caused by fungus! It can be difficult for algorithms to distinguish the music from such low-level artefacts, and it's therefore even more important to monitor what de-crackle plug-ins are removing.
Some plug-ins, such as the Decrackler built into Virtos' Declicker, Steinberg's Clean!, and Sonic Foundry's combined click and crackle remover, use modified versions of the declicker algorithm, but the best decracklers use advanced algorithms that retain the useful information contained in the original signal, and use this to replace the crackles. Examples include Waves' X-Crackle and CEDAR Audio's top-end professional restoration tools.
Digital Transfers
If your source is already in a suitable digital format such as DAT or Minidisc it makes most sense to transfer it digitally, since with most soundcards this will create a bit-for-bit-accurate copy inside your PC. Notable exceptions include Creative's Soundblaster Live!, which converts every analogue and digital signal to 48kHz using its asynchronous sample-rate converters on the way in. Make sure your soundcard is set to external clock so that it locks to your source machine, since otherwise you'll get clicks during the transfer. I discussed all these issues in some depth way back in SOS July 1999 (see www.soundonsound.com/sos/jul99/articles/pcmusician.htm). With suitable settings in your soundcard's utility software, you may also be able to retain the original track ID marker information, which will save lots of time later on.
Unfortunately, the most typical sample rate for DAT is 48kHz, so these files will require resampling to 44.1kHz before you can burn them on an audio CD. I've seen some experts claim that if you have to do this you might as well transfer via analogue, but personally I think many sample-rate converter algorithms are capable of excellent results, and they certainly beat having to set up recording levels.
Hum Removal
Aside from clicks and crackle, analogue transfers can suffer from continuous background noise such as hiss, hum, and buzzes. Of these, it's often best to deal with hum first, as there are ways to surgically remove it without taking much else away. This is also a suitable time to deal with any unwanted low-end rumble and DC offsets; if you've made a vinyl transfer without RIAA de-emphasis, now's the time to restore the frequency balance, and hear how much bass is really there once it's been raised by up to 17dB at 50Hz.
If your hiss-reduction plug-in needs to capture a noise print, like Sonic Foundry's Noise Reduction shown here, don't trim off the start of your tracks until you've done this.
Hum is nearly always caused by mains interference, and the simplest way of removing it is using a narrow notch filter tuned to either 50Hz or 60Hz, depending on the frequency of the mains supply in question. However, hum very rarely exists only at the fundamental frequency, so recordings will nearly always need additional treatment of the first few harmonics at 100Hz and 150Hz in the UK, and 120Hz and 180Hz for the US.
You can try using a general-purpose parametric EQ, but these often don't provide high enough Q values and end up removing the surrounding bass frequencies as well as the hum. Waves' Q10 Paragraphic provides a useful preset for US use, which you can easily tweak for UK frequencies, and the TC Native EQ has a notch option that provides a very high Q value. Waves' X-Hum can cut up to 60dB with its eight notch filters.
Don't assume that the hum will be exactly at the frequency you expect — mains frequency can vary quite a bit, so fine-tune the filter frequencies to get the best hum reduction. Don't get carried away in your hum removal, and listen carefully when you switch your filter in and out, since there may be side-effects at the bass end. In particular, some high-Q notch filters have a tendency to ring, which may cause more disturbance at the bass end than the hum they cure. Once the hum is gone you can also roll off any unwanted rumble. Normally there's little useful information below about 20Hz, while rumble may require rolling off at 12 to 18dB/octave at between 30Hz to 80Hz, which is easily done using many EQ plug-ins.
Transferring 78rpm Records
If you're transferring recordings from 78rpm discs and your deck only offers speeds of 33rpm and 45rpm, there are various ways to overcome the problem, although in all cases you should choose the 45rpm speed, as this is the nearest.
The first solution is simply to record at a non-standard sample rate such as 25442Hz, and then change this immediately after recording to 44.1kHz (44100 x 45 / 78 = 25442). However, not all soundcards support such non-standard rates, so it's easier to record at 44.1kHz and then use the resampling functions in your audio software to change the pitch. Some dedicated restoration software like Dart Pro has such features built in, while in Cool Edit Pro, for instance, you would use the Time/Pitch Stretch function of the Transform menu, select Resample as Stretching Mode, and then enter 57.692 as the Ratio (45/78 expressed as a percentage). In Wavelab you would use the Time Stretch process and enter exactly the same ratio, but untick the Preserve Pitch box. In both cases you should always use the highest quality setting, although this may take a long time to process. You should ideally correct the pitch after removing any clicks and pops, because these will be easier to detect beforehand.
Bear in mind that if you're using a RIAA-equalised preamp, the tonal balance will be incorrect after these shifts (there'll probably be too little top end), although you'll be in the right ball park, and most 78s are likely to require some manual EQ massaging anyway. Bear in mind also that some old records were cut at speeds ranging from 68 to 84rpm, although you may be able to calculate the correct pitch shift if you know the original key of the music.
Removing Hiss
By this point, the remaining offender is likely to be background hiss. Most cassette tape recordings will have plenty of this, and many reel-to-reel recordings will also suffer to a certain extent.
Although you can reduce hiss by rolling off high frequencies, this will have a disastrous effect on the music, while simple expanders can improve matters slightly at the risk of noise pumping. Dynamic filters that open and close according to signal content can also work, but still tend to dull the music and remove transients. Modern noise-reduction plug-ins work by splitting the audio into hundreds of narrow frequency bands, analysing their continuous content, and then using expansion to pull each band down when no additional musical information is present. This process can also deal with other continuous background noises such as air-conditioning, hum and rumble, although the last two can often be more easily removed using filtering as previously described.
If you've transferred old 78rpm records at 45rpm, you can correct their pitch using settings such as these with Cool Edit Pro.
Most noise-reduction plug-ins need to analyse a short section of the noise by itself, to make a 'noise print' or 'profile'. You can nearly always grab a suitable section (half a second or so is fine) between songs, or right at the start of an album. For best results, try to avoid any plops or thuds, and aim for a smooth section.
There are now many noise-reduction plug-ins available for the PC: examples include Arboretum's Raygun and Restoration-NR, Cool Edit Pro's Noise Reduction, Sonic Foundry's Noise Reduction, the Virtos Noise Wizard, Wavelab 4.0's DeNoiser and Waves' X-Noise. Most use the noise profile approach, although DeNoiser has an 'adaptive' design that constantly analyses the noise floor and updates its profile automatically, and Noise Wizard provides this as an option. Whichever type you use, there are various things to watch out for, the worst being 'chirping' artefacts if you attempt to reduce the noise by much more than 20dB. You may also notice a metallic side-effect when using cheaper plug-ins, and as with de-clickers, auditioning of the removed portion by itself greatly aids control setup. Even if you don't notice unusual side-effects, make sure you don't strip the life out of your material — overuse of noise reduction can cause transients to disappear along with reverb tails.
The Final Touches
Noise-reduction algorithms can suppress high frequencies slightly, so you may want to combat this with a tiny amount of top-end EQ. However, a better alternative is often a little enhancement to synthesize new frequencies at the top end, particularly in the case of cassette tapes which had little high-frequency content in the first place. There are now lots of enhancers available, and I've been particularly impressed by BBE Sonic Maximiser, PSP's MixTreble and Steinberg's Spectraliser. Low-end enhancement shouldn't normally be necessary, but BBE Sonic Maximiser or Waves' MaxBass and Renaissance Bass can add this if required.
The final stage before burning a CD is to remove any unwanted sections and deal with the beginning, inter-track and end sections. I tend to add a short fade-in of about 10ms before the very first note of each track to ensure that there are no clicks, and then make a longer fade of a second or more beyond the end of the final note of each track. You can then fill the inter-track gaps with digital silence. However, with old recordings where the remaining background noise is still reasonably audible, it's sometimes better just to leave the inter-track noise alone, as the ear will find this less obtrusive than a more obvious drop to silence.
A two-second pause is always required before the first CD track, but any material before this can be trimmed, including any vinyl lead-in groove noise. Similarly, you can now trim any material beyond your fadeout on the final track, including any vinyl 'clunks' as the pickup arm lifts off. I always add a couple of seconds of digital silence beyond the final fade, so that there's a gap before the CD player spins down.
If your audio editor has integral burning functions, you can make final tweaks to your track marker points, and audition the start of each one to check how it will sound on a CD player. Finally, don't forget to use DAO (Disc At Once) mode rather then TAO (Track At Once) mode when burning your audio CD, since this will avoid a nasty click at the end. If you're using a dedicated audio CD burning package, or one built-in to your audio editor, this will be taken care of automatically.
No comments:
Post a Comment