Transferring LPs to CDR: Some Advice
Preliminaries
This page of notes is a distillation of my experiences
in transferring LPs to CDR. I offer it as hopefully unbiased advice
to others wishing to do the same thing. I have attempted to address
the whole process at a reasonably deep level. Some parts of it are bound
to be of no interest to some readers; I hope the sections are obvious
enough that you can skip those parts which don't apply to you.
My standards are high. I have a stereo system which all
but a small minority of audiophile extremists would regard as "very
high end". I'm not interested in putting anything onto a CDR unless
I can get fairly close to perfection. The advice contained in these
notes is empirical, and results from my own personal experiences.
I am not a digital audio professional. Transferring LPs
to CDR is my hobby, born of the desire to preserve those parts of my
LP collection which (i) are unavailable on CD, or (ii) aren't essential
enough to me to be worth spending the money replacing them on CD.
I work on a PC, and can only offer advice in that arena.
I have no knowledge whatsoever of doing this kind of work on Macs or
Unix machines.
I get quite a few email enquiries from people who have
found this page, and some questions tend to be asked repeatedly. They
aren't directly relevant to the task of transferring LP to CDR, but
are more general about audio on a PC. Therefore I've put together a
separate page (Clive's FAQ) to answer these.
Before emailing me with a question about audio on a PC, please check
this page to see if it's answered there.
Obviously I can't write
down everything I've ever discovered about this process, so if you have
any other specific questions, feel free to email
me. But before you do, I would ask you to check
the FAQ carefully to see if your question is already answered there.
Please also note that changes in my circumstances mean that I will not
be able to respond to emails as quickly as I have done in the past,
but I will try to continue to help wherever I can.
And finally, a disclaimer about Adaptec/Roxio software.
I sometimes receive emails asking for help regarding the Adaptec/Roxio
products Easy CD Creator and Spin Doctor. As far as I can make out,
it is possible to get to this page via links that start on Roxio's website,
and so some people arrive here under the mistaken impression that I
am in some way connected to Adaptec/Roxio. Let me make my position clear.
I have absolutely no connection to Adaptec/Roxio. I have
evaluated both Easy CD Creator and Spin Doctor in the past, but do not
use either of them and am unable to offer specific advice about them.
At the end of this document is a list of useful
links. (I will just point out that I do not read German, so cannot
comment on the content of Joerg
Eisentraeger's site).
Slightly off-topic: why use a computer?: I got
some email from someone who asked me to add a short paragraph discussing
alternative methods of archiving LPs without
using a computer, so I have done so at the end of these notes.
Overview
Recording the LP to Hard Disk
Clean the LP
Before you start, you should do your best to clean the
record as thoroughly as possible; getting dust and gunge out of the
grooves will eliminate a fair amount of the lower level noise that would
otherwise require a very time-consuming job to remove at a later stage.
Ideally use a vacuum device; even better get them professionally cleaned
by someone who has access to a Keith Monks cleaning machine.
If you don't have access to a vacuum machine, and the
record is very dirty, then I've recently tried out a fluid from Australia
known as The
Vinyl Solution which is inexpensive and works well on very dirty
records. I have to say that I do not know what is in this fluid, and
so cannot take responsibility for any long-term effects it may have
on the vinyl. I suspect, but cannot say with any certainty, that it
may have the same alleged drawbacks as wet playing, but if all you want
is to make one good transfer of an LP and can't afford to invest in
a vacuum cleaner, it seems like a good approach.
Playing the LP
Use a good quality turntable to play the LP. The pickup
cartridge should be properly aligned and the stylus should be in good
condition and clean.
A turntable is a mechanical device which is vulnerable
to airborne and floorborne vibrations. Such vibrations can degrade the
playback quality quite noticably. Even the very best turntables can
suffer in this respect. Therefore, while recording the LP, keep surrounding
sound levels as low as possible. Ideally you should not use any kind
of monitoring at all; don't worry about knowing when the music starts
and ends, just start recording from before placing the stylus on the
lead-in groove and continue recording right into the run-out groove
(trimming off these extra bits later is easy).
The Need for a Preamp
I will assume that you will use a moving magnet or moving
coil cartridge (all high quality cartridges are one of these two types).
The signal from such a cartridge is both low in level (typically <5mV
for moving magnet, <0.5mV for moving coil) so must be boosted to
about 100mV needed to drive line level inputs, and is also RIAA equalised
(applied when the LP is mastered to get around mechanical limitations
of the LP system) which must be reversed so as to provide a flat frequency
response for the line input. Both of these necessary operations are
achieved with a suitable preamp. If you have a receiver or stereo amplifier
with a "phono" input, that input will perform this task, and the resulting
line level output will be available on the receiver/amplifier tape output.
Tom Weber informs me that Radio Shack sells an inexpensive amplifier
(model number SA-155) for about $65 which is suitable, and I pass on
his recommendation here while emphasising that I have no personal experience
of this device. I believe that Radio Shack also sell a standalone phono
preamp for about $30. Reports from various sources lead me to believe
that the quality of this preamp is no more than "servicable". NAD make
a well-respected preamp, the PP-1, for about $60. It is also quite possible
to build your own if you're at all adept at DIY electronics; There are
a couple of suitable circuit diagrams on
Mike
Richter's site, and
PAiA Electronics
sell a DIY kit. If you have higher quality equipment such as a separate
preamp and poweramp, you'll know what I'm talking about and I leave
the choice of a quality phono stage up to you.
I see that a few companies, including Stanton
and Denon, make turntables that have
built-in phono preamps, and a few even have SPDIF digital outputs. That
said, these are DJ type turntables, which tend to be built for ruggedness
and reliability rather than outright sound quality.
How about "wet playing"?
One option you might like to consider is "wet playing".
The idea here is to flood the LP with a suitable liquid while it is
being played, in the hope that dirt which normally sits in the grooves
(and will therefore be tracked by the stylus) will be lifted into suspension
and therefore will not influence the stylus.
Other people have different views as to why wet playing
works. There is a school of thought that it's not really anything to
do with lifting dirt into suspension, but rather that it alters the
damping of the stylus and/or allows the stylus to aquaplane over minor
imperfections.
I should also report that some people state that wet
playing actually damages the groove wall. The argument is as follows.
When you play an LP, the (hard) diamond stylus deforms the (soft) vinyl
groove. When played normally (ie. dry), the friction causes the vinyl
to heat up, which allows it to deform and return to its original shape
after a while. If, however, the LP is played wet, the fluid acts as
a coolant which prevents the deformation, allowing the stylus instead
to carve slivers of vinyl off the groove walls.
Others have reported that wet playing can cause a kind
of sludge (ie. the dirt from the LP suspended in the liquid) to accumulate
on the stylus. This would then dry and harden, be very difficult to
clean off, and cause subsequent mistracking. I've not noticed this myself,
but would guess that the scale of this problem would be affected by
how dirty the LP is, the exact composition of the liquid used, and how
highly polished the stylus is.
I have experimented with wet playing, and it does indeed
reduce some (but by no means all) forms of surface noise. There are
a number of issues, though:
- What liquid should be used? You certainly must not use tap water,
as it is full of contaminents. Distilled water is the obvious base
liquid, but it is not a particularly good solvent, and so any dirt
which is stuck to the grooves is unlikely to be lifted. Adding an
amount of alcohol can help matters here, but it tends to evaporate
off the LP surface before the side has finished playing; closing
the turntable lid while playing may help reduce the evaporation
rate.
- After playing an LP wet, the dirt in suspension drops back into
the grooves at very inconvenient places. This results in the surface
noise being significantly increased if you subsequently play the
record dry. Therefore, after playing a record wet, you really need
to continue to play it wet thereafter.
- The most successful wet-play method I've ever come across is a
product called "Lencoclean", which tracks the LP like a second pickup
arm, applying the solution locally to just that part being played.
I haven't used it for over 20 years now, but it is still available.
Lenco have a website (in German),
but it's a little difficult to navigate, and the last time I visited
it I couldn't find Lencoclean mentioned anywhere. An online store
that can supply Lencoclean is Conrad
Electronics.
I must point out that I have no authoritative knowledge
about wet playing, but feel that if I discuss it at all, it is important
to mention all the opinions of which I am aware.
Recording to Hard Disk
Once you have a line level signal, it needs to be digitised
and recorded onto the computer's hard disk. Note that for a typical
40 minute LP, you'll need about 500Mb of disk space, and depending on
what PC editor(s) you use, you might need another 500Mb for temporary
files. Count on needing about a gigabyte in total.
The standard Sound Recorder utility that comes with Windows
is not suitable for this task because it records to main RAM and only
writes the results to hard disk when recording finishes. This means
that recording time is limited by the amount of RAM in the PC; since
16 bit stereo at 44kHz uses about 10MB per minute, there's no way you'll
be able to record an entire LP side using Sound Recorder. What is needed
is a utility that can record direct to the hard disk. There are a large
number of such utilities available as shareware. My own shareware Wave
Repair supports hard disk recording, has rather better record level
metering than most other packages, and if used as a simple record utility
it is freeware. Of the other packages around, the cheapest two I am
aware of are CD Wave and RIP
Vinyl. Another fairly straightforward and inexpensive recording
package is PolderbitS. Although
I've not had a chance to evaluate it properly, a number of users are
very impressed by it, so for that reason it is worth a mention. Two
very good general purpose WAV file editors, CoolEdit
2000 and GoldWave, also offer
direct to hard disk recording.
Discussion of Soundcard Types
Once you have a package capable of recording direct
to hard disk, there are two basic approaches to digitising the line
level signal depending on the quality you require:
- Use a "normal" computer soundcard. You need a soundcard capable
of 44.1kHz 16bit stereo. All modern soundcards are able to do this.
The (analogue) line level signal should be fed into the line input
of the soundcard, and the signal can then be recorded straight to
the hard disk.
- Use an outboard analogue-to-digital convertor, and feed its (digital)
output into a soundcard capable of receiving digital signals. As
with using a "normal" soundcard, the analogue signal must be sampled
at 44.1kHz stereo. If you use an A/D convertor which is capable
of more than 16 bit resolution, you might wish to find a soundcard
that can maintain those bits (although you will of course have to
convert to 16 bit before actually burning a CDR). However, my own
experience is that even the most perfect of LP sources will be hard
pressed to deliver a signal with more than 12 bits resolution, so
plain vanilla 16 bit is fine, and even truncation without redithering
won't harm the LP signal.
Using a "normal" soundcard is certainly a much cheaper option, but the
inside of a PC is a very hostile environment for analogue signals. Some
soundcards are better than others in their rejection of this noise pollution.
My experience is that currently available soundcards generally fall
into four categories:
- Cheap cards with unknown brand names. It is of course possible
that some of these may be fine, but in general they have fairly
dreadful sound quality. Avoid.
- Low end PCI-bus Soundblasters. The old ISA-bus Soundblasters were
frankly not very good, but the more recent PCI-bus offerings are
much better. An OEM version of the PCI128 can be had for about $30,
and has quite reasonable sound quality. The noise floor is about
-65dB (although of course this might be influenced by the particular
motherboard in use), which is better than you'll get off even the
best LP played on a high-end turntable, and the frequency response
is perhaps a little bright but otherwise not bad at all. Note that
the PCI128 has recently been replaced by the PCI512, and there are
reports that the 512 has marginally worse sound quality than the
128.
- Mainstream cards in the sub-$200 range, such as the Soundblaster
Live, Turtle Beach Montego II, and similar. These cards tend to
have much more sophisticated extra features for game support, sound
effects and multi-channel surround sound, but their basic sound
quality is usually no better than cards like the PCI128. An exception
seems to be the recently introduced Turtle Beach Santa Cruz: although
I've never used one, technical reviews consistently report that
its raw sound quality is a notch above the Soundblaster Live. I
have also seen reports that some of the Terratec range are good
performers, but I have no personal experience of them.
- Serious cards aimed at the semi-professional, such as the Echo
Gina, LynxOne, and DAL CardDeluxe. These typically cost around $400-$500,
and can give sound quality as good as entry-level external A/D converters.
The following tip comes courtesy of Richard Melton: When using an analogue
soundcard, it is often a good idea to mute all inputs and outputs that
aren't actually being used (eg. mic, MIDI, etc). This will improve the
noise performance of many soundcards. Note that some soundcards have
a "calibration" function, which should be re-run after changing the
input configuration.
www.pcavtech.com/soundcards/index.htm has a lot of good information
about analogue soundcards.
As for soundcards which can receive a digital signal,
there are a host of options:
- The cheapest I know of that's widely available is the Midiman
DiO2448, with a typical price under $150.
- There is another card called the Zoltrix Nightingale which uses
the same chipset as the DiO2448 which supports SPDIF I/O and has
a price under $40. However there are two drawbacks: (i) it's very
hard to actually find one; (ii) it only has optical (Toslink) input,
although there is a simple DIY modification to add COAX input.
- Most of the higher-end analogue soundcards such as the Gina, LynxOne
and CardDeluxe also have digital I/O capabilities.
- A couple of older Turtle Beach ISA-bus cards, the Fiji and Pinnicle,
have daughter board options to provide SPDIF I/O and are known to
work well, but may be getting difficult to acquire now that they
are out of production.
- Three inexpensive mainstream cards can have an SPDIF daughter
board added: Turtle Beach Montego II, Creative Soundblaster Live,
and Hootech ST Digital XG. However, all three resample the SPDIF
input to an onboard clock rate, so they do not allow for bit-perfect
transfer of external digital audio to the hard disk.
- I personally use a Midiman Audiophile 2496, which can be had for
under $200 and which has excellent analogue I/O (not that far off
the quality of cards like the CardDeluxe) as well as bit-perfect
SPDIF I/O.
- And finally, you may hear stories about $20 soundcards with SPDIF
I/O. Various people have tested these cards and in the vast majority
of cases could not make the SPDIF I/O work properly. Even those
few people who did have success needed to build extra bits of circuitry.
www.digitalexperience.com/cards.html
has a lot of good information about digital soundcards.
As for the outboard A/D converters themselves, I don't
really have a lot of experience in this area, but can report that some
low cost converters which have good reputations are the Midiman Flying
Calf (about $200), the Lucid ADA1000 (about $500), the Symetrix 620
(an old design at about $600, still highly regarded and something of
a de-facto standard in semi-pro studios), and the Lucid AD9624 (about
$800). Above this we start to get into serious professional territory,
and I'm not qualified to offer advice at that level.
An Alternative Approach that By-Passes the Soundcard
In the last year or so, consumer grade audio CD-RW recorders
have come on to the market. These devices can be hooked up to your stereo
system like a tape deck. This presents a new approach you might like
to consider: record the LP to a CD-RW disc using an audio CD recorder,
then transfer that to hard disk using digital audio extraction with
your CDROM drive. The CD-RW disc can then be reused for the next transfer.
This has the following advantages:
- The A/D converters in audio CD recorders aren't at all bad. They
are typically better than the converters on mainstream soundcards
like the SB Live and Montego II. (They are not as good as the converters
on cards like the DAL CardDeluxe, LynxOne, etc).
- Since the CD recorder isn't inside the PC, any possibility of
noise pickup from other subsystems such as the video card is eliminated.
- The CD recorder can sit alongside your stereo system and record
from any of the sources you might wish to transfer (eg. LP, cassette,
radio, etc). You don't have the problem of how to hook the stereo
up to your PC (which in many homes will be in different rooms).
- The only thing you need the soundcard for now is monitoring your
work in progress, and even quite cheap soundcards (eg. $10 SB16
clones) are good enough to identify vinyl damage that needs cleaning
up.
Humming Along with the Music
It's quite possible that when you hook up the line level
output from the LP playback system to the input of your PC's soundcard,
a hum will result. This is due to problems with the ground connections
on various parts of the whole setup (usually the PC's ground and the
stereo system's ground are at different voltages).
Trying to rectify this kind of hum can be problematic,
but a few things to try are:
- Check that all the components in your stereo system and the PC
are actually grounded; leaving off a ground can easily cause hum.
While checking the grounding of the components, don't forget to
check the wiring inside the mains plugs.
- Plug all mains electrical items (ie. all stereo components and
the PC) into the same mains outlet, using multiway adapters as necessary.
(The power consumption of all these kind of devices is so low that
this will not overload the outlet). This forces all the components
to share a common ground, which might fix the problem.
- If sharing the same outlet doesn't help, you can try disconnecting
the ground connections from some components. Start with the turntable
itself. If this doesn't work, check to see if the turntable's connection
to the preamp/amp/receiver has a separate ground wire and try the
effect of attaching/removing it from the preamp/amp/receiver's grounding
post.
- If you still have no luck, try disconnecting the preamp/amp/receiver's
ground (both with and without the turntable ground re-connected).
However, beware of any dire warnings in the manufacturer's manual
about this: in many cases it will be safe, but some components might
not have insulated cases.
- Rod Smith tells me that cable TV connections can sometimes cause
a ground loop if they are fed into the stereo system. In this case,
the simplest solution is to disconnect the cable TV output from
the stereo system while recording into the computer.
- George Koulomzin suggests another experiment. If you have mains
plugs that can be inserted either way round, try the effect of reversing
one or more of them. This alters the relative phase of the mains
to the components, and it may cause the hum to cancel out.
- Keith Christensen tells me that if you live very close to an analogue
TV transmitter mast, the vertical sync pulses can cause interference
that sounds very much like mains hum. Unfortunately I don't know
of any solution to this.
- Stores such as Radio Shack sell small transformers known as "ground
loop isolators". These can cure some types of hum problem when connected
between the stereo system and the computer. One person (Mike, he
never told me his surname) discovered that the isolators sold by
Radio Shack as "Home/Automotive Ground Loop Isolator" (cat. no.
270-054) introduce a serious cut in the bass. Given their low price,
these could simply be high-pass filters rather than transformers
(ie. they solve the hum problem by just throwing away all the low
frequencies along with the hum). Even if they are transformers,
it sounds like they are of insufficient quality to do a good job,
so they are not recommended. Unfortunately good quality audio frequency
transformers are not cheap.
- If all this fails, then I'm at a loss as to what you might try,
short of removing all grounds from all components, bundling them
all together and strapping them to a copper pipe buried in the (real)
ground!
Setting Record Levels
Whether you choose to go with a "normal" soundcard or
external A/D convertor, it is important to set the input levels accordingly.
The aim here is to get peaks as close to 0dB as possible without exceeding
that level. This is for two reasons:
- It uses the maximum resolution available with the 16 bits. Having
said this, since an LP signal manages about 12 bits resolution at
best, you could in theory afford to leave a huge headroom (eg. peak
around -12dB or even less) without losing resolution.
- (Less obvious but probably more important) If you record at a
low level and burn a CDR, you will end up with a CD that plays too
quietly in comparison to your other (commercial) CDs. While it is
possible to bring the level up using the "normalisation" feature
of many WAV file editors, such normalisation is never perfect (the
arithmetic that is performed results in rounding errors), and so
it is better to avoid this if possible.
As a guide, I tend to pick what I think is the loudest part of the LP
(if you know the music well, you'll know where the loud bits are; if
it's an unfamiliar record, visually inspect the grooves for the ones
that wiggle the most), and set record levels to register about -3dB
for that part, which leaves a little headroom in reserve.
Unlike recording to analogue tape (where pushing the
signal level well past the nominal maximum level can sometimes be a
valid approach) it is absolutely crucial that you never exceed the 0dB
level. If you do, the result is digital clipping; an extremely unpleasant-sounding
type of distortion.
Track Splitting
When recording an LP to hard disk, don't be tempted
to try and split the tracks at this stage. You will want to retain the
correct timing of inter-track gaps on the final CD, and it is much easier
to split the tracks and retain the correct length gap using a PC editor
later. (Indeed, you may use a CDR burning package that doesn't need
the tracks to be in separate files anyway).
Cleaning up the Recording
Once the signal is on hard disk, the hard work begins.
Before we start, let me state from the outset that many people hope
to find a single software package that will do everything they need.
This is an unrealistic expectation; in general you will need a toolkit
of various packages. In the notes which follow I will point out the
strengths of those which I have personally used, and in passing will
mention other packages which I have not used but nevertheless have a
good reputation.
Mono LPs
Most people will be recording from stereo records. However,
if you want to transfer a mono LP there are a few extra issues to consider
which I'd like to deal with first. A CD cannot be mono; you must record
it in stereo. The ideal situation is that the two channels are identical,
but if you just play a mono LP on a standard stereo turntable, the chances
of getting identical left and right channels is virtually nil. It may
well be that they are close enough that the results sound fine and you
don't feel the need to change anything.
However, what if the two channels are sufficiently different
that the results are not really acceptable mono? To arrive at two identical
channels, there are basically three options:
- Merge the two channels together and duplicate the results to both
left and right. You may find that some surface noise cancels slightly
this way.
- Duplicate the left or right channel to both channels. This would
be appropriate if you decide that one of the two channels is markedly
better than the other. The most likely reason this might be the
case is when one of the groove walls has worse damage than the other,
allowing you to select the channel with less surface noise and/or
clicks & pops.
- Some intermediate stage between the two above options, whereby
the channels are merged but a weighting is applied so that a greater
proportion of one channel is used than the other.
Either of the first two options can be achieved with creative wiring
of the turntable's cartridge, but a weighted merge is only really possible
using a mixer. You can of course deal with it all in software once the
signal is on hard disk, and this is the course I'd recommend. One advantage
to recording the two channels to hard disk as stereo is that if there
is a click on one channel only, you can copy over a clean section from
the other channel. Only after this stage would it then be appropriate
to start mixing the channels to mono.
CoolEdit
2000 is probably the best affordable tool for this kind of channel
merging operation, using its channel mixer.
Trimming Out Unwanted Sections and Fading In & Out
These are essential steps, and can be done easily using
a wide variety of WAV file editors. Two affordable shareware editors
that I strongly recommend are
GoldWave
and
CoolEdit 2000. My own shareware
Wave Repair can also trim off
unwanted sections. Try to get the start of the WAV file as close to
the beginning of the music as possible, leaving perhaps a quarter second
in reserve. Once you've trimmed this excess at the start, edit the first
few samples to make sure they are zero on both channels and then fade
in the next few samples (making sure you get to full volume by the time
the music starts); all this messing about is to get a nice clean start
to the CDR without a click. The same procedure is required at the end
of the LP, although here you should aim for a longer, gradual fade out.
I like to add a little extra silence at the end; this is because some
CD players make quite a bit of mechanical noise at the end of a CD (eg.
relays switching, laser sleds parking, etc), and I prefer this not to
happen the instant the music finishes.
Reducing Constant Noise
The amount of constant background noise (eg. hiss) is
quite low on vinyl LPs. (I do not include "crackle" in this category:
vinyl crackle isn't really a constant noise, and it is usually better
removed using the methods described in the next section). Constant noise
is usually only significant on historic records, so unless you find
it particularly objectionable it is probably best to leave it alone.
That said, it is rather easier to deal with than random noise such as
clicks, pops and crackle.
Many audio editors include a broadband noise reduction
feature. They usually operate by first taking a "noise fingerprint"
from a region that contains only noise, then removing that noise fingerprint
from the music using a process known as "Spectral Subtraction". This
process can work well, but for tape hiss can cut out quite a bit of
the high frequency programme content. It should also be used in moderation,
since it can impart a sort of "metallic, robotic" sound to the music
if you use it too enthusiastically.
My opinion is that CoolEdit
2000 has the most sophisticated broadband noise reduction available
in an inexpensive product, in that it has a wide range of adjustments.
That said, this extra flexibility often doesn't seem to achieve significantly
better results than the more basic facilities available elsewhere, but
if you're prepared to put in the effort and experiment with different
settings it can sometimes deliver slightly better results.
Another package which deserves a mention is DCart.
This has a fairly good dynamic noise limiter, which varies the amount
of hiss reduction based on the amount of high frequency signal that
is present. When there is a lot of high frequency energy, the amount
of hiss reduction is small; this takes advantage of the fact that the
high frequencies that are present mask the hiss. When there is little
high frequency content, the amount of hiss reduction is high, and the
loss of what little high frequencies there are isn't very noticable.
(This technique is basically similar to the the old Philips "DNL" hiss
reduction system, as used on their cassette decks back in the 1970s).
It can work remarkably well, especially on "busy" music, although it
pumps badly on some kinds of signal (eg. solo piano).
Removing Clicks, Pops and Crackle
Even the best LP will have some minor clicks that you'll
want to remove. Some records in bad shape will have a constant background
of crackle that you'd like to reduce.
Removing these kinds of noises without adversely affecting
the music is difficult. There are a number of packages on the market
which claim to do so automatically, and their number seems to be growing
on an almost daily basis. I have tried the following: DCart,
DART Pro, CoolEdit
2000, CoolEdit Pro (a more
expensive version of CoolEdit 2000), Sound
Forge, Sound Laundry, Spin
Doctor (bundled with full versions of Roxio's Easy CD Creator, but
absent from most OEM versions that come with CDRW drives), Groove
Mechanic, WAVclean, Wave
Corrector, Wave Repair (written
by me, so take what I say about it with a suitable dose of suspicion).
They all suffer from the same basic problem: they sometimes work very
well, and other times they actually make things worse.
You can try fiddling with the parameters, but this rarely
results in any significant improvement. Some of them (DART
Pro, and especially CoolEdit
Pro) have so many configuration parameters that it's well-nigh impossible
to try them all out, especially since they perform their processing
so slowly.
DCart, Sound
Laundry, and Wave Repair
are better in this respect because they have a realtime preview mode
which allows you to adjust the parameters while listening to their effect.
WAVclean, while not
working in real time, allows you to listen to the results so far while
it's still processing the remainder of the file.
Wave Corrector
(which again doesn't work in real time) has detection and repair algorithms
which seem to be rather more effective than many, so I feel it is well
worth investigating. A useful feature is that it allows the user to
review and adjust the correction of each individual click.
Groove Mechanic
is another tool which doesn't work in real time, but whose results are
sufficiently good to warrant a recommendation. Fortunately it has few
adjustable parameters, so you can't waste huge amounts of time fiddling
with them in a futile attempt to find the best settings.
CoolEdit 2000's
Audio Cleanup plug-in and Sound
Forge's Noise Reduction DirectX plug-in seem to give the best results
of them all. Unfortunately CoolEdit's plug-in processing is painfully
slow, but the factory defaults appear to be well suited to LP cleanup.
(Curiously, the presets in CoolEdit Pro are not nearly so effective,
unless they have been updated since I last evaluated it). Initial impressions
of Sound Forge's Noise Reduction plug-in were very good (because I happened
to try it first on a WAV file that had defeated all other tools), but
on further detailed evaluation I've come to the conclusion that on most
material the Vinyl Restorer and Click/Crackle Reducer perform about
on a par with CoolEdit2000's Audio Cleanup plug-in. Given that Sound
Forge itself is currently $350, and the Noise Reduction plug-in is a
further $280, this can't be considered good value from a purely vinyl
restoration perspective. One way to reduce the overall cost would be
to find a cheaper editor that can accept the DirectX plug-in. I see
that CoolEdit 2000 now supports DirectX plug-ins, but have not yet had
a chance to experiment with this facility. Note that the low-cost version
of Sound Forge itself (Sound Forge XP) cannot accept DirectX plug-ins.
Setting aside the price for a moment, on the plus side Sound Forge's
Noise Reduction plug-in executes very much faster than CoolEdit 2000's
Audio Cleanup plug-in. It has a realtime preview feature which should
make setting the parameters much easier. However, you do need a fairly
fast CPU to run it, and based on the time taken to execute the processing
I'd guess that something like a 500MHz CPU is probably required. In
summary, if you have a fast enough CPU to use realtime preview, and
are able to afford the price, Sound Forge's Noise Reduction plug-in
will probably enable you to achieve results as good or perhaps better
than CoolEdit 2000's Audio Cleanup plug-in, and with much less effort.
Another product with a good reputation is Steinberg's
WaveLab (some professional users
regard it as superior to Sound Forge), but I have not evaluated it.
The bottom line though is that there is as yet no automatic
way to remove all the clicks and pops without also affecting some aspect
of the music. I perform this step manually in most cases, by listening
to the waveform, homing in on the clicks, and redrawing the wave shape
with the mouse, interpolating out the defect, or pasting over a replacement
section of wavform from nearby. When doing this, it is best to monitor
on headphones as they are far more revealing of clicks and pops than
loudspeakers. In order to manually redraw a waveform, both GoldWave
and CoolEdit 2000 have the
ability to zoom in to individual samples and move them. However, their
interfaces for this operation are clumsy, and Wave
Repair is a much better tool for this task.
Many people have asked me for tips about how to find
defects in a waveform. The best advice I can give you is that you'll
learn through experience. This may sound like buck-passing, but actually
getting in there and playing around with WAV files teaches you more
than anything you can read. You need to acquire a gut feel for what
the various audible defects look like on the waveform, and the best
way to do this is with practice. As you gain more and more experience,
progress becomes faster and faster.
Having said all that, I can pass on a few tips:
- Obvious narrow spikes that shoot up and down very quickly are
usually easy to spot and fix. Bear in mind that the audible effect
of a big spike may include some "ringing" for several dozen samples
after its visual effect on the waveform seems to have disappeared
(in other words, you might find that it's necessary to extend the
repair further along the trailing edge than seems visually necessary).
Smaller spikes tend to be much more localised, and you can often
repair just the obvious visual damage.
- "Dull thuds/plops" and "smeared splats" tend to be much longer,
often extending over several dozen or even a few hundred samples.
Visually they often look like normal waveforms that just seem bigger
than their surroundings. Other times they are caused by a shift
up or down of the general waveform over the course of a few dozen
samples (a sudden shift up or down within a few samples gives a
much sharper click). When repairing these, it's often surprising
how much further you need to extend the repair both in front and
behind the apparent visual damage. I think this is probably because
it's much harder to be certain where the damage really starts and
ends just by looking.
- It is intuitive to try and repair as few samples as possible,
but as pointed out above, the audible glitch often extends beyond
the visual glitch, so don't be afraid to try longer repairs. When
you're looking at a waveform at a high degree of zoom, it's easy
to forget how short a time period you're dealing with.
- Visual appearance doesn't always correlate with audibility, especially
on minor damage. Things that look quite alarming can often be inaudible,
while you can spend ages trying to track down an elusive "tick"
that turns out to be an almost invisible pimple just one or two
samples long on the rising or falling slope of a much larger waveform.
Another type of vinyl artifact you might want to remove
is distortion due to damage caused by previous mistracking. Manually
redrawing waveforms certainly doesn't get you very far with this. I
have found that this kind of distortion can sometimes be removed quite
well by two of the packages mentioned above. Sound
Laundry's de-scratcher (using only the de-crackle facility, leaving
de-click switched off) can give good results, with only subtle artifacts
(the worst aspect is that vocal sibilance tend to be emphasised). WAVclean
usually removes even more of the mistracking distortion than Sound Laundry,
but its artifacts are rather more obvious, and I can only describe them
as imparting a "hollow" sort of characteristic. I have also on occasions
been able to reduce mistracking damage using parametric EQ with a very
deep notch filter at a fairly high frequency (eg. around 15kHz). This
dulls the frequency balance, so a compensatory lift somewhere around
4kHz is needed to restore some of the lost "sparkle"; it's not perfect
but it can be an improvement. My opinion is that GoldWave
has the best parametric EQ at an affordable price.
Recommended Automatic Declickers
If you insist on being lazy and using an automatic declicker,
then here is a list of packages that I consider are worth investigating.
I'll divide the list into two types. Those capable of real-time preview
have the advantage that you can listen while you adjust the settings,
which makes finding effective settings that much less frustrating:
- Sound Laundry is very
good at reducing constant "crackle", but can introduce artifacts
(hollow-sounding "pops", and an emphasis of vocal sibilants).
- Later versions of DCart
(DCart32 and later) are pretty good for general vinyl-type "hash",
and have surprisingly low artifacts. (My experience with an earlier
version of DCart was not so impressive).
- Wave Repair is more aimed
at individual distinct clicks which, provided they are detected,
are usually repaired with little or no audible artifacts. Since
it attempts to repair without artifacts, it does tend to miss clicks
that other more aggressive packages pick up. The latest version
has a decrackling feature which is fairly effective at reducing
hash and crackle, but this particular function unfortunately has
no realtime preview option.
Other packages which don't have real-time preview can be frustrating
to use, but some give results good enough that the effort is worthwhile:
- CoolEdit 2000's Audio
Cleanup plug-in gives better results than other packages, so I recommend
it despite the fact that it operates perhaps an order of magnitude
slower than other packages. The only noticable artifact is a roughening
of loud "raspy" instruments (eg. sax).
- Sound Forge's very expensive
Noise Reduction 2.0 plug-in works about as well as CoolEdit 2000's
Audio Cleanup plug-in but operates significantly faster. On fast
CPUs it is possible to do realtime preview.
- Groove Mechanic works nearly
as well as CoolEdit 2000, and has the advantage that it runs very
much faster. It has far fewer adjustments, so you can discover whether
it's going to work very quickly.
- Wave Corrector is an effective
all-round declicker with minimal artifacts. Unlike the other packages
mentioned, it allows you to manually review and adjust the correction
applied to individual clicks.
None of the packages successfully repair really big pops, which are
best tackled manually. Note that it is best to do this manual fix up
before running an automatic declicker. This is because big pops can
confuse the declicking algorithms, often resulting in their replacement
with dull thuds and splats which are far more difficult to isolate than
the original pops, thus making them harder to fix in the long run.
An Interesting Approach to Decrackling
I recently came across a suggested method of decrackling
that is definitely worth passing on. I take no credit for this method;
I first saw it described on the AudioForums website by someone calling
themselves "Younglove". It appears that the original thread is no longer
online, but Tom Sherman has found
an
archive of it and kindly passed on the link to me.
The procedure briefly is this:
- Get a noise fingerprint from the WAV file to be decrackled. In
other words, find a section that contains only noise and light crackle,
but no music.
- Use the noise fingerprint to do a noise reduction over the whole
file, but keeping just the noise (rather than the music minus the
noise). You need to do a fairly brutal noise reduction (eg. 100%
in CoolEdit). What you end up with is the noise, the light crackle,
and a bit of the music.
- Save the noise that's just been isolated, either in a file or
a clipboard.
- Run a declick operation over the noise. Be fairly aggressive,
so as to find all the crackle. You now have the noise minus the
crackle.
- Mix-paste the noise that you saved in step 3 over the result of
the declick, but invert the saved
noise that you're pasting. This causes the noise to cancel out,
leaving just the crackle, but inverted.
- Now mix-paste that inverted crackle back into the original file.
Because the crackle is inverted, it cancels the crackle in the original
file.
It really does work rather well, but remember that it only works for
background light crackle; it doesn't deal with big pops and clicks.
So how does this work? My view is that the reason decrackling
is so hard to do is because the clicks that constitute the crackle are
of low amplitude, and are easily lost within the surrounding music:
this makes identifying them very difficult. Once you've isolated just
the background noise & crackle, the click detection algorithms have
a much easier task, so they find the genuine clicks more successfully.
The quality of the noise reduction really isn't that important, and
the declicking algorithms can be less sophisticated since their task
is greatly eased.
Younglove described the process using CoolEdit, and it
is certainly very straightforward using that package since you can set
up a script to do it, but it is a very slow operation. (My old machine,
which had a 350MHz AMD K6-2 CPU, took about 6 hours to decrackle one
LP using this method. I now have a 1.2GHz Athlon CPU, and an LP can
be decrackled in about 50 minutes). In principle the technique will
work with any packages that support the necessary steps. Since CoolEdit
processes so slowly, you might find that using a combination of other,
faster, packages might be quicker in the long run, even if you have
to switch between different packages for the various stages. The only
program I know of which has a built-in decrackling facility based on
this technique is Wave Repair.
Equalisation
Some LPs do suffer from high frequency dullness, and
it's worth giving the top end a little boost. I keep on hard disk a
short section of music (extracted digitally from a CD) which I consider
to have ideal tonal balance and dynamics when played back on my stereo
system, and use this as a reference against which to compare work in
progress. The most I've ever put on is about +6dB from 5kHz upwards;
this is usually only necessary on reissue LPs that were probably pressed
from "high-mileage" stampers. In general, it's best not to fiddle too
much with the balance chosen by the people who originally made the LP.
Normalisation and Compression
Normalisation is a procedure that ensures the WAV file
peaks at the maximum possible value. If for some reason you recorded
at too low a level, then normalisation is probably worth doing.
Note that normalisation does not guarantee that all tracks
will sound equally loud; the perceived loudness is equally influenced
by the dynamic range of the music. Compression can be used to squash
the dynamic range which makes the music sound louder. It also tends
to sound more "punchy". Applying differing levels of compression can
be used to balance the loudness of tracks from a variety of sources,
but be aware that excessive compression, while sounding initially impressive,
can rob the sound of its subtlety.
Some packages I can recommend which provide equalisation,
normalisation and compression include GoldWave,
CoolEdit 2000, Wave
Repair, and DCart.
One situation in which normalisation and compression
are likely to be useful is when you are compiling a CD from a variety
of sources and wish to make all the tracks sound equally loud. The approach
to take here is to normalise all the files, select one as a reference,
and then apply the appropriate amount of compression to the other files
so that they sound as loud as the reference. This is easier said than
done; finding the appropriate compression settings is largely a matter
of trial and error. I have written a shareware program called Volume
Balancer which automates this process. I've deliberately made it
easy to use; having selected the files to be processed, a single button
press does the job. There are some other programs around (eg. Audiograbber)
which give the user more control, but require that you adjust a variety
of settings.
Burning the CDR
Splitting Tracks into Separate WAV Files
Depending on your choice of CDR burning software, you
may need to split the individual tracks into separate WAV files. This
can be done with a wide variety of WAV file editors (eg. GoldWave and
CoolEdit), but the task is much simpler using either
CD
Wave or
Wave Repair.
Using a Cue Sheet to Identify Tracks
On the other hand, you may have a CDR burning package
such as GoldenHawk's
CDRWin
that will place track (and maybe index) marks within a single WAV file.
In this case, you will need to prepare a suitable definition of where
those marks should be. As with track splitting,
CD
Wave and
Wave Repair make
the creation of CDRWin cue sheets simple. Note that if you wish to set
indexes as well as tracks, Wave Repair supports them but CD Wave doesn't.
Stripping Out Headers and Trailers, Padding Blocks
I've never come across one, but have heard rumours that
some CDR burning software fails to ignore the WAV header, which must
be stripped from WAV files before burning. More likely is that an incorrect
WAV header might not be noticed by the CDR burning package, which thinks
it is audio data and puts it on the CDR. Some WAV file editors place
housekeeping information at the end of the WAV file, and this too may
need to be stripped depending on the burning software you use. There
is a utility called StripWave which can help here, and can be found
at
Mike Richter's site.
The Issue of CD Block Size
CDs are organised in "blocks". Each block is 1/75th
second (which equates to 2352 bytes of data). A track on a CD must contain
an exact number of blocks when it is written. Therefore, if the audio
data in a WAV file is not a multiple of 2352 bytes, the end of last
block might be left as garbage, resulting in a small click on playback.
Most burning software will pad out the last block with zeros. The way
to avoid any problems due to this issue is to make sure that all your
WAV files are an exact multiple of the block size. Packages which split
tracks into separate WAV files know about this, and ensure that the
splits are made on block boundaries.
The Actual Burn
This is pretty straightforward. Just make sure you don't
do anything that might interrupt the data flow to the burner. Things
like screen savers, auto-answer modems, email servers and the like should
be switched off for the duration of a CDR burn.
The latest hidden menace is FastFind. This is a little
applet that comes with recent versions of Microsoft Office that wakes
up now and again and scurries around your hard disk building info for
later use. Needless to say, FastFind consumes vast amounts of resources
while it's awake, and naturally it'll be your luck that it will kick
in while you're trying to burn a CDR. Kill it, by removing it from your
system's Startup group.
If you've read earlier versions of this page, you'll
know that I used to burn CDRs from DOS. I took the view that burning
a CDR is a realtime operation, and DOS is the only (vaguely) realtime
operating system available for the PC. Lately, however, I have started
burning direct from Win9x (using CDRWin) and not had a failure yet.
IDE hard disks are perfectly capable of delivering the
data rate required to burn a CDR, even if you burn at 4x speed. Take
no notice of those who say you have to have SCSI disks (not that I have
anything against SCSI). Likewise with so-called "A/V" disks, which are
unnecessary for this kind of work; in any case the distinction between
"ordinary" and "A/V" disks seems to have disappeared in the last few
years. Some people will claim you need to defragment your hard disk
before burning, but modern disks have such low seek times that I don't
bother with this any more, and I've not had a coaster yet. You may hear
horror stories about "thermal recalibration" interrupting the data flow.
This is largely a thing of the past (having been superceded by "embedded
servo") so you probably don't need to worry about this. I'm not aware
of any current production hard disks that still use thermal recalibration.
I'd like to pass on a tip from Martin Audet. This is
only relevant if you use an IDE CD writer. (Since my CD writer is SCSI,
I'd never considered this issue). Martin advises that you should keep
the source hard disk and the CD writer on separate IDE buses. If you
put them on the same bus (eg. hard disk as master, CDR as slave), then
the chances of data interruption and hence buffer underruns is higher.
Track-at-once burning can be used, and with the variable
gap capabilities of some hardware and software can be made to approach
disc-at-once results, but frankly all this fiddling about is just skirting
around the basic issue, which is that audio discs are best made in disc-at-once
mode, period. So don't buy a writer that doesn't support disc-at-once,
ok?
Regarding CDR burning software: for audio CD creation,
I very strongly recommend CDRWin
(and its DOS version, called DAO). CDRWin does one thing very well:
it writes CDRs in disc-at-once mode, and allows you to lay out the tracks
and indexes exactly how you want. Unlike other packages, it doesn't
make any decisions for you. I like to stay in control. A word of warning
is in order, though. Some people have found that they are unable to
get CDRWin working on their particular systems. This is usually a SCSI
configuration problem, but the real issue is that Goldenhawk's support
can sometimes be rather unsympathetic. Because of this, my advice is
that you should download the demo version of CDRWin and check that it
works in your system before actually buying it. (This is the rule for
any shareware: don't pay for anything you've not been able to evaluate
and verify works as you require).
Roxio's Easy CD Creator
is bundled with many CDR drives, and its basic disc writing engine is
very solid, provided it works on your system. (There have been many
stories from people who were never able to get Easy CD Creator working,
and the general conclusion seems to be that it will either work straight
out of the box, or it'll never be right). Although it is somewhat lacking
in flexibility, it is fine for creation of straightforward audio CDRs.
Two other packages which have good reputations are Nero
and Feurio, but I have no experience
with them.
More in-depth discussion of a wide variety of CDR writing
packages is at come.to/satcp.