You are here

Improve Your Stereo Master With Post-Production Sweetening

Tips & Techniques By Paul White
Published August 1995

Paul White presents a few low calorie tips on how to sweeten and edit your sounds after mixing them.

It's tempting to think that once you have mixed your stereo master, that's it — you can go ahead and put it on a record or CD, but that's not the end of the story. If you're recording an album then the tracks have to be compiled in the right order, with the right gaps between them, and at the correct respective levels. That part of the process has been covered in SOS before, and if you're working from DAT (or even analogue cassette), hard disk editing is the most practical way to do the job. If, on the other hand, you're working with an open‑reel mastering machine, you'll need to physically splice the tracks together inserting a couple of seconds of blank tape or leader tape between the tracks. Other than rearranging the songs into the chosen order, what else can be done at the post‑production stage to actually improve the sound of the recordings?

One of the problems with masters made in small private studios is that most have less than optimum monitoring environments, and this in turn leads to mixes which require further work on the tonal balance. To improve on this, you ideally need a good parametric equaliser and a monitoring system that you know to be accurate; one that you are used to working with. There's no reason at all not to use an analogue equaliser for the job, and some of the more esoteric tube models (like Drawmer's 1961 valve equaliser, reviewed in May 1994's SOS) can inject a little of their own personality into a mix. Since setting up my own Sound Tools hard disk editing system, I've been using the Waves Q10 software parametric equaliser quite a lot and find that it's every bit as easy to use as the hardware equivalent. However, what's really important is not which equaliser to use, but what to do with it.

In the case of mixes done in rooms with poor monitoring systems or acoustic problems, the most serious errors tend to occur at the bass end of the spectrum, so you're likely to end up with a stereo master that has too much or too little bass. In a typical pop record, the bass drum sits in the 80Hz part of the spectrum, so you can do quite a lot to balance the subjective bass drum level using a parametric (or even a sweep filter) tuned into this region. However, there's quite often stuff going on below 80Hz, and if there's too much down there, you can end up with a very ill‑defined, rumbly bass end that's best treated using a shelving filter to attenuate everything below 60Hz or so.

For certain musical styles, dance music being one, your mixes may actually rely on deep bass sounds, but when you try to emphasise them using a simple shelving control, the lower mid starts to sound a bit boxy. A parametric EQ tuned to between 40 and 70Hz may produce the desired result, but if you have to use your mixing desk equaliser, try adding around +8dB of shelving bass boost and combine this with around ‑8dB of sweep lower mid cut, tuned to between 150 and 250Hz. This lower mid cut counteracts the bass EQ's tendency to spill over into the mid range, which enables you to add a lot of deep bass without swamping the lower mid range.

Attention To Detail

At the high frequency end, lack of brightness or detail can be a common problem, especially if the recording was made using analogue multitrack and several tape bounces were involved. It's tempting to simply turn up the treble control, but this will also bring up the noise and may not produce the tonal result you need. What can work better is to identify the high frequency sounds that need help, and then use a parametric EQ or sweep mid EQ to lift these slightly. The sound of cymbals or the pluck of an acoustic guitar can be enhanced by boosting between 4 and 8kHz, and because you're only boosting a relatively narrow band of frequencies, you bring up less background noise. Be careful when adding high frequencies (or 'top') as it is too easy to make a mix sound harsh or aggressive. As a rule, the higher the boost frequency, the more 'air' you get around the sound. Boosting lower down, say at 3 to 4kHz, can bring out the bite in electric guitars and similar sounds, but it also tends to make the top end sound harsh.

The other popular way to enhance the top end detail is to use an exciter or enhancer of some kind. Most models work well if used sparingly, but my favourite is still the SPL Vitalizer (reviewed in May 1992's SOS), which includes a very effective bass enhancement system as well. I can't emphasise too strongly that enhancement needs to be used with restraint, so keep switching the process in and out as you're setting up to make sure you haven't gone too far. Comparison with records and CDs is recommended at this stage as your ears easily grow used to treble, and you could end up with a lethally bright mix without noticing that anything is wrong.

Compression & Limiting

Some producers like to compress their finished stereo mixes while others claim that this always makes things worse. The truth is that it all depends on the type of material you're working with. If you want to maintain a high energy feel, then a few dBs of compression won't go amiss. Soft‑knee compressors seem to give the smoothest results on complete mixes, and if you can get hold of a model with an auto attack/release option, you may find this works better than setting up a fixed attack and release time. If you have to use manual settings, use a fast attack time (between the fastest available and 1 millisecond), and try release times of between 0.2 and 0.5 seconds to see what sounds best. Keep in mind that compression will bring up the noise level during quiet sections.

One of the problems with masters made in small private studios is that most have less than optimum monitoring environments, and this in turn leads to mixes which require further work on the tonal balance.

Limiting is also useful on finished mixes, because you'll very often have a few really high level, short duration peaks that force you to keep your average signal level low to avoid clipping. Because these peaks are so short, they can often be limited quite severely before any audible difference is evident, but you'll need a limiter unit with a very fast attack time and a fast release time to achieve this. You can usually pull the peak levels down by at least 4dB without changing the sound, which means your average signal level can be increased by the same amount.


If you're working with a hard disk recording system, you may be tempted to 'normalise' the data to ensure that the highest signal peak comes right up to the digital maximum. If you do intend to normalise, there are a couple of points you should keep in mind. Firstly, always normalise after you've done everything else, otherwise you may perform a subsequent process that increases the gain further, thus forcing your signal into digital clipping. Even an EQ setting that produces only cut can increase the signal level slightly, simply because you may end up reducing the level of some frequency components that were previously helping to cancel out peaks at other frequencies. Apparently this is also true of sample rate conversion, because of the filtering involved.

Secondly, if you have the option, try to normalise to a dB or so below the maximum peak level. For reasons similar to the previous example, clipping may occur when a normalised signal is passed through subsequent digital systems (such as oversampling DACs). In practice, this source of clipping rarely causes audible problems, but if you have the option, then why not be a perfectionist?

Every time you manipulate the level of a signal in the digital domain, some resolution is lost; for example, if you add EQ, the signal level is likely to be increased, so it has to be scaled down again to fit into the original 16 bits. The practical outcome is that if you have several stages of digital processing to go through, low level detail may suffer, resulting in reverb tails becoming less smooth or the stereo imaging becoming blurred. This problem may be countered by adding 'dither' to the signal, which effectively adds a low level of noise, with the end result that low level resolution is improved, but the signal‑to‑noise ratio suffers slightly. A more sophisticated way to do this is to use so‑called 'noise‑shaped dither', which is mathematically designed so that the added noise components appear at the high end of the audio spectrum where the human ear is relatively insensitive. Sony's Super Bit Mapping system is an example of noise‑shaped dither. Once again, if you're using something like Digidesign's Sound Tools, the L1 plug‑in limiter from Waves includes a very sophisticated dither system (developed by AES guru Michael Gerzon) that does the job very nicely. The limiter section of the L1 package is also very effective in trimming short duration peaks down to size, as discussed earlier. As an added bonus, the L1 can normalise the gain of the signal being processed to any maximum level at the same time as limiting, which can save you a lot of time.

Reducing Noise

Good recordings shouldn't be noisy in the first place, but budget effects, noisy synths, and ground loop hums in home studios invariably mean that your recordings will contain some audible noise during pauses or quiet passages. A great improvement can be made by ensuring that your track is completely silent until the music starts.

If you're splicing analogue 2‑track tape, this simply means taking care to make your splice half an inch or so before the song starts. Rocking the tape over the heads manually is the best way to identify this point, and you can mark the back of the tape with a wax pencil to show you where to cut.

If you're using a hard disk editor, life is rather easier because you can actually see the waveform of the first sound in the song. It is then simply a matter of selecting the appropriate area and using the 'silence' command to replace it with digital silence. At the end of the song, you can also perform a digital fadeout at the tail end of the natural decay of the last sound, so that the song fades into true silence rather than hiss.

Noise which occurs during the track is less easy to deal with, and to tackle serious noise effectively, you really need a specialised digital noise removal system such as CEDAR or Sonic Solutions' NoNOISE. Less severe noise contamination can be dealt with in the analogue domain using a single‑ended noise reduction processor such as the Drawmer DF320, Symetrix 511A, or one of the Rocktron, Behringer or dbx units. All these units work by monitoring the input signal's level and frequency content; when a low level signal with little or no high frequency content is detected, a variable frequency low‑pass filter moves down the audio band to filter out the noise. Obviously the filter has some effect on the wanted signal, so the trick is to set up the unit so that it's only coming into action on very low level signals. By switching the bypass switch in and out, you can hear if the sound quality is suffering and adjust the threshold level accordingly.

Sound Tools users have access to noise removal software such as DINR, which isn't as sophisticated as a CEDAR or NoNOISE system but is rather more effective than analogue, single‑ended, noise reduction units. DINR behaves like a bank of expanders, each covering a very narrow part of the frequency spectrum. When the signal in a particular band falls below the noise threshold, the expander mutes the signal. Because this happens over many independent frequency bands, it's quite possible to expand out the noise in one part of the spectrum leaving other parts unaffected.

To use DINR most effectively, it's best to let the software analyse and 'learn' a section of noise‑only recording from immediately before the start of your song. The resulting noise spectrum is then used to set the correct threshold for each of the multiple expander bands, though the user can alter these parameters if required. In practice, DINR is good for between 5 and 8dB of noise improvement before any side‑effects become evident, and though that might not look much on paper, it can make a huge difference to the subjective sound. Be careful though; over‑processing can cause the noise to take on a ringing character as frequency bands in different parts of the spectrum turn on and off.

Artistic Considerations

When compiling an album, there are artistic as well as technical decisions to be made. For instance, the relative levels of the tracks need to be balanced so that they sound right — if you simply normalise all the tracks individually, any quiet ballad will sound unnaturally loud next to a heavy rock track. The only way you can get this right is to use your judgement, but a good guide is to make sure the vocal levels are subjectively similar from track to track. Once again, the old trick of listening from outside the room as one track plays into another will help you recognise any obvious level problems.

A related problem arises if you're using tracks recorded at different sessions or even in different studios. Do you want them to sound different or do you want to EQ them to make them all sound similar in tonal character? Again it's your choice, but you'd be amazed how different recordings of the same band can sound.

Finally, judging the length of the gaps between songs is an art in itself. The usual gap is between two and four seconds, but a lot depends on whether the previous track has an abrupt stop or fades out, and on how different the mood of the two songs is. Fortunately, the correct gap is usually pretty evident, and I've found that even when several people are involved in a session, the chances are that they'll all agree on the optimum gap length to within half a second.

And that's about it. Don't forget to make a backup copy of the tape before you send it away for duplication, and if you're making a cassette master, it helps if the two sides are as equal as possible with side one being the longer if necessary. Leave a gap of at least two minutes between the two sides of the cassette, and if you're mastering to DAT, don't just wind the DAT on, leave silence, otherwise you could confuse the DAT machine at the duplicating plant. If you're mastering for CD, then you don't need to leave a gap, though you should always leave a minute or so of recorded silence at the beginning of any DAT tape, just to ensure you don't get any dropouts or errors. And most importantly, try to record the master DAT at 44.1kHz sampling rate. If your machine only works at 48kHz, then mark the tape clearly as such and you should have no problems. However, if you send away a tape for CD mastering which contains mixed 44.1kHz and 48kHz sample rates, you only have yourself to blame if some of the tracks on your CD album play back 10% slow!

Avoiding Sample Rate Problems

Never mix sample rates on a DAT because a digital editing system will simply take in all the data at a single sample rate, resulting in noticeable speed errors on the offending tracks. I run my own Sound Tools editing suite and I've been caught out before when bands have brought me a tape made at other studios. You ask them what sampling rate the DAT tape is recorded at and they usually reply: "What's a sampling rate?"

If their DAT contains material from two or more studios, there's a good chance that the sample rates could be mixed up. I get around this problem by connecting an Alesis AI‑1 sample rate converter between my DAT machine and the input to my Sound Tools system, so no matter what is on the DAT tape, I end up with a 44.1kHz master running at the right speed.

Another potential pitfall presents itself when you get a DAT tape to edit that was made on a Casio or an older Tascam portable DAT recorder. These two machines record using pre‑emphasis (a simple kind of noise reduction), which means that the digital data on tape actually has a significant top boost which is filtered out again on replay. If you transfer tapes made on these machines into Sound Tools via the digital domain, the emphasis flag is stripped out and you end up with a horrible, toppy sounding master. I asked Digidesign's engineers why the ability to pass on the emphasis flag (present in their old Sound Accelerator system) hadn't been retained, but their reply was, "You're the tenth person we've told today, there's simply no call for it." I also asked them why the system couldn't put up a warning if it saw the input sample rate change, but they said this wouldn't be easy to do.

Until now, the only solution to pre‑emphasis has been to transfer your DAT recording via the analogue inputs so that you're loading in the de‑emphasised audio. However, there's now a de‑emphasis setup in the Waves Q10 digital EQ library that can undo pre‑emphasis entirely in the digital domain. If you have a Q10 plug‑in, I'm sure they'd be delighted to send you the setup via CompuServe.

Useful Contacts

DINR Digidesign UK. Tel: 0171 494 2949.

WAVES PLUG‑INS — Natural Audio. Tel: 0181 207 1717.

SPL VITALIZER — Beyerdynamic UK. Tel: 0800 374994.

DRAWMER 1961 — Drawmer Distribution Ltd. Tel: 01924 378669.

PHONIC PCL 3200 — Audio Awareness. Tel: 0181 598 8081.

CEDAR 2 SYSTEM — Cedar Audio Ltd. Tel: 01223 464117.

SONIC SOLUTIONS NO NOISE — Tyrell Corporation. Tel: 0171 287 1515.