Recommended discs: HHB silver, or Tayo Yuden. Always use CDRdiscs, not CDRWs.

You should write CD masters at 1x or 2x speed, and certainly never faster than 4x. Audio masters should be written from a disc image file if possible.

First, Some Background...

Briefly, what happened was that three tones were recorded at 0VU: these were generally at 100Hz, 1kHz, and 10kHz, though there were some really keen types who would include additional frequencies. The playback deck was lined up to the EQ and azimuth of these tones, and everything was generally tickety-boo. In a perfect world these tones would have been unnecessary, but since no two analogue tape machines have identical record/playback characteristics, they were a godsend. Well, they were just as long as the engineer actually balanced the recording up to them (broadcast engineers hardly ever did it properly, of course!) 

When it came to mastering, all this lining-up took forever: imagine making a compilation of fifty tracks from fifty different artists, recorded in fifty different studios on fifty different tapes with fifty different test tones... Some tapes would be wound head out, others tail out; many would be incorrectly marked, some would have 0VU set differently to the Dolby tone, some were marked Dolby A but were clearly non-Dolby, others were the other way round, and a few would even come in double-Dolbied. Ah, those were the days! But the trouble was that it took three or four minutes to load each tape and line up the machine, which added a considerable time to mastering sessions. These days it's much quicker and easier for us, as most clients come in with digital masters. The move from analogue tape to digital (first F1, then DAT, and now CDR) instantly meant two things: firstly, the recordings would sound basically correct on any playback machine (though there were minor oddities like CTC on the F1, and rather dodgy circuits and transports on some DAT machines), and there was generally no hiss worth mentioning of course. Secondly, engineers had to quickly learn what volume really meant. 

Well, which meter is accurate? Which one best resembles the human ear's response to the sound of a triangle? In my experience a PPM gives a more accurate indication of the loudness of something very percussive like the triangle, while a VU meter is far better for more general sounds like an orchestra or band. A professional digital meter will always make instruments like a triangle appear far louder than they are to the ear, but with digital you just can't push the levels into the red... 

There is often a real difference between perceived volume and absolute peak level, as the triangle example shows; this is further complicated by the variation in the ear's sensitivity across the frequency range. When you record onto analogue, you should try and keep the peak level to within 8dB above 0VU: as you go above this, you first get some gentle tape compression (quite useful, actually!) but it soon deteriorates into a nasty saturation distortion as the levels increase. With digital recordings you don't have any luxury: instead there is a very clear limit to the maximum recording level -- by and large, if you pop it into the red you'll get some very alarming distortion. 

Human ears aren't digital: they're pretty insensitive to changes in level (which is why the decibel is a logarithmic measurement), so the trick with digital recordings is to choose a suitable operating level on the meters which corresponds to 0VU analogue. This level will, ideally, give you a headroom of 8dB PPM before the red clip lights come on. Remember, though, that because of the almost instant rise time of digital meters, you will need to set the operating level below -8dBfs (dB full scale). For most music and speech recordings -12dBfs should do fairly well unless it hasn't been limited at all, when -15dBfs might be better. Conversely, heavily-limited material can be easily set to -5dBfs. The BBC specify -18dBfs for transmission tapes. 

Many engineers deliberately don't set their original masters to peak up to 0dBfs: this leaves space for more sophisticated limiting or compression by the mastering engineer. Without this room for manoevre he may have to drop the absolute peak level of a heavily-compressed track in order for it to match the perceived loudness of other tracks. The general rule is don't over-compress the whole mix: far better to use your compressors on the channels that need them, and leave the final dynamics to the mastering session. 

While we're on the subject of levels, it's not a good idea to take the absolute peak level of digital recordings quite up to 0dBfs: go for -0.5dBfs if you can, to allow for the slight differences between machines: this safety margin prevents unnecessary false alarms over levels. At RPM, using our 24-bit digital feed-forward limiting, we can automatically control peaks to within this level: something which was undreamed of just a couple of years ago. 

48kHz is the standard for "consumer" DATs and "professional" DATs, while 44.1kHz is the rate used by CDs. "Professional" in this context mainly means broadcast, film and video, and "consumer" means any recorder that doesn't have XLRs on the back. CDs, apparently, are neither consumer nor professional. Quite how this ridiculous state of affairs came about eludes me: I'm sure someone had a reason for it, but it seems nuts (a bit like the XLR fiasco: pin 2 hot sometimes, and pin 3 hot at others: woe betide you if you don't have a phase meter on the console!) 

By the way, consumer DATs generally only record from analogue at 48k; they may also have an emphasis circuit which you can't disable, and either copy protection or SCMS. Some of them have little surprises as well, like not recording from a professional digital source (they need the consumer flag set in the subcode data), and their transports often aren't up to much, which is a pain when you try and play the tape in another model. Oh well, they're a lot cheaper than the pro DATs, which is why so many studios use them, so we keep a few of the better consumer decks in the studio here for playing back difficult tapes. Because of this cheap price, many DATs come to us recorded at 48k, which of course means they have to be converted to 44.1 in order to go onto CD. We convert sample rates digitally using state-of-the-art equipment, but even then you can hear a slight difference in the converted sound: it's far better if you can record at 44.1kHz right from the start. 

If you're setting up a small studio where the budget won't stretch to a professional DAT machine, do consider MiniDisc instead. We find they give far fewer problems than dodgy DATs, and the slight drop in quality you get from the lossy compression system is certainly preferable to DAT glitches! On the whole, though, more and more systems now assume that you'll record your final master to CDR: choose the most expensive model that your software will support, use the best media, and write in Disc-at-once at 1x or 2x speed.

Many smaller studios have only two or four channels of compression available at the time of the mix. Under these circumstances it's better to use the compressors to control individual tracks on the desk (eg drum submix, bass and vocals), then leave the overall mix compression to the mastering engineer. We can't easily save a mix which has been over-compressed generally but has a vocal that goes up and down like a yo-yo. If you do decide to compress the full mix, don't overdo it (see how loud is loud? above). 
 

2. Mix instrumental, b/track, and acapella versions.

It makes any subsequent editing so much easier to have four versions of each song: a. full vocal, b. backing track (same as a. but no lead vocal), c. instrumental (no vocals at all), and d. acapella (only vocals). 
 

3. Make a safety copy.

Never come away from the studio with only one copy of the mix. With an automated mixing console, you should be able to make two original masters; if not, then make a safety copy of it. It's a good idea to make the two on different formats such as one on DAT and the other on CDR. 
 

4. Record some room tone.

Room tone, ambience, (and applause in case of live recordings), is an essential component when it comes to editing many types of material, especially classical and jazz music, and all live recordings. It means the mastering engineer can compile the project in a far more natural way than if he has to fade quickly to silence, or if he has to insert ambience from another recording altogether. The ambience should be recorded with the same microphone setup as the programme itself, or the perspective may be difficult to match up. 
 

5. Should you tidy up the original master?

There are two schools of thought here: one says you should present an original master at the CD mastering session, with everything sounding as much like the finished project as possible. The other says you should leave many things up to the mastering engineer: fades, gaps, edits, levels, balances, dynamics, the lot. Well, both have advantages and disadvantages, not least in terms of cost! My advice is this: you should do as much as you can, but should leave anything dodgy for us to sort out. If you're unsure about an edit, for example, then do it anyway; if you also have another recording of the two edit sections (a minute or so either side of the edit), so we can have a go at re-editing it using the sophisticated crossfade techniques at our disposal. Do not top and tail tracks too tightly, and ideally please allow fadeouts to run on for a couple of seconds longer than you would like on the finished master. 

We have a philosophy of trying to save our clients' time and money: if they come in with a master 99% fnished, that's fine with me. On the other hand, we have the equipment and the experience to sort out many disasters! Don't be overawed at the prospect of a mastering session: it's not a black art these days, and clients are far more knowledgeable about it than they used to be. Certainly, if it were *my* music, I'd want to remain in control. Thank God mastering engineers no longer wear white coats! 

 Robin Springall