You are here

Making The Transition To Computer-based Digital Recording: Part 2

Tips & Techniques By Paul Wiffen
Published January 1999

If you're monitoring the audio output from your computer while recording tracks, latency can be very disconcerting.If you're monitoring the audio output from your computer while recording tracks, latency can be very disconcerting.

After the general guidelines on the approach and attitude you should adopt to a computer‑based audio system last month, Paul Wiffen gets down to the nitty‑gritty of spec'ing your system and setting it up for minimum hassle and maximum results. This is the last article in a two‑part series.

The first thing to look at when moving over to a computer‑based digital audio system is what you expect it to do for you. This may seem obvious, but all too often peoples' expectations are unrealistic, ill‑defined, misguided or unfounded. Depending on which adverts and brochures you have perused, who you have talked to and (with a bit of luck) which articles you have read carefully, you will have a set of expectations of how this technology is going to help you in your music‑making. Try making a list of what you expect from your new system; it concentrates the mind beautifully and it will give you a specific set of questions to ask salesmen, product specialists and manufacturers' representatives.

Some of the things that you may have heard might be dismissed as insignificant by those of us who are already familiar with the technology, but if getting rid of tape‑rewind time is important to you, write it down (it certainly seemed to feature enough in New England Digital demonstrations of the Synclavier). But every digital audio system will give you this, so do try to find some more stringent requirements to go with. This will help make sure that you don't end up with a system which seems less suitable for your needs than whatever you were using before.

If you are now staring at a blank piece of paper, unsure of exactly what benefits and features you are expecting from a computer‑based digital audio system, then I suspect you are like the majority of people making the move. The marketing men have done a great job of persuading people that this is the way forward, without being too specific about what the exact benefits are. For those of you who find themselves hard‑pressed to define their precise expectations, the headings in this piece will help you (or at least flag up some of the issues you need to be aware of). Those of you who have been more succesful in formulating your needs can use these headings as shortcuts to the most important issues for you. However, it is worth reading all the sections at some point, because issues which were simple and obvious in the world of tape may or may not be more complex with computer‑based digital audio. The first one I want to deal with is a case in point.

Track Count Versus Numbers Of Inputs/Outputs

Korg's 168RC digital mixer offers 16‑channel ADAT‑format connections as standard.Korg's 168RC digital mixer offers 16‑channel ADAT‑format connections as standard.

In the good old days, you could be pretty sure when buying a 16‑track tape machine that you would get 16 inputs and 16 outputs. Given the hardware of such machines, it was the only way to do things. But because the datastreams inside computers do not require separate pathways and can be routed around at will, gone is the old exact correlation that used to exist on tape machines between number of tracks and the number of inputs and outputs. The average computer (be it Mac or PC) comes with just two inputs and outputs, usually on mini jacks or, at best, co‑axial connectors. Despite this, the latest generation of digital audio sequencers are notching up potential track counts (in conjunction with the most recent processors and clock speeds) approaching 100.

Now this is not necessarily a problem, as programs like Cubase VST or Logic Audio probably now have more EQ and effects‑processing potential than any analogue mixer you ever owned (and the buss structure of Cubase VST v4.0 probably gives you greater routing flexibility as well) — but if you are going to limit yourself to doing everything inside the software, it will require a fairly major upheaval in your working methods. Some people take to it like a duck to water (usually those who have never had a large physical mixer available), but others never get used to controlling software faders with a mouse, or not being able to use their favourite existing effects and/or mixer. If you don't want to get rid of those favourite analogue effects (or for that matter, digital effects with only analogue ins and outs) or the traditional mixing desk you have just got your head round, you will have to consider expanding the number of inputs and outputs on your computer.

Of course, you can always send out just the track (or tracks) which you want to EQ or add effects to, route the signal through the analogue mixer channel or effects unit and then re‑record it back into the computer. Provided you have enough hard disk recording space available, this is a perfectly acceptable way of working, especially as you can keep the original recordings in case you change your mind (unlike on analogue systems where you would have had to erase the original if you wanted to use that track to record something else). The main reason to go with some expansion of the number of outputs in your system is if you want to be able to use external effects and channel EQ at the mixdown stage, keeping your options open to the very last. In other words, if you have a Mackie 1604 and you want to be able to mix down 16 of the tracks in your computer through it, you need to find a way to add 16 outputs to your system.

If your interest is more in the ability to alter mixing and effects parameters instantly using real knobs and faders, you needn't worry so much, as it is perfectly possible to use hardware faders via MIDI controllers to alter mixer or FX parameters. However, you then need to budget for a hardware controller from the likes of JL Cooper, Peavey or Penny & Giles in addition to the basic cost of your computer, and software to enable you to do this.

There are many cards available on the market which offer eight or even more analogue outputs, but before you rush out and buy the one with the most outs, stop and consider what you need in the way of inputs. If you suspect you may ever want to record more than two tracks at once, how can you do this if you've only got a stereo input?

When you process digital audio through the main CPU of a computer, this introduces a delay, generally referred to as 'latency'. The first time you might have come across this term was when you were ringing up tech support to find out why you are hearing a delay on your guitar when monitoring through your software.

The answer, of course, is that you can't! The solution to inputting multiple channels parallels that for outputting multiple channels through an external mixing desk — hardware cards, and those with as many inputs as outputs tend to be somewhat more expensive (mainly because the analogue‑to‑digital converters required for inputs tend to cost more than the digital‑to‑analogue converters required for outputs). You will need to make sure that if you need this facility, you have the extra money in your budget — and the extra power in your CPU (see the section on 'How Much CPU Power?').

Remember also that if you want to integrate external effects units (either analogue or digital) into your system, then you need to allocate as many inputs to the computer as outputs (twice as many if you have mono in, stereo out processors). For this reason I personally favour cards with as many inputs as outputs, as much for the ability to set up effects loops at mixdown as for recording multiple tracks simultaneously.

Bear in mind that if you suddenly decide you want to record a drummer across multiple tracks, you will need a lot more than just multiple inputs. With the need for numerous different microphones, possibly with phantom powering, compression (and the expertise to set it all up), not to mention a suitable environment to record in, you might feel that the best solution is to go into a studio with either a Digidesign system or digital audio tape for a few hours, and then transfer the results across to your system as AIFF files on CD‑R or removeable disk, or via a digital interface (see 'Keeping Analogue Audio Outside The Computer'). However, you probably don't want to be forced into a studio every time you want to use external effects in a mixdown, so a multi‑input/output card can still be useful even if you don't want to record multiple sources simultaneously.

Avoiding Latency Problems Whilst Recording

This is probably the least talked about aspect of using recording software which runs on the main processor of the computer (as opposed to more expensive hardware systems which only use the computer as a user interface). When you process digital audio through the main CPU of a computer, this introduces a delay, generally referred to as 'latency'. The first time you might have come across this term (if you weren't reading this, that is), was when you were ringing up tech support to find out why you are hearing a delay on your guitar when monitoring through your software whilst recording (similar to monitoring off the playback heads whilst recording on an old analogue multitrack). I know many people who have ended up in this situation.

A hardware controller like the Kenton Control Freak can serve as a substitute for a traditional mixer control surface.A hardware controller like the Kenton Control Freak can serve as a substitute for a traditional mixer control surface.There are two way of dealing with this; monitoring at source or using a low‑latency digital audio card with an ASIO driver. The first is probably the cheapest and easiest, especially if you already own a mixer with recording busses or direct outs, and involves listening to the signal before you send it into the computer to be recorded (and disabling monitoring in the software). However, this doesn't allow you to hear some of the glorious plug‑in effects which are now available whilst you are actually recording. This can be a major problem with some effects (the most obvious being guitar amp simulators), as playing without the effect makes it difficult to play the part.

The more permanent solution is to get a digital audio card with a low latency (these take on some of the work of the CPU to throughput more quickly) and, more importantly, its own ASIO driver. ASIO is the system developed by Steinberg to allow cards with multiple inputs and outputs to talk directly to VST without having to pass through lazy software interfaces like Windows MME, or ones which currently only allow stereo in and out like Apple Sound Manager, which were never designed to cope with the real‑time needs of the recording musician. Originally available only when using Cubase VST, ASIO is now supported by all the other major digital audio sequencers on the Mac (or soon will be, in the case of Logic Audio 3.6, which should be shipping by the time you read this). By using an ASIO driver, cards like the Korg 1212 I/O or Digidesign's new Project II PCI card reduce the latency down to 512 samples or 11.6mS. One point of information: as a general rule, latencies always seem to be higher on the PC, so if this is critical to you, you may want to go for a Power Mac G3. Remember, though, that this problem only occurs when monitoring through the software whilst recording. If you can monitor at source, you don't need to take latency into account.

How Much CPU Power?

Almost every one of the wonderful features of digital audio sequencers (such as very high track counts, up to four EQs per channel, multiple plug‑in effects algorithms, multiple input and output bussing, and so on) cause a noticeable drain on CPU power. In the more expensive systems where additional hardware is provided to do the recording and playback, the CPU power of the host computer only needs to be powerful enough to run the minimal demands of the user interface software. But when the digital audio sequencer is running just on the CPU of the computer, then the power of that CPU becomes critical.

Clearly the number of audio tracks you are attempting to play back is a major factor in establishing your CPU power requirement (if you want to get anywhere near the maximum of 96 audio tracks which the Mac version of Cubase VST/24 boasts, for example, you had better get yourself the 333MHz G3). But the demands of playing back tracks can be reduced by getting yourself a faster hard drive and more memory (provided of course you remember to increase the size of the memory buffer for each track in the software). The same cannot be said for the EQ and effects processing. Each EQ module and effects processor requires a fixed amount of power which cannot be mitigated by faster drives or increased memory. Of course some plug‑in effects take less power than others but in general, the more power they take the better they sound. Whilst you may be happy to put on the basic reverbs supplied free with your main program on the less important elements in your mix, do you really want to have to use them on the lead vocal or snare drum simply because you don't have the power to run a better sounding algorithm which you have paid for?

Multiple inputs and outputs are another source of CPU power drain. For each, the CPU has to generate a separate datastream to go to the PCI or ISA buss. There are some cards out there now (both ISA and PCI) which provide an awful lot of additional inputs and outputs and some very low latencies. But you don't get anything for nothing in this world. The more channels of I/O you want and the lower the latency you require, the more demands this will make on your CPU power, which you might have been reserving for some splendid plug‑ins or tons of EQ.

As a general rule, the newer cards coming onto the market now are much more efficient (they won't hog as much processor time as those that have been around for a couple of years), but you still need to make allowances for the fact that the CPU has to handle an extra datastream for each extra input and output. Cubase VST, unlike some other digital audio programs, allows you to switch off inputs and outputs you don't need to allow more power to be directed elsewhere. So if you need eight inputs whilst recording, switch off any additional outputs (and any plug‑ins you are not actually recording through) and just monitor through the stereo Master Out. Once you have finished recording, turn off these inputs and then you can go mad with separate outputs and plug‑ins.

MOTU's Digital Timepiece sends both SMPTE timecode and word clock, so that MIDI and digital audio can be synchronised.MOTU's Digital Timepiece sends both SMPTE timecode and word clock, so that MIDI and digital audio can be synchronised.However, it pays to investigate the power drain of the card you are looking at before you buy (preferably before you buy the computer as well, so that you can make allowances for the power the card needs in the spec you go for). That way, you won't find yourself lacking the power you need when you've got the whole band in and try to record them live across 16 tracks!

Just when you think you have enough power for all the tracks, EQ, effects and I/O you want, a new plug‑in comes out or a software revision which adds some extra capability (like 24‑bit recording which also increases the CPU power requirements by 50 percent — see the next section) and suddenly you find yourself lacking in processor muscle. This all boils down to a fairly simple rule: buy the most powerful CPU you can afford. This, of course, means that the computer may well be the most expensive part of your entire system, but when you consider that it will be handling all your recording, mixing, effects, I/O, (and, increasingly, moving into areas like synthesis with such programs as Rebirth and Retro AS1) then maybe this is how it should be. In these days of replaceable processors boards, you may well be able to upgrade your processor power at a later date, whether it is to a Pentium II on a PC or a G3 on an older Power Mac, but check this upgradability before you buy.

...try not to keep too many different projects on the go at once, don't record more audio than you need and then recover the unused portions too often, and don't drop in and out too much...

16‑ Or 24‑bit Recording

For recording of vocals and other fluctuating‑level signals, a compressor like this LA Audio GCX2 is a worthwhile investment.For recording of vocals and other fluctuating‑level signals, a compressor like this LA Audio GCX2 is a worthwhile investment.

I have chosen to cover this topic immediately after processor power, because bit depth is perhaps the single most important factor in the whole power‑consumption equation. Most people have heard or worked out by now that going up to 24‑bit recording will increase your need for hard disk space by 50 percent (or200 percent if you make the jump to 96kHz sample rate at the same time), but not everybody has made the leap to understanding that the same 50 percent increase applies not only to the data transfer rates between hard drive and computer (so you better make sure you have left ordinary SCSI for one of its more powerful descendants like Ultra Wide, or at least SCSI 2 or 3), but also to the drain on CPU power, because of the extra bits the CPU has to process. Of course, some programs have been outputting 24‑bit (at least as an option) for a while, as most of the processing of EQ and effects is carried out at this level, but it is when the all the individual audio tracks are being held as 24‑bit files that the processor demands become really significant. I have recently talked to people who are finding that whereas at 16‑bit they have been comfortably managing 20 to 30 tracks of simultaneous playback, this can drop to below 10 on the same CPU when they start recording at 24‑bit.

So with this sort of potential reduction in the number of tracks (plus the same reduction in separate inputs/outputs and EQ/effects), you had better be sure why you are using 24‑bit recording! In some fields, the answer is clear. If you are recording acoustic instruments in a very sparse setting, as is often the case with classical music or jazz, then the extra dynamic range is very worthwhile (provided you have the delivery medium to get this extra to your listener). If, on the other hand, you have the sort of dense sonic blanket you get in rock music or dance tracks, it is difficult to see what advantage 24‑bit is going to bring to your finished recordings.

One area in which I have found it useful is when recording lead vocals or other variable level signals without a compressor, where individual lines end up being recorded at 30dB or more down from the optimum level; normalising 24‑bit signals up to the same level as other parts of the recording does not result in the noise floor becoming as frighteningly high as with 16‑bit files. Of course, this depends on the entire signal path being clean, so don't expect this unless you have balanced XLR connectivity, analogue signals kept outside your computer (more on this in a second), and all other components in the signal path of a very quiet nature.

Keeping Analogue Audio Outside The Computer

Floating about inside the main casing of a computer is one of the largest concentration of digital signals you will find anywhere; if these bleed through on to analogue audio lines, the signal‑to‑noise ratio will be severely compromised. Many cards promise levels of quietness which just can't be achieved inside most computers. Insist on hearing the analogue ins and outs on a card on a decent pair of speakers in a quiet listening environment before you are persuaded to buy it. In particular, watch out for those cards which provide multiple analogue outputs directly on the back of the computer. The signal‑to‑noise level may be perfectly acceptable when just listening to stereo out. But if you sum together eight outputs through an analogue mixer (as you would in a normal final mixdown) that noise floor could be up to four times higher. Make sure you hear all eight outputs running together.

Alternatively, you can plump for one of the cards that comes with a breakout box containing all the A‑D and D‑A converters, thus keeping all analogue audio outside the computer, or go for my method — I use only digital interfacing to send critical signals in and out of my computer. The ADAT Optical I/O on the Frontier Designs Wave Center, the Korg 1212 I/O, and Sonorus StudI/O is the safest, as eight channels are sent down a single optical fibre (which can't pick up any interference outside the computer either), but there are also S/PDIF optical or co‑axial connections on many cards these days. If you are an aficionado of Tascam interfacing, the new Soundcape Mixtreme features TDIF.

All of the new digital mixers coming out can send and receive eight or 10 channels in some digital format or other (usually using an optional card) so they can interface very neatly with the Korg 1212 I/O or the Frontier Wave Center, but the the new Spirit 328 and Korg 168RC (now being sold off extremely cheaply) have 16 channels of ADAT optical I/O as standard. If you need more than this, VST can now support multiple 1212 I/Os or StudI/Os to give up to 32 channels of direct digital communication with Yamaha 02R, Mackie Digital 8‑buss or the new Panasonic Ramsa digital desks.

You will, of course, get an extra benefit from these newer desks which have 24‑bit A/D and D/A converters. Using the Lexicon Studio or Sonorus StudI/O, which are 24‑bit ready, will allow you to make 24‑bit recordings inside VST 24, the new top professional‑level version of Cubase. But whichever card you use, you will of course maintain an all‑digital signal path and avoid the repeated A‑D and D‑A conversions which degrade the signal path very quickly.

Properly Sync'ed Sample Rates

The single most common problem I had to deal with in my past life manning the Korg Technical Support Line (and my colleagues at other card distributors tell me the same story), was the clicking which happens more or less frequently when the word clock of a device like a digital tape machine (DAT or ADAT) or digital desk is not properly sync'ed to that of the digital audio software. Another common problem is distortion caused by having the source at one sample rate and the software set to another.

This only occurs when the digital interfacing described in the previous is being used. For everything to work properly (and silently), the sample rate of any digitally connected device has to be synchronised with that of the software (or vice versa). It is not enough to set the sample rate of your software to the same number as your desk or tape machine: you have to make one actually derive its sample rate from the other! The most common way to do this for incoming signals is to set the software to take its word clock from the S/PDIF or ADAT input which the digital audio is coming in on. But in more complex set‑ups involving a digital mixer, you can also do it by setting the mixer to take its clock from the computer's output via whichever card and interface you are using. Experiment to see which gives the best results (listen not just for clicks but graininess in the high end if the word clocks are not properly sync'ed). Some desks (Yamaha's in particular) sound better when using their internal sample rate to clock the DSPs. On the Korg 168RC, if you want to work at 44.1kHz (as would any sensible person wanting their work to end up on CD — see next section), then you must slave it to the software or other clock master.

Sample Rate Conversion: Do It Early On

If you want your recordings to end up on CD one day (and who wouldn't), it is much better to start projects at 44.1kHz, rather than have to convert them just before the CD mastering stage. The more frequencies are present in a signal, the more likely the conversion process is to produce unwanted artefacts. So use 44.1kHz wherever possible. If you use a DAT machine, buy one which samples at both 44.1 and 48kHz,and leave it set to 44.1kHz. If using the Korg 168RC, set its word clock to sync to DIG‑IN A or B (wherever your PCI card is connected). Set other digital desks to work at 44.1kHz before you start a project.

If you're making use of recordings which have been made at 48kHz, sample‑rate convert them as soon as you have transferred them into your digital audio software rather than after mixing down. If your software doesn't have this facility, there are now several inexpensive converters on the market — but they will sound much better if used on individual instrument recordings rather than the final mix.

Derive The MIDI Tempo From The Word Clock

Another common tech support call concerns MIDI sequences going out of time with the digital audio data. This is caused by the MIDI tempo being set to follow something other than the word clock timing. The most common occurence of this is when sync'ing to an external source other than word clock, such as MTC. The MIDI clock follows that external source, and the audio goes with the word clock.

In general, sync'ing digital audio sequencers to analogue tape is like trying to tow a sports car with a donkey and should be avoided. With track counts in modern software now reaching 64, you should ask yourself why you are linking these state‑of‑the‑art programs to antique technology.

Of course, the one time you can't really get around it is if sync'ing to a VTR to compose music for film or TV. In this case, you will need to invest in a professional synchroniser like MOTU's Digital Time Piece or C‑Lab's new Time Machine, which send not only MTC or MIDI Clock for the MIDI to sync to but Word Clock so that the audio playback can vary as well. Otherwise, wherever possible derive the MIDI Tempo from the sequencer's word clock (in the case of Cubase VST, this is done in the Audio System Dialogue box).

Avoid Hard Disk Fragmentation

I recently spent four hours at a studio trying to trace the source of some very annoying digital clicks with VST and a Korg 1212I/O PCI card. It sounded like the word clock problem described earlier, but that was all set up fine. It turned out to be a very fast hard drive, which had become so fragmented that it was slowing down so much that it caused the audio playback to click.

Hard disk fragmentation comes from the way computer files are stored and erased and, without getting too technical, causes the heads to have to jump around more and more on the platters to find the required data, thereby slowing down the access times. Other symptoms include stuttering in audio playback and error messages saying the hard drive is too slow to play back a certain number of channels when yesterday it played that many with no problems.

To avoid fragmentation, try not to keep too many different projects on the go at once, don't record more audio than you need and then recover the unused portions too often, and don't drop in and out too much (record a new track and then edit together the required bits instead). Best of all, record all your audio data to a separate hard drive and clean it off and reformat it before starting a new project. Backing up to another drive or DAT (see next section) and then reloading after initialisation is good way to get rid of fragmentation if it rears its ugly head in the middle of a project. I always think complete re‑initialisation is better than using defragmentation software, but if you use the latter, make sure you have a backup before you start as you will lose everything if the power gets cut in the middle of de‑fragmenting!

The biggest single cause of frustration in digital audio systems is lost work. However careful you are and however bug‑free the software, sooner or later, computers and hard drives crash.

Back Up, Back Up, Back Up

I have left this point till last, although it really should be first; it's certainly the most important! The biggest single cause of frustration in digital audio systems is lost work. Because digital audio gets recorded direct to the hard drive (unlike MIDI data which has to be specifically saved from time to time), it is very tempting to think that it is safe. It is not! However careful you are and however bug‑free the software, sooner or later, computers and hard drives crash — and always at the worst possible moment, like during a disk read/write. Chances are that your audio sections will be more difficult to recreate than your MIDI, especially if they involved a session singer or a particular recording location, so take extra care to protect them.

If you have two hard drives in your system, make a regular icon copy on the desktop from one to the other, and if you have a DAT machine or ADAT, then backup to that at the end of the working day. This won't protect you completely, but it will minimise your loss should the worst happen (and sooner or later it will!). It will also help keep fragmentation at bay (see above).

I'm not saying that if you follow all my suggestions your digital audio system will never give you a moment's trouble; anything based on a computer has a certain degree of volatility. However, you will certainly minimise the risk of wasting huge amounts of time in non‑productive trouble‑shooting, or ending up with unusable recordings because they contain clicks or other audio problems. In time, you'll wonder how you ever made do with the restrictions of a linear medium. Honest!

Analogue Compressors Improve Digital Recordings

Although many current digital desks offer an embarrassment of dynamics processors (up to 50 in some cases) which are extremely good, they cannot fix a poor low‑level recording once it has been converted to digital. If an A‑D converter is presented with a low‑level signal, it cannot perform to anywhere near its theoretical capabilities. Every 6dB below full‑level wastes one bit. So if an incoming vocal phrase is 30dB down (not an uncommon thing) when arriving at a 16‑bit D‑A, you will only use 11 bits to record it, and instead of a 90dB+ signal‑to‑noise ratio, you will typically get around 60dB. There is nothing a digital compressor can do to restore this. If it brings the level of the signal up 30dB to compensate for the lack of level, the noise floor will rise 30dB as well.

The only way to deal with this is to record all fluctuating‑level signals (vocals or practically anything else recorded through a microphone, as well as unprocessed guitars) through an analogue compressor. I bought an Aphex 106 Easyrider to use with the Korg 168RC on the insert points because it has four channels, but I don't think I have ever used more than two at once. Spending a few hundred pounds on a good dual‑channel compressor (preferably one which allows you to link and unlink the channels) is the best investment you can make to ensure you achieve, in practice, the sound quality that most digital audio cards are theoretically capable of delivering.