Studio SOS: Incorrect Monitor Placement

Dorian's monitors were set up on shelving, which was not rigid enough for the purpose and which therefore compromised the bass response of his Spirit Absolute II monitors, speakers which already have limited low-end extension.

The SOS team ride to the rescue of a budding media composer who's having trouble with his mixes.

Dorian Kelly has long-term plans to compose music for TV and film after successfully completing a media composing course (the home-study Music For The Media course run by Guy Michelmore from Atlantic Studios), but in order to gain experience he has set up his own Steinberg Cubase SX-based home studio where he works on every type of music possible, from classical to pop. This seems to be a very sensible and practical approach, but although Dorian's room has had the benefit of some acoustic treatment, he was worried that he couldn't hear enough low end in the studio, which left him with mixes that sounded bass heavy when played elsewhere. He was also keen to try to improve his mixing technique, so he called SOS.

Low-end Monitoring Problems

When we arrived at Dorian's Reading home, where his studio is located in a converted garage, the kettle was boiling and Dorian had assembled a vast plate of chocolate digestive biscuits and croissants — so he was clearly aware of all the necessary Studio SOS protocols! While attempting to reduce the size of the cookie mountain, we played a CD of various test tracks over Dorian's system, which comprised a small Behringer mixer to handle monitoring sources, a Samson power amplifier, and a pair of passive Spirit Absolute II speakers. The recording system was based on a purpose-built PC running Cubase SX with another small Behringer mixer to manage the input signals. An Emu Proteus 2000 provided the only external MIDI sounds.

As we have discovered on previous occasions, the Absolute II monitors don't have a great deal of low-end extension. This, combined with the room size and shape plus 'less than ideal' speaker mounting did indeed produce a bass-light result, though the mid-range and high end were fine. While Dorian's room is adequately long, it is only a couple of metres wide, and in a room this shape there's little alternative than to work across the width of the room, as there would be insufficient space to set up the equipment in an ergonomic fashion if working across the narrow axis of the room.

Even when Paul switched the Absolute IIs for his own Mackie HR624s, which have a better bass response, some adjustment of the speakers' onboard frequency tailoring switches was required to give a representative low-end picture.The wall behind the mixing position was already treated with some acoustic foam tiles and a few small areas of thick carpet tile, while the ceiling was largely covered in fibrous office-style acoustic tiles — the overall result was reasonably well controlled, without any obvious boomy bass notes. The plasterboard walls, a cupboard, the doors, and two windows all contributed to some fortuitous low-frequency trapping, so the main problem was the reduced level of bass from the monitors, rather than uneven bass.

Dorian had his speakers set up on foam speaker pads that rested on the free-standing pine shelving he used to store all his studio accessories, both above and to either side of the mixing position. Using these foam pads is preferable to standing the speakers directly on non-rigid shelves, but there's no real substitute for solid stands or wall brackets. However, in this case these couldn't be accommodated for space reasons. To evaluate the room further, we set up our Mackie HR624 active monitors in place of the Absolute IIs and played the test CD again. The result was better, but we still weren't hearing as much low end as we are used to from these speakers, so the room was clearly a contributing factor. Setting the bass switches on the monitors from their half-space mode to full-space mode delivered a better balance, even though the speakers were being used close to a wall where the half space setting was technically correct. The full-space setting is normally only used when the speakers are set up away from walls, but as it worked we weren't going to knock it! After all, getting any kind of decent monitoring system working in a room this narrow is inevitably a compromise.

After hearing the system with the Mackies in place and after playing some of his own mixes through them, Dorian agreed that he should change his monitors as soon as possible, but because of budget constraints he felt he had to look for something rather cheaper than the Mackies. He wanted to go for active models, and after looking at the options he narrowed his choices down to Alesis, KRK, and Fostex units.

On the subject of mounting the monitors, the existing position was just a little too high, as the tweeters should ideally be aimed at the engineer's head and not a few inches above it. This could be remedied by either lowering the shelves or by angling the speaker downwards and inwards, but it would also be beneficial to provide a more stable mounting platform. You could use a metal bracket to secure the speaker shelf to the wall behind it and then position a small paving slab (around 300mm square) as a base for the speaker — the additional mass of the slab helps to damp and control vibrations. You might also want to stand the speaker on the slab using blobs of Blu-Tac under each corner rather than relying on foam. In a fair and just world, the 'slab' approach should tighten up the bass end to a worthwhile extent.

Vocal Recording & MIDI Tips

Dorian played us some vocals he'd recorded where the singer had stood at the opposite end of the room from the computer and sung with the back of the mic facing the mixing position to minimise noise pickup from the computer. The results were reasonable, but some room coloration was evident and, as vocals often need compressing to make them sit properly in a mix, the room ambience is usually emphasised further. Dorian had bought one of those Auralex foam gizmos that clip over the back of the mic to attenuate sound coming from that direction, but the wall directly behind the singer was fairly bare and had a mirror mounted on it, all of which conspired to bounce sound back into the microphone.

A combination of two reverbs in series was required to get the best out of Cubase SX's built-in plug-ins: the first (left) added a short brighter ambience to give a sense of roominess; and the second (right) contributed a subtle reverb tail without clogging up the mix.A practical solution to this is to hang a double duvet from a rail fixed to the ceiling, so that it is directly behind the singer when recording. In this way, ambient sounds no longer reflect back into the front of the microphone, and the room coloration is greatly reduced. If the duvet can be hung from a curved rail so that it also comes around the sides of the singer slightly, then so much the better. Dorian liked this idea, as it was simple and inexpensive and the duvet could be pushed to one side when not needed provided that it was hung using curtain rings.

Next Dorian played us some of his classical-style compositions, which had been created using Steinberg Halion samples combined with some tracks of him playing real trumpet and French horn. Overall, the sound was well balanced, though we did identify three specific problem areas. The most obvious problem was that some of the parts sounded quite stilted, and this turned out to be because Dorian composed his music using Sibelius before transferring it into Cubase SX to mix. As Sibelius creates MIDI files based on the bars and notes of the score, the data is inherently tightly quantised, but in addition Dorian wasn't making use of those instruments that supported legato mode to allow overlapping or consecutive notes to flow smoothly into each other. Selecting the desired MIDI notes in the Cubase SX grid edit page and then choosing Legato from the MIDI menu removed spaces from between the notes and allowed the sound to flow more naturally, without constant re-triggering of samples. This is particularly evident on string sounds (which 'suck' unnaturally if legato isn't used correctly), but solo wind instruments and bowed solo strings also tend to play legato a lot of the time, so trying to simulate a performance from distinctly separate notes will invariably sound mechanical. We tweaked a couple of parts to demonstrate the difference when parts were played legato, and Dorian immediately saw the advantage.

Some instrument patches are created specifically with legato playing in mind, and some instruments have dedicated legato trigger modes which prevent the envelope from re-triggering when a new note is played, provided that it overlaps the previous one. These techniques are all worth exploring, as is the trick of using a MIDI volume pedal (transmitting MIDI Continuous Controller number seven) when recording strings to create more natural-sounding swells and crescendos.

Reverb Overhaul

The other obvious shortcoming in Dorian's tracks was the rather limited reverb that comes with Cubase SX. Although the mix was obviously reverberant, there was no sense of space or acoustic — an obvious failing in the context of orchestral music. To improve on this, Dorian has the choice of buying a convolution-based reverb such as Altiverb (which will eat up a lot of processing power), adding a hardware DSP card such as the TC Electronic Powercore or Universal Audio UAD1, or plumbing in an external hardware reverb processor such as a TC Electronic M*One XL or a Lexicon MPX550. Because Dorian's M Audio Firewire Audiophile interface has two analogue inputs and outputs, plus S/PDIF in and out, a hardware reverb with digital I/O could be plumbed into the system via S/PDIF to save unnecessary conversion. Any latency affecting the external reverb would be negligible compared to the usual pre-delay added to reverb — provided that the reverb was used in an aux send/return loop and set to 100 percent wet.

However, my own preference would be for a card-based system, as you get the quality of hardware but the effect settings are stored as part of the Cubase SX song. So when you revisit old projects, you don't spend half the day trying to get back to where you started. You can also run multiple instances of the reverb plug-in with a card-based system.

As an experiment, however, we decided to see what could be achieved using the Cubase SX native Reverb A, and the strategy we tried was to use two reverbs in series, the first set to as short a decay time as possible to simulate an early-reflections/ambience type of treatment. This was adjusted by ear to give a coloured roominess to the sound (we ended up with almost as much reverb as dry signal) and a short pre-delay of between 10-20ms seemed to help. The ambience reverb was followed by a longer concert hall effect with a 70-80ms pre-delay to provide the tail of the reverb. We set this at a much lower level so that it added a subtle reverb tail without choking the gaps in the music, and the result was surprisingly good.

Obviously it was not as smooth as a real high-end reverb, but at least we managed to create a sense of space and depth without clogging up the mix with an excessively loud reverb tail. The use of an ambience-like reverb to create the basic sound character certainly helped the overall sound become more homogeneous, and the performance now sounded as though it had taken place in a real space.

Processing The Final Mix

The final challenge was to process the final mix to create a more commercial, 'TV-friendly' sound. Dorian had tried some fairly radical EQ that combined shelving bass boost with a wide mid-range cut to generate something like the traditional smile curve (more of a manic leer really!). However, Hugh and I felt that using only the two parametric sections of the Cubase four-band EQ would give us more control, so we added a gentle 80Hz hump balanced by a broad 13kHz boost to add air and sizzle to the top end without making the mid-range sound harsh.

With the Mackie monitors in place, Paul had a look at the processing Dorian had used on his final mixes. His EQ settings in particular were felt to be a little extreme, so a gentler 'smile curve' EQ was used instead.This was teamed with a compressor setting using a low 1:1.2 ratio, a fairly fast attack, and the automatic release setting. The threshold was turned down until the gain-reduction meter showed around 5-6dB on the louder passages, which left us with a threshold setting of around -35dBFS. This strategy reduces the whole dynamic range of the material in a very gentle way, and differs from the more usual tracking approach where a higher threshold is set and then everything above the threshold is squashed much more aggressively.

Dorian then played us a pop song he'd recorded with a female vocalist. Her performance lacked a little conviction, but was basically sound, and we tried to treat it with a more robust 8:1 compression ratio, the fastest attack, and the automatic release mode. Again, the threshold was adjusted to show the desired amount of gain reduction, in this case around 8dB on peaks. This worked fine, making the vocal sit better in the mix and also making the sound appear more confident. Dorian was a little unsure about the best reverb treatment to use so we again tried the approach of layering a short pre-delayed bright setting to create an almost slapback ambience together with a longer reverb tail at a lower level and with more pre-delay. After a little tweaking we arrived at something more contemporary sounding, where the vocal took command of the mix, rather than apologising for being there as it had before! A little high EQ to sharpen it up and we were done.

When discussing how the vocal was recorded, Dorian also admitted that he hadn't added any reverb to the monitor mix, though Cubase SX does allow this and it can help make a singer feel more confident. However, you do need to check with the singer how much reverb they like to hear in their cans, as everyone seems to have their own preferences in this area. Too much is just as bad as too little.

It can also make a difference to reverse the polarity of the headphone feed while recording, which can be achieved by using the phase-reverse button in the dynamics section of the vocal channel in the case of Cubase SX. This affects the way the singer hears the sound from the headphones when it combines with what they hear of themselves directly, and usually one polarity sounds more comfortable than the other.

Optimising The PC

Towards the end of our visit, Dorian asked a few questions from a checklist he'd been compiling before our visit. The first was about latency and buffer sizes. He wasn't quite sure what to adjust or what the best setting was, but he was aware that it was important. Latency is usually defined as the time delay between a signal being piped into a computer workstation and that same signal appearing at the monitor outputs. It is most important in the context of software instruments, as the monitoring workarounds that can be tried when recording (monitoring the vocal mic at the input to the system rather than at the output, for example) can't be used with software instruments. If the latency is too long, the time lag between pressing a key and hearing a sound will put you off playing, but on the other hand if you try to set up your system for a very low latency, the essential data buffers in RAM may be too small to allow a continuous flow of audio data, and that's when you get clicks and pops.

On a modern computer like Dorian's (which has 3GB of memory), it is usually OK to set the buffer sizes to 128 or 256 samples, which at a 44.1kHz sample rate will give a low enough latency that most users will be unaware of it, yet the buffers are still large enough for reliable operation. Of course the stress on the system gets worse as you add more plug-in effects and instruments, so when the computer is pushed close to its limit, it may be necessary to increase the buffer size to avoid glitching.

To minimise the CPU load, a handy tip is to scale the song display so you can see it all on one screen, which avoids the computer having to work on screen re-draws during playback and when mixing. You can also increase the buffer size when you're mixing, as latency isn't an issue unless you're recording. The other useful strategy is to render instrument tracks as audio files, which saves on CPU load at the expense of a little more hard-drive activity, but in my experience most modern systems are capable of handling far more audio tracks than any sane composer would ever need unless they're scoring a Hollywood movie and creating all their ensembles by layering individual monophonic tracks. There are many articles discussing the causes of latency and its workarounds on the SOS web site, so it isn't necessary to go into more detail here.

All Of A Dither

Dorian's final question involved dither. Again, he had heard that it was a good thing and was beneficial when reducing the resolution from 24 to 16 bits, but he wasn't sure where or when to apply it. In this context the reality is that dither is only of benefit if it is the very last thing you do to a piece of audio, so it's no good dithering your mixes into 16-bit files, putting them into a CD playlist, and then changing their relative levels — the act of adjusting the levels will destroy the dithering benefits. My own strategy is to record and mix everything at 24-bit resolution, then use a CD playlist compiling program (Roxio Jam for Mac in my case) that automatically applies dither during word-length reduction as the very last stage after any other adjustments you may have made before burning the CD-R.

At the end of the afternoon, we felt we'd made a worthwhile amount of progress. We'd demonstrated that the room acoustics were reasonably good but that the current choice of monitoring wasn't ideal for the room. We'd also come up with some cheap and cheerful ways of improving the recorded vocal sound, and our impromptu dual-reverb treatments had worked out better than expected. We had also covered some useful compression and EQ basics, as well as made some suggestions for tweaking MIDI files to make the playing sound more realistic — specifically by using legato. The job was done, the plate of biscuits was looking decidedly well hammered, so we packed up our toys and made for home.