The SOS team visit Dave Stevens' home studio to sort out a chesty vocal sound, investigate a mystery digital buzz, and hand out some mastering tips.
Dave Stevens has a home studio in Gloucestershire where, rather than simply set up a desk for the gear like many of us would, he's divided one of the bedrooms in his bungalow to provide a control room and drum/vocal booth. The soundproofing between the rooms is very effective and the application of Primacoustic foam panels and bass traps in both areas has avoided the boxiness that can occur in small studios. The live room contains a Roland TD8KV drum kit, which David has learned to play remarkably proficiently in a relatively short space of time, plus he has a decent collection of guitars and basses which he tends to record via a Line 6 Pod Pro and Bass Pod Pro.
The Control room is a compact 11.5 x 7 feet, with a window at the back of the room and the speakers set up on the long wall dividing the studio and live room. The Yamaha NS10M studio monitors are on stands either side of a computer table which supports a pair of 19-inch CRT monitors and a Houston moving-fader control surface. All the outboard gear is in a tall homemade rack in the middle of the right-hand wall of the studio. A double-glazed control-room window provides a clear view of the live room where David has a Neumann U87 microphone set up for vocal recording (with a homemade support constructed from a piece of nylon tube!). Recording is to a PC, which David built, running Steinberg Cubase SX and using Yamaha DS2416 and SW1000XG cards. The analogue and digital inputs and outputs from his soundcards are wired to a patchbay and digital router respectively, allowing signals to be processed externally and returned to the Cubase environment. For example, Dave has Cubase SX set up to make use of his Lexicon MPX550 hardware reverb, which then feeds back into Cubase's virtual mixer. A Spirit analogue mixer provides input mixing and routing for the various sound modules to Cubase, and acts as a monitoring controller.
Dave contacted us because, although he felt he was recording everything pretty well, his mixes didn't have the sparkle he felt they should have. He was also having problems getting the vocal sound he wanted. Furthermore, it transpired that he was suffering from a low-level buzz which he believed came from his SW1000XG soundcard (and which disappeared temporarily when scrolling graphics on the screen!), but was at a loss as to how to cure it — more of that later.
When we arrived, Dave produced a vast quantity of Marks & Spencer's executive-grade chocolate biscuits and countless mugs of coffee, so we felt we had to give it our best shot! After locking Dylan the dog out of the studio, we set about listening to a song David was currently working on and which featured his own vocals. Everything was recorded cleanly, but the vocals had a very chesty, congested sound, so in the first instance I tried to find a parametric EQ setting (using a Waves plug-in) to fix this. However, even though notching out the worst low mid-range peaks helped, it was clear that this wasn't a proper solution, so we asked David to explain how he'd recorded the vocals.
His system was to use the U87 at around six inches, with a pop shield fitted between him and the mic. The mic fed through a Focusrite Platinum Voicemaster Pro, where he applied what he felt was a gentle amount of compression, a little mid-range cut to remove some of the chestiness, and the simulated tube section to warm the sound up a little. So far it all made sense, but then David said that he also passed the signal through his Antares ATR1 (the hardware version of the Auto-Tune plug-in) to tune it before recording it (though he monitored the non-corrected signal, as it's almost impossible to sing when monitoring the corrected output). He'd also used the Cubase SX host-powered reverb plug-in (which didn't have the right quality for his voice) rather than his Lexicon. Finally, Dave was also using Cubase's True Tape emulation mode with 12dB of overdrive, to try to create a more analogue quality.
The only real way forward was to get David to sing the track again with us at the controls and see what we could do at the tracking stage. To this end, I bypassed everything, removed the ATR1 from the loop, and then asked David to sing while we listened to him in the control room. Dave had left the voicemaster set up exactly as he had it when recording the original track, and the first thing I noticed was that the level settings David had on the Voicemaster Pro were a bit on the high side, so I backed them off to leave us with a reasonable amount of headroom.
With no compression or tube simulation, the sound was much less congested, but it still sounded too chesty to me, so we asked David to move back to around 18 inches from the mic. This brought about a further improvement, but not a complete cure, so the next step was to drop the mic height so that David was, in effect, singing towards the top edge of the pop shield. By using this mic position and working at between 12 and 18 inches from the mic, we achieved what we felt was closer to David's natural vocal sound, and it was certainly a lot nicer to listen to than his 'before' version!
Though I often add a small amount of compression while recording, I felt that in this instance David's levels were well enough controlled that we could allow ourselves the luxury of recording completely flat, leaving ourselves the option to use plug-ins to do whatever was necessary after recording. I also suggested that Auto-Tune should be applied as a plug-in after recording, as that gives the engineer a chance to optimise the settings for the song, or even to divide the song into sections and process different sections separately where appropriate. If it is used when recording, inappropriate settings could destroy an otherwise good take.
A final tip, and nothing to do with vocals, concerned the Roland TD8KV electronic kit Dave was using. The sounds that David had chosen were probably fine in isolation, but in the mix they lacked definition, so it pays to choose your sounds in context rather than just picking what you feel is a nice sounding kit. In particular, the kick drum was not very well defined, so I suggested trying the one from the heavy rock kit, which has more weight and definition.
My other tip was to buy and use a real hi-hat — the hi-hats are the one thing that lets down all the electronic drum kits I've played. The physical noise from the mesh heads used on these drums is low enough that it shouldn't interfere too much with miking a hi-hat.
We also avoided using the tube simulation, because, although this can be most effective for thickening up thin voices, it can have a congesting effect on a vocalist who already has a strong lower mid-range — which Dave did. Using this setup, we recorded a new vocal part and then tried to sit it in the mix.
As it turned out, no EQ was thought necessary, though compression was added using the Waves Renaissance Compressor with a ratio of around 6:1 and the threshold set to give about 8dB of gain reduction on the loudest notes. A fairly fast attack time was combined with a medium/fast release time — the exact settings can be seen in the screen shot. For the finishing touch, we added some reverb from the Lexicon MPX550's Warm Plate preset, tweaked for more pre-delay (around 80-90ms worked fine) and the decay time was set to 1.6 seconds.
Hugh was concerned that David had also been recording with the Cubase True Tape facility active, which essentially provides soft saturation on higher-level sounds (he had the threshold set at 12dB from maximum). It was just possible that this degree of saturation, combined with the valve emulation on the Focusrite preamp, might have contributed to the thickening of the vocal sound, so we ran two vocal recording tests, one with and one without True Tape. As it turned out, there was no material change in timbre so True Tape was exonerated!
Background buzzes and ticking sounds that are affected by mouse movements and drive accesses can be caused by a variety of mechanisms, so it pays to be systematic. I started by checking that Dave didn't have an unusual graphics card, since a few types (Nvidia is one manufacturer that springs to mind) have been known to induce soundcard interference, as have some Adaptec SCSI cards. In these cases it's worth trying to move your soundcard to a slot as far away from the offending item as possible. However, he no longer used SCSI, and his Matrox dual-head graphics card (identical to mine) has an extremely good track record for musicians. Dave had also taken the often wise precaution of placing his Yamaha DS2416 and SW1000XG soundcard pair well away from the AGP slot anyway, leaving two empty slots in between, so I didn't need to try this either. I next checked that all cards had been well seated in their slots and the backplate screws tightened down firmly to give them a solid ground connection, but again everything was in order.
Interconnected Yamaha DS2416 cards employ a ferrite RF filter on their short interconnecting ribbon cable to minimise interference pickup from other internal components, but the internal cable used for connecting a DS2416 to the SW1000XG doesn't, so I'd taken the precaution of bringing one along to try. However, it made no difference to the audible interference after clamping it in place, and neither did an impromptu screen placed round the soundcards — kitchen foil inside a polythene bag to prevent it shorting something out, with the aluminium foil temporarily connected by an earth wire to the PC metal work. Even temporarily disconnecting all Dave's USB peripherals (which have been known to cause the occasional hum) made no difference.
We hit a glitch when we rebooted Dave's PC with his Houston controller still switched on — the computer suddenly became unable to see his RAID (Redundant Array of Independent Disks) setup and his system hung. However, he'd already suffered from the same problem on his IMSI 845E motherboard a couple of times previously, and passed on details of how to cure it just in case anyone else ever experiences anything similar. You turn off Houston, reboot, and then enter the RAID BIOS during startup. Then you delete the current drives and set up exactly the same RAID array once more. It seems that the order in which you switch on peripherals is crucial for this particular system.
We continued checking at the PC end for possible reasons for the annoying amount of background noise, including the SW1000XG's inputs being accidentally left at mic sensitivity and confirming that the external cables had been correctly wired with good shielding, but eventually we were all satisfied that there was nothing else to try at the source end, and we could only assume that this was a more general ground loop problem. Martin Walker
Having obtained a better vocal sound, we set about finding a strategy to more effectively master David's mixes using the available plug-ins. However, in the course of our listening tests a rather annoying digital-sounding buzz which emanated from his DS2416 soundcard's analogue outputs had to be addressed. This initially appeared to be a software driver problem, particularly since the noise would disappear momentarily when the computer was doing other things (such as updating the screen), although Dave had already updated all his video and soundcard drivers.
As our colleague Martin Walker has experience with such problems, and because he only lives around 20 minutes from David's house, we asked him to come round and check over the system. While we waited for Martin to arrive, we started checking to see just which outputs suffered the noise and discovered that the digital outputs from the card seemed clean, so we started to think the problem might involve ground loops, rather than software problems. To test this idea we connected one of the sound card's analogue outputs directly to the monitoring input on David's TD-8KV drum kit — the handiest source of headphone monitoring which was also isolated from the rest of the rack's earthing. Sure enough, this turned out to be clean too, so there was obviously a ground loop issue here.
Making up balanced cables to feed the power amplifier from the Spirit mixer's monitor outputs improved the situation noticeably, but the buzz was still there and, perversely, it disappeared if the audio was routed via the outputs of the SW1000XG, even though the grounding regime was essentially identical. We tried all the standard ground loop fixes, including unplugging all the mixer connections and then putting them back one at a time, but we couldn't find a definitive cause. I even tried the old 'resistor in series and lift the screen' trick on the soundcard outputs, which ended up being like an episode of Junkyard Challenge because nobody had brought any resistors. I toyed with the idea of fixing two electrodes to a known volume of coal, but eventually settled on using the resistance of a lead pencil — all of which went to prove that the problem was still there!
When Martin arrived he checked all the computer's internal connections, tried installing a ferrite clamp and adding additional screening between the cards (see the 'Dave's Soundcard Problems' box for details), but David had already done just about everything that Martin could recommend trying. Then disaster stuck when David patched everything back in after Martin's exploratory efforts, because the whole system went into a high-frequency howlround. Before we could turn down the volume, jets of smoke shot out of both of David's NS10 tweeters. One had died altogether and the other, while still appearing to work OK, was clearly not a happy tweeter! Fortunately, Peter Peck at Yamaha Kemble was able to save the day, because, when we phoned him to enquire about replacements, he graciously arranged for a pair to be sent to us free of charge. They arrived just two days later, so a big thanks to Yamaha for bailing us out!
Having sorted out the levels and checked the wiring, we continued with the one injured, but still working, tweeter to find a mastering solution. As mastering with one blown and one lightly toasted tweeter isn't to be recommended, we also made CDs of our efforts and checked them on David's hi-fi system in the lounge. What we came up with was predictable enough, but, although David had been using a similar collection of plug-ins to do the job, he hadn't appreciated the benefits of using low-threshold, low-ratio compression when mastering to reduce the overall dynamic range in a subtle way before applying a peak limiter. He'd treated his mixes in a similar way to his individual tracks, and had also been a bit heavy-handed with the limiter — while his mixes didn't actually sound that bad, they didn't have the transparency and sparkle he was looking for.
I assembled a Waves C1 full-band compressor, a Waves Renaissance EQ, and a Waves L1 Ultramaximizer limiter, with dithering set to 16 bits. We also tried the Steinberg Spectralizer harmonic enhancer and found that it could produce useful results in some mastering situations — to get the benefit of the dither and limiting, however, we patched it prior to the L1. The settings we tried included compression ratios of 1.1:1 and 1.2:1 with the threshold pulled right down to give a level reduction of around 6dB on the loudest sections. The attack was fairly fast, with the shortest release time we could set without invoking pumping — you can see the settings we ended up with for all the processes in the screenshots.
The EQ was set up to give a very gentle 'smile' curve, by adding a hint of broad boost to the frequency extremes, while the middle was dropped by a decibel or two using a very wide, gentle setting. This is a common ploy to add clarity and sparkle to a mix. Where the Spectralizer was used, the amount of harmonics added was carefully monitored to ensure that only a slight brightening effect was achieved — it is so easy to go too far with these things. To finish off, the limiter output target level was set to just below maximum (around half a decibel is a common setting) and the input gain was adjusted so that the amount of limiting on the peaks was only between three and five decibels. As we had already reduced the dynamic range with the compressor, most of the time the L1 gain-reduction indicators showed little or no gain reduction taking place — whereas when David had used it, it was showing gain reduction almost all the time.
The result of this simple plug-in setup was a general levelling of the track and an increase in density, but with more transparency and detail in the mix. The final track was also hugely louder than the original stereo mix, even though that was peaking close to full scale on several occasions. But, importantly, the new mix didn't sound squashed, even though it was louder. The tip here is that, when mastering, don't try to get all your level gain by limiting, as that just squashes the peaks and does nothing to improve the density of lower-level sections. By using low-ratio, low-threshold compression, the whole mix sounds more homogenised and, as a bonus, it leaves the limiter with less work to do. This technique also shows that you don't have to have a multi-band compressor for mastering and, indeed, many mastering engineers seem to prefer using a full-band compressor anyway. Multi-band compression does help get the maximum possible level, but it can be at the expense of tonality or balance if used to excess.
At the end of our day with Dave he asked for any other advice or comments, and Hugh came up with a couple of suggestions to improve the acoustics in the control room. The first was to replace the large CRT monitors with LCD screens — although this would be a fairly expensive option because Dave was so used to his 19-inch monitors that only 17-inch flat screens would probably do! The reason for changing the monitors was primarily because of their intrusion into the monitoring area between the NS10 speakers, causing reflections and scattering of sound which had a harmful effect on the stereo image. This situation would only be exacerbated when Dave installed his keenly awaited Mackie HR824 full-range monitors alongside the nearfield Yamahas. Using flat-screen monitors would also benefit his guitar recording by reducing the amount of electromagnetic radiation likely to be picked up by the guitar.
A similar acoustical problem was caused by the placement of the equipment rack, which was very close to the right-hand speaker and caused more reflections and sound scattering. By placing the rack closer to the rear corner, this situation could be remedied very easily, and Dave actually achieved this shortly after we had left.