Here is the solution we have implemented (see picture).  We have six separate 20-amp AC lines coming into the studio from the main power service.  First they go to Tripplite power conditioner-line stabilizers.  These maintain the incoming voltage at a steady 120 V and also filter out noise, spikes and other undesirable artifacts.  Then they go through a metering panel, which shows amperes – volts – watts.  This gives a ready indication of each circuit’s capacity and status.  Then they go to balancing transformers manufactured by Equi=Tech.  Usually AC line voltage is 0 V on one leg and 120 V on the other.  What balanced power does is change this to 60 V on both legs.  This promotes common-mode rejection (analogous to balanced audio cabling) and lowers the noise floor by at least 10dB.  It also eliminates ground loops (audible as 60 Hz hum) and complex star-grounding schemes.  We can turn our main mixing system all the way up without audible hiss or noise.  Oversized cabling is used throughout to reduce impedance and promote headroom.  We also have uninterruptible power supplies for computers (not depicted).

Here are a few other observations:

1.  Most studios have at least three separate cabling systems (AC, audio, MIDI).  Power cabling should be physically separated from audio and MIDI.  If they run parallel, there should be space between them; if they cross, they should do so at right angles.  These same principles apply to digital cabling, computer cabling and time-code cabling.

2.  Power to anything but sound-producing keyboards or modules must be on a completely separate AC circuit.  Lighting and computers, for example, always should be on their own lines (as seen in the picture, four circuits are devoted to audio and two circuits to non-audio).  It is difficult to over-emphasize the importance of this step.  Back in the days of 24-track analog tape recorders, we even went so far as to rewire them, segregating power to the electronics from power to the motors.  This considerably improved the sound of the electronics, because they did not have to deal with motor factors such as torque and slew.  It also improved the performance of the motors, because now they could have a consistent power source without also worrying about powering electronic components.

3.  Multiple wall-warts of the same power type (e.g. 12 VAC) easily can be replaced with a single, larger AC transformer.  This provides consistent power to all of them and results in significantly improved sound quality.  Many modules are 9 or 12 VDC.  We have two large transformers (made by Acopian) to make these conversions, then distribute the power along separate lines, to the same effect.

In conclusion, studios typically spend thousands of dollars on gear while at the same time ignoring basic principles of how to make it sound good.  This doesn’t make a lot of sense because the way things sound is constrained by the least-satisfactory variable.  It’s important to optimize every step of the process in order to achieve a desirable outcome.

Publison Infernal Machine – Manual

Publison Infernal Machine

I recorded this track to demonstrate their capabilities in three passes, 1:30 long with minimal processing.  It includes sounds from the dreaded “Super Space Drum” (left), the “Electronic Crash Drum” (right) and the “Clap Track” (towards the end).  The Mini-Synthesizer on the left sounded better with bass and the one on the right sounded better with lead, so there you have it.

ElectroHarmonix

Over the years the studio where I produce many records has accumulated a variety of MIDI sound modules (“ROMplers,” so called because their sounds typically are stored on a single ROM chip) and samplers.  The MIDI sound modules include older ones made by Roland, E-Mu, Yamaha and Alesis, together with others.  The newer ones are made by Access Virus, Nord, Novation, Waldorf and others.  The samplers are Akai, Kurzweil, E-Mu and Roland.  At one point this studio probably had 200 sound modules and samplers in a dozen floor-to-ceiling racks.  All of them were/are MIDI-controlled and stereo out.

The last generation of sound modules made right before the advent of VST instruments have SPDIF digital outs.  These are intriguing because they suggest the possibility of by-passing a stage of digital-to-analog (“DA”) conversion before introjecting the sound into ProTools or Logic.  Most digital converters (such as the Apogee Ensemble and all of the other Apogee converters) have a SPDIF digital in.  The problem is there is only one of them, which creates a challenge to mix all of these separate digital sources. 

Surprisingly there are not a lot of SPDIF digital mixers.  Almost every mixer now made claims to be a “digital mixer” and they are in the sense that they take analog signals, mix them and then convert them to SPDIF or AES-EBU digital.  None of them however mix signals in the digital domain, which is what one might think a “digital mixer” does.  After exhaustive research we concluded that the best (and possibly the only) solution to mixing actual digital signals is the Roland M-1000.  It accepts four (stereo) SPDIF ins and also a (stereo) analog input.

We took 24 modules and connected them to six M-1000s as depicted in the accompanying photo (one M-1000 not shown).  We established a limit of 24 modules for this exercise because we devised a way to address each of them separately in Logic (which will be the subject of a separate post).  The problem with this set-up was that the Ensemble (like all other Apogee converters) only has one SPDIF in.  Thus it was only possible for one M-1000 at a time to address Logic.  The other ones had to go through a DA conversion phase out of the M-1000 and then another analog-to-digital (“AD”) conversion phase through the Ensemble.  Although these were high-quality conversions it still basically defeated the purpose of remaining exclusively in the digital domain.

We experimented with ganging four M-1000s into one M-1000 as a kind of digital summing mixer however this lead to audible jitter and other undesirable digital artifacts.  To stay exclusively digital we ended up disconnecting and reconnecting M-1000s so as to address that elusive but desirable single SPDIF port.

After a while we decided that this system was a lot of trouble.  So we decided to utilize the on-board DA converters on each of the modules and mix straight into Rane SM82 mixers.  These are analog but then went straight into the AD conversion phase of the Ensemble.  The results are depicted at the second photo.  We also retained one M-1000 so we could continue to use the Ensemble’s sole SPDIF port.  The Ranes are stackable and in principle we could have tripled the inputs for this revised system.  We retained the 24 module limit however for further interface with Logic’s MIDI control system.

We were not able to discern any sonic difference between this set-up and the initial one using the M-1000s.  We also were able to stay with a single AD-DA conversion cycle.  From this we concluded that the DA converters on the last generation of ROMplers and samplers are as good as (or at least indistinguishable from) their respective SPDIF outs through a digital mixer and then into Logic through Apogee converters.

A post on addressing 24 MIDI channels in Logic simultaneously will follow shortly.

The studio where I do a lot of work has a variety of synths and synth modules.  As a matter of performance ethic we decided we wanted to keep all of them on, all of the time, and accessible via a master stereo buss.  Over a period of time we attempted to devise a variety of ways to accomplish this result.  For different reasons the solutions we implemented were unsatisfactory.  Our Solid State Logic console lacked sufficient inputs.  The new SSL XLogic SuperAnalogue X-Rack has an XR624 eight channel input module going to a summing mixer, which allows for 28 stereo inputs in four rack spaces (60 stereo inputs with two racks).  To be sure this is a high density count however all of the inputs come up at unity gain (none of them have variable gain controls).  Furthermore the on-board computer (which adds considerably to the cost) doesn’t do anything for this module configuration.  The Speck Ultramix is an elegant solution however even with its high input count it taps out and was unable to put enough inputs on a single buss.

We therefore determined to use the Rane SM26 for the mono synths and the SM82 for the stereo synths and modules.  We acquired 24 of the SM26 and 24 of the SM82.  We rechipped several of the older units.  A few of them did not function properly and in lieu of trying to fix them ourselves we sent them off to Rane for repair.  We would like to complement Rane’s product support department, which restored them to pristine working condition quickly and inexpensively.

The SM26 has six input channels and the SM82 has eight stereo channels (actually nine if you use the stereo return as an input channel, which we did).  Both have a feature where you can chain several modules together on a single buss, which we did in combinations of two or three.  We evaluated several master buss modules, including the attractive and desirable Neve 8816 summing mixer.  We ended up selecting the TubeTech SSA 2A, which has eight stereo pairs and four mono channels.  I have the feeling these master buss summing mixers are marketed as a tool for digital audio workstations such as Logic and ProTools, so one can take separate stems from the DAW and mix them down “outside of the box.”  Although our purpose might not be the one initially intended, it works great. 

All of the modules terminate at the back of the racks on EDAC/ELCO 90-pin connectors.  The snakes then lead to (a) stage boxes throughout the studio for the keyboard synths, or (b) EDAC/ELCO 90-pin connectors on the back of the racks for the samplers ROMpler modules.  This was a considerable wiring project, as one might imagine.  Here’s a photo of the front of the racks:

This supplied us with a total of 576 channels.  The stereo buss is whisper-quiet, even when pushed to higher gains.  We attribute this primarily to (a) plenty of AC headroom (900 amps coming into the studio); (b) use of Equi=Tech balancing transformers, reducing the noise floor by about 10db; (c) we tied all of the power supplies together to a single transformer; (d) careful wiring practices such as separating all power from all audio; (e) progressive optimization of the chips inside of the mixers; and (f) over-all component excellence.  We sincerely appreciate the great modules from Rane and Tube-Tech which made this project possible.

BTW we still need some more channels.  

The first string synthesizer I recall was the Mellotron as heard on early Moody Blues and King Crimson records.  This would have been around 1968.  Of course the Mellotron isn’t a synthesizer at all.  It ingeniously plays back tape loops like a gigantic cassette deck. 

There are two ways of mixing the Mellotron: one where it hovers over the entire sonic landscape and a second more subtle variation where it simply is another localized instrument.  King Crimson provides illustrative examples of both techniques.  Its first two records “In the Court of the Crimson King” and “In the Wake of Poseidon” feature the overarching Mellotron.  Whereas on the much-maligned “Islands” its presence in the soundfield is reduced considerably, it is panned off-center without much reverb and it basically sits in a corner minding its own business. 

These two contrasting philosophies pervade much subsequent thinking on the subject.  Most of the dreaded “pad” backgrounds on 1980s records are wide stereo with varying levels of dynamic modulation.  As a result they tend to dominate the mix.  They typically are uninventive and musically disinteresting.  Almost without exception a better use for pads is if they are recorded dry, reduced in scale and presented to the side.

I remember seeing The Strawbs in concert and they actually were using two Mellotrons, which even then seemed like a bit much.

Robert Fripp once offered to sell me both of the original King Crimson Mellotrons.  He said they were sitting in storage and not getting much use.  I told him I thought he should get busy with them.  Even though he likes to eschew that part of his career they were integral to his musical destiny.

For a while I had a rack of Mellotron tapes that were custom-recorded for Black Sabbath.  I bought them from Peter Forrest in a VEMIA auction.  I wonder where they went off to.

The next string synthesizer I remember with clarity was the Freeman String Symphonizer.  This would have been around 1973.  The Peter Sinfield record “Stillusion” captured my imagination with its beautiful music and evocative lyrics.  The Freeman was one of the key elements in its musical presentation.  Even now as I listen to that record I am instantly transported back to a specific moment in space and time.  Music has a capacity to energize texture in a way that no other sense does.  The Freeman has a sweet, almost wistful sound.  One can manipulate the controls for reverberation, sustain and animation in real time, like playing the drawbars on a B3.  I hear tell that its inventor Ken Freeman was one of the few then-contemporary musicians who could play it properly, not just sitting there idly holding chords but interacting with the instrument as he performed.

Then along came many others, including the Logan String Melody and the ARP String Ensemble (this latter entrant in the string synthesizer demolition derby pretty much remaining in post position throughout the later 1970s and well into the 1980s).  Here’s a picture showing all three of them set up, which makes for a formidable combination.  We would have included a Mellotron M-400 as well but from a logistics standpoint it was too daunting.

All of these instruments now pretty much are obsolete for all but aficionados.  Samplers such as the AKAI S1100 and the Kurzweil K2000R took over for about a decade from the early 1990s to the early 2000s.  Mike Pinder (of the Moody Blues) released a sample CD of Mellotron sounds in AKAI S1100 that was very nice.  On the heels of the samplers came a horde of sample players (“ROMplers”) from Roland, E-Mu, Yamaha, Korg and others.  They captured and retained string synthesizer sounds with varying degrees of success.  Now the field is dominated by VST instruments such as the GForce Virtual String Machine, which all things considered is impressive.  It’s amazing how all of those sounds can live on a small shiny DVD and then come alive in a program like Logic.

None of them however sound remotely like the real thing.

I keep getting asked questions about the transformer-based processing arrays I designed in the mid-1990s, so here’s the information.  The line output of most synthesizer modules sounds pretty lame.  It might be characterized as “thin” or “weak.”  It lacks spatiality and dynamics.  In the case of IC or chip-based samplers and romplers (ROM-based player devices) this often is due to lack of integrity in the underlying sample (poor recording technique).  Use of inadequate or inexpensive components in the output stage also is a culprit.  The first can’t be fixed and the second requires internal modifications.  These become increasingly elaborate with the complexity of the device and frequently aren’t worth the effort.

Another problem though is concealed by the nature of line output itself.  While it varies with frequency for most synth modules it is around 1K Ω at A = 440 kHz.  This level of impedance inhibits the blossoming and florescence of the sound.  The outputs also are unbalanced, which is problematic under the best of circumstances.  So I speculated what would happen if the impedance was stepped down to around 50 Ω, which is about the output impedance of a typical microphone.  It then would be possible to use a mic pre to bring the sound back up to line level for injection into whatever recording media one was using.

Jensen Transformers makes a transformer that does just this, which is the JT-DB-E.  Many eschew transformers on the grounds they undesirably interfere with the sound.  They must be listening to inferior makes and models because the Jensen Transformers flawlessly accomplish their assigned tasks and additionally enhance the overall sonic presentation.  I hypothesized it would be possible to design a simple circuit and array the transformers in a way that would facilitate the accomplishment of this objective.

The studio where I did most of the records I produced acquired a large quantity of the JT-DB-E transformer – approximately 96 channels if I recall correctly.  I then went to work on interconnections and here is the result:

By way of explanation the inputs and outputs both are on ELCO/EDAC 90-pin connectors.  The sound comes out of the modules on ¼” unbalanced connectors, which then go to an ELCO on the back of the rack where they are situated.  Those outputs are cabled over to the transformer inputs; processed through the transformers; then routed to the inputs of a separate rack of mic preamps.  From the mic preamps the sound then goes directly to the tape inputs (by-passing the console, which was used only for monitoring and mixing).

This particular studio conveniently had a large quantity of outboard mic preamps, facilitating this outcome.  It would of course be possible to route the output of the transformers into the console’s mic pres or (more contemporarily) into the inputs of an A/D converter for routing to one’s preferred digital media.  In such event the results would depend significantly on the circuit and component quality of each of those electronic intermediaries.  With Apogee converters into Logic the outcome is as good as one would get to 2” Ampex 456 tape at 30 i/p/s with Dolby SR, which was my preferred recording combination.

We discovered these steps significantly improved the quality of the overall sound.  In fact it improved it so much that we became somewhat known throughout the industry for the superb sound we were able to extract from otherwise poorly-performing samplers and other ROM-based devices.  People constantly were asking what kind of secret devices we were deploying to achieve these results.

In addition to the mundane step of balancing the line output this technique also enabled one to pick up the mellifluous tonality and sonorous timbre of whatever mic pre one was utilizing.  We experimented with a wide variety of mic pre combinations with various modules.  Neve 1272s and GMLs typically emphasized higher frequencies and had a lot more “sizzle” around 10 kHz.  API 512s typically supplied mid-range boost around 2.5 kHz.  Others such as the John Hardy M-1 excelled across the entire frequency range.  I could go on and on about these and similar nuances, which needless to say were dependent on a variety of other factors including monitor positioning and room acoustics.  We did the best we could (and if I say so, succeeded) in controlling for these variables.

Thanks to Jensen Transformers, Bill Whitlock and Dave Hill for these amazing devices.