Pushing Back Frontiers?
Emulator III Sampling Keyboard
If there is one company which has made its name on sampling that company is E-mu Systems. The term 'Emulator' is now synonymous with sampling and now the third instrument in this distinguished series is on the market. Paul Wiffen checks it out and finds E-mu bang up to date.
Unfortunately, you cannot monitor your sound source directly through the E3 when sampling, which is the only facility the new Emulator has lost from its predecessor, so A/B comparisons take a while to set up at home (still, I guess anyone who can afford eight grand for a sampler can probably also afford a high quality mixing desk to use with it!). Fortunately, 23.6 seconds of sample time (@ 44.1 kHz) does give you a long enough stereo sample with which to assess the sound quality (the shorter samples on cheaper machines are often over before you've had a chance to hear them properly). Perhaps it is because of the 2 times oversampling used on the E3's outputs (extra sample points are plotted to give a higher playback rate), but I really couldn't hear any difference between even the most testing of compact discs and the resulting samples taken at 44.1kHz. At 33.1kHz, I did notice a slight loss in the highest frequencies (really only discernible on hi-hats and cymbals) and in the quietest passages a barely detectable increase in the noise floor, but less than you'd expect when reducing the sampling rate by a quarter. Maybe some kind of conversion is being used on the 33.1 kHz rate, which would account for the higher sound quality. Whatever the process, the E3 passed this test with flying colours.
I couldn't spot any phase problems when monitoring a stereo sample in mono. My guess is that E-mu are using two separate analogue-to-digital convertors phase-locked to achieve the 'true' stereo sampling they talk about in their publicity. Once again, however they are doing it, it sounded pretty damn good to my ears.
The actual procedure for sampling is as easy as ever. Sample module 5 gives you all the set-up parameters: Sample Rate, Time (with remaining time also shown), stereo/left/right Input Select, and Trigger Threshold, as well as the stereo VU display. The threshold is actually shown directly above the VU display in the LCD, so you can tailor the record trigger to a particular event in the incoming music or to just above ambient noise. The threshold level is set using the data slider, whereas the incoming sample level has its own dedicated slider. The left and right sample inputs seem to be able to cope with all variations between mic and line level thanks to this slider (which removes the need for a mic/line switch, or the 20 and 40dB pads that the E2 had), making adjustments to the incoming level quick and easy.
Extra parameters are also available in the latest version software (V1.13). Here the E3 can be set to automatically Truncate (shorten) and Normalise a sample, saving you precious time doing chores a machine can perform much faster. You can set the Auto-Truncate to act on the sample Start, End or both, and the Normalise for absolute or relative. As there was no documentation with the latest software (I think they trusted me with a prototype version), I can only hazard a guess at the difference between these two normalise functions: presumably 'absolute' takes the highest level sample and uses the ratio of that to the full range available to adjust the level of all the others, whereas 'relative' deals with small sections and adjusts them independently to give some kind of compressed effect which can then be enveloped with the VCA. I'm sure all will become dear when the owner's manual arrives.
Version 1.13 also has an Auto-Placement function which could be set to either white keys or any number of keys between 1 and 24 (ie. two octaves). The first option means that, when multisampling, subsequent samples are automatically placed on successive white notes (useful for sound effects or percussion that have no particular key-related pitch). The second allows you to quickly sample the whole range of an instrument in a repeated interval (as low as a semitone for those who want to get every note authentic, and as high as two octaves for the adventurous and those in a hurry). It takes the bottom C note of the keyboard as the location for the first sample, and then moves up from there in whatever interval you have selected, saving you all that tedious mucking about with sample mapping. The only place where this system seemed to fall down was when the instrument I wanted to sample didn't sit naturally at the bottom of the keyboard (flute, for example). I could find no way to tell the machine to start somewhere other than bottom C, and so had to resort to mapping out samples long-hand.
The all-important Digital Processing module is where your initial recordings are turned into something more suitable for playing from a keyboard. Here you can Loop and Truncate samples, as well as Cut and Paste sections. A nice safety feature is that whenever you do anything which will permanently alter the sample data, the Emulator automatically makes a back-up copy on its hard disk. Then if you don't like what you have done you can get rid of it using the wonderful Undo function. Truncation (if not already done for you by the E3 in the Sample module) is a fairly simple procedure, where you use either the data slider or the keypad to enter the sample number for both Start and End points, shown in the LCD. A handy third line keeps you in touch with the resulting Size of the sample, and you are given the Start and End times in seconds. You can hear the results of each edit without actually performing the truncation, but when you are happy with your truncation you can press Enter to make it permanent. The E3 then asks if you want the sample 'Auto-Correlated Y/N?', which presumably means that the Start and End points of the sample move to zero crossings so that you don't get a click from any jump in amplitude. If you press Yes, then it moves the Start and End points. It's a shame there isn't a mode which would let you step through zero crossings automatically, as this would eradicate the need for auto-correlation (if that is indeed what it's doing!). When you are happy with the sample, you press No. The E3 then makes a back-up copy and performs the truncation.
If there is insufficient room on the hard disk to create a back-up (perhaps because it's a long recording, or you have the disk chock full of goodies), then the E3 gives you the opportunity to have the courage of your convictions and press Yes (whereupon it will perform the truncation anyway) or wimp out and press No (leaving the sample as it was before you began messing with it).
The same procedure applies to Loop Start and End, Auto Loop and Crossfade, except in this case you move the Loop Start and Size parameters and End is moved automatically. It's a pity you can't move the End point directly yourself, especially as Loop Size overrides the Start point (in other words, as the size of the loop defaults to the size of the sample you have to make Loop Size smaller before you can even change the Start point at all). It would be much better if you could move Start and End points independently, so that the loop End didn't move every time you altered the Start. If enough E3 owners want it the other way, I'm sure E-mu would change it.
Once you have set your loop points manually, you press Enter. Again, there's the opportunity to Auto-Correlate. I began to notice that every time I pressed Yes for Autocorrelation, only the loop End point moved. This means that the Auto-Correlation cannot be working on zero crossings (as I had assumed), but rather on matching sample values and slope gradients. I eventually discovered how to step between zero crossings for single cycle loops on brass and woodwind sounds, by using the left/right cursor buttons. Once I had figured this out, I was well away. I then discovered that the E3 way of selecting Start and Size allows you to slide a good length loop (say one cycle) around, so you can create really short samples by moving the loop nearer the front.
Not that there's anything wrong with using Auto-Correlation for longer loops; if anything, it works better than zero crossings on instruments with a changing waveform that need loops that use hundreds of wavelengths rather than one or two, like pianos, strings etc. But nine times out of ten you're going to need the Crossfade loop function which E-mu thoughtfully supply.
After you press No for Auto-Correlation, you are given the opportunity to crossfade loop (either Linear or Equal Power types). Linear tends to work best for those sounds without too much of a regular harmonic structure (eg. percussion) whereas Equal Power is better suited for 'musical' notes, where all harmonics are related to the fundamental. Both options seemed to work as well as any other crossfade looping system I've ever come across. You set the crossfade length in sample units, and the time in seconds is also shown. Once you have settled on type and length, pressing Enter backs up the sample to hard disk and then asks if you want to 'Truncate After Loop Y/N?'. If you press Yes it saves sample memory but means that you cannot loop during a note's release phase, which is often necessary for an authentic dying away of the sound once the key or sustain pedal is released. Again, if you don't like the results you can Undo them. As well as Loop in Release, you can also specify the loop type to Forwards or Forwards/Backwards, and choose Forward or Reverse playback of the whole sample.
From here on in the E3's Digital Processing features go much further than anything available outside computer-based sample editing packages like Digidesign's Sound Designer and Blank Software's great new stereo editor Alchemy, which will shortly be working with the E3. Here you can Cut, Copy and Paste samples like on a Macintosh or Atari. Cut and Copy both set Start and End points for the section to be cut/copied and allow for Autocorrelation before the section is placed in the E3's clipboard (sample number 00). Paste lets you set the sample number where the clipboard data is to be inserted or mixed, with or without crossfade, just like in looping. Wherever possible, the E3 backs up data automatically to hard disk so you can undo any cock-ups.
Next comes Sample Rate Convert. This allows you to specify a new sample rate which the E3 uses to recalculate the sample data. Choosing a lower rate will save on memory or allow you to cover those pitches which the original rate sample cannot reach. But a word of warning: you can tie up the E3 for a very long time if you don't know what you're doing. I mistook which sample I was working on and the Emulator was not available for comment for almost half an hour whilst it produced an unlistenable version of a 15-second piece of music from CD, resampled from 44.1 kHz down to 10kHz. However, once I'd worked out what I was doing I achieved some very good results, most spectacular of which was a trumpet sample which I resampled at half frequency (22.05kHz), taking just over two minutes, and indistinguishable from the original except that it now replays up to three octaves above the original pitch.
The final Digital Processing section is called Digital Effects. It is a cover-all for anything else the clever chaps at E-mu come up with. The 1.01 operating system offered only Digital Tuning, which recalculates the sample data to change the tuning. Again, this is not a very speedy routine, but it's a permanent process for changing the sample's pitch, unlike the tuning function in the Analogue section which only applies to the appropriate zone of one preset. This is most appropriate when a sample is out of tune, and you can rectify the problem in cent and/or semitone steps. In Version 1.13, however, Digital Effects has expanded to include Taper, Gain Change, Reverse Section, Stereo-Mono, and Sample Calculator.
Taper allows you to fade the level of the sample up or down from a definable Start point to a definable End, either of which can be changed by up to +/-96dB. The curve between these two points can be set to linear, or three different degrees of exponential. The best use I found for this was to fade out any noise at the end of percussion samples, but the way it is formulated is so open-ended as to be applicable to many different problems. And with the hard disk back-up, you can have as many goes as you need to get it right.
The same is true of Gain Change. Having specified a Start and End point for the change, the E3 then tells you what sort of volume increase is needed to normalise the sample. You can then use this amount or any other between +/- 96dB to alter the level of the sample digitally. If not altering the level of the whole sample, you can specify a fade to smooth out level fluctuations.
The Stereo-Mono function lets you make mono samples of stereo ones and vice-versa. Mono-to-stereo conversion involves duplicating the signal for both sides, which can be done fairly quickly provided there is sufficient free space in the bank: this is useful as a starting point for analogue processing or as a way of ensuring the sound comes from both left and right outputs. Stereo-to-mono is a more complex conversion process, as the left and right signals have to be summed. The 12-second stereo sample I experimented on took around 20 seconds to be converted to mono.
Sample Calculator doesn't actually modify samples in any way, but it's ideal for working out the change in sample rate required to re-pitch a sample a fifth lower, for example, or to change the cycle length. There are two parameters you may alter: the original sample pitch and the sample frequency. The E3 then tells you the frequency in Hertz and the length of a single wave cycle in sample units. Besides anything else, it's very informative to see how the numbers change, especially as you can vary pitch very quickly from the keyboard and frequency from the data slider. These Digital Effects are all extremely useful for tidying up sounds and making good samples from bad, but they are not exactly what most people would understand by 'effects'. I look forward to further additions which will give you the sort of signal processing effects everyone is crying out for - phasing, delay, and maybe even reverb. Even so, this is certainly one of the most forward-looking areas of the E3.
This module allows you to Load Zones from previously recorded banks, Edit their assignment to the keyboard, Copy or Erase them, and thus build new presets from old with the minimum of fuss. Copy Zone, in particular, allows you to double or chorus sounds very quickly.
Like the Emax, the E3 can assign two samples to a note within a Zone. These can then be used for doubling sounds or for Velocity Crossfade or Switch. Preset Definition also allows for MIDI parameters like Channel, Mode (Omni or Poly) and Overflow to be set and stored for each preset (by placing each preset on a different MIDI channel, the E3's SuperMode can be used to sequence different sounds from different MIDI channels on an external sequencer). Also set from here is Pitch Bend Range (max +/—7 semitones - a perfect fifth) and the choice of 14 different Velocity curves. There's even a versatile arpeggiator, with some 15 different parameters - anyone still using arpeggiators will love this one.
But the most interesting Preset Definition facility for me is the matrix assignment of Real-time Controls. Here you can route the Left or Right Wheel, Pressure, Foot Pedal, or two MIDI controller numbers to ten different destinations including Pitch, VCF, VCA, LFO to the previous three, Pan, Attack or Crossfade. Similarly, two footswitches can be routed to control such functions as Sustain, Sample Switch, Sequencer Stop/Start/Continue, Arpeggiator control or Preset increment/decrement. If you can't do it here, chances are there's no call for it.
This is an area of great interest to me, as the analogue parameters of the Emulator II were what lured me into sampling in the first place. Needless to say you can do all the standard ADSR envelope shaping of VCA and VCF, but with the addition of a Hold parameter which sits between the Attack and Decay and allows the sample to sound at full volume/brightness for a specified time. This is particularly useful on samples, as you often don't want the Decay to start until the loop does. If only the VCA had keyboard tracking, like the VCF, then you could really fine tune your VCA envelope to match the decay phases to the loop.
All the envelope times are displayed in seconds and hundredths for accuracy, and with decays and releases like 163.69 seconds available, the New Age music fans should have a field day! There is also a third envelope (also AHDSR) which can be routed to a whole host of things (Pan, Pitch, and various LFO functions) and what you can't route Velocity to isn't worth trying (Pitch; VCA Level and Attack; VCF Level, Attack, Cut-off and Q; Pan; and, best of all, Sample Start - my favourite parameter for avoiding Velocity Switch/Crossfade by having quieter keystrokes miss the emphasised attack of a loud sample). The LFO can be routed to Pan, VCF Cut-off, VCA or Pitch, simultaneously, but with different amounts if required, and you can even introduce a percentage deviation from the programmed amount for adding extra humanity to ensemble sounds! There's Delay and Chorus as well, which first appeared on the Emax and which add to the formidable range of effects that make richer, more complex, textures quick and easy to achieve on the E3.
Four keyboard modes allow you to select between Gate and Trigger, Transpose and Non-Transpose, Arpeggiator Enable/Disable, and Solo/Polyphonic performance. You can also enable or disable any of the available real-time controllers, and select which of the 16 individual outputs are used. Any of the above Analogue Processing parameters can be set independently for each Zone on the keyboard, and parameter 0 lets you choose which you are working on at any time. This is achieved in the fastest way possible, by simply selecting the low and high keys of the Zone you wish to work on. This means that you can modify just one note of one sample, just one sample, several, or the whole keyboard range simultaneously, however the fancy takes you.
That just about covers all the things you can do with the Emulator III as far as sound shaping goes, but we still haven't mentioned the onboard sequencer, the SMPTE features or the SCSI possibilities. I am happy to say that E-mu have learned from their mistakes and the sequencer on this machine is as flexible and editable as the E2's was rigid and awkward to use. You can use the sequencer with looping and auto-correct like a drum machine, with cut/copy/paste like a computer sequencer, or with SMPTE as a Cue List. You can step edit sequences by using the » and « locate buttons to move through the events (hearing them as you do so), and alter the parameters with the cursors and keypad. The sequencer can be clocked through MIDI, SMPTE and, in the near future, MIDI Time Code. Events can also be tied directly to timings for SMPTE or MIDI Time Code triggering (for those who want to sync sound effects to picture etc).
Besides MIDI and SMPTE, the Emulator III also features RS422 and SCSI interfaces. I would imagine the RS422 port (which is now starting to look dated as an interface) is primarily for interfacing with Optical Media's CD-ROM system (they are already preparing an E3 CD) and with computer editors like Sound Designer and Alchemy. The SCSI interface is the forward-looking side to E3 expandability. This is bidirectional, allowing simultaneous communication in two directions and up to eight devices on the buss (if the E3 is the first, the internal hard disk the second, that means six more hard disks could be used simultaneously). SCSI is also ideal for communicating with other mass storage media like writable optical disks - Optical Media have a WORM (Write Once Read Many times) drive coming later in the year - so there are no fears for the future expandability of the E3.
E-mu themselves are careful to keep the E3 an ever-growing system. The 4 megabyte RAM expansions are available this month, as is the 8 meg Emulator. Either way the additional cost will be about £1500. In three to four months time E-mu will be releasing the E-Rack, a 300 megabyte hard disk which communicates with the E3 via SCSI, and this will give the sort of on-line memory you need for direct-to-disk recording which I'm sure E-mu are working on. By the end of this year the E3 system expander should also be ready, enabling users to jump to 32 voice channels/outputs and giving another 8 meg of sample RAM. All in all, it looks like the E3 has got the edge in terms of expandability.
But even the basic unit is the most compact solution to the demands of 16-bit stereo sampling. You can't really use the amounts of memory you need to do sampling of this quality and length without a hard disk, and the E3 comes with a 40 meg unit built in. Before using the E3, I was as much in favour as anyone of taking advantage of the cheaper prices of hard disks in the mass-market computer industry, but having seen the advantages of an internal hard disk, I'm a convert. I'm not just talking about not having to connect up separate boxes every time you want to use them (major consideration though this is). I'm referring to more subtle advantages, like automatic back-up of samples before making permanent changes, and the fact that the machine doesn't have to keep loading the operating system from floppy disk every time you change module (admittedly only an E-mu disease).
It strikes me that 4 meg of sample RAM is the minimum requirement for stereo sampling at 44.1 kHz and I can see the need to expand this impressing itself very quickly on users. By the time the system expander appears, I suspect people will be screaming for it. The same goes for 32 voices with separate outputs. With a multitimbral system (which sampling has always been), you can never have enough voices.
I am pleased to see that E-mu are staying away from fads which appeal to the spec hunters, but have little practical application. I refer principally to the trend started by Casio for numerous loop points - the FZ1 offers six loop points per sample. Big deal! After you've spent six times as long setting up these loops, all you get is six different timbres stepping through the loops. Who needs it? One well executed loop with the option to loop in release or read through to the end of your sample data is all you need, and this is what the Emulator III gives you. It is this sort of approach which distinguishes the E3 operating system throughout. It gives you the tools you need to do the job and doesn't waste your time with impressively named gimmicks which may read well on paper but add nothing to your sound.
E-mu pioneered the field of user-friendly sampling, with data compression short-cuts when memory was expensive and storage capacity limited, so it's good to see that now cheaper memory and greater storage capacity have reduced these constraints, they have gone on to greater fidelity, improved save and load times, made back-up easier and more reliable, and left the way open for future expansion. With all the things you need in one box (large sample memory, hard disk, 16 voices, sequencer) the E3 makes a very neat package. I'm pleased to see them back out in front: 'The Emulator is dead. Long live the Emulator!'
Contact E-mu Systems (UK) Ltd, (Contact Details).
Review by Paul Wiffen
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!