Alesis made their name with their reverbs, but really came of age with the revolutionary ADAT digital multitrack. Paul White talks to the company's founder and President Keith Barr about the impact of ADAT, and the future of technology in music.
Despite the achievements of his company Alesis, Keith Barr seems less than comfortable with the crowds at music trade shows. Whenever I meet him, I always get the impression he'd rather be back at base, working on a new design. However, he's always ready to talk shop, so at the winter NAMM show in California, we escaped to the tranquillity of a back room for an informal chat, and what follows is the result of that conversation.
Alesis was founded when Keith Barr turned his back on designing MXR effects pedals, and instead developed the XTC reverb unit and the world's first ever ultra‑low‑cost digital reverb, the Midiverb. The last time I spoke to Keith in any depth was a year or two before the announcement of ADAT, and at the time, he saw tape as having a secure future as an audio storage medium for many years to come. Even with the recent dramatic falls in the cost of hard disk storage, tape is still significantly cheaper than disk as a storage medium, and ADAT has obviously been one of the big success stories in recent music history. I asked Keith if ADAT's success had taken him by surprise.
Do you realise it costs a billion dollars to make a plant that produces eight‑inch silicon wafers in sub‑micron geometry?
"It's overwhelming. I hadn't realised until this trade show just what a profound effect ADAT has had on the recording world. When I was younger, playing guitar and going into a studio running two inch, 16‑track tape, a big board full of knobs and huge speakers — that was the ultimate. The notion that we would be building machines to bring that kind of capability into people's homes was astounding."
Last time we met, we were looking at the life expectancy of recording tape. Do you still think its future is secure?
"We researched various removable media concepts, and at present, it seems that only through very aggressive compression techniques would you be able to use today's alternative media technologies in a way that would allow them to compete with tape. I can't predict how long tape will last, but I would be surprised if it wasn't still going strong even ten years from now — no kidding. There's a great advantage in an inexpensive medium which you can just fill up with data and stick on the shelf."
Did you design the new ADAT XT so that the tape transport could be replaced by a disk drive of some kind, should the technology become available?
"Products like the XT are synergistic, because we find that's the only way to make a cost‑effective product. If we did build a hard disk system, we'd have to take the same approach and build it from the ground up — we couldn't really modify ADAT technology."
Now that the ADAT XT has replaced the original ADAT, can we expect to see a new BRC coming along, or will the existing one do the job okay?
"System synchronisation issues, which is largely what the BRC deals with, are complicated, and it's probably a good idea for us eventually to think about a more up‑to‑date BRC — but I can't divulge any plans for one yet. You don't need to worry about selling your old one though — it will work fine with the XT!"
Digital Audio
How do you feel about the analogue‑versus‑digital argument: the notion that digital sounds clinical?
"Introduce me to that guy! It all has to do with sample theory, information rate and a little bit of maths, I suppose, but ultimately you use your ears, and if you find something you don't like, you track it down and solve it. In a digital system, it could be a crummy filter here or a little extraneous noise there, or even something happening in the speaker system."
Is 16 bits enough, or should we be looking towards higher resolution systems in the future?
"It depends on what you need. There's a big difference between active signal‑to‑noise ratio and full‑on dynamic range, and if a person wants to be a bit sloppy about recording levels, then a little extra dynamic range would be helpful. The current signal‑to‑noise ratio is pushing 90dB — the extraneous distortion and noise that introduces when a signal is present in a channel is so insignificant that I'm not sure, for any practical purposes, we need to go much beyond 16 bits. In terms of dynamic range, I'm sure 20‑bit would be helpful, but it's not so much a question of how you print 20 bits to tape, it's how do you convert audio to 20 bits of digital? There are a lot of converters that claim to be 20‑bit, but they are barely 17‑bit in terms of dynamic range."
Do you think noise shaping techniques are helpful in extending the dynamic range of 16‑bit systems?
"The results we've had from experimenting with noise‑shaping systems are mixed. The results for some kinds of sounds yield an improvement, but for others, not necessarily. No system we make right now has a noise‑shaping system in it; we've done some experiments in that area, and merely shifting where the noise pops up in the spectrum causes a change in the way the sound is perceived. It hasn't been demonstrated yet that it offers any great advantage.
"When you're talking about signals down in the least significant bits, you're not talking about meaningful information in the first place. Some background noise will always be present, upon which we place sounds that are, hopefully, pretty far up in the dynamic range compared to the noise. If you actually try to resolve tiny signals down in the least significant bits, where noise shaping really could be of value, you're really asking a lot of a system."
In my own experience, far more noise gets onto the recordings from the source material than can be attributed to the recording machine anyway. Would you go along with that?
"Well, you have to be careful about levels when you're tracking, but in general, I think you're probably right. You're going to get noise from mic amps, and from the electronic instruments you're recording. Perhaps we could go a little way beyond 16 bits in the future, but I can't see a need for 24 bits — which is what some people are now talking about."
On The Level
Perhaps you can help to clarify some confusion over digital metering? I've had readers phoning up asking why, when a steady tone on their console meters reads 0dBu, ADAT is reading it at around ‑15dB. I think I know why this was done, but I'd appreciate hearing your reasoning.
"With digital audio, peak reading meters are used, because everything is linear up to digital saturation, and beyond that we have clipping. Most mixing boards are set up with meters that are best suited to analogue machines, which means that you'll often go beyond 0dB to push the machine a bit harder, and you'll suffer a little more distortion due to tape saturation. In other words, mixers have a certain amount of headroom above 0dB, and most people are expected to push their signals into that region. Even when the meters are reading below 0dB, transient peaks may still be well over 0dB.
"Headroom, when applied to digital recording, makes no sense, because you can't push the signal beyond 0VU. If you put a sine wave into your mixer and it reads 0dB, then you will get a reading of around ‑15dB on your ADAT — but if you put in a microphone and your drummer hits his sticks together, where the meters on the board go and where the meters on the ADAT go will be different. The meters on the desk will barely move, but you may still have clipped the ADAT."
So if you had made the ADAT read 0dB at +4dBm (or ‑10dBV), the user would either always be running into clipping, or be forced to use the mixer at an unusually low level, which would probably compromise the gain structure of the mixer?
"It's really a matter of education, as the industry outgrows the older notions of nominal level, and enters the digital world — where only peak levels matter."
And it doesn't help that different manufacturers of digital equipment tend to settle on slightly different levels — there's no real standard, from what I can tell.
"It all comes down to what you want to record — if you want to record sine waves, you don't need any headroom, but if you want to record kick drums, you need lots. I don't know if a standard is actually required, because one day it's likely that everything will be done using peak reading meters. Perhaps the only place we need a standard on nominal levels is at the mastering stage, so that when we play back any two pieces of music over the same system, they come out at the same perceived acoustic level. That's the only place, other than broadcast, that I see standards being beneficial."
You can talk about modelling all day, but a good keyboard with nice sounds in it is something that most people will want for a long time.
People often tell me that they push their ADATs into clipping, but they don't hear any distortion. Is there any form of analogue limiting in there, or is it just because short periods of clipping are difficult to hear?
"There's no limiting in ADAT, but different digital systems clip differently, and some digital systems clip terribly. The way numbers are represented in digital audio systems means that if the signal clips and there is no provision for what we call saturation limiting, then the peak of a sine wave, instead of just flattening a little bit, will jump to its opposite extreme level for a brief moment, resulting in a huge glitch which sounds terrible. We don't allow that to happen in ADAT or any of our signal processors, but some older digital products suffer from this problem — the one that helped give digital audio a bad name."
Conjugating 'Verbs
Alesis has a strong name in the effects market, but are there any new effects out there to create? Are we destined to keep finding new ways of combining and refining the old ones?
"We walk into a room, we clap our hands and we get this marvellous echo coming back, so we work on putting it in a box. Or you move your head and experience this funny flanging sound — we realise what it is, so we put it in a box. Effects don't start here from people imagining what could be generated by electronic processors — we hear natural phenomena, which we try to duplicate. All today's effects have grown from that."
I was raised on vacuum tubes, and I felt like the luckiest kid on the block when my daddy bought me a bag of 100 silicon transistors.
Initially, your fascination was with reverb and how to make it available cheaply. Has everything been done in that area, or would you like to explore further?
"There's no end to reverb. It's like a violin — you can always make a better one. It remains an intellectual pursuit, but it's also a matter of trading off cost and performance, and refining the involved elements to come up with the best and most cost‑effective solution. This is, by and large, a cost‑driven marketplace, and I think we do very well in offering the right combination of quality and price."
That seems to be evident from your new effects processors. When you take the top off a Microverb IV, there's very little inside.
"You should look in the new Nanoverb — that's got a few of my analogue tricks in it [see picture — Ed]. The sound quality of our effects is much better than it used to be, before we found some conceptually simple ways to improve the algorithms — but that still involved a lot of hard work and sensitivity. Reverb algorithms are still hand‑crafted things — I couldn't hand you a formula that would predict how a reverb would sound, although there are people who claim they can. The parameters that are hidden from the user are what make the difference to a reverb sound: their values are really important. When an impulse goes into a reverb, the order in which the successive impulses exit the reverb says everything about the tonality that you perceive, the impressions of repetition, of smoothness... And those things are very difficult to hand‑craft, to get them to sound the way you want them to sound."
Related to reverb is the concept of perception. Do you see any mileage in any of the 3D sound processes currently available?
"Let me talk a little bit about that. I want you to try an experiment someday with a white noise generator — get it from a synth, or tune an FM radio between stations. Connect to a single in‑ear headphone to get a point source, and place it about 12 inches in front of your face. As you move it around or up and down, notice how the frequency response seems to change due to the shape of your outer ear. You'll hear resonances, high frequency losses and very abrupt transitions — moving it just one inch can cause a huge difference in frequency response. It's clear we get clues as to where sounds are coming from due to the expected frequency energy spectrum, and the perceived energy spectrum when we move our head. Very often, we'll move our heads just to see how a sound changes.
"You could try to replicate these tonal changes using filters, but you'd need to know exactly the shape of your head and ears to do it. I could build a set of filters and apply them to a sound in such a way that you would perceive the source as moving all around your head, but if you were to do the same experiment with someone else who had a different‑shaped head and ears, it would be a different experience."
So do you think the difference between individuals is so great that we'll never get a system that works reliably for everybody?
"I don't really know enough about the differences between different people's hearing responses, but I will say that I cannot imagine in my wildest dreams that it will ever be universally effective. Maybe, to get futuristic for a moment, you could have your hearing analysed and the data put on some on some sort of card that you could swipe through a slot in your hi‑fi system that would move the sound around just for you, but it might not move around so nicely for somebody else listening with you."
Keyboards
I notice you have extended your keyboard range, so it's obviously an area you're pursuing. Is there anything else out on the market that interests you, or any new keyboard technology that you'd like to explore?
"We have excellent keyboards, but more than that, we have some excellent sounds, because we have some really neat people who just live their lives to make these sounds. They're the most inspired, creative people I've ever met and they're a joy to work with. To me, keyboards are good, solid instruments that humans play, and I'm not certain that anything wild or esoteric needs to be done. It needs to feel good and the sounds need to be satisfying. You can talk about modelling all day, but a good keyboard with nice sounds in it is something that most people will want for a long time. There is joy in playing a keyboard and being satisfied with what comes back; then there's the more technical experience of examining how sounds can be manipulated.
"Real‑time control of things like filters can add life to solo parts without resorting to modelling, which is why the microprocessors that run these keyboards are much more important, ultimately, than the signal processing chip that does the number‑crunching work. They have to scan the keyboard to see what's being played, they run the displays and keep track of all the voices to see where its filters are going — it keeps them busy.
Which of the products you've developed, other than ADAT, do you feel best about?
"You have to understand that I don't get fully involved with all these products, but I really like the original Microverb and the SR16 drum machine. The SR16 was a joint effort between Alan Zak, Marcus Ryle and myself; we did it in a very short period of time, and I think I can speak for them when I say it was thoroughly satisfying for all of us."
Machine‑Gunning A Feel
On the subject of drum machines, is there anything that can be done to make fast‑playing sounds more natural? Even on my Alesis D4, I still can't get tom fills to sound quite real. It's something about the way the samples interact and retrigger that isn't quite right.
"The pads that we've developed will help enormously from the playing point of view, but to get a really natural sound, you have to use multisamples. ROM is getting really cheap, and I think you're going to be able to look forward to sample playback devices that have more and more in them. For example, you might have ten snare drum samples under one key, so that each beat and each different amplitude of hit is slightly different. We'll certainly be looking at this area. Sample playback machines — be they keyboards, drum machines or whatever — I wouldn't ever like to see the company give up doing something like that.
"It's very satisfying to make a musical instrument, and clearly, all the things you're alluding to — multiple samples and different amplitudes, being able to play rapid rolls that don't sound like a machine gun — all that comes from the higher‑speed microcomputers that are available now. Am I defining a new product here?"
Digital Mixing For Humans
Contrary to all predictions, you've yet to come up with a digital mixer to accompany ADAT. What are your reasons for holding fire in this area?
"That's the way of the future. The concept of digital mixing offers many advantages to both the user and to the manufacturer. One of them is economy; you can get more at your command, at a lower cost. The mixer can also be smaller, and the ability to integrate effects and multiple routing paths without needing a lot of patch cords is attractive.
"I feel that the biggest problem with digital mixing consoles at the moment is the human interface — the controls. With current technology, this might best be implemented using big CRTs or maybe colour LCDs, but these aren't cheap. There are pots that move, but you have to think about what their function is at every moment — it becomes a struggle trying to relate how you move this control and what the numbers changing on the screen mean. I feel that digital mixers will only become really practical, and ultimately dominate all mixing possibilities and make analogue obsolete, when we have controls and visual feedback mechanisms which are both very low‑cost and responsive.
"The user has to be able to interact with the controls, rather than have to displace himself intellectually by thinking, 'I turn this knob, this number changes, therefore such and such happens'. The control surfaces and feedback surfaces must evolve before digital mixers dominate — as they ultimately will. As you might imagine, we are dedicated to that technology."
Does that mean that the electronics is developed to the point where it can equal the performance of a good analogue console?
"The signal processing chips could be built with absolutely mind‑boggling capabilities, and they could be made at an unbelievably low cost. The problem is not in the the electronics, it's in the controls."
One of the pitfalls that I can already see is that you can get a well‑thought out console like the Yamaha 02R, where all the usual controls are fairly easy to get at, but because there's so much more you can do with digital mixing systems, there's the temptation to add all kinds of new features that can't be accessed so easily. How do you avoid that pitfall?
"The grand promise of digital mixing is that you can multiplex a limited number of controls over a wide number of channels, yet it takes a lot in your head to keep track of everything. On the other hand, if you were to allow all the controls to be immediately available, you'd have a very large and very expensive control surface. If you want to multiplex a large number of functions via a small number of controls, your brain is going to become involved."
Do you see outboard gear remaining in separate boxes but digitally linked to the console, so that when you save a mix, you also save your effects and outboard settings?
"I don't think so. Today, with Windows on your computer, you can call up many windows, allowing you to access several applications simultaneously. Even with a traditional studio setup, you can't focus your attention on more than one task at a time, and on the board, you can only focus your attention on one knob at a time, so the idea you'd want separate devices present all the time will wither away, driven by economic demand. The challenge is: how do you organise the control surface to be so incredibly, intuitively responsive that you can at one moment be controlling a mixer channel, and the next an effects patch or a tape machine? It's like virtual reality without wearing the goggles."
Happiness Is A Bag Of Chips
You've obviously been very successful, to the point where I suspect you would no longer have to do this for a living if you didn't want to. That being the case, what keeps you going and what fires your enthusiasm the most?
"I love doing this, but what I'm really excited about is designing integrated circuits. I was raised on vacuum tubes, and I felt like the luckiest kid on the block when my daddy bought me a bag of 100 silicon transistors. It wasn't until I was about 20 that there were such things as analogue ICs. There were also simple op‑amps, which you could wire up to make filters and other fun things, and I immediately went off making fuzzboxes and phase shifters. But in the entire world of electronic design today, nothing, nothing comes close to the playground that is afforded by integrating analogue functions onto silicon. It is the most fascinating, wide‑ranging, elegant and beautiful medium that any electronic engineer could be involved with.
"I do a lot of work in this area, and it shows up in a lot of our products in very strange ways. I'm not going to go into detail, but we can take a logic CMOS processor with its logic‑type transistors on silicon wafers, and can configure devices transistor by transistor... to produce operational amplifiers, resistors, capacitor networks, temperature sensors and a hundred other things. You're in a medium where the electrical leakages are so close to zero, they bear no relationship to the leakages you see on a printed circuit board. When you're dealing with capacitances in the range 0.001pF, or current in the pA range [one million, millionth of an amp — Ed] scientific notation is required to get a grasp on how small things are, and how vast the capability is. We have only scratched the surface of what is possible using CMOS FETs on silicon. It is the most wonderful playground, and I'm standing on the diving board, leaping in and having a blast with it."
Does that mean you're still focusing on the music market, or have you looked at other IC applications outside music that you'd like to get involved with?
"The cost of the computer tools for developing and processing ICs is so great — and the size of the music industry is so small — that it's not cost‑effective to get really deep into this kind of design work exclusively for music. There are many applications outside of music, and I wish to produce devices for those commercial applications. However, there will certainly be ideas that can be transferred to music and paid for in the process. Our efforts in the music business can only be improved by expanding into commercial integrated circuits. Do you realise it costs a billion dollars to make a plant that produces eight‑inch silicon wafers in sub‑micron geometry?
Fantasy Island
A few years ago, Keith Barr came up with the 'Dream Studio' concept, where Alesis would eventually produce all the major items of equipment found in a typical studio. I wondered whether or not this vision had faded.
"Studios are evolving, and the technology is always changing direction — it's like chasing a chicken. We'll keep coming up with new products that we feel are of value to people working in this industry, and the direction it takes in the next year is likely to be slightly different from the direction we've plotted this year. The Dream Studio concept still seems valid, but it's hard to predict things too far ahead."