The Sound Archeology Digital Remastering Process : Digitally Rewriting History

An Interview with Sound Archeology's Founder Robert Ball

One evening in mid-June, I caught up with Robert Ball, Sound Archeology's founder and chief technical engineer, and had the opportunity to get some insight into the technological and philosophical ideas behind his remastering process. I began the interview by asking Mr. Ball how he got involved in the black art of CD remastering…

Andrew Chasin : How did you first become interested in developing a process which improved the sound of the compact disc?

Robert Ball: It's kind of entertaining. I've been doing computers since I was 12. I built my first stereo when I was 7. I've always been into audio and electronics. My father brought me into it. In the '70s, I did some work at Atlantic wiring some boards and other types of things. But mainly, I went to college and then I got into the computer industry and I began working on Wall Street. I used to do a lot of driving from job to job and from system to system and I'm very eclectic and I love music. I had a very high-end system in my car. I wanted to hear really great sound. I went out and bought these CDs of stuff I really loved. I used to buy first generation records. I literally made a little pencil mark on the leaf for each time I played it [ laughs ]. And then I heard CDs and I was actually quite disappointed in them. So that really got me on the road as a hobby. So in my spare time I started building sound cards. I didn't like what I was getting. So I got my hands on an AT&T prototype board that they were using for long-distance communications, and I modified it up to 128-bit/96kHz. I started redoing music and then producing cassette tapes to listen to in the car.

AC : That was an interesting application of that technology!

RB: Yeah, John Tabacco [ John Tabacco is a composer and mastering engineer, and was a significant contributor to the development of the Sound Archeology process ] basically says "he's crazy!". You know, it turned into like a passion, you know people build boats in bottles and stuff like that. I would spend hours and hours redoing these mixes of these CDs so that they would sound like I remember the records sounding, you know that real true sound. So I was writing more and more software and more and more stuff, and about six years ago, I met John Tabacco and I played him some of the stuff and he just went crazy and said "this is unbelievable."

AC : Is John a mastering engineer?

RB: He's a composer and mastering engineer. He just went nuts, he'd never heard anything like it, he couldn't believe it. So , I said "yeah", I'd been working on this but it takes thousands of hours, you know, for the machine to grind through it because there's so much data. So I made three major scientific discoveries when I realized that the problem with digital audio is the sampling process itself. Even now, for example, they're going 24-bit, 96kHz, and it's terrible. Basically they've only added eight more bits and they've doubled the sampling rate and everybody focuses on these numbers because they're the buzz words because "It's 96kHz, it's 96kHz". But AT&T, between 1985 and 1987 published, articles and studies, where they found, and I found the same thing by the way through experimentation, that you need at least 128 bits to get the type of linear resolution and the curve to produce a 96kHz sampling rate.

AC : Really.

RB: Oh yeah, this is fact this is not fiction. And what happens when you go to 128 bits? Your data rates go through the roof! I'm running four Unix towers in parallel with 132GB of storage so I can deal with that, but most people, we're talking ten years ago, did not have the capabilities to deal with that data. However, if you're amongst the studio people, you know that some of the best digital delays, which give you one second of digital delay, are 128-bit 96kKHz and they sound incredible. The whole thing about going to 24-bit 96kHz, is that everybody is thinking that the problem is in the sampling rate, but it's not. If you actually analyze what's going on, the problem is that, in any audio you have a wave, and the frequency is a finite amount. However, you have two infinities to deal with - zero to one and zero to minus one. And what happens is that in there, they only allow 35,757 spots to the top and the same number in the bottom, and that's the problem. Now when you go to 24-bit, 96kHz, they haven't changed the top or the bottom, so now these triangles are twice as long, that means you have twice as long a flat line to deal with, that means that the arc that you're cutting out is even greater, and that's the ripping that you hear. When you listen to digital stuff you know that ripping that you hear, the of loss of essence and nuances? Well that's where it's coming from. So the problem is, how can you convert an analogue signal to digital and get rid of this, and also how can I take this 128-bit, 96kHz stuff I have and convert it down to 16-bits without all these aliasing problems that I was having?

AC : That's the tough part.

RB: And that's when I invented the SA3, that's the third generation of hardware and software. The first two were not totally successful but they weren't failures either. They led me in the right direction, the basic theory works. It's basically three distinct areas of scientific discovery. I would like to publish this stuff but I can't at this particular point in time because you can't patent scientific discovery. Because my technology is based on that, and we have not as yet patented it, we're not sure if we're going to yet. I've invented a whole new A to D that doesn't do sampling so we can use the full 16-bits and we have a higher frequency response. That's why it's so much louder, I guess you probably realized that like "how come it's so much louder?"

AC : Sure, that's the first thing I noticed when I listened to the processed tracks.

RB: It's because we don't throw away half our data [ laughs ]. The other thing is that all of the manipulations are done in a 512 bit format, so that's why you hear such smooth transitions and good dynamics. I don't know if you noticed the depth and everything that's there.

AC : Anthony Kershaw and myself sat down and listened to the CD the night it arrived, and we were just kind of blown away with the two Steely Dan tracks and the Frank Sinatra track, which I wasn't too familiar with in its original form but Anthony was, and his jaw just dropped when he heard it. We were both quite amazed. It just sounded more analogue-like, the depth was there, timbres were right…

RB: Yeah, it sounds like you're there, you can hear Count Basie's foot right?

AC : Yeah, exactly!

RB: And Frank snapping his fingers in the background when the solo was happening, right?

AC : Definitely!

RB: And you even hear when Frank comes in and he moves really close to the mike and the guy dumps him right into the reverb, because they were mixing that live. You know, sometimes it's just easier for me to tell people that I built a time machine and I went back in time and recorded him, rather than for me to tell them that it's just physics, mathematics - just science. We're doing 6-8 trillion calculations a second here and we basically have 85 processors in parallel that are doing that.

AC : So, that answers one of my other questions. Obviously this process isn't computationally simple, but do you think it's possible that this sort of thing could be done in real-time.

RB: We are doing it in real time!

AC : But I mean could it be done real-time in a commercial form?

RB: Yeah, we are doing it in a commercial form!

AC : Okay, let me rephrase, could it be done in a consumer-based commercial form?

RB: Yeah! Yeah, right now we have two prototype systems. The two systems I designed and programmed myself can be turned into chips.

AC : So do you think that it could be done cheaply enough such that consumers could buy a black box to hook up between their transport and D/A converter and do this on the fly?

RB: Oh sure, but that's at least three years away to be honest with you. However, as a mastering facility, we're open for business now, not to mention that I have a catalogue of over 800 albums that I've done over the past ten years which is stuff that you've got to hear to believe. I thought you'd be impressed with the Frank Sinatra stuff and the Aja tracks.

AC : I was definitely impressed. The first tracks, the Mobile Fidelity reissue…

RB: If you actually A/B them against the Mobile fidelity originals you'll actually find that ours are better than theirs. Look, I come from the computer area but I also come from the audio end. I'm into digital integrity. Everybody is into the mentality of a limited loss. Every piece of equipment that's designed has a limited loss. All the stuff you're hearing, we did not have the masters to it.

AC : That was one of the most amazing aspects of your stuff, that it was done with only the commercially available recordings..

RB: Thank you. The reason we can do that, specifically with the older CDs, not these new ones that they've recently digitally remastered, we can't do anything with those because, believe it or not, all of that noise they're getting rid of, that has all of the information that we need [ laughs ]. Just because you don't hear something, doesn't mean that it doesn't exist. In order for you to hear it, it has to be at least a cycle, actually a few cycles long, and has to start propagating a wave. Some of the signal strength on these recent remasterings isn't that strong and some of the waves aren't complete. What happens is that the D/As determined it as noise or the systems on which the process was done determined that it was a quantization error. It's not. What it is, is transient signals which are part of further waves which were not propagated because of either a) the original analogue recorder could not capture the frequencies or b) the A/D they were using couldn't capture the frequencies. Because they're cycling at 44,100 times a second, they still measured that voltage and put it down as a sample in that spot. Our process knows the proper math and can predict those values. It takes a tremendous amount of computing power, but you can do it. It's all physics, it's all physical stuff, and I've developed proprietary hardware and software that understands it. When you listened to Frank Sinatra and heard the bass, and heard his hands, and heard all that stuff, you didn't hear it in the original, you could turn it up from here to oblivion. Believe it or not, my source was a third generation reel-to-reel, but we didn't put that down since nobody would have believed it. The reel-to-reel was thirty years old when I did it. Who's going to believe that? We don't bother saying that because people just don't believe us. They're like, "what are these guys doing, they're compressing it or something like that". No we're not compressing it.

AC : People might be attributing the difference in volume between the A/B tracks to compression.

RB: Listen to the soundstage alone. We have 125dB of separation. You might ask "how can there be more separation than my hardware is capable of?" Well, we've found things that people aren't using in the hardware.

AC : Such as?

RB: Well, I can't get into that, I can't give that away. But, that's why we specify that you can't run digital out, you have to run analogue out because if you run digital out it becomes nasty. It won't destroy anything, but the whole thing just collapses. We aren't using tricks or anything. We've developed an entirely new technology based on new scientific discoveries that I've made. It's taken me the last three years to perfect the real-time process - now we can do anything real-time. As a matter of fact, the Sting tracks only took ten minutes. We can now do anything with the process. We can make different mixes of things. For example, the SA3 stuff that you've been listening to, that is what was coming out of their monitors when they were listening to it. The SA3ri, well, that's what the stuff actually sounded like when they were recording it. For example, the piano notes at the beginning of Aja…

AC : The piano on the original sounds like a toy.

RB: The original is very muted and monotone. You know that each piano note has two strings that are struck. On the SA3, you start to hear that, but the ri totally restores that. Not to mention, the cymbals in Deacon Blues, where you just have to fall off your chair over it!

AC : The documentation you sent along doesn't make clear the distinction between the SA3 and the SA3ri.

RB: Everything is based on the SA3 technology. The ri is additional hardware and software that I've come up with over the past three months, and what it does is it is a reintegrator. It figures out what the instrument actually sounded like. I'll explain it to you this way. If you create a sound, the tones of that sound oscillate at a given amplitude and diminishment on a very predictable level mathematically - every oscillation has to do that. Recording technology, mixed with everything else, screws around with that and screws it up. I've developed a circuit and software that corrects for that.

AC : So you can more closely re-derive the original.

RB: Well, not the original master, the process goes beyond the original master tape. That Frank Sinatra you're listening to, that's much better than the master tape. I can tell you that right now. That's from October 1962! You can take an HDCD recording that you have right now, and ours will blow it away. That recording I have of Frank Sinatra will blow away a 24-bit sound. When those horns come in, they take your head off! It's like you're there. We didn't correct that, that's just the way it was. We could have corrected that, and made it "nice", but that's not what we're after. We're after transparency, and when you were listening to that stuff, your whole system just disappeared didn't it?

AC : A heck of a lot more than it did with the original, that's for sure!

RB: Yeah. And that's what we're after. The thing is, you didn't need to buy any equipment.

AC : That's the one major advantage you guys have over something like HDCD. Consumers can fully realize the benefits without the need for special decoding hardware.

RB: And we can run it on anything, old or new. And that's a big advantage.

AC : And I assume that that's a big part of your marketing strategy to the record companies.

RB: Yes, yes. That's the idea. Now, right now, we're the only ones who have this technology, but eventually we will be licensing it. We'll be doing this in about 2-3 years.

AC : John Tabacco mentioned that Sony and a couple of other companies are interested in it.

RB: They are interested in it.

AC : Have any of the audiophile reissue companies, like Classic Records or Mobile Fidelity, any of those guys, expressed interested in it?

RB: Yes, they have. We're currently talking to Mobile Fidelity.

AC : Do you think that with things like HDCD becoming more firmly entrenched that there's room for yet another new process?

RB: You see everybody is really going down the wrong path. They don't realize that the essence of the problem is sampling itself. What they're doing is taking a strobe and flashing it at a dancer and they're getting movement of that dancer. Then they're interpolating the movement between those frames, and that's why you lose all the nuance. That's why nothing that's digital that you've heard, and even our stuff has limitations, I'm not going to tell you that it doesn't, it does. But I think our stuff is more analogue than anything else you've heard in a digital format.

AC : I tend to agree.

RB: Because we write the waves and they're properly written. We have a proprietary way of creating our SA3 files with more data than anybody else, and that's why the frequency response is through the roof. That's why when you go back to the originals, it sounds stupid. And that's not even our best stuff, but we figured that it was probably the best stuff to send out in a sampler.

AC : And it's probably music that people are familiar with.

RB: That's right, it's probably stuff that people own.

AC : How much better do you think your process can get?

RB: How much room for improvement? I think we're at the line right now as far as dynamics and frequency response. As far as 16-bits, we can't do anything more in that respect. But the thing we can do more with is the way we're handling the EQs, the way we're dealing with certain transient waves etc. There's still room for improvement.

AC : How do you think your stuff compares to analogue?

RB: It's not better than analogue, and by analogue, I mean, the best I've heard which is an Ampex 1" half-track running at 30ips, first pass. Nothing has compared to that, but, you lose 30% of the signal after the first pass, and we're better than that. We're not as good as the first pass, but we're better than the master tapes because of the recording technology and the way they did things. If I was doing the mix now, and I was using the kind of equipment that I would want to use, then we would be better than the analogue tape because it's not realistic to say you can only play it once.

The other thing I wanted to cover with you, and that I don't think anybody really realizes, is that S/PDIF, AES/EBU and all these other protocols used in the current mastering houses, there are articles [about these] that are being written that are incorrect. First of all, all of these people in the industry have no idea what digital is. What I mean by that is, you can download a file from NASA's web site five times and it will always come out the same way. That is a true digital communication. Even though you're going through phone lines, and you're going through routers, there is an error protocol built into that, and that is the type of thing that we're doing here, which means that whether it's a digital DAT tape that we're mastering or we have an analogue tape that we're putting into our system, once we put it into our system, it never leaves the computer domain. We do not lose one bit, not one, in anything that we do. That's where I come from. I come from the world in which I will not tolerate one bit of difference. Everything has to be exactly the same, and that's why we store everything on digital tape, true digital tape using error protocols, and CD ROMs.

AC: And that's not the case with the rest of the industry?

RB: Oh no, no, they master every onto 1630 tape. Now 1630 tape is basically ¾" video tape. Video tape is worthless - tape is worthless as far as I'm concerned, because if you're not playing it on exactly the same machine, at exactly the same time, then you have interpolation. There's no error protocol on that PCM, none. That means when it hits a problem, it interpolates. So people love the sound of it but that's not what they got. So you can tell me that you're a purist, but that's not what it sounds like, that's interpolation. That's the way the industry is running right now. They're in a limited loss mentality.

AC : Do you think that the average music consumers will a) care and b) appreciate the difference? Most of them are very content with 16-bit CDs.

RB: Well, I don't think they care about 24-bit, I don't think they care about DVD, but I think they'll care if they can go out and buy something that sounds a lot better. I don't think that most of them can hear, say, the difference between a Mobile Fidelity disc and the original. Everybody we've played our discs for, on anything, on a boom-box or whatever, the first thing they've said is "those are really great speakers." One of our guys said we should put up a sign at demos saying "No, it's not the speakers!" [ laughs ]. We feel that there's a market for it for new releases, and there's definitely a market for it for the past stuff. I made a disc of the Sinatra stuff for my parents and they went nuts! They love it, they think it brings them back in time. They love the instrumentation, and the energy that came through.

AC : I suppose the hardware manufacturers aren't going to be too thrilled about a process like yours which promises to keep the 16-bit digital format alive and well. They're probably looking for DVD to stimulate sales.

RB: Well, the thing is, the record companies will like it. First of all, Disney is sitting on the fence with DVD. Second of all, the record industry put off, for another three months, deciding on which format they're going to use for DVD. And DVD has been broken up into six different regions which means you have to make six different masters of a disc, and even if you didn't have to do that, anything that's done on a DVD has to be done using special compression which takes almost 30 minutes of computer time for every minute of audio.

AC : I was hoping that the regional break-up would only apply to DVD video.

RB: No, it's for everything. It's a very expensive process. I've seen the DVD picture and sound and I don't think it's any better than a laser disc. Sure the line resolution is better but I've seen the same with a line doubler. I saw pixelation in the colour and you don't get that with laser disc. Now, for the computer industry, I think it will really take off. You're talking about 4.5GB of storage, and we're definitely due for that, there's no question about it. As far as the audio end of it, I'm not impressed. I'm not impressed with 24-bit/96kHz, I'm not impressed with Dolby digital, I'm not impressed with AC-3.

AC : This has been very insightful. Thank you very much for talking to me.

RB: Anytime.

3 Although the demo disc contained tracks remastered using the SA3 process, this discussion will focus exclusively on SA3ri since that is likely the only version of the process which will see the commercial light of day.