Stalking the Magnesaurus^TM
by Mike Joseph
Recording Engineer-Producer, Feb., 1991.
Copyright © REP, 1991. Lightly edited by JB, 12/99.

[Specifications of
Magnesaurus^TM 351.5 All-Tube Analog^TM Master Recorder]

 
Less than common in our industry is the individual who combines...musical ability, proven technical prowess and...scientific achievement. James Boyk - owner of and producer at Performance Recordings in Los Angeles; concert pianist; and director of the music lab, pianist-in-residence and lecturer in music in the department of electrical engineering at Caltech, in Pasadena - is such a person.
      For seven years, Boyk has been developing one specific analog tape deck into an ultimate analog performer. The platform he chose, in one sense, is surprising. In another sense, it's quite interesting. The details follow:

REP:  What led you to the Magnesaurus project?

JB:  As a concert pianist, my interest has always been to record the piano sound as legitimately as I could. Along the way I've learned something about recording and something about making vinyl records, an art which is now completely obsolete. I have been privileged to co-engineer recordings for Sheffield Lab, including ones of the Kodo Drummers and the Los Angeles Philharmonic. I also co-engineered and co-produced a recording by my friend Lincoln Mayorga; solo classical piano works. I've had quite a bit of experience in the purist sort of stuff, none at all in multi-mic studio work. My goal has always been to record live material better and better.

REP:  Tell me a little bit about the tape deck itself. It's an early Ampex?

JB:  Originally a 15 ips Ampex 351. Our name for it is the Magnesaurus 351.5, the ".5" referring to 1/2-inch tape. Ampex called it a portable machine. I think the transport weighs about 60 pounds and the electronics 30 or more. The transport's in one box and two signal chassis, each with one big meter, in another. All tube; in fact, all miniature triodes.
      The series began around '57 or '58. The date codes on some of the parts in mine are actually as late as '62, which I was pleased about, because it meant that I had very late electronics. The basic deck is a very good machine. In my opinion, it's one of the best off-the-shelf recorders that's ever been made.

REP:  How did it perform in stock condition?

JB:  The transport tended to have wow rather than flutter, unlike the modern servo decks. The electronics all had some 120Hz ripple. But basically a good machine. In fact, three of our records, our catalog numbers pr3, pr4 and pr5, were made with a stock 351. It was one that had been tweaked and maintained immaculately by Michael Fraser, who was my recording engineer before he retired from the business.

REP:  What led you to owning and improving your own 351?

JB:  Back then I had already started thinking, "How could we do this substantially better?" When Fraser retired and sold the machine, I was galvanized. I was in a position where I had to have something, and I decided that it had to be an all-tube machine. I'm simply not happy with any solid-state recorder that I've ever heard personally - none of them touch what in my opinion can be done with tubes. I find a lot of agreement about this among the professionals I talk to, professionals in recording and also professionals in equipment design.

REP:  What do you think was the magic in the older tube 351 series Ampex machines?

JB:  The more I look into the 351, the more I realize that it was a brilliantly engineered machine. Not that it was perfect, because it wasn't, but that it was the balance among the various aspects of its performance that was quite remarkable. It's been very hard to make it better without an enormous amount of work, which we have put in.

REP:  Detail, if you would, what have you done to your machine to improve the performance?

JB:  The head nest was replaced by John French of JRF Magnetic Sciences in New Jersey, with a custom-modified Sony MCI headnest, simply because it gives you control over all the axes of the head, which the Ampex nests do not. John made a little hand-operated gate for the playback head and got custom record and play heads designed and made by Greg Orton, senior staff engineer of Ampex. The heads were custom-made because you can't go out today and buy 1/2-inch, 2-track heads with the appropriate inductances for tube electronics. They're not ferrite; they're solid metal heads, and have nothing weird about them. They work wonderfully. We did find an appropriate off-the-shelf erase head.
      Caring for the transport has been done by Shelley Herman of Acoustronics Sound. Herman is a well-known audio guy in the L.A. area. He used to be at Coast Recording and is very knowledgeable. The motors, bearing rebuilding and the true-ings were done by Jeff Gilman of MDI/Precision Motor Works in Hudson, MA.

REP:  Did you stick with the original Bodine motors?

JB:  The original capstan motor was a Bodine, which we replaced, along with all of the motor-run capacitors and so on. All three motors have been rebuilt to within an inch of their lives. I gave MDI carte blanche. I told them, "When the machines leave your shop, I don't want you to be saying to yourself, "Gee, if I had $50 more to work on it, or five more hours, I could make them better:' I want you to say, "I don't know how to make them better:" So, they're all quite marvelous.
      The capstan shaft now has the ceramic face on it. Every moving part has been analyzed with fast Fourier transform analysis (FFT) to find out whether there were any rotationally caused sidebands. The flutter idler has been trued. The whole assembly has been dynamically balanced.

REP:  Did you custom-fabricate any components?

JB:  We now have a custom flywheel made of a special material that won't resonate, so we get rid of the 480Hz ring in the standard flywheel. We were damping the original with rubber bands, which, by the way, are very effective! Just put a rubber band around the perimeter, and thread another one around itself in the four round holes that are in the middle of the flywheel. You can reduce the ringing 10dB with just two rubber bands. It's quite amazing, the performance that the transport design was capable of. I'm not saying that it ever even reached that performance level in those days, because FFTs weren't thick on the ground.

REP:  Did you learn anything about tape path design, or mechanical performance, in the rebuilding process?

JB:  We found out a lot about such things as pinch rollers. For example, the capstan motor mounts to the top plate with four screws, and the common way to remove one of the screws is to lever the pinch roller out of the way. Well, the moment that you lever it out of the way with your screwdriver, you've just ruined it. You have to then replace the whole pinch roller assembly, not just the tire, once you've bent the little vertical shaft it's on. If you don't, you've just increased your wow and flutter tremendously.

REP:  Something tells us you learned that the hard way.

JB:  Exactly! But we also did comparisons. We tried a stock Ampex pinch roller and a brand new Electrosound. We found out that at least with the two we tried, the Electrosound was not only less expensive, it was better. It actually gave us smaller and fewer sidebands. I'll say, by the way, for anybody who's doing such analysis, that often you'll see a wow and flutter analysis published on a tape deck. They'll show you a 3k tone and its sidebands. The resolution they use commonly is 50Hz per division.
      It turns out that's totally useless and completely meaningless. It will not show you what's going on. You really want to use 10Hz per division. One Ampex ATR that I looked at had a bad capstan sideband that was down 13dB relative to the 3k tone. That definitely affected the stability of the signal. At 50Hz resolution, you simply didn't see it. It looked perfect. So we used 10Hz resolution, and there it was.

REP:  Going through the machine, you obviously used a lot of FFT analysis and transducer-generated measurements. How much of your discoveries were just good basic sleuthing, figuring out the relationships between the rotational speeds of the capstan and flywheel vs. the flutter scrape filter or reel motor?

JB:  A lot! You ask yourself: What is it that goes around at that rate? "Well gee, that sounds like it's about 5 per second:' Where the FFT helps is that you can say it's 4.8Hz, so you can look for something that goes around 4.8 times a second. Measure diameters. You know, there are a lot of things similar in size, for instance, the capstan and scrape flutter rollers, or the pinch roller on one side and the flutter idler on the supply side.
      You ask yourself, What is it that has a perimeter, a circumference of x inches, therefore a diameter of x divided by pi? Well, it's this roller; it's not the capstan because that's only 1/4-inch, and so on. So FFT is enormously helpful. I don't mean a little bit helpful; it really makes a qualitative difference in your ability to track down these things.
      The net performance result for us was DIN weighted wow and flutter better than 0.03%, and, unweighted, better than 0.06%. That first spec is as good a number as any manufacturer has ever claimed for a commercial tape deck, even a modern servo type. I'm not aware of any manufacturer's spec better than that. It sounds solid as a rock, even on piano.

REP:  And the electronics?

JB:  It's a line level 15ips machine. The electronics originally had three stages to the record head, four including the mic pre, which we eliminated. We now have two stages from Line In to the record head. No transformers at all; we just come into the unbalanced input. The electronics were the purview of Steve Haselton, the chief engineer of Sheffield Labs. He has gone through them extensively, simplifying the circuitry in a number of ways, such as eliminating that one input stage in the record circuit.
      Let me give you an example of the remarkable design of the original, which we only discovered in the painful way. First, we eliminated all of the record amplifiers except the actual single triode, which drives the record head. We thought we'd hit the single amp record driver flat, no EQ, very hot from outside, and run it in some approximation of constant current. We found out that it's not capable of driving the head flat.
      It turns out that the EQ pre-driver is doing two things: It provides EQ, the NAB or CCIR curve, and it's also providing EQ for the non-flatness of the record amp. We are now going right into the front of the final buffer before the EQ stage of the record amp. We have changed the EQ stage somewhat. But as it stands, once you get rid of the ground loops and hums and so on, it's quite a remarkable recorder. People can judge that themselves just by hearing the record. It's Performance Recordings pr7lp or pr7cd, released in October last year.

REP:  Were there any other modifications?

JB:  We have an outboard power supply made by Richard Maurer of Musical Fidelity in Culver City, CA, that provides much better power. It has on-board regulation for the filament supply, but not for the B+. I don't like regulated B+; I don't think it's done right in 99 cases out of 100. I think it really degrades the sound.

REP:  When you put this all together, what do you end up with?

JB:  Well, we took our reel of tape made on the Magnesaurus over to Doug Sax at the Mastering Lab, and we played it back and measured things. Their standard mastering playback machine, the one with the preview heads, has its own custom tube electronics, which Haselton also designed. We measured pitch stability from one end of a 10 1/2-inch reel to the other. It was 0.15%. Damn good. Very impressive, considering the changing tension. That's two and a half cents, two and a half hundredths of a semi-tone.
      People who know about such things tell me that this stability is good enough that you can edit from one end to another; but I never do that, because no matter who I'm recording, I'm always recording in concert, straight through. I only edit from part of one reel to a similar part of another reel. We can make that spec even better if we continue playing with the tension adjustments.

REP:  How does the noise measure out?

JB:  Noise measurements are enough to drive anybody nuts. I can't come up with anything meaningful. What I want to do is the CCIR/ARM measurement that Dolby has promulgated, but I don't have the proper filter for that. My guess is that it's 67dB below 250n/W.  (Complete specifications.)  I'll tell you what I've learned, and I've looked into each of these questions elaborately: The most meaningful thing you can say about measuring noise is that with blank tape on the machine and the tape stopped, you read the noise off the playback electronics. Then you run the blank, unmodulated tape. Your tape-stop noise, playback electronics only, should be better than 10dB below the silent tape run noise at every point in the spectrum, looking at it with an FFT analyzer.
      That's the goal. Then you turn on Record with the level pot down all the way. Now you have bias noise on there, and the record electronics, of course, and ideally it should not go up more than 4dB or 5dB above the blank tape playback noise.

REP:  And that's mostly bias and tape modulation?

JB:  Yes. Now that's the ideal. The theoretical is that it goes up 3dB. If you get 4dB or 5dB, you're doing great. But let's talk about dynamic range for a moment. The stock, factory 351 gives you a signal-to-noise of 60dB, very roughly. When you go to 1/2-inch tape, you gain 5dB. You ought to gain only 3dB because you've doubled the tape width. But the relevant thing is not the width of the tape; it's the width of the track. The 1/2-inch, 2-track tape uses the tape very effectively; 1/4inch, 2-track does not. When you compare track width instead of tape width, you see that 5 dB is what you should expect.

REP:  What is the dynamic range of the machine?

JB:  Using a peak meter at the Mastering Lab, we actually measured transients off tape, which are 14.6dB above 250n/W, clean. Absolutely not getting into the tape. Nobody would listen to it and say it was overmodulated, compressed. 14.6 above 250, which means it's 16.6 above 200, or 17.2 above 185n/W. That's almost unbelievable, and frankly a level that I didn't think tape could take. It's hot! With piano transients! I think part of that possibility, that advantage, is tube electronics, much more headroom and much more transient capabilities. It sounds marvelous.

REP:  And the frequency response?

JB:  We did a test tone run before this latest piano recording session, measurements of extreme breadth from 12.5Hz to 40kHz, to learn something about the machine. We were looking for the shape of the response curve. We recorded lots of tones, lots of points, measured with our Fluke meter, being sure that we were putting them onto tape extremely flat. Recorded on our machine at 0 level, not -10, at 250n/W, and played back and measured off the playback head of the tube MCI at Mastering Lab, which is realistic because that's where we play back for mastering the disc.
      The -6dB points were 13Hz and 30kHz. At 15ips! The -3dB points, still at 250 nW/m, were 18Hz and 22kHz, and ±1 dB was from 20Hz to about 17kHz. ±0.5 dB was from 22Hz to just under 15kHz at 250. And the bottom end! I've never heard a bottom end like this. That's the reason we went 15, not 30. No headbump. We don't see one. It was ±0.1 dB at 25, at 50, at 100. Doug Sax gets credit for that, as headbumps are a function of the playback head.

REP:  Tell us about the perceived sound quality of the Magnesaurus.

JB:  Speaking as a pianist now, I'm hearing attacks like I've never heard before off tape, and certainly never off digital tape, which completely screws up piano attack.  [Dec. '98: For my views of the current digital audio situation, see my review of a recent digital recorder with usable sound quality. --JB]  When the hammer hits the strings, it's struck; it shouldn't sound plucked. There are sharp leading edges, with an adamant quality, which is quite marvelous. I'm hearing the tone much better here than I've ever heard before, except off direct-to-disc recordings.
      It has all the virtues of tubes, such as high resolution, where you can really listen in and hear the true piano sound in a way that I have simply never heard on any solid-state recording. I can hear the harmonic development of a note as the note sustains and dies.
      I can hear the change in the note and the overtones off the tape the way I hear them on the live piano. I've never heard that on any digital, including the DAT recording we made simultaneously at the same session (the pr7 recording), splitting the Blumlein ribbon mic pair signal from the custom-built tube mic pre-amps, through a borrowed Sheffield tube stereo splitter, to the Magnesaurus and DAT recorders both.
      On the pr7 CD, we have the analog master tape followed by the digital master tape so people will be able to compare directly. Does the analog have more noise than the digital? Some. But the sound of it is very tube-y. Very smooth. It is very airy, more tube-like than tape hiss-like. It's really quiet, even in direct comparison to the digital.

REP:  Any final comments?

JB:  Only that analog recording hasn't nearly reached its true potential. It really hasn't been touched compared to what it's capable of.