Once upon a time, analog ruled. There’s a lot of nostalgia for it now, but maybe, just maybe, some of that is caused by perhaps, just perhaps, the music being better then, eh? OK, OK, I’ve got that off my chest – old timers are always nostalgic for the old days, but how was it really? Why did it seem to work, and when didn’t it? We’ll work our way through from pickup to recording, and see what some of the problems were for the engineers. This is a tutorial for young people who never faced – and probably never will face – these problems; and it’s a reminder for those a little further along in their careers of just what we put up with, say, 20 years ago.
First was getting the gallon of dynamic range output from microphones into the quart jar of microphone preamps, and then into the pint jar of the medium. Sure we had pads and gain controls, but these can limit dynamic range since we turn the microphone level down, but don’t change the noise. It is still a reasonably rare microphone preamp that can handle the full dynamic range of microphones on the market. Duh. It would be done routinely if it were easy, but it’s not, and it isn’t. We’ve covered this one before, though.
Second was getting the wide dynamic range of the output of the preamp through the rest of the console. The usual culprit was equalization. In order to provide, say, 12 dB of boost EQ without clipping the maximum signal from the microphone preamp output, either the equalizer has to have 12 dB more headroom than the preamp, or the signal has to be reduced 12 dB. Since the first isn’t usually practical, it has to be the second: the net result is putting in EQ usually raised the noise floor 12 dB. Ugh. Then we send the signal out an insert point for external processing and merrily cascade a number of devices in series: compressors, limiters, custom equalizers, etc. There’s no problem with this because they’re all nominally unity gain, or can be set that way, eh? Wrong. Your headroom is now controlled by the item in the chain with the least headroom, and the noise is controlled by the item in the chain with the worst noise! Ugh. Years ago a rack of gear was delivered to me by a sound contractor for one of the best-known venues around, and I went through it optimizing dynamic range. It took 6 dB pads, 14 dB gain stages, etc. sprinkled around to get the most dynamic range through the system, all worked out on the bench with test gear. But what studio ever does that? (I can hear the screams now: “WHAT, you put MORE things in the chain – that can’t be as good.” Well it was needed to get this particular chain of equipment to have inaudible noise and no problem with headroom, and this chain had caused trouble on the high notes before, so let’s leave it at that.)
Now we sum signals together. We’re told by the manufacturer that this is done at nearly the theoretical noise floor. So what do I do? I measure the noise, of course, do the calculation, and find out the console is 12 dB noisier than theoretical! Take it apart, re-work it for weeks, mostly in grounding, and get the noise floor down to within 2 dB of the theoretical. Then Studer and others come along and build differential summing busses and get the noise floor down another 3 dB, and, by lowering impedances, lower the “theoretical” noise floor, too… So much for theory, huh?
Now we put the signal on tape or mag film. Let’s start with the influences on frequency response. At DC, there’s no output, of course, in other words when the tape is stopped, there’s no voltage coming out1, but getting from no output up to full output level things doesn’t just come up to flat at 1 Hz, but instead the playback head causes a low-frequency cutoff. It is partially caused by the full face of the head, and not just the gap, capturing the long wavelength flux. That’s why bigger heads are better at low frequencies, and the biggest heads, found on some of the oldest machines, were the best. It’s also what limits quality at high tape speeds like 30 ips, or the 22.5 ips of 70mm film: the low-frequency cutoff could be as high as 50 Hz or more, and lose an octave of bass.
Next we get into a frequency region where the response undulates up and down about the desired flat response. This region is also caused by the interaction between pickup by the exposed laminations on the head face and at the gap, the wrap angle, and how the shields between channels conduct flux around. Ugh. When I started at Lucasfilm and looked at my first mag recorder, its response was +4, -7 dB in the audio band. My experience at that time had been in cassettes, and that made the first steps easy, since it was so much harder to get things flat on a slow-speed cassette machine than on a high-speed dubber. The +4 dB was a “head bump” caused by using the wrong head shape. After a few years of work, three generations of heads, two of mag film, two of noise reduction systems, and changes to the electronics, the response could be ±1/2 dB from 100 Hz to 20 kHz and ±1 dB from 20 Hz to 20 kHz. But it was hard, and it took vigilance.
By the way, the undulations move around with head wear, just to keep things interesting and make electronic solutions difficult. They are also potentially non-minimum phase, which, to electronics engineers, mean they are not simple to equalize.
In the midrange, the biggest perturbation was to reference levels. Music and film treated this completely differently from one another. As stock improved, music chased the improvement upwards, increasing the reference level for 0 VU from 185 nW/m, old Ampex reference standard, through 250, 355, to even 500 nW/m. That’s because the stock really did improve that much over time, but music studios generally took all the advantage as signal-to-noise improvement by increasing level and keeping distortion constant. Film engineers, more concerned with day-to-day interchangeability in libraries and so forth, kept 185 nW/m alive forever, but allowed for higher peak level as time went by, kind of doing the same thing of chasing level upwards as music, but also leaving the noise floor alone, primarily because virtually all film applications used Dolby A or dBx companding at first, and later virtually all switched over to Dolby SR, thus keeping noise under control.
The other midrange level problem was due to head height. The heads have to be mechanically aligned with the tracks for “height,” and this wasn’t always easy on film transports, where the heads were floating up in the air on a bunch of screws. Multitrack tape recorders had the easy life for this problem, but not so some others.
At high frequencies, lots of things come into play. First is spacing loss. Anything on a head that gets between the play head gap and the tape dramatically muffs the highs, and here’s why. The equation for spacing loss is: Loss = 55 d/l dB, where d is the distance caused by the obstruction, and l is the wavelength being reproduced. So let’s take a 1/1000 of an inch obstruction. And let’s say we’re on mag film operating at 18 ips. The wavelength of the recording is: Wavelength = Tape speed/frequency.
At 18 kHz and for a tape speed of 18 ips, we get a recorded wavelength of 1/1000 of an inch. Dividing 1/1000 of an inch obstruction by 1/1000 of an inch wavelength, and multiplying by 55 gives the answer: 55 dB loss! That’s why cleanliness is next to godliness in tape recorders, and the first thing you do is clean those heads (comes right after demagnetizing them). At slower tape speeds, of course, the above equations predict things are much worse, and they are.
Next is head alignment. Spacing loss happens if the head face isn’t perfectly in the same plane as the tape or film. This is usually called zenith. Heads also might be rotated so that the gap isn’t putting its best foot forward, that is, if the tape or film contact footprint isn’t centered on the gap. Then there’s azimuth. That’s the perpendicularity of the head gap or recording to the edge of the tape or film. Think of it this way. If the head is off azimuth, then the signal off the tape is being smeared out in time: this is equivalent to a high-frequency loss. Gap scatter occurs in multitrack formats if the gaps are not perfectly in a line – and all at the same angle.
I remember visiting a famous tape machine manufacturer who didn’t show me the secret facility where they made their heads, but they told me rejects were far higher on Mondays and Fridays than in the middle of the week. What that means is that those pesky 24-track heads were really hard to make, and azimuth is one major problem.
Interestingly, azimuth became an even greater problem for LtRt-encoded masters like Dolby Stereo. The matrix was more sensitive to high-frequency errors than even listening to stereo, apparently because the electrical sums and differences of the matrix were even more sensitive than listeners. This caused us to mill the two inertial rollers in the film path to the width of the film so that the azimuth stability was improved, and, audibly, then fewer “s’s” wound up in the surrounds!
Heads exhibit high-frequency losses other than azimuth. That’s because when one wavelength of a sine wave just fits into the gap measured along the length in the direction of tape travel, there is no output: the positive and negative parts of the sine wave add together, and thus cancel. This frequency is about 44 kHz for one of the mag film heads operating at 18 ips that I measured. This is a loss that has the form called sin x/x. This means it rolls off before reaching a null, bounces back up, goes down to a second null, etc., the response bouncing away to infinity. We only care about what happens below the first null, and it is the playback gap that causes this loss: a bigger gap captures more signal and higher level, while a narrower one has better hf response. The response rolls off in the audio band because of the sin x/x null at higher frequencies.
There’s a trick, though. The initial rolloff of a sin x/x loss can be compensated by a single RLC resonant circuit out to the point where the response would otherwise be down 3-4 dB by resonating at the right frequency with the right Q. And where is such an RLC hanging around? Why it’s the head inductance, cable capacitance, and amplifier input impedance, and those are all there already. Remember that first bad machine (+4, -7 dB)? As I moved from channel 1 to channel 6, the hf response got worse and worse. The culprit: the head cables were longer and longer, thus resonating lower and lower, and the whole system wasn’t tuned to compensate for the head loss. By substituting low capacitance cable and fixing the amplifier input impedance interaction, and “tuning” the load for the head sin x/x loss, the high-frequency response was made much flatter. Those of you with Revox machines may wonder why they switch a load resistor across the head to terminate the head differently when they change tape speeds? That’s why. There’s a zillion other things to talk about: headroom and its change with frequency, noise and its reduction, wow and flutter, scrape flutter and modulation noise, transport mechanical azimuth stability, etc. Audibly speaking, however, in multi-generational film dubbing, the next problem in importance was the accumulation of distortion over the generations. While this wasn’t audible, not much, at least, in any one generation, there was audible distortion on an A/B of 70mm prints of movies circa 1983 with the earlier generations, and it wasn’t the fault of the 70mm generation, but the accumulation. By 1984, the problem, at least audibly for the material we used, was solved. The difference was changing mag stocks from one that had been around for years to one that had the latest gee-whiz oxide. You wouldn’t hear the difference in one generation, but one of the toughest problems for audibility of IM distortion is chorus with strong bass, and the heart-wrenching scene from Indy II provided both in spades, and stayed audibly undistorted, despite the prints being around 5th generation. While there’s probably a lot more to say about the old analog days, you’ll perhaps feel partially relieved that we’ve moved on to a digital world. “Or will you?” said in the voice of Bela Lugosi.
Next time: Tom Holman’s Digital Chamber of Horrors.