namely the motor pulley and the turntable platter. Since both these constraints are rotatable, the shrinking fourth section of belt tries to rotate them. The shrinking fourth section of belt actually tries to rotate the motor pulley backwards. But since the motor pulley is being actively driven forward by the motor (of hopefully adequate power), this might not be a serious concern during the active pole drive portions of the motor's rotation (but, between poles, while the motor is coasting, note that the valleys of the kick and coast problem are actually made deeper and worse by the belt shrinking back to its normal length after it is stretched).
     On the other hand, the platter is not being actively driven by anything other than the belt, and the platter speed is what matters for the time axis of your music, so what the fourth section of belt does to the platter speed when it shrinks back to normal length is of crucial concern.
     When the fourth section of belt shrinks, it causes the platter to speed up. It's effectively yanking at the rotating platter, saying come on, let's go faster (much as the multipole motor's kicking jerks had yanked at this same fourth section of belt). The platter can't stretch in response to these yanks, but it can rotate faster, so that's what it does. And, if these yanks from belt shrinking are in any way irregular over time, then the platter's speed rotation in response to these giddyap yanks from the shrinking belt will be irregular. But speed irregularity is what we want to avoid, and we came to belt drive turntables with the hope of getting regular and accurate speed. Can we get it from belt drives or not?
     As we discussed above, it is virtually if not actually impossible to get a belt's stretching behavior, in accepting the kick jerk input from a multipole motor, to perfectly offset the kick and coast problem. Now we see that the belt stretching, when it accepts the input from a kick jerk, and acts to slow down the platter, could perhaps be offset against the opposite tendency, namely that the belt acts to speed up the platter when it releases the potential energy of the stretch and shrinks back to normal length. Might these two opposing tendencies perhaps be precisely balanced against one another, to produce exact and constant speed control from a belt drive turntable?
     Indeed, we could envision that the fourth section of belt, having been fed several stretching kick jerks within its length, might even contain within itself the mechanism for potentially averaging out the kick and coast problem, introduced at the motor pulley end, before it even reaches the platter rim at the other end. If the belt material is constructed so that the stretches from each kick are quickly converted from potential to kinetic energy, then the irregularities of stretching and shrinkage, which affect the platter speed, might average themselves out over the length of this fourth section of belt, and thus effectively never reach the platter rim, thus never causing speed irregularities at the platter. Can we count on the fourth section of belt to simply average out, within its length, the kick and coast problem of every multipole motor?
     Nearly every belt drive turntable designer has thought that the answers to the above questions is yes, yes, yes, and yes. But, if you want truly accurate and constant speed, if you want truly clean, distortion free music, the answer is no, no, no, and no.
     Most belt drive turntable designers have blithely assumed that you can slap together any old reasonably compliant (stretchy) belt, any old reasonably heavy platter, any reasonably smooth bearing, any old motor, and presto, you have a belt drive turntable with enough speed accuracy to bring to market. There's other design work to do, such as vibration isolation and packaging aesthetics, but they assume that the speed accuracy and constancy problems have been solved by the simple use of a belt. These designers have been ignorant of the weaknesses, compromises, and limitations of belt drive, and they have also been ignorant of the engineering required to at least get the best speed accuracy out of a belt drive design. There are more belt drive turntable designs than there are engineers who know how to design a belt drive turntable for the best possible speed accuracy. That explains in part why belt drive turntables sound so different from one another, in a diverse range of sonic aspects including pitch (speed being slow or fast), pitch wandering, rhythm and pace, solidity of bass impact, etc.
Turntable designers who blithely assume that the answers to the above questions are all yes are in for a rude shock when their belt drive turntable doesn't work as well as they hoped. They face a number of shortcomings in speed accuracy performance which they can't explain, and perhaps don't even know how to adjust for.
     First, there are issues of absolute speed accuracy. For example, if a designer calculates the theoretically correct ratio for his motor pulley to his turntable platter rim, he'll probably find that his belt drive turntable actually runs at the wrong speed (usually too slow, giving a poor sense of pace and rhythm). One problem is that, when he counts on the belt to average the peaks and valleys of the multipole motor's kick and coast phenomenon, he doesn't have a profile map of the motor's speed and energy output over time (this depends on the actual geometry of the poles in the particular motor he selected), so he doesn't really know what the average input to the belt is. A second, related problem is that there is some energy loss in the belt itself (to heat), as it first stretches in response to the motor's kick jerk input yanks and then converts most (but not all) of this potential energy to kinetic energy as it shrinks in length and in turn yanks on the platter rim. This energy loss will make the turntable speed slower than predicted.
     Designers of belt drive turntables usually respond by machining their pulleys to an ad hoc diameter, rather than a scientifically predictable diameter, in order to get the correct platter speed with a given belt. Such ad hoc designing is already a warning flag, alerting us that the belt drive design approach is fundamentally compromised in terms of providing truly accurate speed.
     Then there's a third, related problem. What if the belt manufacturer slightly changes the exact way he makes the belt, without telling the turntable manufacturer? The speed of the turntable will change, because the amount of this energy loss to heat depends on the precise material composition of the elastic belt, and how its long chain molecules intertwine. A fourth, related problem is that this energy loss, and so the turntable speed, will be affected by the aging of the belt and by changes in the ambient temperature of your room.
     Second, there a re issues of speed irregularity. The belt releases the pent up potential energy of its stretch as output, converting it to the kinetic energy of shrinking back to normal length. But it releases this output in a completely different time frame than it accepts the input from the motor's kick jerks. In other words, the energy output of the belt (which tends to speed up the platter) has no temporal relation to the energy input to the belt (which tends to slow down the platter). If these two effects are temporally unrelated, then they can't truly offset each other in real time. It's as if one boxer with his hands freely swnging rapidly through air were opposed by a second boxer whose hands had to slowly work their way through molasses; the second boxer couldn't keep up with the first in real time, and thus he couldn't truly offset the efforts of the first boxer, even if he were just as strong.
     In this case, the belt shrinkage back to normal length (which tends to speed up the turntable) operates in a much slower time frame than the belt stretching (which tends to slow down the turntable). Why is that? On the input side, the motor kick jerks are usually relatively powerful, and the belt is usually quite compliant (easy to stretch), so the belt responds quickly, stretching quickly in response to m motor kick jerks. It's like the quick boxer. On the other hand, when that fourth section of belt tries to shrink back to its normal length, it has an enormous load on its back that it must tug along, this load being a relatively heavy turntable platter (with a high moment of inertia). This fourth section of belt is trying to do the 100 yard dash in shrinking back to its normal length, but it must tug along a giant river barge, and that slows down its response dramatically. It reacts like the boxer with his hands stuck in molasses.
     The belt's shrinking response is so much slower than the belt's original stretching that there's no hope of the two truly balancing or offsetting each other in real time. Thus, the belt drive turntable designer cannot rely upon any such offsetting to obtain constant speed. The only turntable to even try this design approach was the Weathers turntable of the early 50s, in which Paul deliberately mated a weak motor (thus slowing down the response of the belt stretching) with a very light platter (thus speeding up the response of the offsetting belt shrinking). But it had other problems, including poor immunity to automodulation distortion from loud music passages (see discussion above).
     We saw earlier that it was probably impossible to engineer a true offsetting match at the input to the belt, where the belt accepts the kick jerks from a multipole motor, and we now see that it is also probably also impossible to engineer a true, real time offsetting match at the output from the belt, where the belt releases its stored up potential energy to the platter.
     Grim news indeed, for belt drive designers who hope to achieve regular speed constancy. So what do they do? What can they do?
     They throw up their hands, throw a Hail Mary pass off into space, and pray for the best landing. They fall back upon a saving grace of the gross temporal mismatch we discussed above, between the belt stretching and the belt then shrinking back to normal.
     How does this saving grace work? Recall that the quick response occurred at the input to the belt, as it was being stretched by the kick jerks of the motor poles. And the slow response occurred at the output from the belt to the platter, as the belt tried to shrink back to normal length, tugging the heavy platter along behind it. It so happens that, with the slow response being at the output and the quick response being at the input, the output simply won't see the quick things happening at the input. It's as if you had a power amplifier with a quick input stage and a very slow output stage; the output stage simply wouldn't respond fast enough to let you see, at the output of the amplifier, the quick things happening at the input. In other words, the amplifier as a whole would respond as slowly as its output stage (or indeed as slowly as the slowest, weakest link in its internal chain).
     This means that the belt drive turntable designer can cheat by using this saving grace. He can get away without understanding any of the other intricacies and tradeoffs of turntable design discussed above. A Hail Mary pass can win a game even if a team doesn't understand how to design football plays and is inept at actually executing them; likewise anyone can "design" a belt drive turntable by using this saving grace, even if he knows little about turntable design.
     When this saving grace is relied upon to "design" a belt drive turntable, the kick jerks from the multipole motor are not truly offset in the belt itself, or in the fourth section of belt that we have focussed our recent attention on. These undesirable kick jerks actually reach the platter. But, thanks to the saving grace fact that the belt's output is so slow in tugging at the heavy platter, these quick kick jerks don't affect the platter speed much.
     In the power amplifier analogy, the slow output stage could also be likened to a filter introduced in the amplifier circuit, to filter out most higher frequencies (i.e. most faster changes in the signal). Likewise, the slow response of the belt, trying to shrink back to normal length while being forced to tug at a heavy platter, can be likened to a filter introduced in the turntable drive circuit. This filter would filter out most of the rapid speed changes from the multipole motor, such as the dreaded kick and coast problem from each pole. This filter effectively integrates or averages the rapidly changing speed input from the motor, to produce a more constant average speed output. So this averaging process can indeed be effective in reducing the kick and coast problem coming from a multipole motor.
     But notice that neither the belt nor the motor actually control the accurate speed of the turntable any longer, especially not in real time. Instead, the motor and the belt merely make small repeated contributions, which are deliberately slowed down so as to minimally affect platter speed in real time. The designer settles for an averaging process as a creator of speed constancy, and settles for an average as a measure of speed constancy.
     To make this averaging process effective, the designer needs a powerful averaging tool. The motor and the belt have been discarded as useless in this quest, so that leaves the platter as the only other physical element remaining that could be used as an averaging tool. This explains why turntables platters are made heavy (actually, heavy at their rim, so they have a high moment of inertia). The higher the platter's moment of inertia, the more powerful it is as an averaging tool (and it also has more value in being immune to automodulation distortion from stylus drag during loud music passages, as discussed above).
     To say the same thing another way, the platter as an averaging tool works like a slow, low pass, integrating filter, which lets very low frequency variations (including speed variations) through, but which works to filter out higher frequency, more rapidly changing variations (including speed variations, such as the kick and coast variations from multipole motors). Like every other filter, this platter filter has a certain fixed slope, so it doesn't totally eliminate unwanted higher frequency, more rapidly changing speed variations such as the kick and coast variations. Instead, it merely reduces these unwanted higher frequency speed variations by a certain amount, in proportion to how low in frequency the filter starts operating. The lower in frequency such a filter starts operating, the more effective it can be at reducing the unwanted higher frequency speed variations by an even greater amount. To be maximally effective at reducing the higher frequency unwanted speed variations from the multipole motor's kick and coast problem, the filter should start operating at the lowest possible frequency. This means that the platter's moment of inertia should be as high as possible, consistent with product cost and some other technical factors such as load on the main thrust bearing.
     Indeed, the sum total of design work going into may belt drive turntables, with respect to speed accuracy and constancy, has been simply to combine a belt of some indefinite and unknown compliance with as heavy a platter as the product budget can afford. This is the Hail Mary pass that most designers throw into the air, both praying and knowing that they'll get halfway decent speed performance, thanks to the saving grace of any old heavy platter acting as an averaging filter with any old compliant belt. But is halfway decent good enough for you?
     Furthermore, there are also some problems and penalties with this Hail Mary design approach. Firstly, by making the platter heavier in order to improve its filtering performance, the designer is actually worsening the disparity we discussed above, wherein the energy output from the belt is much slower than the energy input to the belt, so one cannot offset the other in real time. Making the platter heavier gives the belt a heavier load (higher moment of inertia) to tug on when it tries to shrink back to normal length and thereby feed kinetic energy into the platter speed. Thus, making the platter heavier actually works to worsen the belt's ability to correct the platter speed expeditiously (and there a number of reasons why it might need correcting, some of which are discussed below, others of which relate to correcting for stylus drag slowdowns as discussed above). Making the platter heavier further gives up control of the platter speed by the other two elements of the system, the motor and belt.
     Making the platter heavier also further reduces the belt's ability to play any significant role in being an internally self correcting mechanism for speed constancy (especially within that fourth section). There might have been some hope, with a belt carefully engineered and matched to both motor and platter, that the belt stretching from the motor's kick jerks (which tend to reduce platter speed) might have been at least crudely or partially offset by the belt shrinking back to normal length (which tend to increase platter speed), thus providing at least some measure of speed constancy benefit or improvement within the belt itself and by the belt itself. But, with the platter being made much heavier, the heavy platter dominates so there is no matching among the three drive elements, and the temporal disparity between belt output and input is worsened to the point where the belt itself cannot contribute significantly to speed constancy.
     Designers typically throw away all hope of matching the belt to both the motor and the platter (or they don't even know enough to know what such matching might mean). Instead, they fall back upon the simplistic Hail Mary game plan, relying on the saving grace filtering feature of any old compliant belt with a very heavy platter. In so doing, they effectively surrender control of speed constancy to the platter.
     But what's so bad about that? Why not surrender speed constancy to the platter? It turns out that there's a very nasty penalty for doing this. The compliant belt and the heavy platter form a reactive tank (LC) circuit, in which energy is exchanged back and forth between the belt and the platter. Another name for such back and forth exchanges of energy is oscillation. These energy exchanges occur in their own time frame, thereby producing speed oscillations that occur in their own time frame.
     Speed oscillations are of course speed irregularities, just the opposite of the speed constancy we are seeking here. These speed oscillations typically occur at a low frequency, sounding like wow or pitch uncertainty (unlike higher frequency speed irregularities, which produce flutter and the grundgy or fuzzy sound of FM distortion). Making the platter heavier places these oscillations at a lower frequency, perhaps even to the point where a belt drive turntable would not have audible wow but could still make you seasick. Making the platter heavier also can make these speed oscillations larger (worse), since the tank circuit is now even more reactive.
     And furthermore, in a cruel twist of irony, making the platter very heavy, which was the simple saving grace, Hail Mary filter tactic designers used above to reduce the multipole motor's kick and coast problem, now turns around to bite them in the butt. By making the platter very heavy, they have reduced the motor's influence, and have made the platter's own tendencies dominate the situation. Now they suddenly discover that the platter, rather than being the prefect passive partner that tends to run at a constant speed, is instead engaging in active hanky panky of its own, oscillating with the belt in its own dance to its own rhythm. But now it's too late. For they have weakened the motor's influence to the point where it cannot effectively correct the oscillations of the heavy, domineering platter.
     That's why belt drive turntables tend to exhibit their speed irregularity problems as wow rather than flutter. They are still severely compromised in speed constancy, but simply at a lower frequency of variation, of modulation distortion, of sidebands, than rigidly coupled drive turntables. Belts don't

(Continued on page 16)