Here is Anatoly Markovich Liknitsky article where he compares idler and belt drive.
Sorry for a bad translation from Russian language.
"Why is an idler drive better than a belt drive? If you focus only on reducing rumble, the Belt really has an undeniable advantage. After all, he has more flexibility than a movie. True, this is true only as long as the passage of vibrations along the second path is not taken into account. If we compare these drives with a predisposition to detonation, then all the advantages are on the side of the idler drive. It remains only to answer the question, what bothers us more: rumble or detonation? I personally am of the opinion that rumble is an unfortunate hindrance to the perception of music, while detonation of sound, especially low-frequency (with modulation frequencies below 10 Hz), even inaudible, destroys the integrity of music to the ground. It is because of this integrity that an idler drive should be preferred. I’ll try to explain why the idler drive has advantages in terms of detonation. In a belt drive, the moment of inertia of the disk and the moment of flexibility of the belt (rotational flexibility) form a low-pass filter [6] of the second order, which frees the rotation of this disk from irregularities. The source of irregularities can be a drive motor, as well as mechanical transmission elements of this rotation (idlers, belts, gears, etc.). It would seem that cleaning rotation from irregularities is very useful if you do not take into account that due to the absence of losses in the belt, a pronounced resonance is formed with a Q factor of 20-30 at the cut off frequency of this filter. This resonance, as it turned out, does not weaken, but rather enhances the irregularity of rotation. Due to insignificant mechanical disturbances in the drive, caused, for example, by slightly uneven friction in the axis of the rotary disk or by slightly varying thickness of the belt, a rotational “swing” of the disk occurs at the frequency of this resonance. We call this phenomenon, similar to the rotational oscillations of a pendulum in a mechanical watch, rotational resonance. Swinging of the disk in a belt drive is usually observed at frequencies of the order of tenths of a hertz and therefore causes low-frequency destructive music detonation of sound. For similar reasons, the actual, that is, not weighted, low-frequency detonation of sound in the B1-01 Electronics player reaches 0.5%. It can only be reduced by damping rotational resonance. However, in a belt drive, this damping is practically not feasible. Nobody has yet succeeded in making a flexible belt with the necessary internal attenuation, and adding viscous mechanical resistance to the axial bearing of the rotary disk, although partially correcting the situation, will increase the load on the drive motor to unacceptable limits. Trying to solve this problem, some companies tried to apply a belt, inflexible in the longitudinal direction, for example, waxed cotton thread, and immediately faced a new problem: how to hold such a belt in tension? After all, without tension, he will not be able to rotate the rotary disk and in the end it will simply fall down. The way out of this predicament is to pull the thread with a passive idler held by a spring or elastically suspended by a drive motor. By the way, in a similar way, that is, with the help of a thread stretched by a spring, the unit of variable capacitors was rotated in the radios of the 30s. But what have we come to? A thread drawn by a spring is a thread with flexibility introduced into it, and without mechanical losses! So, it turns out that we are back to where we started. The “inflexible belt” turned out to be a beautiful myth, which allowed for one audio season to hide the problem of the belt drive under the carpet. There is also rotational resonance in the idler drive, however, it is well damped and therefore does not enhance the detonation of sound. Good, that is, critical resonance damping is achieved in this drive in a natural way, due to the successful combination of flexibility and mechanical resistance of the rubber ring nozzle on the idler. No wonder the drive of this type was originally called friction."
Sorry for a bad translation from Russian language.
"Why is an idler drive better than a belt drive? If you focus only on reducing rumble, the Belt really has an undeniable advantage. After all, he has more flexibility than a movie. True, this is true only as long as the passage of vibrations along the second path is not taken into account. If we compare these drives with a predisposition to detonation, then all the advantages are on the side of the idler drive. It remains only to answer the question, what bothers us more: rumble or detonation? I personally am of the opinion that rumble is an unfortunate hindrance to the perception of music, while detonation of sound, especially low-frequency (with modulation frequencies below 10 Hz), even inaudible, destroys the integrity of music to the ground. It is because of this integrity that an idler drive should be preferred. I’ll try to explain why the idler drive has advantages in terms of detonation. In a belt drive, the moment of inertia of the disk and the moment of flexibility of the belt (rotational flexibility) form a low-pass filter [6] of the second order, which frees the rotation of this disk from irregularities. The source of irregularities can be a drive motor, as well as mechanical transmission elements of this rotation (idlers, belts, gears, etc.). It would seem that cleaning rotation from irregularities is very useful if you do not take into account that due to the absence of losses in the belt, a pronounced resonance is formed with a Q factor of 20-30 at the cut off frequency of this filter. This resonance, as it turned out, does not weaken, but rather enhances the irregularity of rotation. Due to insignificant mechanical disturbances in the drive, caused, for example, by slightly uneven friction in the axis of the rotary disk or by slightly varying thickness of the belt, a rotational “swing” of the disk occurs at the frequency of this resonance. We call this phenomenon, similar to the rotational oscillations of a pendulum in a mechanical watch, rotational resonance. Swinging of the disk in a belt drive is usually observed at frequencies of the order of tenths of a hertz and therefore causes low-frequency destructive music detonation of sound. For similar reasons, the actual, that is, not weighted, low-frequency detonation of sound in the B1-01 Electronics player reaches 0.5%. It can only be reduced by damping rotational resonance. However, in a belt drive, this damping is practically not feasible. Nobody has yet succeeded in making a flexible belt with the necessary internal attenuation, and adding viscous mechanical resistance to the axial bearing of the rotary disk, although partially correcting the situation, will increase the load on the drive motor to unacceptable limits. Trying to solve this problem, some companies tried to apply a belt, inflexible in the longitudinal direction, for example, waxed cotton thread, and immediately faced a new problem: how to hold such a belt in tension? After all, without tension, he will not be able to rotate the rotary disk and in the end it will simply fall down. The way out of this predicament is to pull the thread with a passive idler held by a spring or elastically suspended by a drive motor. By the way, in a similar way, that is, with the help of a thread stretched by a spring, the unit of variable capacitors was rotated in the radios of the 30s. But what have we come to? A thread drawn by a spring is a thread with flexibility introduced into it, and without mechanical losses! So, it turns out that we are back to where we started. The “inflexible belt” turned out to be a beautiful myth, which allowed for one audio season to hide the problem of the belt drive under the carpet. There is also rotational resonance in the idler drive, however, it is well damped and therefore does not enhance the detonation of sound. Good, that is, critical resonance damping is achieved in this drive in a natural way, due to the successful combination of flexibility and mechanical resistance of the rubber ring nozzle on the idler. No wonder the drive of this type was originally called friction."