The Suzhou Bafang hub motors tend to make the most noise under load as in acceleration or hill climbing, generally a curious combination of whine and growl. The lower the speed the higher the growl content and vice versa.
At full speed they are generally at their most quiet, just a sort of gentle whine left.
They vary quite a lot, both by model and individual motor, but I don't find any of them objectionable. They are all much quieter than the old Powabyke motor and even the Heinzmann motor.
The Panasonic crank drive units are quieter, but differ by model. The old unit with steel helical gears made a distinctive low level "whoosh" with each pedal downthrust but not particularly noticeable and well below normal road traffic noise levels. The latest nylon geared unit is very quiet, almost silent in most road conditions.
I have an article on motor noise on my website, but that's offline at the moment. I'm therefore copying it below:
All electric motors have an intrinsic physical reason to create noise due to the way in which they work, and the amount varies according to the motor type, it's tolerances, it's revolution rate and the amount of work being done. In the case of Hall effect designs they are potentially a little more noisy than some other types, but any disadvantage of this is outweighed by the power to weight ratio and high efficiency. Conversely, the quietest motors are the alternating current synchronous types like those in fans and fan heaters, but they aren't very powerful for a given weight. Furthermore, the main reason for their quietness is that they run at their constant maximum and technically ideal speed, typically about 1450 rpm. As you know, if you run your bike at maximum speed it's also at it's quietest. How much average noise any one motor puts out depends on tiny variations of tolerances in bearings, gears and controller electronics, the precise alignment of the hub components, and the interacting and therefore varying relationships between those.
In our hub motors, the armature is fixed to the spindle and consists of electrical windings around iron cores, called poles, arranged around the spindle. Surrounding that is a metal drum with bar magnets mounted all round the inner surface of the drum, which is free to revolve with the magnets just missing the armature. It's the revolving drum that drives the gears which in turn drive the hub and wheel. The passage that follows is a simplification of how the motor turns, not precisely accurate but easier to understand. As a magnet approaches one of the armature poles, the controller delivers current to that pole which temporarily magnetizes it, tugging the bar magnet towards it and thus rotating the drum. As you can see, the pull isn't only around the armature in the direction of rotation, it's also inwards, trying to pull the bar magnet into the pole as it passes.
Since the current turns off as the bar magnet passes by, the drum is exposed to a series of in/out jerks all around it's circumference as magnets pass poles. These are rapidly rocking the drum about it's bearing, vibrating it, and transmitting the rocking vibration to the gears and then the hub. Because there are many magnet and pole interactions per revolution, the minute knocks are at high frequency and therefore produce a form of whine with an odd vibrating component in the sound. The exact note and the quantity of it depends on all the slight dimensional variations and relationships I've mentioned. Just turning off the throttle and opening it again can change the sound as the twitch produced by that action changes the internal dimensional relationships.
As a bike goes faster and the drum rotates more quickly, there's less time for current to go into each pole winding so the consumption and power per pole reduces. This means less magnetic push pull, therefore less vibration in each pulse from a pole. In turn, the centrifugal force of the faster spinning magnet drum starts to overcome the now weaker inward pulls of the poles, so the operation is smoother and the noise reduces to a minimum at maximum speed.
It might seem that a motor assembly with minimal clearances would be best for the least noise, but that's not necessarily true, minimising the clearances initially just increases the frequency or pitch of the sound, the harmonic energy of which can set up worse vibrational noise elsewhere in the assembly. That's why an assembly can get quieter as it wears, and also why it's so difficult to manufacture this type of motor with an absolutely consistent pitch and quantity of noise.
To understand why a small clearance produces a higher frequency or pitch, think musical instruments. Short strings in a piano produce higher frequency notes, that's because a short tight string is limited in it's travel from side to side so it does more return trips in a given time. Likewise the short path across a small clearance, means more return journeys of component vibration in a given time. Quite separately, the components vibrating in our hubs each have their own harmonic vibration rates, just like a tuning fork, so want to vibrate at those consistent rates, but the vibration speed imparted by the magnetic pulses will almost always be different. This frequency mismatch means that the two clash to produce sound distortions which are made up of odd harmonics having very harsh natures. That assists in producing the curiously characteristic harsh growling, deep whining quality of a hard working electric motor such as heard when an electric milk float struggles on a hill or when we snap open the throttle at low speed, and the edginess of the sharp sounds in the whine when the motor spins at higher speeds.
I've only touched on this subject, for as you can see, it's immensely complex and potentially an engineer's nightmare. It's important to remember that it's an electric assist bike we're riding, and if we pedal the bike along and use the motor to help rather than the other way round, we reduce the noise to an optimum. Letting the motor do most of the work maximises the amount of sound produced in any given circumstance.
.