Concise and clear explanation of digital bit depth.

Great article.

The fact is, that when listening to recorded music at anything near reasonable levels, no one is able to differentiate 16-bit from 24-bit. It just doesn't happen. Our ears, brains and bodies just can't process the difference.
Click to expand...

Isn't this because speakers max out at something like 14 bit resolution (IIRC), even very good ones? Bob Stuart of Meridian has written some interesting stuff about this, and the purpose or otherwise of increasing resolution with respect to home audio and playback.

S
 
Tenson said:
Interesting. What do you mean by speakers having a bit-depth?
Click to expand...

I don't think Stevod said loudspeakers had bit depth - that is not a quality they can posses!

But he did say that former research had shown that once a digital electronic system achieved 14 bit resolution it would be impossible to hear any further advancement - because of their physical limitations the loudspeakers would not project any more improvement, and human ears would not hear any difference.
 
What is it about a speaker that would stop it playing very quiet signals, though?

I can see that if a speaker has distortion around -60dB then signals below that may be inaudible, but that is not quite the same thing. In a digital system you could have 16-bit performance (i.e. -96dB noise floor) but distortion elements may be higher. Distortion seems to be ignored in digital systems when describing bit-depth. If not, then something like the Fio D3 would have 13-bit performance. due to IMD being around -75dB.
 
Tenson said:
What is it about a speaker that would stop it playing very quiet signals, though?
Click to expand...

The bit depth is about much more than the noise floor, is it not? With the sampling rate, it's about how jagged the smooth edges of the analogue signal are. The equivalent in digital photography is "pixilation".

Distortion seems to be ignored in digital systems when describing bit-depth. If not, then something like the Fio D3 would have 13-bit performance. due to IMD being around -75dB.
Click to expand...

Of course! The bit depth is the bit depth! But if you say such-and-such a box is 16 bit but because of poor engineering it performs no better than an optimally designed and constructed 16 bit device, that is a less precise use of the technical vocabulary.

I think Merdian said 14 bits were good enough - we can't hear improvement beyond that. The BBC agreed, and the aging digital feed to the FM Radio transmitters is 14 bit. (Let the analogue/FM devotees take note!)

The first generation CD players were only 14 bit, and many were fine players. As 16 bit chips became easier to manufacture domestic digital products settled to that, giving a very comfortable margin over the 14 bit quality threshold. Professional studio and production equipment went to higher bit depths so digital domain editing could be done. The rounding up and rounding down errors in the data manipulation took place at higher than domestic playback resolution.

Conclusion: 14 bits is good enough, and 16 bits generous. It is capable of better resolution than the analogue side of the replay chain, so why get fussed?
 
Last edited by a moderator:
I'm not fussed about it, I just wondered how speaker performance could be equated with bit-depth. I can't understand how it could be thought of in the same way at all since speakers don't have a noise floor. I suppose you could arrive at a dynamic range by taking the maximum level the speaker can produce and subtracting the rooms noise floor. Say 110dB - 30dB to give a dynamic range of 80dB or about 14bit.

I'm not sure sample-rate has much to do with jagged waveforms you know. After reconstruction of the data to analog and low-pass filtering, sample-rate really just sets the upper FR limit and how far out of band oversampling noise and aliasing artefacts are pushed.
 
Tenson said:
What is it about a speaker that would stop it playing very quiet signals, though?
Click to expand...

Thinking aloud....

Given we have a moving mass driven by a movement generated via a motor, would there be a minimum input required to set the 'motor' into motion?

Think of a conventional motor. If you slowly reduce its input power, there comes a point where it stops completely even though it still has current flowing in the coils.
....and a MC driver is effectively a motor. Therefore, if I'm thinking this through correctly, there will come a point with a driver where signal level reaches a point at which cone motion either stops or becomes very non-linear. Theoretically this could limit resolution.
 
I was thinking the same, Rob. "The rubber effect".

Very low energy pulses will not move the whole cone/dome - they will just distort it - and the energy will be lost in heat rather than sound.

Anyway, the effect is to do with materials science, not digital electronics.
 
I don't think that follows. In a rotational motor there are discrete steps because the coils are attracted to independent magnet blocks with alternate polarity. This happens in an AC or DC motor, the only difference is the mechanism which inverts the direction of current flow in the coils.

As far as I understand it, a speaker uses a fixed magnetic field and when the current flows one way in the coil it attracts and when it flows the other way it repels so there is no discrete step from one magnetic field to another as there is in a rotational motor.

I'm not 100% sure about that though. Can anyone confirm if a speaker motor has a single polarity magnetic field, or if there is a point in the middle of the speaker motor where the magnetic field inverts? If there is a point of field inversion in the middle of the speaker motor (driver rest position) then it follows there may be some 'crossover distortion' where the field is basically nil and a certain level of signal is needed to kickstart the motion.
 
Labarum said:
The bit depth is about much more than the noise floor, is it not? With the sampling rate, it's about how jagged the smooth edges of the analogue signal are. The equivalent in digital photography is "pixilation".
Click to expand...


This is a false concept.

As explained in the demonstration by Mr Montgomery from Xiph, below . . .


Quantisation and it's effect on resolution


JC
 
Tenson said:
I don't think that follows. In a rotational motor there are discrete steps because the coils are attracted to independent magnet blocks with alternate polarity. This happens in an AC or DC motor, the only difference is the mechanism which inverts the direction of current flow in the coils.

As far as I understand it, a speaker uses a fixed magnetic field and when the current flows one way in the coil it attracts and when it flows the other way it repels so there is no discrete step from one magnetic field to another as there is in a rotational motor.

I'm not 100% sure about that though. Can anyone confirm if a speaker motor has a single polarity magnetic field, or if there is a point in the middle of the speaker motor where the magnetic field inverts? If there is a point of field inversion in the middle of the speaker motor (driver rest position) then it follows there may be some 'crossover distortion' where the field is basically nil and a certain level of signal is needed to kickstart the motion.
Click to expand...

I was thinking something a little different.
For example, if I gently push the coffee cup in front of me with my finger, it doesn't start to move until the force exerted reaches a certain point. Not only that but I can feel the tip of my finger compressing until it becomes sufficiently stiff to transfer enough energy into the cup for it to move.

If you now look at a driver, its held in place by the spider and surround in much the same way that my coffee cup is held in place by the table on which it sits.

At very low energy input levels it follows there will be with no motion, or the transfer of energy from the voice coil will be very non-linear. I'm just not sure at what level this happens and the relevance to dynamic range.

The relevance to bit dept is performance of the driver at extremely low power levels and therefore very small signal levels.
 
Hmm.. I was thinking of that too, but I've no idea if it works in the same way as the coffee cup or not. The coffee cup obviously needs to have force that exceeds the frictional losses before it will move. In the case of the driver, the suspension parts are bending, not moving over other surfaces. Even if the suspension were made of steel plate it would still bend just a tiny imperceptible amount with a small force applied. Using a materials 'Young's modulus' that amount can be calculated for any weight and various mounting conditions. This might be fun to play with (http://www.engineeringtoolbox.com/beam-stress-deflection-d_1312.html

That said there is 'shear' friction internal to the material itself where the crystalline make-up of the material needs to side around itself. So perhaps there is on a really small scale, a level of friction that needs to be overcome. I really wish I knew the answer, materials science is very interesting!
 
To correct myself on the earlier post about a drivers magnetic field, there are in fact two poles (north, south) in the driver but they are not above and below the centre rest position of the driver as I had imagined they would be. They are on the inside and outside of the voice coil. That's because the movement force is generated at a right-angle to the direction of current flow. So there is certianly no 'crossover distortion' at the centre position of the driver but not for the reason I had imagined.
 

Latest posts

Back
Top