# Topic: what is velocity meant to be (quantitatively)?

1. ## what is velocity meant to be (quantitatively)?

Hi everybody,

just wondering what sort of quantitative measure velocity should be:

a measure proportional to...

- sound pressure
- something logarithmic (like decibels)
- perceived loudness
- something mechanical (like force by which a key is pressed)
- something that reflects the intensity resolution of the device
- a fuzzy combination of the abovementioned
- something arbitrary, purely up to the manufacturer

any ideas?

-mat123

2. ## Re: what is velocity meant to be (quantitatively)?

Hello Mat123,

I'll take a shot at answering your question, from the point of a pianist, piano technician/tuner, teacher and retired former engineer in the steel industry.

Velocity is a numeric representation of 128 values of downward key speed (not force) that a virtual piano has available to it to create its sound. The 128 values are numbered zero (off, obviously) through 127. The choice for this number 128 was conceived in the early 1980s when three companies, Roland, Yamaha and Korg developed the MIDI standard. By the way, 128 is represented in ones and zeros by means of the equivalent of 7 decimal places, only in Base-2 rather than Base 10. Restated: the number 128 is represented as 1111111 in binary code (when the only allowable numbers are ones and zeros).

Now, every vendor of virtual piano software has a number of design choices to make. The vendor can record the piano once, and then use velocity to control the loudness and use some algorithms to vary the EQ (brightness) as one plays with higher velocity. As you are well aware, literally none of the major vendors of piano software record and encode a piano from just one set of samples.

Some vendors record the piano at as few as 3 or 4 loudness levels (of the original real piano), and then combine those 3 or 4 recorded levels in ways they see fit ... to render a smooth transition from softest to loudest sound as divided among the 128 MIDI velocity levels.

Other pianos are recorded with approximately 6 to 10 loudness levels as captured from the original piano. It stands to reason, the transitions between MIDI velocity calling up any of the 6 to 10 recorded loudness levels of the piano, are finer when more samples are chosen. Restated, there is less chance of the listener discerning any great changes in volume or timber between, say, velocity 87 and velocity 88.

Still other pianos have 18 velocity levels to be divided (at the vendor's will) among the 128 MIDI velocities.

I was at the NAMM show last month, and heard the Vienna Symphonic Orchestra's soon-to-be-released new version of the Bosendorfer wherein something like 100 individual velocity levels were recorded on their piano. They also state there are as many as 1200 individually recorded samples PER NOTE on their piano -- they did not state to me the computer requirements in terms of CPU, hard disk streaming or RAM requirements.

They played a very convincing recording of Rachmaninov's Prelude in G Minor Opus 23 No. 5 to demonstrate the sound of this piano. They also had a prototype of a hardware piano to be used with their software. When I played their prototype, it was not balanced velocity-wise -- and the VSL representative told me this was only a prototype of their piano.

Regarding a measure proportional to sound pressure, perceived loudness and/or something logarithmic e.g., decibels -- the answer is a partial "yes", because when a real piano is played with greater key velocity, it gets louder. And human hearing is a logarithmic function regarding decibels -- but this isn't the whole story.

Regarding your statement of something like mechanical (force by which a key is pressed). My answer is that it is a qualified "yes" -- but be aware that it is not key force that controls loudness--- it is key velocity! I know they seem to be the same thing at first glance -- but you can press very slowly and very heavily (and bottom out the key), but the note will still be soft. It really is key depression speed, hence velocity, that controls loudness, timbre, etc in a real piano -- and this is the stuff that a software vendor wishes to capture in a sampled piano.

Regarding a fuzzy combination of the above and something arbitrary by the manufacturer: Again I answer a qualified "yes" -- Depending on the make and quality of the given real piano, depending on the recording equipment, depending on the length of each sound, depending on whether the manufacturer samples the note in its entirety until the sound dies away (or whether he samples, say 1 second of sound, and then loops it), and the vendors patience and expertise in capturing the sound, etc. etc. all are controlled and affected by key velocity.

Cheers,

Joe -- jcfelice88keys

3. ## Re: what is velocity meant to be (quantitatively)?

what velocity means quantitatively, especially with respect to loudness still remains a mystery to me. Suppose we have a reference note/sound correspoding to velocity 20. What should be the velocity value for another note/sound that is perceived twice (fourfold, resp.) as loud? Is velocity 2 really twice as loud as velocity 1?

Are there so many "that depends on" that it doesn't make sense to give an answer, or is there an idealized, simplified answer?

On a side note: I'd like to defend my notion of mechanical force being a relevant parameter: please imagine two balls of equal diameter but different weight that are being dropped onto a keyboard from the same distance, say from 2 meters above. Both will hit the keyboard with the same speed, but with a different energy. The heavier one will sound louder. [Edit: both will have the same initial speed until they are touching the key, and the same final speed (i.e., zero), when the key is completely down. The question, of course, is: what is the speed in-between? And this speed in-between is directly correlated with the force between the pressing agent and the key]

cheers,
mat123

4. ## Re: what is velocity meant to be (quantitatively)?

well, from what I understand, apart from what I think velocity "should be", is that most keyboards simply use the time it takes for a key to start moving, and for the key to bottom out. So, if the key started to go down, and the time is 0 milliseconds, and bottomed out at 20 milliseconds, then the velocity would be in some relation to 20 milliseconds. (I'm trying not to use technical terms) This is not accurate at all, since the velocity "should" be the speed of the hammer when it strikes the string. Kawaii's keyboards I think measures it this way. Only problem is that since there is a physical object striking a sensor, I've heard of people complain about the sensors going "off" after lots of use.

Roland's v-piano uses an interesting approach, in that it continuously tracks the position of the key, and continuously feeds the position info to the virtual hammer. This much more correct. Infinite response's VAX77 also achieves its 16,000 levels of velocity through the same method. (continuous tracking)

I think the bigger problem is that there really is no standard for velocity. If everyone can agree upon a standard method of deriving velocity, then it'll be easier to optimize sample libs.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•