need for larger velocity scales, 0-128 to small

peter_schwartz · Post by **peter_schwartz** » Mon Jan 31, 2011 6:17 pm

Agreed, velocity is a much more difficult topic to resolve (pun intended) in terms of the practical application of having higher resolution (and editing those value - OMG!). In terms of control changes, MIDI already supports 14-bit CC messages, though few instruments respond to it. The fact that the Kronos provides parameter interpolation between "coarse" CC values is fantastic, and allows people to edit the more manageable number of 128 CC values than the 16384 that 14-bit MIDI affords. Anything that's going to reduce zipper noise or give controls a smoother feel is certainly a welcome thing. Increasing velocity resolution would be welcome too, but as I said, there's a limit to the practical application of it, at least at this time.

StephenKay · Post by **StephenKay** » Mon Jan 31, 2011 6:32 pm

Lougheed wrote:
StephenKay wrote:It is not possible to divide the full range into more than 128 steps, for higher-resolution.
Sure it is. The Avant Grand N3 sitting in my studio is proof!

I don't know this device, but yes, it's quite possible that *internally* a device can have more than 128 steps of velocity control. However, it's not possible for that to be sent over MIDI, and since everything uses MIDI these days, 128 steps has become the default. If a company wants to make a device that has 1000 steps of velocity internally, it's possible....but that cannot be represented in MIDI so no external sequencers could record it, for example. It could be used in an entirely self-contained system, however.

peter_schwartz · Post by **peter_schwartz** » Mon Jan 31, 2011 6:43 pm

Hi Stephen!

Here's a linky where you can download a PDF about the HD velocity spec:

http://www.midi.org/techspecs/index.php

maphill · Post by **maphill** » Mon Jan 31, 2011 6:52 pm

Rather than the resolution of MIDI velocity, the perceived problem is likely due to how that control is mapped to a sound. Obviously a piano program with 2 or 3 velocity layers will have very audible tone differences that don't smoothly vary.

Or if the physical velocity curves are mapped poorly for a given keyboard/performer, they may end up constantly playing at 127 or jumping between extremes.

However, if the control is set up well and the patch responds evenly, your ear will have a VERY hard time hearing ANY limitation due to the resolution of the velocity control.

I've written scripts that generated MIDI at all velocity levels and then play the notes sequentially. Or if you use an automated sampler like EmulatorX you can do something similar. The difference (on a good patch) between most incremental velocities is down around perceptual JND (just noticeable differences). Meaning, in a double-blind test on a program that is only scaling the volume and not switching layers, you will just barely hear the difference in volume between a velocity of 115 and 116.

The difference can be more pronounce on some synths at low velocity, but this is an implementation issue, not a velocity issue. (Running down to an unusable 0 amplitude with 0 velocity)

One other dimension to consider: How accurate is a player's velocity resolution? If a person can't consistently play at the same velocity level at the current resolution (can they play a 78 in velocity 3 times in a row?), then the current resolution is likely to be sufficient to capture a player's performance (the extra resolution isn't needed).

Food for thought.
Mark.

EvilDragon · Post by **EvilDragon** » Mon Jan 31, 2011 7:05 pm

maphill wrote:One other dimension to consider: How accurate is a player's velocity resolution? If a person can't consistently play at the same velocity level at the current resolution (can they play a 78 in velocity 3 times in a row?), then the current resolution is likely to be sufficient to capture a player's performance (the extra resolution isn't needed).

QED.

synthjoe · Post by **synthjoe** » Tue Feb 01, 2011 12:26 am

Some great discussion. I like most of it and agree with a lot of stuff, particularly with the velocity curves being cr@p.

And it also means, unfortunately, that the 127 values are not divided up in dB linear values, hence sometimes you get much less liberty than you _could_ control. Add to this all issues about samples, filters and envelope generators mentioned earlier and you got yourself quite a pretty mess.

So I agree that the biggest issue these days is probably not the 128 theoretical values you can have in MIDI (e.g. velocity), but saying that 128 would be sufficient for all purposes forever is about as bright a statement as 'No one will need more than 637 kb of memory for a personal computer' or that '320 kbps MP3 will satisfy even the most audiophile listeners'.

I think with most of today's digital keyboards these values are just fine, but for a realistic piano emulation, for example, I think we'll need to go beyond - and serious instruments demanding such solutions are coming up.

Finally, just for the record, 1 dB is an awfully bad granularity. When mixing, I'm often fiddling on .2 or .3 dB's on certain frequencies. Go and ask a mastering engineer about 1 dB granularity. You'll hear wonders...

Our ears are much more sensitive at certain frequencies and volume ranges than at others. If 1 dB was good enough, CD's would play scaled (velocity-curved) 7 bit data and not linear 16 bit.

synthjoe · Post by **synthjoe** » Tue Feb 01, 2011 12:38 am

peter_schwartz wrote:Hi Stephen!

Here's a linky where you can download a PDF about the HD velocity spec:

http://www.midi.org/techspecs/index.php

I like that, it looks good! Hope to see it implemented in some top-notch keyboard soon - I'd love to try that with the SG-1!

jemkeys25 · Post by **jemkeys25** » Wed Feb 02, 2011 11:41 pm

me too.

SpIdErWeB · Post by **SpIdErWeB** » Thu Feb 03, 2011 12:23 am

StephenKay wrote:I don't know this device, but yes, it's quite possible that *internally* a device can have more than 128 steps of velocity control. However, it's not possible for that to be sent over MIDI, and since everything uses MIDI these days, 128 steps has become the default. If a company wants to make a device that has 1000 steps of velocity internally, it's possible....but that cannot be represented in MIDI so no external sequencers could record it, for example. It could be used in an entirely self-contained system, however.

Hm... actually you can use 1000 steps velocity internally (or even 32 bits floating point velocity if it makes some people happier) and send them on MIDI, but the velocity "resolution" will be scaled down to 128 steps "only".

In the other hand, you will need a physical ivory keyboard good enough to detect and transmit more steps...

Most of the time, people hate 127 steps because their keyboards aren't good enough, the velocity curve doesn't match to their sound engine and aren't right calibrated, and the sound engine does only have 2, 4 or even 8 layers and some steps could be heard....

Just record 12 notes and try to hit the value 5, 31, 32, 33, 62, 63, 64, 65, 112, 113, 114, 125... etc... I will pay you a beer if you hit these exact values at the first try. What the point? The point if the current Midi 127 steps is already good enough for keyboards today. The limit isn't the electronic, but the physical proprieties... including the capacities of the human being to be precise enough in that case.

So, before to try to increase the Midi velocity, we should find better keyboard mechanics and have better sound engine with more layer or no layer at all (modeling). Until then... it would be just useless

Phil

StephenKay · Post by **StephenKay** » Thu Feb 03, 2011 12:55 am

SpIdErWeB wrote:ust record 12 notes and try to hit the value 5, 31, 32, 33, 62, 63, 64, 65, 112, 113, 114, 125... etc... I will pay you a beer if you hit these exact values at the first try. What the point? The point if the current Midi 127 steps is already good enough for keyboards today. The limit isn't the electronic, but the physical proprieties... including the capacities of the human being to be precise enough in that case.

So, before to try to increase the Midi velocity, we should find better keyboard mechanics and have better sound engine with more layer or no layer at all (modeling). Until then... it would be just useless

Maybe...but keyboards are only *one* type of MIDI controller. There are guitar-MIDI devices, violin-MIDI devices, Continuum, Laser harps, etc. Some of these devices are much more sensitive than a keybed, and capable of, in the hands of a skilled musician more accuracy that could benefit from HD MIDI velocity.

SpIdErWeB · Post by **SpIdErWeB** » Thu Feb 03, 2011 1:03 am

I agree and that's what I wrote Stephen... good enough for Keyboard.

I don't mean HD Midi is useless... I'm even curious to know more about it when it will be time. But for keyboard, I don't see lot of benefit for the velocity. (But I'm convinced there are other benefits somewhere else such channels, midi CC, etc...).

VSTs do use more than 128 steps internally and it doesn't make them more accurate, because the midi controller keyboard, even in USB, isn't good enough to transmit more steps... that's my point.

Phil

Timo · Post by **Timo** » Thu Feb 03, 2011 1:10 am

Bleh. 0-127 on a decent keybed with various curves is more than enough for me.

Something of more substance is the 0-127 resolution for filtering. That's where problems in resolution can really be heard.

On the Radias, the patented ultra-low aliasing algorithms and pristine high-frequency response captures the stepping artefacts perfectly.

http://www.infekted.org/korg/Rad-stepping.mp3

At least the Kronos/Oasys interpolates between MIDI/knob values, the Access Virus too. Great shame that the Radias didn't.

Shakil · Post by **Shakil** » Thu Feb 03, 2011 9:17 am

Somehow YAMAHA Motif series has 0-255 range for filter cut-off parameter.

ozy · Post by **ozy** » Thu Feb 03, 2011 9:39 am

Shakil wrote:Somehow YAMAHA Motif series has 0-255 range for filter cut-off parameter.

it DISPLAYS its range in 0-255 fractions.

That doesn't absolutely mean it has a resolution better, equal or worse than 1/225th.

The DSI analugues filters have 0/127 display resolution, and the filter sweep has 1/8000th resolution. between 32 and 33 there are various levels, and you can get them while sweeping.

real resolution
midi-transmissable resolution
internal scanning resolution
display scale

are. not. the. same. thing.

synthjoe · Post by **synthjoe** » Thu Feb 03, 2011 11:20 am

SpIdErWeB wrote:...So, before to try to increase the Midi velocity, we should find better keyboard mechanics and have better sound engine with more layer or no layer at all (modeling). Until then... it would be just useless

Phil

Exactly my point. And on that road somewhere lies the need for a higher velocity resolution (and other controllers). Plus, as Stephen wrote, some MIDI controllers and synths could benefit from it today, already.