Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Casio Privia 350 and "hi-def" midi

The Casio Privia 350 (and the 150) have a very nice feature which it sends the regular range of midi velocity values (from 0 to 127) with each note on AND it also sends additional velocity information on CC#88 (also from 0 to 127).

So the full range of dynamic values the Privia can transmit via MIDI is 16,384 (128 values that get attached to each note-on and then each of those note-on velocities is further qualified by the additional 128 values that can transmit on CC#88

This is a nice feature when using the Privia as a controller for Pianoteq which can use that full velocity range when set to receive "Hi resolution" midi.

I've been playing with this a lot and exploring it and have come to a problematic conclusion, which is all that extra resolution doesn't seem to have that much effect on Pianoteq. In fact, having done A/B tests with a several listeners (to see if they can identify when hi-def midi is on and when it's not), what I've found is no one can consistently identify high-def from listening alone.

I THINK when I'm controlling Pianoteq with the Privia 350 I can sometimes hear/feel a difference. But only sometimes. And although I can see via PIanoteqs MIDI display that it's receiving and interpreting all the extra CC#88 data I can't consistently do anything at the keyboard that demonstrates the difference between regular midi and "high def" midi.

I'm wondering, is anyone else using the Privia (350 or 150) as a controller because of the extra data that's available on CC#88? If so what have you found? Is there a situation where you can clearly and always hear the difference? And what exactly is the Privia measuring to produce that extra data on CC#88?

Intuitively, it seems to me that there should be a HUGE difference when playing with 16,384 velocity values instead of the more normal range from 0 to 127. But in actual, I'm not seeing anything that's really really significant or clearly noticeable.

can anyone clarify?

Au contraire ...

Intuitively, it seems to me that there should be virtually no difference when playing with 16384 velocity values.

Analytically ... Given that the dynamic range of a piano is generally no more than 60 dB, then the standard 127 velocity levels could produce increments of 60/127 = 0.47 dB. That's under half a decibel! Way too small for the ear to detect. To further divide the range in 16384 values is beyond absurd.

Now, there's more to piano sound than just volume/dynamic range. There's timbre. But do you think your ears can distinguish even 127 different timbres? Doubtful. So once again, 16384 levels would be pointless.

Conjecture: The 16384 spec is pure market-speak, meant to attract numbers people.

It's like they say in the trash bag commercials: "Our bags are now six layers thick!"
To which I reply: "Mine are 19,000 layers thick! (Each layer is one atom!)"

Numbers can be used for good purpose. And they can be used to deceive.

How can you notice difference ?
8 is enough for me :-)
ppp,pp,p,mp,mf,f,ff,fff

jarosuju .. thanks for taking the time to post. but my interest really is in hearing about experience from anyone who's actually explored and used the technology and can document some actual examples of when and where it works (or vice versa, if it simply doesn't work). for example, i've seen that there's at least one casio rep here on the forum, so i'm hoping this is a thread he'll find interesting ...

in my experience i can definitely tell that there are instances (not always reproducible) when the high res does make a difference as regards to what i'm hearing in relation to how the keys feel. so, for me, it's a question of trying the privia with pianoteq and seeing how that works.

so, i'm taking in what you're saying (and i do appreciate what your saying about dbs, etc.). .... but i'm really hoping that a few or more on the forum hav own hand-on experience to discuss. and in that case, if hands on experience verifies your explanation, that would be great! we'd have a very solid answer put together from practical experience and theoretical understanding (or intuition, for that matter).

thanks much! anyone else have some experience to speak of?

eeks! jarosujo! so sorry .. most of my comments were to the macmacmac. my answer to your question is yes more than 8 dynamics are desirable and useful.

but really, as i'm saying, and to repeat the obvious again. it's question of experience. who's actually USED that extra controller information, whether successfully or not, from a keyboard that can generate it (like the privia 150 or 350) with a dig. piano that can interpret it (like pianoteq).

I must admit.. I don't have any input on this.. But the digital forum gets some very in depth threads.. I'm amazed at the ideas and questions posted here.

Ps: not being sarcastic at all.

Have a great one everyone!

Though that assumes that the 127 levels of force detectable by the mechanism are assigned evenly over the 60 dB audible range... and that doing so results in a touch-to-volume correspondence that approximates that of a real piano. I think that's a lot of assumption!

That said, while I don't think your math proves it, I'm still inclined to be skeptical of any real benefit to more than 127 velocity levels, as you are.

Yes, I neglected the possibility that some parts of the range could be given up to coarser increments, and others to finer increments.

But does that matter? I'm sure 127 increments is fine enough. It might do no harm to make SOME of the increments larger (in the 16384 scheme). But what would be the point of making SOME of them finer?

"Velocity" (playing) precision mainly depends on the interface/hardware. If the keyboard/controller uses the usual approach of defining velocity between the closing of two contacts (like Fatar and many others), then it doesn't matter if you subdivide every 127 by 127 or increase bits - you're limited by the interface and how it creates velocity.

The Bosendorfer CEUS, which uses advanced optical sensors, could probably give the player a better impression that he/she has control on more than 127 velocity levels. Certainly not a Casio, since the hardware alone implies expensive stuff.

Besides, a real acoustic piano mechanism can't be adequately reproduced by velocity alone. The great old Kurzweil MIDIboard used an interface much closer to the real thing in terms of approach (even though still different from an acoustic piano action) with its elastomeric hammers hitting rectangular surfaces. This also could provide other useful outputs for polyphonic aftertouch and release velocity.

The problem today is nobody wants to pay $3500 or more for just a controller. Fortunately, you can find good new controllers today like the VAX-77. So to go back about a new MIDI implementation, enhanced velocity could be useful with controllers equipped with appropriate sensors and velocity approach. Also, more velocity values can certainly be useful to program better velocity maps, since many cheap controllers send only a limited percentage of what's available (often not going lower than 40). Just like old controllers like the Yamaha DX-7 stopped at 100 on 127, you could then use the "hi-def" MIDI velocity to improve the limitations of the limited hardware/interface and end up with more expression/dynamics while playing.

So many skeptics! Haha.

Actually the question of whether 127 velocity levels is sufficiently granular or not has been open and posed on this forum a number of times. As a rule, we've only been able to conjecture whether it makes a difference because until now it appeared that the only DP or controller with higher resolution MIDI was the VAX77, which apparently no one owns.

I'm very glad Casio put this feature in their piano because a number of forum members have this piano and can test it themselves. If possible you should give yourself a few blind trials to see whether you can tell if it's engaged or not. I would be willing to believe either case.

I've sort of had the same intuition as MacMacMac, that 127 is probably enough levels for the tolerances of our ears, fingers, and DP sensors. However, since we often remap the velocity curves in the software, I can imagine cases where it would be advantageous to have higher resolution coming out of the piano. Kind of like we can't really tell the difference between an 8 bit image and 16 with our eyes, but if you edit it and tweak the color curves it's easy to get posterization and other artifacts with 8 bits that you don't see with 16.

In either case, this feature doesn't hurt anything and it will be interesting to test. Thanks for bring it to our attention and kudos to Casio for thinking of it.

Now if they just made a free-standing progressive pedal that would work with their pianos!

Bosendorff ... the idea of a better velocity maps is part of what i've been exploring ... with max/msp where the incoming values from the privia map to exponential curves, etc. for sure, the best overall response i've gotten out of pianoteq has been when mapping standard midi velocity between 0 and 127 to high-def curves between 0 and 16383 ... which pianoteq can respond to.

i don't see it as a question of imitating a real piano or even getting closer in sound to a piano. i'm more interested in getting what i hear (whatever it might be coming from speakers) closer to what i feel (on the keybed). and one way to do that would be with high-def velocity maps.

i think the privia has 3 velocity sensors rather than two. so i'm curious to know what exactly what that 3rd sensor is measuring and how whatever extra is - at least in theory - is supposed to give extra (or extraordinary) control. if it's just generating extraneous values that have nothing to do with anything i'm happy to know that too! i just want to know what exactly is being measured ..

i don't think the cc#88 values are mentioned in the the privia manual ... i found out about that possibility on the pianoteq forums. casio, to the best of my knowledge, doesn't mention that particular feature in privia marketing.

so, again, just curious to know what's going on. if it turns out the extra-resolution of the privia isn't useable, whether for quality of the hardware of the limits of human perception, i've still found nonetheless that there's some potential with the use of high-def velocity maps ... to which pianoteq can respond.

gv ... in blind trials (the few i've done), i and a sales person and someone who just happened to be in the store did sometimes hear a difference. and sometimes we guessed wrong on that difference. the closest i've come to being able to identify where exactly the difference happens is when playing softer rather than louder. some chords just seem to articulate in a more detailed fashion ... or so it seemed at the time.

the progressive pedal ... the solution i'm using now is a roland rpu-3 plugged into the expression pedal jack of another device. it works well that way and feels good as well. now, having said that, there's still more that could be done to make it feel even better. but that's for later

... i have no idea of whether or not the privia itself responds to high-def input. i use it only as a controller for pianoteq.

Two sensors give the time it takes the hammer (or key, in the new privias) to pass by, giving the velocity. The other sensor controls the damper, essentially. This allows the possibility of playing the same note twice without lifting key high enough that the damper is lifted and first note is cut off. This mimics the behavior of a grand with double escapement and can mean faster repetition, which can make it feel more responsive. In addition it's theoretically possible that the two velocity sensors can be placed closer together since, unlike in a two sensor action, the top sensor need not be positioned at the location where dampers are lifted. I don't actually know how much closer together (if at all) the two bottom sensors are in practice and what the tolerances are, but if they are closer together and closer to the end of the stroke, it would more precisely estimate the terminal velocity.



Shame, seems like a great feature to me. Don't know how much I'd use it (not a bit PianoTeq fan) but it does enhance the value of those Casios in my mind, so I guess it's working!

Depends on the spacing of the original 128. If the curve is not ideal then 256 or 512 would make a difference, but somewhere it becomes meaningless long before 16384. 1024 has a nice ring to it ...

MIDI supports 128 values per parameter. The way to get more than 128 values is to use two parameters. Then you end up with 128 times 128, which is 16,384 possible values.

IMO simply false. Listen to the DPBSD MIDI File in the sequence of C4 over all of 127 velocity levels. You will hear out the loudness difference between two neighbouring sounds out of practically all of tested instruments (even ones with smaller dynamic range.)

To percieve doubling of a loudness level psychologically you have to add an increase of ca. 3dB.
Considering this one half of a decibel is not a small difference at all, isn't it?

Articulating a melody is mostly playing with these subtleties - these are the means of expressive playing. Qualities of big/professional artistic performances are determined probably on this level.

On the other side, there exist probably even on acoustic pianos some inconsistencies, adding to some other randomising effects from the finger, and the neurology/muscles/joints of player.
Because of these effects I guess that the granularity defined by the 127 MIDI velocities are similar to that of a real thing. But it is a guess only, therefore I am very curious, if this effect could be verified by some professional (blind test is easily to arrange with a professional plyaing while someone turning extra layers off - the player should be able to distingiush between the different velocity sensitivity levels in a consistent manner).

Perhaps this is one important contributory factor to the superior accoustic experience we all will confirm.

I am very interested hearing the conclusion to this topic too.
Attila

This is not true:A change of 3 dB is a doubling of power. But it's only a subtle change in apparent loudness.

It takes 10 dB to perceive a doubling of apparent loudness. That's a 10x power increase.

Further ... measured results conducted by audiologists show that this is not correct:
The evidence is that a 1 dB is VERY difficult to hear, even at midrange frequencies. It's considered the threshold of distinguishability. And that ability diminishes as you move toward higher/lower pitches.

So 0.5 dB is not distinguishable to humans, only to suitable test equipment.

To my original point ... to divide that 0.5 dB further into 128 finer increments is quite pointless.

That is an oft-repeated but, as far as I've been able to see, dubious assertion. Though you are correct that a 3 dB increase, while clearly louder, is not generally perceived as being nearly as much louder as its double-the-power would intellectually imply.


I think 1 dB difference is pretty easy to hear. I score 10/10 on this one:

http://www.audiocheck.net/blindtests_level.php?lvl=1

It would be even easier if the two sounds had different harmonic content, which is what happens on a piano (i.e. the overtones don't all increase together in a linear fashion, so that there are shifts in timbre, not merely level).



For many people, .5 dB actually is distinguishable in the midrange frequencies, even without a timbral (harmonic) shift. You can test your own perception here...

http://www.audiocheck.net/blindtests_level.php?lvl=0.5

I think it's easier to hear when a sound increases by a half dB than when it decreases by a half dB. Again, it would presumably be easier to tell if there had been a change in tone as well as level, as there would be on a piano.

Just to pull back to my original query, which is about experience with the Privia and Pianoteq (and before pulling back, I should say it IS REALLY REALLY interesting to hear different opinions about what's audible or not ...I'm learning quite a bit from I'm reading here and it's helping me to think about how to test this stuff ... what to look for, etc.)

But, having said that, I'd say that direct experience is essential. By direct experience, I mean working with a keyboard like the Privia (which outputs that extended range of values) and Pianoteq (which reads and interpret that extended range). I'm not saying that Privia or Pianoteq is better - or worse - for being able to put out and take in the extended values. For me it's just that the these kinds of features seem to allow possibility that gear without them doesn't. So we can really only know for sure if that possibility exists if we have a chance to check it out and see ..

II do, I think, hear a slightly audible difference between the extended range of 16384 values and the standard range of 127 values. Sometimes, the difference translates into a more direct relationship between the feel of the keys and sound. Sometimes it's in the smoothness of a transition from one phrase or chord to another. And often, it's just not detectible ...

Some places where that extended range could be helpful are with more detailed velocity maps (curves) than we can usually get. For example a velocity map with a curve that moves quite a bit away from the values along the diagonal. With Pianoteq, for example, you use play on a keyboard that outputs 127 possible velocity steps but, with something like Max (in between the keyboard output and Pianoteq's input) you can map those steps to non-linear curves that fall in the greater range of 16384 values. So this is an example of precision you can get from a digital piano sound engine even if the controller keyboard can't itself generate those values.

Another example is that extended range means we don't all have to play with and within the same range of quantized steps enforced by 127 steps. To apply what Temperament says about this general kind of thing: ... "Articulating a melody is mostly playing with these subtleties - these are the means of expressive playing. Qualities of big/professional artistic performances are determined probably on this level." That's a really good point.


So, going back to the original query ... it would be really interesting to hear from others on the forum - and it seems they're out there! - who've had a chance to work directly and intentionally with the extended range. If you have worked with it and you've found it doesn't make a difference, it'd be interesting to hear how and wh and vice versa. If some one with a direct link to Casio (I know they're out there too!) could explain exactly how the CC#88 values relate to the standard 127 note-on values, that would be good to know too. If you haven't actually tested these extended values but you suspect they probably don't really make a worthwhile difference, well, maybe that's worth testing too?

Well, thanks to all who've been contributing to the discuss!!! Much appreciated!!

Thanks for this - what a good site.

I scored 10/10 too, but IMHO the difference is MINISCULE! Also, I spent a little while training myself too. The very first time I listened to the 1dB differences, I wrote it off as completely negligible.

However, I agree that it would be easier to determine the difference when there is also a timbre change.

At the moment I personally would not pay ANY attention to this extra MIDI resolution when making a purchase decision. In fact, I'd hope that it could be disabled, to avoid sloshing around extra data for no benefit at all. All it does is add a bit of latency. (only over standard MIDI - any extra latency over a pure USB would be negligible) Can it be disabled?

Greg.