Yes, the CA93, CA63, and MP10 all feature UPHI sound technology. However, as I have stated on a number of occasions, the MP10 features additional acoustic effects and new jazz/pop sounds not found on the CA93/CA63, so it's not entirely a like-for-like comparison.

Cheers,
James
x

Are you saying UPHI has multiple variants when it comes to APs?

Sorry, I know I'm being a royal pain for asking questions about KAWAI products, please forgive me.

And if no one else cares, please ignore me.

Not James, but he previously said that the MP10 has a new set of UPHI AP samples for the jazz/pop pianos. He also said that it has additional effects (pedal up/down sounds perhaps, I don't recall the specifics, but things like that not directly related to a note's sound) not present in previous UPHI pianos. So the CA63/93 have the same base samples for all the notes (and same velocity layers, sample lengths, etc) but not necessarily the same secondary effect samples.

Aaron

Thanks! I'll shut up now.

I suppose until someone reverse engineers a new Kawai DP (expensive proposition!), we'll never know the ins and outs of the various forms of Harmonic Imaging.

Personally, I don't see how it could possibly harm Kawai to publish basic data, such as the different memory allocations or sampling levels. Secrecy might be OK if you could wander down to your nearest friendly DP store and see how much you like one model or another, but when you're "flying blind" and ordering on the basis of a few audio clips and some marketing spiel, it's frustrating. I actually think this kind of thing is counterproductive, turning potential customers off.

Corporate secrecy is reaching the point where employees are so afraid of revealing anything that might get them in trouble that even higher-ups won't go on record for telling you essentially nothing about their products. If I weren't shopping and actually trying to buy the very things they can't discuss, it would be kind of hilarious.

For some reason the Cone of Silence comes to mind.

dewster,

I thank you for your thread that dissects the sound of the DP's and analyze them to bits. I think that this analyze is a VERY good idea in that it reveales the DP manufacturers cheats. But are the tech specs (or how-does-it-work explanation) really that important?

Think about it this way:
We have two boxes from two different manufacturers, one white and one black, that plays HD movies from an external harddrive. The black one has a slow CPU and very little RAM and the white one has huge amounts of RAM and a kickass CPU. The software implementation on the black one is so good that it actually exceeds the performance and usability of the white one. If you assume you cannot hack the boxes and put your own software on them the black one would of course be the better buy. The tech specs on these two would be meaningless if there wasn't a universal tech spec unit for general excelence (GE maybe?).


However, I think that your interrogation like curiosity is a good thing. Every customer should be able to ask these questions without feeling like a pain in the butt.

A quick run through of all the AP/EP presets and a presentation of what can be produced with the amp sim and efx. Since the MP10 has so few sounds I think it would be good to show how good the few that are in there really are. I really like the idea of focusing on the AP/EP's and skip the dozens of mostly unusable sounds others have in their GM banks.

Those APs sound great. The EPs sound pretty good too. Makes me much more comfortable about getting rid of the CP1. The CP1 really has an excellent EP section that is going to be hard to beat, but the RM3 action, the USB recording and the AP sounds of the MP10 are just what I need.

I really wasn't asking for anything nearly that deep in the way of info. If they said something like "the latest variant of sound technology on the new brand x model y better captures the full dynamics of playing (implying more layers or something analogous) and includes mechanical noises for extra realism, plus better sympathetic resonance" I'd be fine with that. Roland doesn't explain how they do SN, but I'm pretty happy with the way they describe what SN means to a DP buyer / player.

Why is important buyer info freely given by one manufacturer but need to be pulled like teeth from another?

1. For a company known more for its APs than DPs it boils down to mystique - the more the product is perceived as a rich acoustic instrument the better, so the less talk about a DP line as a pile of plastic keys on a slab of MDF loosely tied to a 10 MIPs processor and 64 MB of Flash and a couple of $5 speakers / 10W amps the better. Any technical details that absolutely must be revealed will be strictly controlled by the marketing department - and this performed as something of a striptease, where the emphasis is on the revealing rather than what is revealed, and in the end you may have only caught a glimpse of the underthings.

2. For a company known more for its prowess in keyboards, synths, and other sound technology, it is in that company's best interests to shout the results of that prowess from the rooftops, though perhaps without coming out and saying exactly how they accomplish it so as to maintain some mystique there.

Companies in category 2 (Roland) raise the consumer information expectation bar for companies in category 1 (Kawai, Yamaha). I also think that the DP departments in category 1 companies don't have to try quite as hard to be ahead of the curve, because their digital products get a contact mystique boost from their AP products.

I believe that's actually what I've been saying from the very beginning.

Cheers,
James
x

You have, but more in terms of the difference between this new "enhanced" PHI and UPHI found in MP6 / MP10 (roughly twice the memory plus sound effects) as compared to the older PHI / UPHI found in previous Kawai DPs. I'm interested in the differences a listener / player might notice between HI, PHI, and UPHI (and also this new EPHI and EUPHI).

Well dewster, I fully respect and appreciate your work in the DPBSD, but it seems clear that KJ can't give away this information. I guess insistence is futile.

So the only way to obtain the data you want is to buy and disassemble the required instruments. How difficult would that?

[Sorry for the OT]

Well, eventually hopefully there will be some MP6 and MP10 owners on this forum... all we need is one of each of them to send Dewster a DPBSD detector file, and then there will be a basis for determining what kind of audible differences there are. Patience...

I have learned from the german Kawai Forum that both CA 13 and CN 33 have a 8 step pedal. So the question arises, if all PHI pianos support 8 steps for the pedal only, and if all UPHI support 24 steps.

Especially for the MP6 -which is often used as Masterkeyboard- this question would be of interest. Also I think, such details cannot been hold as a secret, because they are easily measurable and a customer will know it after buying without problem.

This data that is /easily/ obtainable (measurable) after buy should be public.
(So basic data like levels and impedances should be public also)

I have also a proposal here for the Kawai engineers: Those pianos that only use 8 steps could nevertheless output 24 steps on the MIDI port and record them on the recorder. Normally the pedal is continuous and the steps are generated per software. So this should be easy to do. I think, for example a MP6 will be compatible to MIDI files generated by a MP10, it will simply ignore the extraneous steps and so no problems should arise from that, but the usability as a masterkeyboard would be improved.

I myself are not so much interested about internals of the soundsystem. Data about sample layers and so on can be very misleading if the piano internally uses a combined method of sampling and modelling. (I believe the Kawai's do this) Even the polyphony data is misleading if it is not specified wether voices, mono voices, stero voices or number of oscillators or notes are meant. A piano with a polyphony of 32 stereo notes can easily be better than a pino with 128 oscillators polyphony and a piano with high polyphony but primitive management can be worse than a piano with low polyphony and sophisticated management.


Best,

Peter

Below average marketing mixed with the fact that what's new in UPHI might not sound nearly as impressive on paper as in real life?

I agree. Specs are often missing complete context; and even if they are complete, they give you data only on a very small piece of what makes the thing sound the way it does; and even on that part of the sound, there is not necessarily a direct correlation between a number and subjective quality. As I've mentioned elsewhere, I've heard single layer pianos with more natural sounding dynamics than some multi-layer pianos.
That said, I think to the extent specs are useful, they are more useful when comparing models within a manufacturer's line than when comparing models between different manufacturers.

Precisely. If I'm considering an MP6, I want to know in some quantitative way (so that I can extrapolate some sort of qualitative reference point) how it differs from the MP10.

At the moment, even though this data is not published, we know that the MP6 has only around half the memory of the MP10. That implies that something must be missing or heavily compressed, or both. What that might be is important to a prospective purchaser. There will be other differences too, both in the sound engine and elsewhere in the software implementation, that will affect everyday operation; where such things impact on performance, they need to be spelled out so that the buyer is able to make an informed judgment as to whether the instrument is fit for their purpose.

Maybe it helps: MP5 sustain outputs 24 MIDI steps (0, 11, 18, 21, 26, 30, 37, 44, 50, 55, 60, 65, 72, 79, 86, 93, 97, 102, 106, 110, 115, 119, 123, 127). I wouldn't expect less from MP6.

My guess is that the 24 steps limit may be some kind of limitation of the pedal itself (not enough resolution). If MP6 uses the same pedal, probably it will output the same steps.

Thank you very much!

No, the reason is not the pedal resolution.
The pedal voltage of my Kawai CP136 has infinite resolution.
This voltage is directly measurable at the processor.
The processor outputs only 7 distinct MIDI values (including zero) So the software makes the steps. (Modern processors have 8 Bit analog inputs or better)

The reason is practical: A pedal cannot always produce MIDI events for all 127 steps. This would give several 100 MIDI events per second, if the pianist is very busy with the pedal.
This would exceed the bandwidth of the MIDI interface. Therefore the amount of events must be reduced and the simplest way to do this, is to send only those steps that are recognized by the receiver.
And yes, I think 24 steps are a very good choice, I dont believe more is necessary for any piano library. 8 or less steps are however not enough. In Vintage D the sustain varies between 0-64 and is almost constant above. So, with an 8 step pedal, only 4 would be effective usable.
The steps 64-127 are apparently only used for pedal noises. If a step is too abrupt, the noise will be too loud.

The MP8 II and the MP5 shared exactly the same sound engine.
With the MP6 and MP10 this has changed. It will be interesting to see the pedal output of the MP6.

Best,

Peter

All DPs are tremendously compressed, some are somewhat less compressed than others, but I don't believe any are out of the "heavy" category. Why this is so in the days of ~$1/GB Flash (retail mind you) I have no answer for. Which DP manufacturer will be the first to enter the 21st century? If one did it right now they could clean up. Most likely they will enter it en masse at NAMM ~2020 (so as not to ruffle any industrial feathers / upset any retirement pension applecarts).

Peter, did you see this post?

Cheers,
James
x

I agree with Peter that this is to limit (already limited) MIDI bandwidth from the pedal, and not a limit to the pedal resolution. A better solution might be to sample the pedal at regular intervals and output that value in the continuous range of 0-127, which would be easier for the end sound generator to interpret as a continuous value, and wouldn't look like such a glaring limitation from the controller.

Thank you James,

This is still a little bit unclear to me. I would like to know which steps are sent via MIDI.

BTW, I think Kawai could easily without hardware modification modify all their current models to send 24 levels. This would increase the usability with software pianos.

Or did they already do this? Unfortunately there is no clear and understandable information for all models available...

Best,

Peter
BTW, I dont currently intend to get a new piano. But I always look into the future and I am sometimes asked for recommendations and so Im interested to know this.

I see... a pedal producing all the 127 steps would certainly eat a lot of MIDI bandwidth... if I recall correctly MIDI can transport approximately 1000 messages each second, so heavy pedaling with full 128 steps would be a bandwidth problem. My guess was wrong... once more time



Much neatest solution the way you put it. It's not a big deal as 24 steps are enough for lots of things, but I agree that as a MIDI controller it limits somewhat pedal possibilities (for instance when using the pedal for controlling some other effect via MIDI).

Although reducing cost is the obvious reason to limit the amount of memory, may I offer another reason. VSL Imperial Grand is a huge library. They compress it 10:1. Even compressed, it's too big to read into RAM. It's read from the disk and decompressed on the fly. In their cases, they use a lossless compression similar to zip. Decompressing something with this kind of compression is pretty quick compared to reading the same amount of uncompressed data from a hard drive. So for VSL, this is more a matter of performance (keeping latency down) as reducing the cost.

Other software instruments stream from disk too, but I'm not sure about compression.

Compression by itself is not a bad thing as long as it's lossless. As sample size increases (longer samples, less looping, more velocity layers, pedal noise, hammer noise, resonance, release samples, higher bit rate, higher sample frequency, etc.), it will never be cost effective to store it uncompressed in expensive flash memory versus decompressing on the fly.

Something like Imperfect Samples Fazioli Extreme is 162 GB installed. That much NAND memory would be quite expensive today. A typical SSD of that size is about $2/GB. And that's from manufacturers who buy it in much higher quantities than a Roland or Yamaha or Kawai. Even Apple who gets one of the best deals for NAND is paying around $27 for 16GB (according to iSuppli).

So, until modeling overtakes sampling, if you want something the size of IS Fazioli or VSL Imperial, all stored in flash memory, no compression, it's going to cost you. In the meantime, I'd rather manufacturers compress (losslessly) and increase sample length by 10x. Or up the resolution from 16-bit/44.1kHz to 24-bit/192kHz and add 25% more velocity levels and 25% longer samples.

I think there is too much focus on metrics that aren't always good for apples-to-apples comparisons. My priorities are 1. how does it play, 2. how does it sound, and then 3. how do the metrics that contribute to these compare. The DPBSD project is a good example of 3. The other two are subjective.

The only way the amount of memory comes into play is if it affects one of these things or makes it too expensive for me to afford.

That doesn't sound right. The MIDI baud rate is 31.25 kHz, plenty of bandwidth for 128 steps several times per second. Otherwise, it could also not handle mod wheels and expression pedals, to say nothing of polyphonic aftertouch.

That is not really plenty of bandwidth. It seems a lot but it is not. You need approximately 30 cycles of those 31.25 Khz for sending a single MIDI message (3 bytes), so you don't have so many message slots in a second (1000).

From http://www.bikexprt.com/cakewalk/midiprob.htm#bandwidth



Mod wheels and pitch bends can, in fact, saturate MIDI channel, but a pianist repedaling would generate enough messages for getting the MIDI channel busy for a quarter of a second, and that would generate latencies. And yes, polyphonic aftertouch must be another source of problems... I don't be surprised to discover that devices with aftertouch do something to avoid jamming MIDI with messages, maybe limiting the number of steps (128 steps seem to much for aftertouch), or limiting the number of messages sent per unit of time.

I think the key word in that paragraph is "poytimbral." Yes, if you're running a sequencer and sending data on up to 16 channels simultaneously, and a number of those channels may also be employing continuous effects like aftertouch/wheels/pedals or sending lots of other data at superhuman speeds, it can choke. But I'm talking about just "live" playing, which just isn't nearly as demanding of MIDI as those kinds of applications, since you're basically limited to what your hands and feet can do in real time.

Keyboards are full of 128-step continuous controls (pedals, wheels, faders, knob, ribbons, aftertouch). It just doesn't seem like a continuous sustain pedal would create so much extra data, I'd just be surprised if this is where they drew the line in the sand and said, "no, not one more 128-step controller!"

As long as I can't hear it and it doesn't fatigue my ears, I'm fine with lossy compression. High end piano software companies are playing a bit of specmanship (I never thought I'd complain about this) when they brag about lossless compression and the size of their sample sets - though I probably would too if I were them.


I'm also fine with 16/44.1 samples, though once combined I'd like a bit more headroom. But otherwise I pretty much agree with you. Nord already has something that I think could give SN a run for it's money in the Nord Piano. But they would need to release a single piano sample set that would utilize all of the ~500MB sample memory.

I find it bizarre that tossing a couple of dollars of Flash in a DP, or merely putting a USB or SDHC connector on a DP, and allowing for sample sets in excess of 1 GB would blow every existing DP completely out of the water, including ones that cost many thousands of dollars and that are hyped specifically for their realistic piano sound.