2017 was our 20th year online!

Welcome to the Piano World Piano Forums
Over 3 million posts about pianos, digital pianos, and all types of keyboard instruments.
Over 100,000 members from around the world.
Join the World's Largest Community of Piano Lovers (it's free)
It's Fun to Play the Piano ... Please Pass It On!

Shop our online store for music lovers
SEARCH
Piano Forums & Piano World
(ad)
Best of Piano Buyer
 Best of Piano Buyer
(ad)
Faust Harrison Pianos
Faust Harrison 100+ Steinway pianos
(ad)
Wessell Nickel & Gross
PianoForAll
Who's Online Now
99 members (Amadeus M., Anglagard44, brdwyguy, anotherscott, ando, 26 invisible), 1,068 guests, and 447 robots.
Key: Admin, Global Mod, Mod
(ad)
Estonia Pianos
Estonia Pianos
Previous Thread
Next Thread
Print Thread
Hop To
Page 12 of 23 1 2 10 11 12 13 14 22 23
Joined: Dec 2009
Posts: 4,675
4000 Post Club Member
Offline
4000 Post Club Member
Joined: Dec 2009
Posts: 4,675
Yes, we've been over this before.

You don't want good random access performance. You want good streaming access performance because 99% of the time you want the next sample. Buffering requirements are quite minimal even for lots of polyphony (I've shown this previously as well). I imagine even the OS can stream to RAM / cache and run from there efficiently. It's all about cache lines, real processors don't run directly out of ROM anymore. DRAM is peanuts and you don't need that much.

Honestly, there's no way we could have possibly had DPs in the 80s if the requirements were at all onerous today. It doesn't pass the smell test.

Joined: Feb 2010
Posts: 6,125
A
6000 Post Club Member
Online Content
6000 Post Club Member
A
Joined: Feb 2010
Posts: 6,125
Originally Posted by dewster
Yes, we've been over this before.

You don't want good random access performance.

Random access per se isn't the issue. The way I see it (and perhaps this is imprecise) it's more the ability of the processor to "see" the data as being in "memory" as opposed to on a "storage" device. Basically, it's the XiP (execute in place) functionality I'm talking about... that's not the same thing as random access, but random access is a prerequisute for XiP. I admit I'm at the edge of my knowledge here though, and may not be putting things quite right.

Originally Posted by dewster
Honestly, there's no way we could have possibly had DPs in the 80s if the requirements were at all onerous today. It doesn't pass the smell test.

The smell test? How scientific! ;-) Seriously, it's not a mystery at all. The piano samples were small. The Korg M1 (one of the first sample-based pianos that people found genuinely usable) had all its sounds (piano and everything else) in 4 megabytes of ROM.

Then as now, all you need for a sampled piano is enough memory (not storage... memory!) to hold it, whether that be ROM, RAM (loaded up at each boot), or (more recently) NOR flash. Except that now--at least with Linux/Windows/OSX--we can also get by with a smaller amount of RAM and can stream the rest as needed through fast transfers in and out of RAM, with any storage device that can keep up (and a sufficiently fast processor). But you still need to move the data from the storage device into RAM, you can't "execute in place" from a storage device (hard drive, SSD, NAND flash) as you can from a "memory" device (RAM, ROM, NOR flash). So if you're using NAND, you also need a bunch of RAM, and an OS that manages the real-time data swapping (and a processor that can handle it). You can't simply add $5 of flash to an existing DP design and suddenly have a multi-gig piano.

Yeah, I know, this is what I've always said, and you disagree. It's just that I don't see a rationale for your disagreement beyond what amounts to a gut feeling. The data sheet didn't support it. Your "smell test" or simply what you think should be "incredibly obvious" does not support it. It doesn't exist in any real world form, and you don't have the expertise to design it, and no one who has ever designed a piano has thought of it. But we're supposed to believe that the problem is just one of, what, incompetent engineers at every single manufacturer? Or conspiracies among the manufacturers to not rock the boat? I don't know, but that's not passing the smell test for me. ;-)

The good news is, compared to when we first talked about this 3 (or more?) years ago, NAND flash has gotten faster and cheaper, and an OS like Linux can run more quickly on cheaper hardware as well. Maybe building something off that odroid platform will work, or if not, before long, maybe something similar. But this doesn't mean that your idea for a dirt cheap implementation works... it means that the ability to do it the old, expensive way is getting cheaper. ;-)


Joined: Mar 2010
Posts: 614
K
500 Post Club Member
Offline
500 Post Club Member
K
Joined: Mar 2010
Posts: 614
Okay I'll wade into these waters. To create a loopless sample based piano. You need a CPU that can access large memory (at least 32-bit minmum), enough RAM memory to hold the samples while playing, non-volatile storage and as close to a realtime OS as possible.

Only relatively recently have we had the last part that is affordable. Realtime OS's were the domain of medical equipment, DOD weapons systems and such. They were very expensive. With open source and linux, things changed. The reason you want a realtime OS, is apparent to anyone who uses a laptop to gig. First you boot the laptop, then load the program shutoff anything you can think of that might screw things up and hope the latency isn't too bad. And sometimes when you use it just starts making crackling noises, freezes or some other anomaly until you reboot or restart the program. That's how it is with an all purpose computing machine, it does a lot of things well but its not perfect. Only mostly all good.

For a digital piano you want the finger to ear to be immediate, your boot time as short as possible and it to perform exactly the same way every time you use it. Yes you need to access from RAM, the CPU cache is the first place the CPU looks for instructions and data. Next it goes to RAM and loads the CPU cache from that, last it goes to non-volatile storage (flash, disk, etc). But it always in that order, each transfer takes time you are moving from the slowest to the fastest storage to keep feeding that CPU's Arithmetic Logic Unit the part that does the actual work. Then the results are transferred either back to memory or more likley its I/O section to send the digital audio to its sound modules DAC (digital to analog converters) and finally the amp and speaker. The sound module also has a cache that keeps a queue of audio going to DAC.

In current models, instead of all that multipurpose computing hardware manufacturers use highly specialized DSPs. These are single purpose chips designed to one thing and one thing only. These can run extremely fast and will do that one thing very well. Usually that thing is produce spectacular audio, at least the best the designers can make it. It will have effects processing, high quality DACs and generally be the best they can make within a price point. These chips come with their own advantages and disadvantages and requirements. The cost of these chips like everything else continues to get cheaper while also getting more powerful. Until Moore's law finally looses to physics.

Now given that general purpose computers are more powerful now than ever, and we have an off the shelf real-time OS, and storage is cheaper. It is now possible to take these, marry them to some high quality DSP components and the result is the Korg Kronos and very likely Yamaha Tyros 5 according to some sources.

That is a far cry from a $1k or less digital piano. And all the assumptions, and I guess smelling won't make it so. So when will we have it? When a company believes it can make it at a profit and not before. There is no grand conspiracy behind it. The Illuminati of piano manufactures doesn't exist.

Roland and Yamaha haven't invested all the money and research into instrument modeling technology when all they needed was longer samples. Its not as easy as you assume.

I am not a digital piano designer. I work on space systems chiefly spaced based remote sensing systems. I don't know all the reasoning behind manufacturers design considerations but I am well versed in state of the art computational systems. And cost concerns even when those costs run in the billions.

Last edited by Kbeaumont; 07/21/15 10:48 AM.

A long long time ago, I can still remember
How that music used to make me smile....
Joined: Dec 2009
Posts: 4,675
4000 Post Club Member
Offline
4000 Post Club Member
Joined: Dec 2009
Posts: 4,675
I give up, the "can't do" attitude around here has overwhelmed me.

Everyone back to their caves. Maybe we can at least figure out this thing called "fire" if the gods aren't too mad at us.

Joined: Nov 2013
Posts: 1,174
1000 Post Club Member
Offline
1000 Post Club Member
Joined: Nov 2013
Posts: 1,174
Originally Posted by ElmerJFudd



Nice link. I preferred #1...
was sure that was going to be the Nord, but it's the Casio apparently. Well done Casio!


Kawai CA95 / Steinberg UR22 / Sony MDR-7506 / Pianoteq Stage + Grotrian, Bluethner / Galaxy Vintage D / CFX Lite
In the loft: Roland FP3 / Tannoy Reveal Active / K&M 18810
Joined: Aug 2012
Posts: 736
500 Post Club Member
Offline
500 Post Club Member
Joined: Aug 2012
Posts: 736
Originally Posted by Kbeaumont
Okay I'll wade into these waters. To create a loopless sample based piano. You need a CPU that can access large memory (at least 32-bit minmum), enough RAM memory to hold the samples while playing, non-volatile storage and as close to a realtime OS as possible.

Only relatively recently have we had the last part that is affordable. Realtime OS's were the domain of medical equipment, DOD weapons systems and such. They were very expensive. With open source and linux, things changed. The reason you want a realtime OS, is apparent to anyone who uses a laptop to gig. First you boot the laptop, then load the program shutoff anything you can think of that might screw things up and hope the latency isn't too bad. And sometimes when you use it just starts making crackling noises, freezes or some other anomaly until you reboot or restart the program. That's how it is with an all purpose computing machine, it does a lot of things well but its not perfect. Only mostly all good.

For a digital piano you want the finger to ear to be immediate, your boot time as short as possible and it to perform exactly the same way every time you use it. Yes you need to access from RAM, the CPU cache is the first place the CPU looks for instructions and data. Next it goes to RAM and loads the CPU cache from that, last it goes to non-volatile storage (flash, disk, etc). But it always in that order, each transfer takes time you are moving from the slowest to the fastest storage to keep feeding that CPU's Arithmetic Logic Unit the part that does the actual work. Then the results are transferred either back to memory or more likley its I/O section to send the digital audio to its sound modules DAC (digital to analog converters) and finally the amp and speaker. The sound module also has a cache that keeps a queue of audio going to DAC.

In current models, instead of all that multipurpose computing hardware manufacturers use highly specialized DSPs. These are single purpose chips designed to one thing and one thing only. These can run extremely fast and will do that one thing very well. Usually that thing is produce spectacular audio, at least the best the designers can make it. It will have effects processing, high quality DACs and generally be the best they can make within a price point. These chips come with their own advantages and disadvantages and requirements. The cost of these chips like everything else continues to get cheaper while also getting more powerful. Until Moore's law finally looses to physics.

Now given that general purpose computers are more powerful now than ever, and we have an off the shelf real-time OS, and storage is cheaper. It is now possible to take these, marry them to some high quality DSP components and the result is the Korg Kronos and very likely Yamaha Tyros 5 according to some sources.

That is a far cry from a $1k or less digital piano. And all the assumptions, and I guess smelling won't make it so. So when will we have it? When a company believes it can make it at a profit and not before. There is no grand conspiracy behind it. The Illuminati of piano manufactures doesn't exist.

Roland and Yamaha haven't invested all the money and research into instrument modeling technology when all they needed was longer samples. Its not as easy as you assume.

I am not a digital piano designer. I work on space systems chiefly spaced based remote sensing systems. I don't know all the reasoning behind manufacturers design considerations but I am well versed in state of the art computational systems. And cost concerns even when those costs run in the billions.


Wow -- Thanks for this. Very well thought out. The part about the specialized DSP chips was spot on. There are very complex sound physics at play going inside an acoustic piano and things like string/damper resonance, lid simulations, velocity smoothing, and reverb, can not be done by a simple loopless sample playback system. It would sound dull and lifeless.

To try to bring this back to the actual topic, Casio uses something they call an AiR sound source to describe their specialized DSP chip.

http://www.casio-europe.com/euro/emi/airspecial/air/

http://www.gear4music.com/news/article/Casios-AiR-Technology-Explained/62W/2013-07-26


Joined: Apr 2015
Posts: 588
500 Post Club Member
Offline
500 Post Club Member
Joined: Apr 2015
Posts: 588
Originally Posted by anotherscott
... I still see no evidence of being able to play the samples directly from cheap NAND flash.


I think thats what the Kurzweil Forte does... atleast thats what i understood from it ..

Joined: Jan 2010
Posts: 1,627
1000 Post Club Member
Offline
1000 Post Club Member
Joined: Jan 2010
Posts: 1,627
Originally Posted by dewster
I give up, the "can't do" attitude around here has overwhelmed me.

Everyone back to their caves. Maybe we can at least figure out this thing called "fire" if the gods aren't too mad at us.


Thank goodness you're done, Dewster. Because the argument is over until a prototype appears. Delivering full length, uncompressed, unlooped, unstretched, multigigabyte samples at low latency, high polyphony, and high sample rate is not easy.

What we know is we can build a PC that will do this with over the counter parts, using OSX, Linux, or Windows, Intel or AMD CPU, spinning drive or SSD. It's expensive, unattractive, noisy (needs a fan(s)) or pricey fanless cooling build. Third party software from Synthogy, or VI Labs or something. It's not a compete instrument (still needs a controller keyboard). It's not a viable/marketable/profitable instrument/product.

Korgs sells their Kronos at over $3000. It runs hot, needs a fan, uses SSD which is susceptible to failure, has slow boot time. Biggest complaint is its a PC in a keyboard's box.

Muse Receptor can do this. It IS a PC in a box. You still have to buy software pianos and plugins. It's not a viable design to base a digital piano around.

The hardware in Apple's new Mac Book is coming very close. As is the hardware in the iPad Air 2. It has a street value of $600 with 64gb of storage, which you would need if you planned on offering a few 16-20gb sample libraries. Basing a DP around this type of hardware is much more expensive than what DP companies are doing at the moment and they sell a fraction of what Apple sells in laptops or tablets. This isn't viable yet.

Modelling like the V-Piano is much more feasible, and that's why we actually see a V-Piano and Physis Piano as products. Neither of them sell in large quantities and are expensive.



Joined: Aug 2012
Posts: 736
500 Post Club Member
Offline
500 Post Club Member
Joined: Aug 2012
Posts: 736
Originally Posted by Bachus
Originally Posted by anotherscott
... I still see no evidence of being able to play the samples directly from cheap NAND flash.


I think thats what the Kurzweil Forte does... atleast thats what i understood from it ..


The samples are not directly played from the flash. They are fed into a custom DSP (Mara) chip for processing. The Forte actually has three of these Mara chips: one for the sound engine, one for effects, and one for the organ.

So back to the conspiracy theories... eek

Joined: Feb 2010
Posts: 6,125
A
6000 Post Club Member
Online Content
6000 Post Club Member
A
Joined: Feb 2010
Posts: 6,125
Originally Posted by Scott Hamlin
Originally Posted by Bachus
Originally Posted by anotherscott
... I still see no evidence of being able to play the samples directly from cheap NAND flash.


I think thats what the Kurzweil Forte does... atleast thats what i understood from it ..


The samples are not directly played from the flash.

Correct. A bunch of discussion about it at
http://forums.musicplayer.com/ubbthreads.php/topics/2579835/1

Joined: Oct 2012
Posts: 608
D
500 Post Club Member
Offline
500 Post Club Member
D
Joined: Oct 2012
Posts: 608
Originally Posted by ElmerJFudd
Delivering full length, uncompressed, unlooped, unstretched, multigigabyte samples at low latency, high polyphony, and high sample rate is not easy.


good

hope they can now concentrate on faster CPUs (or even faster and cheap GPUs) for physical modelling solutions to sound synthesis

and more accurate actions :thumbup:


unlocked by keys
wordless poetry sings free
- piano music -
Joined: Aug 2012
Posts: 117
P
Full Member
Offline
Full Member
P
Joined: Aug 2012
Posts: 117
Originally Posted by Doritos Flavoured
Originally Posted by ElmerJFudd
Delivering full length, uncompressed, unlooped, unstretched, multigigabyte samples at low latency, high polyphony, and high sample rate is not easy.


good

hope they can now concentrate on faster CPUs (or even faster and cheap GPUs) for physical modelling solutions to sound synthesis

and more accurate actions :thumbup:


Thumbs up to more accurate actions.


"But its got a crap keyboard action Dave ... no amount of great sounds help that."
Dr. Popper

Piano Student at University of São Paulo - Ribeirão Preto
Music Education Major Diploma at Music Department of FFCLRP - University of São Paulo
Ronisch Upright, P140 and DGX-670

Joined: Dec 2009
Posts: 4,675
4000 Post Club Member
Offline
4000 Post Club Member
Joined: Dec 2009
Posts: 4,675
Originally Posted by ElmerJFudd
Delivering full length, uncompressed, unlooped, unstretched, multigigabyte samples at low latency, high polyphony, and high sample rate is not easy.

Actually, once you've got everything in place to do looped with good sympathetic resonance, it's a pretty small step to fully sampled, particularly with a 32 bit address space. Looping introduces algorithmic complexity that straight sample playback doesn't require.

Once you have a 32 bit processor running in the hundreds of MHz you have something like 10,000 instructions per sample time. And you can bunch 50 or so samples (~1ms) together for further processing savings (loading filter coefficients, etc.). You could probably even employ the video core to do a lot of the calculations, making it even more doable / reasonable.

Really strange vibes around this issue - it's like everyone here has a huge bet riding on it not being the case. People are pointing to workstations to prove inexpensive DPs are impossible, and acting like modern GPUs don't have DSP extensions and SSDs haven't come down the pike yet.

Joined: Jan 2010
Posts: 1,627
1000 Post Club Member
Offline
1000 Post Club Member
Joined: Jan 2010
Posts: 1,627
Lol. Incorrigible, Dewster.
Please build the prototype. I implore you.
The digital piano community is yearning to see your simple design come to life!

Joined: Jan 2010
Posts: 1,627
1000 Post Club Member
Offline
1000 Post Club Member
Joined: Jan 2010
Posts: 1,627
Originally Posted by dewster
Originally Posted by ElmerJFudd
Delivering full length, uncompressed, unlooped, unstretched, multigigabyte samples at low latency, high polyphony, and high sample rate is not easy.

Actually, once you've got everything in place to do looped with good sympathetic resonance, it's a pretty small step to fully sampled, particularly with a 32 bit address space. Looping introduces algorithmic complexity that straight sample playback doesn't require.

Once you have a 32 bit processor running in the hundreds of MHz you have something like 10,000 instructions per sample time. And you can bunch 50 or so samples (~1ms) together for further processing savings (loading filter coefficients, etc.).

Really strange vibes around this issue - it's like everyone here has a huge bet riding on it not being the case.


Can you explain why my dual 1ghz G4 with WD raptor 10k spinning drive and 4gb of RAM from a few years back could not effectively perform a large piano sample library in Kontakt or EXS24 at sufficiently low latency and polyphony to be considered an option?

Joined: Dec 2009
Posts: 4,675
4000 Post Club Member
Offline
4000 Post Club Member
Joined: Dec 2009
Posts: 4,675
Originally Posted by ElmerJFudd
Can you explain why my dual 1ghz G4 with WD raptor 10k spinning drive and 4gb of RAM from a few years back could not effectively perform a large piano sample library in Kontakt or EXS24 at sufficiently low latency and polyphony to be considered an option?

Any number of issues, though the main ones are likely:
1. Non-RTOS OS struggling to meet minimum QOS timing.
2. Gobs of crap software between the application and the metal.
3. Poorly written application.

It's been my experience that most code is pretty horrible, and that many software types don't really know or care about what's going on in terms of hardware. They spend their days writing fancy constructs that don't generalize well. It takes a HW engineer to efficiently employ hardware, and many of them are pretty lame too. Not that I'm god's gift, but at least I try to put in the time to understand and implement well the things I get involved with.

Joined: Nov 2012
Posts: 860
K
500 Post Club Member
Offline
500 Post Club Member
K
Joined: Nov 2012
Posts: 860
I have an idea for the perfect piano:

1. Do a cabinet and put the two 16-cm subwoofers and two tweeter plus two midranges. Put the tweeters in the tubes.
2. Put there a top action - like GFII / PHAIV Concert or NWX. Don't do upright cabinets. They are too big. Current like CLP575 / CA97 / 508 are great. It's good for small flats.
3. Put inside Pianoteq or Top sampled piano, or best, blend it like a Roland (what I though do not like about the Roland, is that sometimes is sounds not like a piano at all... frown especially the newst one, I prefer much more their old HP5x3/5/7 series sound)
4. Give countinous pedaling.
5. Give an operating system to tweak the sound / kind of virtual technian.
6. Put microphone, few organs and other piano sounds, metronome, usb, line in /out, phones out and that's all. We do not need 300 sounds and other fancy useless stuff. We want a piano.

Surprisingly, it's really hard to match that. Instead of producing better pianos, in price we get often only lots of marketing, or we are paying 1000 USD more for two speakers more which costs with assembly not more than 100USD....

And WHAT I REALLY HATE is that both Roland and Kawai are doing their pianos sound straight and bright. My current best choice is Boesendorfer put in Clavinovas. Why it's so hard to the the dark german grand as a standard piano?

Dear ROland, please come back to your previous sound character!
Or Casio... go and to this. It's really simple. and, as I pianist, I really DO NOT WANT in a PIANO a big colorful DISTRACTING SCREEN. the side panel in Kawai and Yamaha is a really great thing.

And work on your desing, it really looks like a cheap piano from Korea, compared to others. Don't get me bad. I just want a perfect DP to be made smile

Last edited by kapelli; 07/21/15 05:57 PM.
Joined: Aug 2012
Posts: 117
P
Full Member
Offline
Full Member
P
Joined: Aug 2012
Posts: 117
Oh my gosh, this thread is not about the Casio Digital Pianos anymore. Is about high tech applied to Digital Piano design and construction. Ok. How about talking about it on a specific thread about it?


"But its got a crap keyboard action Dave ... no amount of great sounds help that."
Dr. Popper

Piano Student at University of São Paulo - Ribeirão Preto
Music Education Major Diploma at Music Department of FFCLRP - University of São Paulo
Ronisch Upright, P140 and DGX-670

Joined: Dec 2009
Posts: 4,675
4000 Post Club Member
Offline
4000 Post Club Member
Joined: Dec 2009
Posts: 4,675
I actually think it's central to Casio Digital Pianos, but I'll try to shut up for now.

Joined: Nov 2012
Posts: 860
K
500 Post Club Member
Offline
500 Post Club Member
K
Joined: Nov 2012
Posts: 860
Originally Posted by Pedro_Henrique
Oh my gosh, this thread is not about the Casio Digital Pianos anymore. Is about high tech applied to Digital Piano design and construction. Ok. How about talking about it on a specific thread about it?


It's about Casio still. It's because Casio did some step forwards here in the new line of pianos. This is not new in the workingstations or some high advanced synths, but in DP's and the way they did it is really interesting.

Page 12 of 23 1 2 10 11 12 13 14 22 23

Moderated by  Piano World 

Link Copied to Clipboard
(ad)
Pianoteq
Steinway Spiro Layering
(ad)
PianoDisc

PianoDisc
(ad)
Piano Life Saver - Dampp Chaser
Dampp Chaser Piano Life Saver
(ad)
Mason & Hamlin Pianos
New Topics - Multiple Forums
Kawai CA93 vs Kawai CA67
by talamundele - 09/21/21 11:57 AM
Viscount Physis V100 and G1000
by David Izquierdo - 09/21/21 10:50 AM
Yamaha CLP 575 VS Kawai Ca48
by Amadeus M. - 09/21/21 09:16 AM
Charles Walter
by Judise - 09/21/21 09:12 AM
Pronunciation .... lol
by cygnusdei - 09/21/21 07:45 AM
Download Sheet Music
Virtual Sheet Music - Classical Sheet Music Downloads
Forum Statistics
Forums42
Topics209,223
Posts3,134,023
Members102,788
Most Online15,252
Mar 21st, 2010
Please Support Our Advertisers

Faust Harrison 100+ Steinways

Dampp Chaser Piano Life Saver

 Best of Piano Buyer

PianoTeq Bechstein
Visit our online store for gifts for music lovers

Virtual Sheet Music - Classical Sheet Music Downloads



 
Help keep the forums up and running with a donation, any amount is appreciated!
Or by becoming a Subscribing member! Thank-you.
Donate   Subscribe
 
Our Piano Related Classified Ads
| Dealers | Tuners | Lessons | Movers | Restorations | Pianos For Sale | Sell Your Piano |

Advertise on Piano World
| Subscribe | Piano World | PianoSupplies.com | Advertise on Piano World |
| |Contact | Privacy | Legal | About Us | Site Map | Free Newsletter | MapleStreetMusicShop.com - Our store in Cornish Maine


© copyright 1997 - 2021 Piano World ® all rights reserved
No part of this site may be reproduced without prior written permission
Powered by UBB.threads™ PHP Forum Software 7.7.5