Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Piano Sampling Question

The interesting thing to me is that evidence points to DP manufacturers skimping on the storage space, not CPU. They have very short samples with looping and blending, etc., instead of a great big samples that just stream with little further processing. Perhaps there is some issue of throughput. I don't know.

In this forum we are not DP designers, but it is hard to reconcile the progress in computers and software pianos to the apparent lack of progress in DP's. I actually wish I had a better idea of what the constraint really is. Presumably with four companies vying for the market, if there were low-hanging fruit it would have been plucked by now.

I have to think there must be some reason commodity hardware isn't as attractive an option as it seems to us.

Indeed it's impractical, for if Yamaha, Roland, or Kawai sold you a DP that sounded like one of them fancy software libraries in 2013, what would they sell to you in 2014?

I've often had this thought (I've also wondered if Yamaha and Kawai don't want digitals competing with their acoustics but that wouldn't explain Roland and Casio).

You may be right. But if so, it's an unstable equilibrium. Any company that deviated from that strategy and actually made a big technological leap would gain a huge chunk of market share at the expense of the others. They would have to coordinate to make the oligopoly work.

More likely in my mind, the problem is the consumers: not educated enough to know/care and comparison shop effectively. That being the case things like marketing and brand name are more important than the underlying tech.

I don't think we usually have any idea what CPU they are using.


It is happening. The Kronos, the Crumar, and the Receptor do exist.

I think the question people mean, then, is why isn't it happening more cheaply. What I've noted in the past is that streaming (or any method of seeing storage space as virtual RAM such that the entire data set for an operation doesn't have to be in directly accessible memory simultaneously) seems to require a sophisticated OS (i.e. Windows, Mac, Linux) and so then also the hardware those OSes require. I suppose a company could write their own OS with this functionality, but it is probably non-trivial and beyond the normal scope of their programming talents, and who knows if it would even run on notably less hardware anyway.

Anyway, if that's the case, it becomes easier to see why keyboard manufacturers can't do it that cheaply, between the fact that instrument companies can't build computers as cheaply as computer companies can, and that the specialty instrument market can't survive on the low profit margins of commodity products, and that an instrument buyer is not going to put up with the things that computer users often put up with (occasional glitches, possibly leading to need to reboot; operations that take too much time; sometimes having to fiddle with settings to get things to work right), so things do have to be optimized for the desired functionality even in a commodity-based system. That is, even Kronos, Crumar, and Receptor are doing more than tossing commodity components and stock OS into a box... though Crumar comes closest by using embedded XP and not having the system do anything beyond piano... and even they're not cheap, especially when you realize they're selling direct.

But maybe we'll see some advancement here at NAMM in a couple of weeks!


Not too many people buy a new piano every year, and I don't think these companies are looking at that as a strategy. They'll probably be happy if you want to buy another one five years from now... and no matter what they could possibly give you today, I bet they would be able to have something better 5 years from now!


I don't believe there is any coordination to limit technology. Apart from my general aversion to "conspiracy theory" thinking, I really think each company wants to do what they can to be at the head of the pack. And arguably, Korg made the technological leap you describe with the Kronos. (And in fact, that hasn't stopped people from buying Nords, Rolands, and Yamahas--even reasonably high priced ones--as they each still have their own unique advantages.) Also, the instrument companies do know that you can put a laptop on any MIDI keyboard, so they must recognize that that is part of their competition as well. And the fact that not everyone is happy with that approach is not simply a matter of it being in two pieces instead of one, so I think that whatever an instrument manufacturer does along those lines needs to address the downsides of the computer approach a little more thoroughly.

I mentioned elsewhere that it took Korg a year to get the Kronos to stream user samples (in addition to the ones they pre-designed for it), and that's with an OS and hardware that already inherently supports streaming. I just don't think this stuff is as simple as many people here seem to think it is, especially if you're looking for the rock-solid operation you'd expect in a pro instrument.

In part, what I'm saying is that I think Ivory et al work because, on the software side, Apple and Microsoft have already done much of the low-level "heavy lifting," so to speak, while also providing a relatively low-cost mass production hardware platform; and they also benefit from the fact that people are okay if they can't just fire it up and have it work perfectly out of the gate "first time, every time," because expectations are different on a non-dedicated system.

People do it all the time on Youtube, and we hear it as such .

My understanding is that the user may put any VST they like on the system. (it comes with a 7200rpm drive, so it should be able to stream samples, too). I think Crumar confirmed this in another thread some time ago. If you're referring to the config as sold though, then yes - I think that's correct.

Regarding the Kronos - do you happen to know whether it streams samples from the SSD in the sense that it has to pre-load the attacks? The reason I ask is that I've just noticed that fast SSDs are very VERY fast now - the first one I looked at can do about 100K operations per second (= 0.01ms per read). The Kronos spec says 100 voices(*), and 100 x 0.01ms is just 1ms. I don't know what latency the Kronos has when using the SSD, but if we could tolerate about 5ms, that leaves 80% of the audio buffer time for processing of those samples, which is a lot. This could mean that it doesn't really need to pre-load at all, which would make a simpler system design. Maybe back when it was designed SSDs were not nearly this fast though.

Greg.
p.s It says 100 "dual stereo voices". If those dual stereo voices are not always co-located in the SSD, then I would need to multiply the voice count by 4 (to get the 400 raw voices they quote), and in that case the calculation doesn't look as attractive. I suspect those dual-stereo voices would be packed together, though, and that's why they quote the true polyphony as 100. Otherwise, they would have simply said "400", without any qualification.

p.p.s When I say "pre-load" - I realise it may well just put the attacks in fast FLASH or something - not actually "load" the attacks.

Quite. Sooner or later someone’s going to blink and DPs will make the quantum jump we’ve all been craving. In any case, the marketers and retailers are unlikely to have a problem persuading some of us to buy a real strung piano for its organic form factor, its physical presence, its sound board and near-inimitable ‘sonic superiority’.

Even with the inevitable lag between the conception and realisation of a new DP in the market, I would be very surprised if we didn’t see something non-looping within the year. 16gb of solid state memory is now a drop in the ocean compared to overall costs.

"This could mean that it doesn't really need to pre-load at all, which would make a simpler system design"

I've no doubt that's true. Not only the read/write rates, SSD access times are absolutely blistering also.

I suppose it’ll be a few years before notebook and desktop HDDs give way more or less completely (apart from bulk storage and archiving) to SSD so software pianos have no choice but to pre-load for now.

I don't know if it's of any interest but I raided the piggy bank and bought a samsung 840 SSD last week and it's been a revelation, for the boot-up and – more importantly - loading into Kontakt. But this blew me away; prior to the SSD my acid test for setting a generous buffer - gliss the entire keyboard while holding sustain – would show a gradual climbing in Kontakt’s disk-usage meter so that after 2 or 3 seconds the bar was fully white = disk overloaded. The same test with the SSD and there’s not a flicker of movement, the meter registers nothing.

Completely disagree. My cheap 8 ys. old Linksys router has a Linux operating system with some elementary web server functionality. Hardly bigger than my pocket wallet, costed some 30$. I was involved in such projects previously for developing such dedicated HW/SW applications, with high reliability specifications - DPs are nothing special in this regard.

I would bet in all of these DPs is running under such an operating system already, probably a Linux derivate.

(I can remember we were astonished to find out that in 1999 to that time revolutionary seeming touch screen application for choosing meals and printing tickets in the staff restaurant of the bank central we worked at was a normal program task running under WindowsNT - a proprietary OS.)

This will now change rapidly I think.Manufacturers could explore halting back development because their targeted buyer in the classical instrument market was to a higher percentage a technically relatively conservative - not to say naive (older people without technical background and primary interests in computers). Completely different with digital photography, where development was some one decade ahead of DPs. But all of this will be changing rapidly. Facebook citizens are sophisticated enough to be able to get user of a multimedia solutions with a master-keyboard, which will become the major concurrency for DPs.

KAWAIs announced new MIDI master KB VPC is such a first sign for this kind of development.

I think even some small companies launching some HQ DPs from their garage could now enforce breaking the hegemony of the restrained development strategy of mainstream manufacturers.

Most of that makes sense. But I don't see "small companies" entering the market and "breaking the hegemony ... of mainstream manufacturers". The cost is prohibitive.

Mac, I don't think the engineering and manufacturing costs are that high at all. A small local player can bring out such a thing easily. A SW piano on the basis of a standard enginge (Kontakt) is a very small project. (You can measure it on their price, even HQ instruments cost <= 100$). A keybed on the industrial scale shouldn't cost much more, just as 40GB SSDs, cabinet is carpenter work, boxes aren't much special either...

Marketing and logistic advantage seems more to be the most prohibitive factor for newcomer.

Looks like the real-world access times are far greater - up around 0.1ms, not 0.01ms: http://www.tweaktown.com/reviews/5128/samsung_840_pro_128gb_ssd_review/index7.html And is it true that the Kronos uses an Atom processor to talk to the SSD? Hmmmm.

Greg.

Temperament: I'm talking about market-entry costs.

How can a newcomer company convince retailers to stock their products when those retailers already have showrooms full of established, well-selling products from the existing market players?

And how can a new company convince an investment banker that the firm can make a go of it? It takes a large investment to startup a company.

MIDI takes ~1ms to send a single note on, so 0.1ms is down in the noise, no?

I don't know exactly how this test works, but it seems time to random first byte, after which sequential reads should be lightening fast.

- for huge corporations, economies of scale and the ability to swiftly tool up a new production line are the main deterrent to small operators, I believe.

That's cheating. I want it to be able to bang out all 100 voices together, like my laptop does. I'm sure I've seen you complain about how slow standard MIDI is, too. Yes, though, if we allow the 0.1ms skew between notes, that makes it much easier.

Greg.

I’m always mystified by the ‘real world’ – I suspect my real world isn’t quite the same as anyone else’s.

I was curious about all this so I’ve just run a trial of HD tune pro for my two resident drives.

1) Bog standard Toshiba HDD, 500gb, 5400 rpm.
2) Samsung 840 ssd, 250gb (not the pro model)

Toshiba:
Max read (periphery of platter?) 76MB/sec
Min read 36.2MB/sec
Access time 17.6 ms

Samsung:
Read rate avg 173MB/sec more or less constant across the drive.
Access time 0.204ms

Coming back to my real world. My system drive is always rather clogged up with nonsense. From the HDD, boot-up used to be circa 3 minutes. With the SSD its 25 secs. I didn't bother to quantify the Kontakt test but I found the results to be stark and impressive.

On the other hand, an access time for the SSD of 0.2 ms is much less impressive than I imagined but, still, approx 80x faster than the HHD (assuming the test to be reliable).

Very true, indeed. We have seen how even big companies like Kawai have sometimes their handycaps because their world wide retail chain is smaller than that of their even bigger competitor.

But I will set up a Reaper/Kontakt/VintageD konfiguration for my friend on his laptop next week with a used Fatar keyboard, and I will be a locally operating direct competitor for all of the big players :-)

You and your friends Cuckos, Native Instruments and Galaxy Instruments to whom you will be paying fees to use their products.