Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Why don't manufacturers record "bad" piano sounds?

My teacher always told me that the problem with digitals is that whatever key in whatever manner you press, it always gives you "good" sound. While on acoustic your sound can be "bad" - too thin, lack in timbre and overtones if you press a key to soft in a wrong manner, or it can be too harsh and knocking if you hit the key to hard and fast (like Heinrich Neuhaus wrote in his book about these two extremities - "not the sound yet" and "already not the sound"). So, for DPs to be more realistic, all these "bad" sounds should be sampled top, I guess. Do any manufacturers do it? Or they consider this as a flaw and thus create DPs with no "bad" sounds - flawless instruments?

I imagine it would be difficult to sell it:

New upgraded version! In this new version we have introduced truly bad sound samples, which will make anyone cringe. You have to play very accurately to avoid them

But then again, Pianoteq does this; they have a slider going from pristine condition to totally wrecked piano, and this seems to increase realism.

But the question remains: when and how should the "bad sounds" be triggered?

The manufacturers are trying to replicate---for most piano sounds---the concert version. That includes resonances and slight detuned, but not poor quality tuning etc. Obviously, sounds like Honky Tonk piano can be recorded on beat up pianos.

You can create flaws modelled pianos if that's what you're after, however, humans are pretty bad at guessing random patterns.

There is one brand focused on imperfect samples… and it is named Imperfect Samples !

But the last time I have tried one, there were huge velocity layer gaps…. Too imperfect for me !

I don’t know if they produce virtual pianos with imperfect samples as described.

It would be interesting to know what should trigger a bad sample. If it is the combination of acceleration and velocity, this won’t be possible since all digital pianos only measure the velocity (excepted the latest Clavinova which measure acceleration too).

Some brands and models can have the volume and timing of individual notes changed via an app. Yamaha comes to mind here. Not quite the same thing perhaps as the OP wants but you can make a honky tonk piano or just a badly tuned upright. Also some poor eq settings help go some way to creating a bad sounding piano IMHO.

If I wanted a bad sounding piano, I would just get a second hand acoustic.(which I had the first 5 years of playing)

I certainly did not buy my DP for it to sound like crap.

I would not be interested in such a thing.


Ron

I think this is due to restricted digitisers in the digital key action, typically merely 2 or 3 on/off switches, all dynamics deduced from the timing between them (some DPs do use different sensors). In AP there is a lot more to position, force and acceleration. No this is not a matter of detuning, those detuned pianos are quite unlike the AP "imperfections". This is surely one reason why piano teachers prefer acoustic pianos. But I'm not sure if the market would reward more realism. Additionally, DPs are based on weak DSP chips, likely insufficient for more complex processing, and VSTs only accept simple MIDI input that cannot register untypical keystrokes.

String resonance with pedal is one thing that is included in better DPs, makes pedal abuse sound muddy, while on entry DPs this is absent or dialled down, because beginners mostly sound bad when playing the real thing. And the pedal in cheapos is an on/off switch rather than continuous.

There is "bad" and there is bad...

It is true that acoustic pianos, even well-tuned ones, react very sensitively to changes in the environment (room temperature, humidity, etc.) which gives them a kind of "life" not easily replicated in digitals. In fact my grand sounds slightly different every day, and changes its character very slightly with the weather, the seasons, etc. - before I call the tuner. It may even go ever so slightly out of tune and then back to better tuning by itself, depending on the conditions.

There are actually digital pianos that try to emulate *some* of this - but never all of it. Listen to the Nord Piano Library sounds, for instance. Most of them - and especially the XL versions - do contain some oddities that are intended to enhance realism. Pianoteq has a detuning parameter that you can set.

What they don't reproduce, so far, is the ever so slightly changing behaviour over time that certainly adds to our affection for real pianos.

The Vienna Imperial for example has 1200 samples per key (pressed ba a robot). Do you (or your teacher) really think you can make more different sounds of one key? My impression is only that there are so many bad pianos out there you get used to it. The sampled grands are really top grands. Of course, there is a big difference between sitting in front of a piano (radiation to all sides, natural reverberation of the room) and listening to a digital grand piano on loudspeakers. Much then depends on the quality of your equipment (and also the ideal position of the speakers in the room) etc etc.

I think that the replies so far do not address the original question... As far as I understand, PianoStartsat33 is asking why DP have less dynamic range than in a acoustic, so they are easier to play without jumpiness in volume from note to note.
There are several reasons for that: in first place speakers cannot easily achieve the range of decibels of an acoustic; second, many DP if you press the key fast or slow, they only change in volume not in timbre (ie, from mellow to metallic); thirdly, because many DP users actually prefer it because it allows for easier play...
This being said, if you don't want easier play, the first 2 problems can be overcome: the top DP have speakers that almost get the full decibel range, and if you use computer VSTs instead of the internal sounds, you will have almost the same control problems of an acoustic.

I am not sure nowadays digital pianos have only a single velocity layer (making the velocity only control the volume). I would think the control of the timbre is a common feature nowadays. However, we may miss a large range of timbre if the samples are recorded with too close velocities.

I spent many years with broken keys and untuned strings [usually weak block pins] in uprights and the one grand I could afford but sold because I moved frequently. IMO DP's are a blessing, maybe not for advanced concert pianists but for folk like me who can do moderately advanced grade 7 some 8 works. My VPC-1 and various very good vst's far outweigh the decades of acoustic instruments I could afford. Not to mention the convenience of not having to tune and regulate or pay to move.

I'm pretty sure the sound on a piano only depends on how fast you hit the keys, how you release them, and how you use the pedals. All the other aspects (like tuning, regulation) are "external" and you can't really influence them by playing. Most modern DPs have decent sampling of all these aspects, and if you want more detail you can always use a VST.

What your teacher is saying sounds like he is pretending you can control the tone of the instrument like on a violin. Can you explain how can anyone produce a "bad" sound if you play with the correct dynamic and articulation? I'm pretty sure almost all modern digitals respond correctly to dynamics and articulation.

Exactly my thoughts @vagfilm! I think (as it often sadly happens) the discussion drifts in a totally different direction irrelevant to the original question, and I find that frustrating in general and extremely frustrating when the initial question has lots of merit such as in this case. Anyway, I think there is more compared to what you said for why it is so, but I cannot point to all of it.


There is indeed a myth that the piano can respond like a violin and have different qualities which one need to learn to bring out. I believe there is a seed of truth in such a myth, but that's just a "gut feeling". You can find some discussions about that in the acoustic and musician subforums, more than here in the digital one. I have experienced such a thing in my playing of acoustic pianos (especially mine), and have been able to produce "good" or "bad" sound -- on the same playing session, without a tuning or change of weather in between. I have been unable to figure out why and how. With digital pianos, for me the sound is always "ok-ish" neither good nor bad. The most that I've been able to accomplish in this regard is with pianoteq, which I am indeed able to make it sound bad!! In fact, I am unable to make it sound "ok-ish" (let alone good!!) with my own playing, however I've heard it sound almost good when played by others (most notably, its embedded blues demo).

So, for what I know, my suggestion to @PianoStartsAt33 is to try pianoteq. The evaluation version is quite complete (with very few limitations) and its free, so you can experiment plenty with it, and perhaps have your teacher give it a try too.

Ask your teacher why do we tune acoustic pianos? and why a quality tuned acoustic piano sounds "good" every time you play it?

Overtones are bad and in order for your acoustic to sound good, you need to eliminate these overtones.

That's probably not his point though...

We have round-robin piano sampls (and instruments) that sound different and "good" every time you press a key.

In addition to that, Pianoteq is an exceptional VST that you can try and see how great it emulates all of these using math and not samples.

When it comes to hardware, we have many limitations:

- The most important one: real-time performance (I don't want to wait a few mins for gigabytes of samples to load)
- Many gigabytes of samples means it requires specifically designed hardware to have reasonable real-time performance
- Bandwidth and processing power issues

It's not an easy design. A sampled piano with multiple layers can easily take up 20-100 GB of space and that's only for 1 sound. Imagine if you have 10 piano voices. that can easily consume 1TB of data.

We are talking about the memory requirements of a PC but yet a simple OS that can fit in a very small micro-controller that can boot up in seconds, It's a very niche hardware.

Korg OASYS was kind of an example of such a device, 8000-9000 USD when initially released.

Expect a similar or even higher price tag for what you're asking

*** So I did some calculations:
OASYS was released in 2005, so in 2022 you should pay 12100 USD given that we didn't have COVID.

Given the situation right now I'd say sth around 13k would be an optimistic price for an instrument capable of putting all VST technologies included and be real-time.

I have other hobbies to follow and my MP7SE sounds darn good (haha)

100GB is the need of a multi-perspective virtual piano like Garritan CFX : 123GB for 7 perspectives. Vienna Imperial : 46GB for 3 perspectives. For these two, there are only 15GB to 17GB per perspective (despite 100 layers with Vienna Imperial). And some good virtual pianos are in the 5GB side. (Galaxy …).

Also, some digital pianos with 10 pianos have only 2 or 3 piano sample sets and use filtering to « create » other pianos.

Then I bet we don’t need 1TB of samples for a good digital piano with multiple layers and unlooped samples. Like Gewa, Korg (Krome, GrandStage…) propose.

I'm not sure if Garritan has round-robin samples (doubt it). So imagine you sample each layer multiple times to compensate for the repetitive playing... that easily multiplies the samples by the number of round-robin dedicated (randomly switching) layers.

So one set for Berlin, one for Humburg, one or two Japanese and then also include the uprights... the numbers easily go to 3 or 4 sample sets. Let's be optimistic:

50GB (which can be unrealistic) for each instrument. 2 germans and 2 Japanese and 2 uprights. 6*50 = 300GBs of samples.

You buy a laptop with 250-500GBs of SSD memory these days and that needs to be embedded in a microcontroller hardware architecture capable of addressing this much memory. Then it comes to the real-time sample switching and voice switching and it's a very long story.

Round Robin sampling itself is a cumbersome chore.

PMFJI --

+1 to Syld.

Your teacher is propagating a belief without any objective evidence to support it.

Basic physics says that the sound of a struck string is controlled by the piano design and mechanism (which doesn't change), and _the velocity of the hammer when it hits the string_.

. . . I've never heard anyone explain how it's affected by _how_ you press the key, to achieve that velocity.

At the _ending_ of a note -- when the key is released, the damper follows it down, onto the string. _That_ process, the player has control over. Similarly, the player's foot has physical control over the dampers, as they release and touch the strings.

. . . But at the _start_ of a note, the hammer is swinging freely, not connected to the key.

The string knows nothing of the hammer's history, only how fast it's moving when it hits the string.

If you want to test your teacher's statement, I can tell you how to do it. But frustration and anger are the likely result.

Some other problematic statements:


For the past 15 years or so, even entry-level DP's have changed their tone quality, along with the volume, according to the key velocity. You can see this in the "DPBSD" thread, which has spectra of lots of DP's, at varying key velocity. (Even synths can do this, by controlling the low-pass filter cutoff frequency from the MIDI velocity.)

If you want a one-generation-old example, find a Roland with SuperNatural sound generator, and strike a key with slowly-increasing velocity. You'll hear the high, dissonant harmonics come out as the volume increases.

. . . ????????

Without overtones, an instrument won't sound like a piano.

There are reasons why digital don't feel, or sound, "just like an acoustic piano". But IMHO, "they can't capture the subtleties of a player's touch" isn't one of them.

Although the review is now a little dated, when I was playing around with low-cost VSTs, I enjoyed Galaxy's "The Maverick" probably more than all the others. It's based on an old Bechstein, that definitely doesn't sound overly sanitized and new:

https://www.pianobuyer.com/article/six-software-pianos-under-150/

@Abdol

I said 2 or 3, you said 3 or 4… this is compatible ! With Yamaha : CFX, Bösenforfer Imperial, a Pop (the 3 AP I have on a N1X), or on a Casio GP : Berlin, Hamburg, Vienna. I think we can agree on 3 !! (Ok, 4 with Kawai : SK-5, EX, SK-EX, upright…) but not 10.

With a Galaxy (5GB) recorded twice to make it « round robined », this makes 3 or 4 x 2 x 5GB = 30 or 40GB. If we take the Vienna Imperial, the number of layers (100!) makes round robin not needed (it is hard to trigger the same sample twice), then 15GB x 3 or 4 = 45 or 60GB.

As I have indicated there are already unlooped digital piano (Gewa, Korg)… and Yamaha has some technology used in the Montage or the ModX which streams samples directly from a NAND ROM to the DSP trough a ONFI bus.

The only virtual piano I have which has round robin is Pearl Concert Grand… I think the extra samples would be better used as extra velocity layers than round robin. (I don’t miss it on other libraries).