If you spend any amount of time exploring the digital piano space, you’ll inevitably come across the terms ‘sampling’ and ‘modeling’ fairly often, whether reading down specs sheets or checking out full-length model reviews.
If you aren’t a professional gigging pianist or haven’t been an observer in this space for at least a moderate amount of time, labels such as these can be quite confusing at best. And for beginners, wading too far into specs sheets and owner’s manuals isn’t always fruitful.
With that in mind, in today’s article, we’re going to go into depth as to exactly what the terms sampling and modeling actually mean, and why this is an important spec to be aware of. We’ll be using the Roland FP-90X and Kawai ES920 portable digital pianos as examples of each.
If you’re looking for some clarity on this subject you’ve come to the right place!
Sampling vs Modeling – Background
How Does a Digital Piano Work?
The concept behind a digital piano is fairly straightforward – the goal is to recreate all of the sonic elements of the acoustic piano playing experience. For a long time, the only approach available was to take a real piano, record the piano sound note by note, and store those recorded samples in a computer so that the samples could be triggered by pressing a key.
Of course, being able to play the piano in a natural, dynamic way means that replaying just a single sample was never going to come close to capturing all of the nuances and possible permutations that make up the acoustic piano playing experience since harmonics behave very differently if it’s a single note being played, versus combinations of multiple notes.
Of course, the limitless number of different velocity pressures needs to be considered as well since the piano tone of a given note changes dramatically depending on how hard or soft you play.
4 Decades of Innovation
Challenges such as those mentioned above are what have fed the constant innovation and creativity of digital piano makers for 40-plus years now. In that time, the technology has come an incredibly long way, to the point that modern sample-based tone engines are able to produce incredibly convincing piano tone.
More recently, however, a new approach called modeling was developed, and instead of relying on a recorded sample, this approach generates the tone entirely from scratch in real time via a series of complex algorithms.
Roland digital pianos have led the way in this regard in the digital piano industry, but this approach is also very popular with software synthesizers, otherwise known as VSTs (Virtual Sound Technology.)
These are the two approaches to digital piano sound in use today, and while both approaches can produce excellent results. the sonic experience is often quite different, to the point that in some cases, each approach has its own very vocal adherents.
Let’s get into a little more depth regarding each approach, and now would be a good time to check out the companion video if you’d like to hear the differences for yourself.
Sampling
Kawai’s Harmonic Imaging XL Sound Technology
To demonstrate what a high-level sample-based engine is capable of, let’s take a closer look at the Kawai ES920, which uses Kawai‘s high-quality Harmonic Imaging XL Technology with 256-note polyphony.
This is an advanced sound engine that uses a combination of both sampling and synth technology added on top which serves to replicate many of the tricky harmonics that emerge when multiple notes ring out at the same time, such as damper resonance, string resonance, cabinet resonance, hammer noise, touch curve and many more.
Sample Layers
Sampling has evolved immensely, from a very small number of samples stretched across the range of the keyboard in the early days, to a stereo recording of each individual note at multiple velocity layers today thanks to modern processing power and memory storage. And in the most extreme instances, you can even find multidimensional, multi-channel sampling which means that the sample has been taken from several microphone positions around the piano.
VSTs on computers take this even further simply because the sample size and sample bank can be significantly larger than anything currently on a digital piano. The most luxurious of VSTs can now get into many gigabytes worth of samples and require incredibly fast processors to run even by PC standards to run.
So while it simply isn’t possible to fit this level of processing power into a digital piano, we’re nonetheless still a very high level of quality achievable by sample-based engines in a digital piano, and the HIXL engine is a prime example.
Musical Impressions
To demo the HIXL sound engine we’ve gone with the default piano sound on the ES920, which is the Shigeru Kawai SK-EX Concert Grand piano sample. This sample and other instruments’ sounds are mated very nicely to the Responsive Hammer III key action (RHIII) with let-off and ivory touch key surfaces allowing for great expressiveness from pianissimo to forte.
Right away you can hear great tonal variety across this sample, with the sound behaving differently across the ranges as it does on an acoustic piano. The sustain is something that’s traditionally been notoriously hard for digital piano manufacturers to convincingly recreate, but in this case, the decay sounds very natural and life-like.
The resonances that a piano creates on its own that are very difficult to simulate such as cabinet resonance, string resonance, damper resonance and more are all produced and added via a synthesis engine. Yamaha (Yamaha P-515) and Casio (Casio PX-S7000) have taken a similar approach in this regard.
In this case, Kawai allows quite a bit of access to the synthesis engine courtesy of the Virtual Technician feature. Thanks to Bluetooth MIDI connectivity, you can access the app (iOS and Android) on your phone and edit 21 of these parameters to really customize the tone. This can be done directly on the instrument as well via the control panel and OLED screen.
As you go down in price points these extra synthetic elements tend to disappear, and the sample sizes also shrink and are generally less complex. But paired with the high-end built-in speakers here there’s no question that this sample-based engine and sound system is producing quality tone.
Modeling
Roland’s PureAcoustic Piano Modeling
We’ll shift over to the Roland FP-90X stage piano and its PureAcoustic Ambience modeling engine with limitless polyphony, which has fairly recently replaced the modeling version of their famous SuperNATURAL engine, though a sample-based version of this engine can still be found on the FP-30X and FP-60X (both equipped with the PHA4 action).
Now, this is a great example at a similar price point in the market where you can actually start to choose between a sample-based digital piano and a modeling-based digital piano.
From a practical perspective, the key distinction between a sampling vs modeling engine is that the modeling engine actually generates the sound in real-time from scratch, and 90X has some fun extras such as a mic input.
One thing to note however is that the engine reverts to sampling when on non-acoustic piano tones such as those from the e-piano category (electric piano), harpsichord, etc.
Editability
When this technology first came out, most examples of it generally offered an incredible level of customization because the user was given access to the algorithms that control all of the various parameters. For example, with the great VST Pianoteq (an early innovator in this area) virtually every single aspect of the piano can be manipulated.
With the PureAcoustic engine and the Piano Designer feature just some of the editable parameters include lid height, key-off noise, soundboard types and various resonance effects. There’s also the My Stage feature which offers a variety of preset configurations. The 90X also has Bluetooth connectors so you can access apps like this and the Roland Piano App (replaces Piano Every Day) wirelessly instead of operating from the LCD screen.
But one of the major complaints with some of the earlier modeling VSTs was that they were too clean and perfect sounding to the point that many of the natural imperfections and anomalies that occur with an acoustic piano were absent, resulting in an obviously synthetic, sterile sound.
Newer modeling engines, including the PureAcoustic engine, have made serious strides in terms of realism by adding some of these natural imperfections and anomalies.
Musical Impressions
When it comes to the sustain on the core grand piano sound, there’s an interesting sort of subtle undulation in the decay that really adds to the authenticity and was missing from earlier versions.
There’s a wonderful range of tonal variety as we go through the different velocity layers, and this is without messing around with the onboard equalizer sliders.
The extreme ranges of the piano really seem to tax modeled tone engines quite heavily as getting the right combination of growl, clarity, and fullness in the bass has been historically very difficult to pull off, and the treble is even tougher with all of the shimmering harmonics. The combination of overtones and attack is just very intricate, but kudos to Roland for really doing a great job here.
The way the tone engine is paired with the speaker system (2 mains, 2 tweeters and 60-watt amplifier) and PHA-50 Hybrid Keyboard hammer action (has escapement and adjustable touch sensitivity for extra subtlety) is also pulled off really well, and this is an area that fans of modelling engines like to highlight since it feels really connective to generate tone in real-time as opposed to triggering a sample.
Roland’s RPU-3 pedal unit (soft, sostenuto and damper pedals) and stylish KPD-90 pedal board (requires the KSC-90 keyboard stand to function) also interact really well with the sound engine, especially when playing with headphones thanks to the Headphones Acoustic Projection effect.
Blindfold Test
At this point, despite hearing very significant differences between these sound engines, we wanted to test whether these perceived differences were the result of genuine musical differences, or the result of preconceived biases.
With a set of decent quality cans plugged into a headphone jack on a mixer, we decided to randomly playback various WAV recordings from both pianos to see if our Chief piano reviewer Stu can tell which one was which. Each of the tone engines was on default settings without added reverb or adjusted band EQ and was recorded direct instead of through their amps.
This proved to be no problem for him as he scored perfectly. The SK-EX sample has a particularly resonant lower mid-range just like a real acoustic SK-EX does as well, so it’s possible his perception of this was a key giveaway.
The PureAcoustic engine is so even in terms of its behavior throughout the various ranges, even though it also happens to be quite dynamic.
Closing Thoughts
We hope you’ve enjoyed this quick explanation and discussion about the differences between sampling and modeling.
From a quality perspective, it’s not fair to say one approach is hands down superior to the other, but since they are quite different, there’s a good chance you may find yourself with an obvious preference.
If you’d like a more in-depth look at either of these pianos where we dive deeper into their keyboard actions, piano features (like split, metronome, rhythm section accompaniment), Bluetooth Audio and more, check out our channel!