I am wanting to learn the piano, and I think its best to start on a digital piano. How many notes should a decent digital piano need to place at once at the beginner up to intermediate level? I can afford 64 notes on the yamaha ydp140 and maybe i could buy something with 128 if i saved some more. thanks.

Hi ZXCZXC12345: Talking about polyphony the most the better but if you are thinking of solo piano, 64 polyphony is more than enough not only for beginner but also for advanced players. Pieces that would demand more than that are very few and far between. If however you are thinking of playing other instruments along the piano using the same digital, then go after larger amount you can afford. My digital has 64 and I have never had a situation where I miss or regret not having more. My level of playing is advanced. Am a fan of Bach, Mozart and Chopin and for my style having more polyphony would be welcome but not necessary. Hope this helps.

Yes, but don't forget that the total number of polyphonic voices has to include every note you might conceivably play during the time the sustain pedal is depressed. Even so, having more polyphonic voices than there are keys might be a bit over the top

For learning beginning piano, 64 is going to hold you over just fine. You'd probably get by on 32 as well if you're particularly budget conscious.

Polyphony is more important to people doing performances and people doing multiple tracks / multiple instruments / multiple people. 64 note polyphony means that 64 notes are sounding at the same time. When you exceed the maximum number, what happens is that the notes that were played first (and thus is the latter stages of their decay) get cut off prematurely.

Keep in mind that instruments with stereo piano samples use two notes of polyphony for each key played. So in reality on an instrument with 64 notes of polyphony a maximum of 32 notes would be available for a stereo piano sound. This number would be further reduced by layering sounds such as piano and strings.

i don't think the polyphony
number is that important. Since
restarting as an adult I've
bought five pianos: a quality
acoustic upright, apparently with
infinite polyphony; a Korg C-800
with 16 note polyphony; a Casio
AP-24 with 32 and an optional
64 voice--I couldn't tell the
difference and so I always just
used the default 32 note grand
piano voice; a Korg SP-250 with
60, and my current piano, a
Williams Overture with 64.

I played the same music on
all five pianos and didn't
see much difference in performance
between any of them. The
16 note Korg C-800 held its
own against all of them.

I've tried much more expensive
digitals with 128 poly, but
I didn't see enough difference
in performance to warrant buying
them.

The polyphony number is just one
element in the total design
of a particular digital piano,
and is not the whole story.
Buying a dp based solely on
the poly number is sort of
like buying a car based just
on the number of cylinders.
For example, an 8-cyl. American
station wagon is far superior to
a 4-cyl. BMW coupe, if you
went just by the number of
cylinders. But the BMW is
a much better driving machine
because of the overall design
package.

64 polyphony is not a problem. Less than that would be bad, though. On the other hand, if you're considering a YDP-140, I would try to upgrade to a YDP-160. The YDP-140 has Yamaha's GHS action, which is not very good. The YDP-160, which is otherwise similar, has the much better GHE action. This is much closer to an acoustic and will be better for learning and more enjoyable to play.

A couple years or so ago, I bought a Casio PX-575 which has 32-note polyphony, but on some tones the owners manual said it only has 10 note polyphony. I was layering multiple stereo tones with the pedal, and was often playing large chords, some of which involved me playing multiple notes with one finger - for example, a D7+9 chord, like C,D,F#,A,C+D+E in the left hand (and yes I can bend my thumb to hit 3 white keys at once) and A,C+D,F#,A,C+D in the right hand simultaneously, and this after having played a chromatic 88-note glissando with multiple stereo samples layered, and while the pedal is still held down. Notes were dropping out faster than my ears wanted to believe (for example, get barely 2 octaves through the glissando and they'd start dropping out, and with the aforementioned 14-note chord, I might only hear 3 or 4 notes sound). I ended up returning it to the store, and eventually ended up settling for two acoustic 1950s Baldwin Hamiltons (after getting rid of an early 1900s tall upright which, in spite of an impressive tone that to my ear ran circles around ANYTHING produced today, including the Steingraeber 138 (and to a lesser extent, the Baldwin 6000), needed to go).

re: Gyro

I have 32 note polyphony (16 with the piano sound) and I found myself dropping notes occasionally. Dont know how you got by with only 16/8...

Thing about polyphony is, its like gas in a car or RAM in a computer, how much you have isn't important unless you don't have enough (and then its really really important haha). I'm no expert on the topic, but it appears as if 64 (32 for stereo sound) polyphony is just fine unless youre doing some layering

the current minimum polyphony standard on DPs is 64, which would mean anything below that is inadequate for anyone, even a beginner. if it's not important, the DP makers would not raise the minimum number of polyphony on all current DPs for no reason.

the point is that polyphony would be used up with stereo sound, pedaling, fast playing, chords and etc. (in double/triple note eating fashion). maybe it doesn't matter when you just start learning, and cannot even play a scale yet, but it would matter after 6-month or a year when you can play better.

Are you sure?

If anything below 64 is inadequate today (because the mfgr's are raising the bar even higher), than what would we have concluded a few years ago when 32 was common? Was 32 "okay" then, and anything less inadequate?

Moore's Law applies to DPs too. The cost of supplying 64-note polyphony is now lower than the cost of providing 32-note polyphony was not that long ago.

But lets not get caught up in a pure numbers competition at the expense of quality of sound. I'd rather have 32 notes of good sound than 128 notes of not-so-good sound.

32 was inadequate years ago--it's just that it wasn't economical to do anything about it. This is like when computers came with slow processors and tiny memories. They were obviously slow and needed to be improved, but it just wasn't possible at the time.

Now 64 is cheap to provide and 128 not much more so. With layering and damper resonance and stereo, you burn through voices faster than you would believe. So 64 as a minimum is reasonable.

Would you buy a new Pentium-II PC with 256 Mb of RAM? So don't buy a DP with a 1995 spec.

I would have thought that, as a rule of thumb, more notes of polyphony than you have actual keys (ie, 88) is more than enough.

Of course if you are combining two sounds at once ie piano and strings, or you are using 'effects' like echo/delay or chorus, these may count as extra notes (ie, with certain effects on your effective polyphony may be lower). Still, if you have 96 notes polyphony you'd very rarely have any issues.

Think about it - would your brain notice if 96 notes were being sustained simultaneously and the oldest, faintest one dropped off to make way for a new note? What about 64? In normal piano pieces you'd never even encounter such a situation, let alone notice it if you did: even a glissando of your full keyboard is rare enough, let alone holding it with the sustain pedal and keeping it sustained into your next however many notes.

First, stereo samples use twice as many voices as mono. So that's half of your polyphony gone to start with. Next, if you layer sounds (e.g. Piano and string), you can cut it in half again. And effects such as damper resonance mean that striking a single string may cause a dozen resonant notes to sound (at greatly reduced volume). I can easily imagine a situation where a 128-polyphony DP still needs to steal notes, because it's run out. Frankly, 256 polyphony or more would be useful, although not strictly necessary.

It seems you have a computer. What kind of speakers do you have? If you have nice speakers, I believe you may start with as low as 64 and, afterwards, fetch a decent sound card (ASIO 2.0) and Pianoteq or Ivory or whichever software/plugin you like.

I made a demo for testing polyphony with Pianoteq. The followings are made of the same MIDI file with different number of polyphony 8,16,32,64. I don't tell which is which now and I'll show the answer one week later.

http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest/A.mp3
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest/B.mp3
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest/C.mp3
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest/D.mp3

# Notes
1.A pair of steleo voices is counted as ONE voice in Pianoteq.
2.Pianoteq seems to allow some surplus polyphony. For example, even if you limit the maximum polyphony to 8, Pianoteq might generate 9 or 10 voices.
3.The Pianoteq indicated that the maximum polyphony is less than 50 in this piano piece.

I'll bite...

I couldn't tell them apart. I guess you can do a lot with 9-10 stereo pairs of voices, given that humans typically have 10 fingers or fewer

It would be interesting to have the original MIDI file for this -- I'd like to see how other MIDI rendering software handles it if you limit the number of voices.

Incidentally, I wonder if part of my inability to tell the versions apart is because of the relatively low MP3 bandwidth? Even at higher bandwidth I probably wouldn't be able to distinguish them because my elderly ears aren't up to much. But people with better ears might.

kevinb, here is the mid file.

http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest/Fantaisie-Impromptu.mid

Now I prepered the demo with non-compressed sound. I used a shoter sample. Their numbers of polyphony are also 8,16,32,64.

http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest2/A.wav
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest2/B.wav
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest2/C.wav
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest2/D.wav

Their mid file is
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest2/PianoCompare.mid

Interesting. With the Chopin I start to hear artefacts when there are fewer than about 14 (stereo) voices, using Timidity++. There's a huge difference between 8 and 16, but none (that I can hear) between 16 and higher. But I can't tell the difference between your various excerpts.

I'd guess that your MIDI rendering software does a smarter job of polyphony reduction than Timidity++ does -- by default T++ just snips the end of the decay of any note it hasn't got a voice for. I imagine that Pianoteq is smarter (well, it should be, for the price :))

I think this particular piece of music would struggle on an instrument with fewer than 32 conventional voices (that's assuming two per note for stereo). But I doubt it would gain much with more voices.

I rather suspect that, with DPs, it's not always the total amount of polyphony that's important, but how it's used. I've played some instruments that simply refuse to play the next note if there are already too many sounding -- which of course sounds dreadful. I would hope that nothing made these days would be that rubbish, but still I imagine that control of polyphony varies from one machine to another.

I do wonder, however, what kind of electronics there must be in a DP that can only offer 32-voice polyphony, with the current state of technology. I wonder if there is some sort of salesmanship going on. You know -- this one only has 32 voices, but for an extra $100 you can have this one which... you get the idea

The reason I mentioned Pianoteq is the fact that it could help zxczxc12345 work around the polyphony issue on a digital piano, once it comes to that.

Kevin, there may be some sort of salesmanship going on, but 32 notes polyphony is not enough to comfortably play a digital piano without a computer. And if someone uses instrument as a MIDI controller, instrument’s polyphony becomes quite dispensable. Asa a controller, piano only has to send proper MIDI messages (key on, key off, velocities, etc.) and let computer do the rest. I guess you knew that, but I wanted to make it clear for anyone observing this thread.

Mezzo-poor, I am wondering how Pianoteq handles damper pedal in terms of polyphony. Would you please let us know does maximum polyphony differ in these two cases:
- two chords with damper pressed
- two chords with damper released

Yes. That's why I said

``I think this particular piece of music would struggle on an instrument with fewer than 32 conventional voices (that's assuming two per note for stereo).''



The issue of salesmanship I was referring to was whether manufacturers really need to sell a keyboard/DP with only 32-voice polyphony, given how cheap electronic technology now is. A long time ago when I worked in the electronics industry, my company used to make and sell products which they never expected to sell, just so customers could be encouraged to buy more expensive ones. The sales pitch generally ran like this: ``That one won't be enough for you, but this one has twice as many whotsits and it only costs an extra £100'' or whatever.

I do wonder if something similar goes on in the DP world, but maybe that's just my natural cynicism.

I've said on other threads that, in my view, a good way to get a decent piano sound without spending a fortune is to rig up a MIDI master keyboard (with no sound generator) to a computer and thence to a home-theatre amp setup. This is particularly cost-effective if you've already got everything except the keyboard. But it's not very convenient if you just want to be able to sit down and play.

I had this in mind...



But, yes, this particular piece.

Makes a lot of sense. There is the other direction of this salesmanship. Imagine, for instance, plain MIDI controller with AWA Grand Pro II or PHA III.

I'm afraid this post become long. I hope this post will be helpful for understanding how Pianoteq (or other software pianos, DP's) handles polyphony.

I am also wondering because I am not a developer of Pianoteq. Just I can do are experiments on my Pianoteq and guessing.

Before experiments, I should explain some basic things.

For a computer, a voice is a task which consumes some resources (memory, time). A voice is invoked when it is required and terminated when it becomes useless (for example, after sufficient decay).

There are two types of resonance, one is global resonance and another is sympathetic resonance. Global resonance represents the resonances of all chords while the damper is on (pressed). For example, press the damper pedal, and next press C4 (center C) key, then you will hear the global resonance. Recently even inexpensive DP's (for example, Privia's) handle this global resonance. Sympathetic resonance represents the resonance among all pressed keys while the damper is off (released). For example, without using the damper pedal, press G5 key with ppp and keep it pressed, and next press C4 key with staccato and with strong accent, then you will hear the sound of G5 (let me call it sympathetic resonance on G5 by C4). Only some high-end DP's handle this sympathetic resonance. Pianoteq handles both types of resonances.

Pianoteq has an indicator which shows the number of polyphony currently used. Let me call it polyphony counter. A pair of stereo voices is counted as ONE voice in the polyphony counter.

Now let me show some experiments and comments (prefixed with #).

Experiment 1) Playing arpeggio with damper off
I used the midi sequence "note-on C4 | note-on E4 | note-on G4 | all-notes-off". Note that each key was kept pressed until all-notes-off came.
Then I got

(0) | note-on C4 (1) | note-on E4(3) | note-on G4(6) | all-notes-off(0).

Each number in ( ) represents the number that the polyphony counter showed immediately after each event.

# note-on G4 invokes three voices.

• voice of G4. Let me call it direct voice
• sympathetic resonance on C4 by G4.
• sympathetic resonance on E4 by G4.

I also tried an arpeggio consists of 10 notes and I saw the polyphony counter exactly showed 55=1+2+...+10 when the last key pressed.

# I also found that no sympathetic resonance was invoked in the case that the interval of two note-on events is small(say less than 100ms). This is a reasonable saving because a sympathetic resonance is easily masked by a laud direct voice.

Experiment 2) Playing arpeggio with damper on
I used the same midi sequence as used in Ex.1. I got

(0) | note-on C4 (1) | note-on E4(2) | note-on G4(3) | all-notes-off(0).

# In Pianoteq, the global resonance seems to be implemented as a special task other than voices. So it does not affect the polyphony counter.

# Since sympathetic resonances are masked by global resonance, no sympathetic resonance is invoked while the damper is on. In other words, the number of sympathetic resonances can increase only while the damper is off. Therefore, as far as you use 10 fingers, the number of sympathetic resonances does not surpass 45=1+2+...+9. On the other hand, the number of direct voice has no limit while the damper is on.

Experiment 3) Same as Ex.1 but the maximum polyphony limited to 2
I got

(0) | note on C4 (1) | note on E4(2) | note on G4(2) | all notes off(0).

When G4 was pressed, one of two sound C4/E4 disappeared. In this case, less laud voice seemed to disappear.

# In this case, just direct voices are invoked. When Pianoteq meats the shortage of the number of voices, it reduces even direct voices.

Experiment 4) Same as Ex.2 but the maximum polyphony limited to 2
I got the same result about the polyphony counter. Disappearance of C4/E4 was same as in Ex.3. The difference from Ex.3 was that I heard the global resonance.

# Global resonance does not consume any voice.

End of experiments--------------------------------

I didn't mention half-damper or quarter-damper because I could not guess what happened. And I didn't mention the sostenute pedal because my silent upright does not have it.

As Kevinb mentioned, any software pianos or DP's should have the policy "Least significant voice should be reduced first." The meaning of "Least significant" depends on the developers. I believe every developer keeps this policy.

Through the experiments I found that in discussion of the number of polyphony, we should pay attention to

• How do they count the number of polyphony, in other words, what does ONE voice means?
• How many voices are invoked when a key pressed?
• When a voice is terminated?.

Let’s have a round of applause for Mezzo-poor!

Refreshingly thorough. Thank you very much.

I assume DP manufacturers count each note we hear as two voices (stereo samples), which means the second experiment would produce significant increase on the counter of any DP, as you go from C4 to G4. Of course, it would be nice to know, not to assume. My assumption is based on common results of competition and power of computers put in these pianos.

I agree with Geoffk on upgrading to YDP160 and now have to direct attention to another difference between these two instruments. The sound system of YDP160 is a bit more powerful - 2x20W compared to YDP140’s 2x6W.

Well, it's certainly worth checking with the manufacturer or vendor how they're measuring polytphony, since it isn't as simple as it might appear. The practice of counting two stereo channels as as two polyphonic voices is widespread and, I can't help thinking, a bit deceptive. It seems to me to be similar to the way that mid-priced hifi manufacturers find ways to express the power output of their amplifiers to allow them quote really huge numbers (e.g., only meausuring with one channel driven). It's not actually wrong, but it's unhelpful.

I noticed some mistakes.

In Experiment 2

is wrong. So please forget it. The number of sympathetic resonances has no limit too. If you keep pressing C4,E4,G4 with your left hand and repeatedly press C5,E5,G5 with your right hand then the number of sympathetic resonances increases 9=3*3 everytime you press C5,E5,G5.

In Experiment 3

Of course "laud" is a typo. "loud" is correct.

Last week, I posted two demos for testing polyphony with Pianoteq. Now let me show the answers.

Note

• A pair of stereo voices is counted as ONE voice in Pianoteq.
• In some cases(I wonder in what cases this happens), Pianoteq uses a few surplus voices. For example, even if you set the maximum polyphony setting 8, Pianoteq might uses 9 or 10 voices. In the answers, the maximum polyphony actually used during a performance is shown in (). However, I'm afraid the least significant digit in () is not accurate, because the polyphony counter was changing very quickly.

Please DO NOT evaluate the sounds after you know the answers.
The answers are here:

http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/pub/ans.txt

And let me show some extreme examples. In each example, the maximum polyphony setting is 2 and maximum polyphony actually used is 5.

http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest/E.mp3
http://cid-902c819bc390f6ee.skydrive.live.com/self.aspx/polyphonyTest2/E.wav

(Hmm... I like this lightsome Chopin. I might use this minimal-polyphony setting as an effect.)