• Register
  • Help
Results 1 to 10 of 10

Topic: Piano Parameter Priorities

  1. #1

    Piano Parameter Priorities

    Hello piano sample users,

    I am trying to decide whether or not it is worth the effort to perform another, more extensive, sampling session on the Steinway-C and Steinway-B pianos (NOT the Oliver Truin CD) that I have previously sampled.
    I am hoping to get some feedback from readers of this bulletin board as to the various \"wish-list\" items that may be incorporated in a future piano sample set, and any weaknesses of current piano offerings that may need to be addressed.
    Before deciding on whether or not to resample the pianos more extensively, I would like to know if there are other specific aspects of piano sounds that a significant number of users would like to see included or improved.

    I originally recorded the gigabytes of samples for use with MIDI rendering software which did not have some of the versatility of the newer soft-synths, like gigasampler and others.
    The type of music that I personally was using the samples for, was mostly ragtime/stride/trad-jazz. This style does not make heavy use of the sustain pedal, and long resonating decays are extremely rare. As such, the lack of the ability to switch to alternate \"pedal-down\" samples was not a big issue for me. Nevertheless, I realize the for other styles, particularly \"new-age\" types of music, these long resonating decays are more important.
    So, since I am now making these samples available to others, I would like to get some feedback on how much the availability of separately recorded pedal-down samples REALLY matters to most people, or is it just a psychological/marketing \"nice to have\" for most people. Is the availability of the separate resonant samples worth the significant price differential for the more extensive sample sets for the majority of users? Do people add so much reverb/equalization to their tracks that the inherent sample resonance becomes irrelevant anyway?

    I have personally found that 3 recorded strike layers are generally fine for softer and more mellow pieces, but for pieces with really wide dynamic range, 5 recorded layers make a significant difference. I would be interested in feedback from others about your experience with numbers of sample layers (I\'m talking truly recorded distinct layers, not artifically mixed/filtered layers) for the type of music you are creating. MIDI file examples would be greatly appreciated.
    My own inclination is to go with 5 or 6 recorded strike layers, but if somebody can make a compelling case for more layers, I would like to hear it.

    The other thing I would like to know is:
    How many people REALLY need samples of duration longer than about 10 or 15 seconds? Including the 20 seconds or longer samples as I did before takes lots of disk space, and may be just a waste for all but a few extreme users who would be just as well served by a 10 second sample properly looped and articulated.

    Are there any other \"secondary sounds\" that people would like to see included? I can include the mechanism noise of the pedal and damper mechanisms as separate samples if many people think this sort of thing is useful (personally I would never include this in my own recordings).

    I would appreciate any feedback and suggestions for piano sound parameters to capture/improve within the next week or so. I hope to decide whether or not it is worth the effort to make another sampling session on the pianos before the end of the month.

    Thanks in advance for any input.


    Warren Trachtman http://www.wstco.com/

  2. #2

    Re: Piano Parameter Priorities

    I am glad to hear that you are considering expanding your Steinway samples. My vote is to include a complementary set of undamped samples. The sustaining tones provide the \"singing\" quality that appeals to many of us. IMHO, the Bosendorfer, otherwise a splendid piano sample, could use more sustain resonance.

    I recommend that if loops are employed that they be restricted to the bass end only and the notes should be allowed to decay to at least -60dB before entering a loop.

    Four velocity zones (each for damped and undamped) is sufficient in my estimation.

    Sampling every note is not absolutely necessary if you are trying to keep the size within reason, although 4 notes/octave should be a lower limit.

    Keep up the good work.

  3. #3

    Re: Piano Parameter Priorities

    I would like to emphasize the need for the pedal down samples. I do a lot of scoring to picture, and for delicate parts up high, the pedal down samples ringing out sound wonderful. The pedal up samples for parts like these sound very stale. Good luck.

  4. #4

    Re: Piano Parameter Priorities

    This is the sort of feedback I am hoping for, so please continue.

    To address a few of the points raised:
    1. >>notes should be allowed to decay to at least -60dB before
    entering a loop<<
    The long sample/looping tradeoff is only a concern for the bass notes, which sustain for an extremely long time on a large Grand.
    For the Steinway-C, low A is only down about -40dB after 20 seconds.
    20 seconds or so is really a pretty long time (try it with a watch), and what I really was hoping to hear is how many of you can actually say that you often hold a note that long when you play.

    2. >Sampling every note is not absolutely necessary <<
    This is one thing I will not compromise on. No matter what other decisions I make, I will always independently sample each and every note.

    3. >>for delicate parts up high, the pedal down samples ringing out sound wonderful<<
    The ringing (or lack thereof) of the treble notes is often a function of improper bankset programming, not independent samples.
    The actual pianos do not have damplers for the upper notes
    (from Ab6 - C8 on the model-C; and from E6-C8 on the model-B).
    These notes ring on the real pianos without holding the dampler pedal down. They should also ring without holding the pedal down
    in a programmed bankset, but many creaters of these sets fail to account for this, and incorrect damp the upper treble in the programming.
    Holding the damper pedal down on a real piano while playing high treble notes often gives (to my ear) an echo-like and possibly muddy sound.
    What you get is the added sympathetic resonsnces of the lower strings which become undamped and can resonate when the pedal is down.
    Still, I am trying to find out if many people find this sound important.
    I do, however, want to clearly distinguish the separate pedal-down sample
    issue from the ringing treble issue.

    4. >>I would certainly use and like soft pedal samples too, though it\'s not
    I am trying to include soft-pedal (and even sustenuto) pedal behavior if
    possible. This too, is mostly a programming issue rather than a separate sample issue. The striking of 2 versus 3 strings when the soft pedal is used does have some effect on the beat patterns, but this is too subtle to warrant completely doubling the sample set size.

    5. >>you can hear two \'splits\' almost, where the tone character changed from
    one key to those that followed<<
    This is a function of the design of the piano. There are \"breaks\" where the
    types of strings change (from the fat wound bass strings to the thinner strings)
    and this will create a tone character shift. The amount of shift, and how
    noticable it is, will depend on the piano design and the \"voicing\" done by
    the piano technician. This can only be improved by sampling a well designed
    and properly voiced piano. That\'s why the choice of which specific individual
    piano you sample is as important as the actual sampling itself.

    6. >>Doing full length samples of 15+ seconds - why not? <<
    Currently, with 5 layers, my set is 1.06 gigabytes. If I double the size
    by adding pedal down samples, I\'m then at 2 gigabytes. To have 6 layers would put me over the 2 gigabyte size. Some computers have problems handling hard disk partitions over 2 gigabytes in size. Others can handle larger disk partitions, but may have to use software drivers which can impact performance.
    So I would like to stay below 2 gigabytes if I can. That\'s why I\'m asking if many people REALLY hold notes for longer than 10 to 15 seconds in their typical playing.

    Again, thanks to all of you for this type of feedback/input.
    I would like to hear from more folks about their actual use/need for the long duration samples and any other particular sound
    characterization issues/desires/comments.


    Warren Trachtman http://www.wstco.com/

  5. #5

    Re: Piano Parameter Priorities

    Great that you are asking for feedback. I think you are right that after 15 seconds or so a loop should not really mather (especially when done right) and that it is VERY important to sample every note individually. What I do miss on all liberies untill now is that three or four layers are not enough. Sometimes you need those really hard or really soft notes and with fewer layers the hardest note is not really played hard and the softest not really soft. It also would be nice to hear the sustain pedal being pressed (for some recordings).

  6. #6

    Re: Piano Parameter Priorities

    re: soft pedal. >>This too, is mostly a programming issue rather than a separate sample issue. The striking of 2 versus 3 strings when the soft pedal is used does have some effect on the beat patterns, but this is too subtle to warrant completely doubling the sample set size.<<
    I\'m usually distrustful of programming to simulate sounds, but maybe this will do the trick. Soft pedal does give a different tone color and if this can be programmed then great. I wonder though if limited sampling wouldn\'t be more realistic. Anyway, glad to hear soft pedal is a consideration for you. Oh, also.. you mentioned sustenuto pedal- I don\'t believe it colors the tone in any way, so I wouldn\'t know what to address with it.

    >>There are \"breaks\" where the
    types of strings change (from the fat wound bass strings to the thinner strings)
    and this will create a tone character shift.<<
    ah that explains the change in tone. I think the gigapiano was close miked and I\'d speculate that this may have accentuated the regions - bass, mid (two strings), and higher(three) - by picking up a higher proportion of string sound and lower of piano body and soundboard. I was just playing our grand piano and the region changes are much harder to hear than I remember with the gigapiano.

    >>Some computers have problems handling hard disk partitions over 2 gigabytes in size. Others can handle larger disk partitions, but may have to use software drivers which can impact performance. ... That\'s why I\'m asking if many people REALLY hold notes for longer than 10 to 15 seconds in their typical playing.<<
    So long as it\'s solely for gigasampler(I am a little biased since it\'s all I have), the os is win98/95, and partition size isn\'t a problem for win98 or win95 osr2. The original win95 does have the 2 gig limitation. It\'s hard for me to say how many people are using gigasampler with orig win95. I don\'t think it\'s a good choice for os, due to the partition problem and to lack of IDE drive dma mode support. They should spend the $ and upgrade!
    All that said, looping after 10-15 seconds should be ok if you need to keep size below 2 gig. I want to be a purist about it but it\'s not all that often that I\'d have reason to let the note ring more than 10-15 seconds. And I realized that when I do, I play other notes above it which reinforce and sustain harmonics in the ringing note, radically changing the character of the note after 10 seconds. There\'s no way sampling a note for complete length is going to reproduce that, although the complete sample may sound a little better than the looped version in those rare times it rings for so long.


  7. #7

    Re: Piano Parameter Priorities

    Glad to hear you might be expanding on your set. Here\'s some input. I play classical piano and for a while used the gigapiano when I didn\'t have a real piano. This sorry pianoless state is going to happen again soon, so of course I\'m interested in a very good set! Judging by the multitude of digital pianos at music stores, perhaps others have decided to use gigasampler as a piano substitute as well. Having a great reproduction in a solo environment is my aim. Damper pedal samples are extremely important to me- probably I wouldn\'t be interested without. I would certainly use and like soft pedal samples too, though it\'s not essential. I\'m just dissapointed that noone has done them. Two or three velocity layers and fewer samples per octave would likely be okay, as I think most people use it as an effect and much less than damper pedal. Damper and soft pedal combined would be cool too.. but I wouldn\'t expect that anytime soon.
    I\'ve used gigapiano (yamaha) in the past, and there was one thing that bothered me. Playing chromatically up the keyboard, you can hear two \'splits\' almost, where the tone character changed from one key to those that followed. If I remember, I think one was awkwardly located around the f below middle C. I don\'t know the reason for it, but of course avoid this if possible. Also it seemed to have a little bit more than natural resonance with pedal down, but not problematic and maybe just my imagination.
    Overall it\'s a great set, but could be improved by fixing the above, and adding soft pedal.
    Addressing your questions- I don\'t see any need for pedal noises since that\'s something I\'d want to minimize on a real piano if I payed much attention to it. Perhaps there is such a thing as too accurate! If the extra work increases cost that\'s understandable. Personally I would pay up to $500 for an extremely well done piano, so there\'s certainly room for more cost as far as I\'m concerned.
    Doing full length samples of 15+ seconds - why not? Looping just adds work and disk space is getting only more and more plentiful. It may help to note that the present gigasampler users needed fairly new drives to meet performance requirements - thus their drives probably average 6 gigs, 2.5 gigs would likely be the very lowest.
    Number of velocity layers- 5 or 6 is probably sufficient for a high degree of realism. I think that\'s maybe more than the gigapiano, which sounded ok to me.
    Hope this helps. Good luck-

  8. #8

    Re: Piano Parameter Priorities

    Like Danny Lux I use GS in scoring, and what\'s lacking most for me in the exisiting pianos is a good pianissimo.

    Would you consider making a separate, \"soft\" piano instrument that has the layers concentrated at the low-velocity end of the spectrm?

    What I mean is, when I\'m playing really softly, I\'d like to have three or so velocity layers all at the p dynamic or softer. Right now, the best piano for me (for that) is still the Gigapiano (Yamaha), and there\'s really only one velocity layer I tend to play in (the one below velocity values of 40 or so). I\'d love to have a ppp, pp, and p velocity layer, with maybe two or three more to cover the rest.

    I\'m probably in the minority on this though.

    One other thought, Warren, about the 2 gig file size issue. Hard drives are going to get nothing but cheaper, faster, and bigger... why limit yourself to 2 gigs? And think of all the time you could save by not having to find all those loop points! Think about it... by the time you\'re done programming your instrument GS 2 (or whatever it will be called) will be out, and will require an even bigger faster machine to run on, and HD space will not be an issue. Think about the future!

    Oh and thanks for putting your time and effort into this... I understand there\'s not exactly a mass market for this stuff.

    - Chris
    - Chris

  9. #9
    Senior Member Bruce A. Richardson's Avatar
    Join Date
    Sep 1999
    Dallas, Texas

    Re: Piano Parameter Priorities


    I\'m glad to see the possibility of more great pianos...my $.02:

    1) Don\'t loop--disk space is cheap, and the purity of every note decaying into silence is a nice discipline to take on.

    2) High Register--include the pedal down, please. I agree with the statement that many sampled pianos are mis-voiced with dampers that don\'t exist. However, there\'s still the difference between just the high register strings vibrating sympathetically and the entire harp. Part of what says \"piano\" about the GigaPiano is that low-frequency resonance in the high register\'s pedal down sample set. This is produced by the impact of felt against wire more so than by the string\'s response, and is a vital part of the illusion in my estimation. Also, from a strictly tone color standpoint, it adds more to the potential palette.

    Discrete Samples--more is better. I would note and agree with the statement that the pianissimo register is often underrepresented. One minor complaint I have about the GigaPiano is its tendency to be almost Rhodes-like in response in the mid high area. You can make it \"bark,\" which is an actual characteristic of a Yamaha grand anyway, but is somewhat more pronounced. Not that I mind it at all--matter of fact it\'s positively excellent for jazz work that rises above the rhythm section without needing a volume boost or a lot of compression. However, it could be a little smoother for solo work, and I think adding at least one more sampling level, as you\'ve proposed, should help.

    Pedal Down--Yes, definitely.

    Release samples--I haven\'t seen these mentioned, and they are absolutely critical. The release samples in the GigaPiano are by and large the biggest \"seller\" of the instrument, IMO. When you hear the overtones choke off/shift on release, it\'s a very palpable illusion. Turn off the release resonance, and it\'s immediately less \"real.\"

    The Acoustic Bass that Giga released is another prime example of the importance of release samples. Turn it off, and it becomes just another very well sampled bass. Turn it on, and you begin to actually visualize a player pouring elbow grease into a large ungainly hunk of wood and wire.

    Thanks for the effort, and for opening up the discussion for input.

  10. #10

    Re: Piano Parameter Priorities

    Well, I have decided to go ahead with the resampling, and the next marathon piano sample session starts tomorrow (Friday October 1).
    Thank you all for the inputs, suggestions and opinions.

    Unfortunately, it is not possible to include everything I (or you) would like to have included in the next release of the piano giga-instrument.

    The basic problem comes down to the fact that the current version of Gigasampler is not adequate to give all the desired permutations.
    With only 32 \"dimensions\" available to be allocated, tradeoffs and compromises are unavoidable, so the various parameters have to be prioritized.
    I will be recording many more samples than the software can currently handle, and expect to be able to release updates with more features incorporated if and when the gigasampler software versions provide increased capability to handle the extra effects.

    My own evaluation of sample performance leads me to conclude that 4 layers are noticably less realistic in dynamic timbre behavior than 5 or 6 layers. With the binary dimension allocation, as long as I need more than 4 strike velocity layers, I plan to record at least 8.
    I will be taking samples with both sustain-pedal-up and sustain-pedal- down for all 88 keys at all 8 strike layers.

    With stereo samples, this uses up all the available dimensions in the current gigasampler version. (8 strike velocities X 2 sustain pedal states * 2 channels = 32 dimensions).

    However, I will also be recording soft-pedal down samples and hammer release samples during this recording session. I believe that these sound parameters are less important to include than the extra strike layers and the sustain pedal samples, so these others will not be included in my next piano CD release. However, if future updates of the gigasampler allow for more dimensions and also the extra polyphony required to support the extra layers/dimensions, then I will be ready with the extra samples whenever the software capabilities can handle them.

    Thanks once again for the feedback.


    Warren Trachtman http://www.trachtman.org/

Go Back to forum

Tags for this Thread


Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts