Sampled Pianos - Mono or Stereo?

Author
markheath
Max Output Level: -82 dBFS
  • Total Posts : 408
  • Joined: 2003/12/30 05:28:55
  • Location: Southampton, England
  • Status: offline
2005/11/08 08:09:37 (permalink)

Sampled Pianos - Mono or Stereo?

This might be slightly off topic, but I am thinking of sampling my piano for use with sfz or Dimension, and was wondering about whether to make some sfz files based on mono samples.

All the sampled pianos I have used provide their samples in stereo. Now I appreciate that for the person playing the piano, they will hear more bass notes in their left ear, and more high notes in their right, but for generally using a piano sound in a mix, why do you need to sample in stereo? After all, if you use 2 microphones to sample an acoustic guitar you don't pan one hard left and the other hard right. Why not mix however many microphones you used to sample the piano together into one mono sample per note? You can always pan the samples, and even apply different EQ to the left and right channels, to create a stereo effect.

The reason for asking this is that if you use mono samples you can double the number of velocity layers (or the number of notes sampled) you use for the same memory footprint. To my mind, this would contribute more to the realism than having a subtle stereo effect.

#1

10 Replies Related Threads

    techead
    Max Output Level: -31.5 dBFS
    • Total Posts : 4353
    • Joined: 2004/01/24 08:40:20
    • Location: Macomb, IL, USA
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 08:48:48 (permalink)
    I know it is a lot of extra work, but perhaps sampling both mono and stereo then building two sets of multisamples would be an interesting exercise. Perhaps you could limit the experiment to a smal playable range. I thought your question was interesting but I have no answer for it.
    #2
    MIDIMONSTER
    Max Output Level: -89 dBFS
    • Total Posts : 64
    • Joined: 2005/01/26 09:55:31
    • Location: Beantown
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 08:49:26 (permalink)
    Hey Mark

    I would use mono samples with multiple layers[velocity groups] and vary pan this way in your SFZ file:

    <region> samplefolder/samplename.wav pan=

    -100 is hard left
    0 center and
    100 is hard right

    so you could create something like this:

    <region> samplefolder/samplenameC2.wav pan=-43
    <region> samplefolder/samplenameC3.wav pan=-10
    <region> samplefolder/samplenameC4.wav pan=10
    <region> samplefolder/samplenameC5.wav pan=40

    writing out pan= for each region may seem a bit too much, we should ask Rene if there is some type of Pan_random opcode. It would be much easier to write something like this:

    <group> pan_random=100

    ???

    anyone want to chime in on how to get this done?
    #3
    René
    Max Output Level: -68 dBFS
    • Total Posts : 1103
    • Joined: 2004/01/06 13:15:57
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 09:34:34 (permalink)
    There are many approaches when mic'ing the piano. The most common, mostly used are 'player perspective' and 'audience perspective', and depending on what kind of room/hall is used to record the piano, there're several 'ambient' perspective positions as well, available in different libraries.

    The 'player perspective' approach is used in most stereo piano sample libraries. This is intended to provide a real piano *playing* experience, rather than a finished mix. As you can adjust the stereo width and position anytime in your mix, it works most of the times (though it's far from perfect). If you pay attention to some ballads, you'll see that the intro (raw piano) uses a very wide player perspective, which quickly mutates into a centered, narrower piano as soon as the drummer <g> starts playing.

    Stereo recordings are tricky though. Unless you use a true-stereo microphone (like a schoeps if you're rich), using two microphones to get a phase and color coherent recording can become a nightmare. Apart from the standard pan effect, there's the phase correlation (read as the time difference between left and right channels) effect, which can create a very unplayable piano if not done properly.

    Here's a nice overview on how to position the transdiucers to achieve a good stereo recording:

    http://www.schoeps.de/E-2004/overview-stereo-techs.html


    If you will finally record in mono, you can use the pan keyboard tracking in Dimension to position the piano in a player perspective anytime. However, the pan position is not the only thing a stereo recording will capture. For comparison, you can test the small mono piano patch in Dimension vs. the small stereo piano patches.

    -René
    #4
    MIDIMONSTER
    Max Output Level: -89 dBFS
    • Total Posts : 64
    • Joined: 2005/01/26 09:55:31
    • Location: Beantown
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 10:05:01 (permalink)
    Rene, what about a Random Pan opcode in SFZ, how can this be acomplished???
    #5
    markheath
    Max Output Level: -82 dBFS
    • Total Posts : 408
    • Joined: 2003/12/30 05:28:55
    • Location: Southampton, England
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 10:46:07 (permalink)
    Rene, what about a Random Pan opcode in SFZ, how can this be acomplished???


    Do you really mean random pan? That would just be weird on a piano. Some kind of key2pan opcode would be more relevant. But if you have enough regions (e.g. 1 for every minor third), you can probably get away with 1 pan value per region.
    #6
    MIDIMONSTER
    Max Output Level: -89 dBFS
    • Total Posts : 64
    • Joined: 2005/01/26 09:55:31
    • Location: Beantown
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 11:01:50 (permalink)
    ORIGINAL: markheath

    Rene, what about a Random Pan opcode in SFZ, how can this be acomplished???


    Do you really mean random pan?


    Not completly Random - More like controlled Randomness

    Use the random opcode to generate a slight variance in the stereo position from sample to sample. Not an extreme pan at all.

    I was imagining there would be a LOW random and HIGH random setting, so each new note would change within a constraint of values. There are other sfz opcodes that work similarly to this idea, such as lorand and hirand.
    #7
    markheath
    Max Output Level: -82 dBFS
    • Total Posts : 408
    • Joined: 2003/12/30 05:28:55
    • Location: Southampton, England
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 11:18:06 (permalink)
    ORIGINAL: René
    If you will finally record in mono, you can use the pan keyboard tracking in Dimension to position the piano in a player perspective anytime. However, the pan position is not the only thing a stereo recording will capture.


    René, if the width opcode is set to zero, does sfz / Dimension store a mono sample in memory? If not, perhaps this could be a future optimisation. Then you could produce 'lite' versions of any stereo sampled piano by setting width=0 and pan appropriately. This would result in a piano that required only half the memory, but would not need the extra disk space of creating mono samples.
    #8
    René
    Max Output Level: -68 dBFS
    • Total Posts : 1103
    • Joined: 2004/01/06 13:15:57
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 12:02:06 (permalink)
    Rene, what about a Random Pan opcode in SFZ, how can this be acomplished???


    // Unipolar random, center to R
    pan_oncc135=100

    // Unipolar random, center to L
    pan_oncc135=-100

    // Bipolar random, L to R
    pan_oncc136=100

    // Keyboard tracked pan L to R
    pan_oncc133=100


    Combing above opcodes with the static pan= value result in a random or keyboard-tracked variation with minimum and maximum adjustments.

    René, if the width opcode is set to zero, does sfz / Dimension store a mono sample in memory? If not, perhaps this could be a future optimisation. Then you could produce 'lite' versions of any stereo sampled piano by setting width=0 and pan appropriately. This would result in a piano that required only half the memory, but would not need the extra disk space of creating mono samples.


    They don't. That would be possible, but the implementation might be tricky, due the sample pool scheme both sfz and Dimension use: if a sample is used by several regions, it's loaded into memory only once. If a new .sfz file uses a previously loaded sample, it won't be loaded again. Even if a new instance asks for a sample previously loaded, sample will be reused.
    It could be possible to flag if the sample has been converted, and load it if so, but well... hard disks are huge and inexpensive lately
    Anyways, make sure you FR it, so it's examined in depth.



    -René
    #9
    Jim Wright
    Max Output Level: -66 dBFS
    • Total Posts : 1218
    • Joined: 2004/01/15 15:30:34
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/08 18:35:35 (permalink)
    ORIGINAL: René

    Stereo recordings are tricky though. Unless you use a true-stereo microphone (like a schoeps if you're rich), using two microphones to get a phase and color coherent recording can become a nightmare. Apart from the standard pan effect, there's the phase correlation (read as the time difference between left and right channels) effect, which can create a very unplayable piano if not done properly.

    I bought a Rode NT4 several years ago for just that reason -- it's a point-and-shoot stereo microphone with fairly good sound quality (not a schoeps, but pretty nice nonetheless). It's great for live concert recording (mostly high school concerts/musicals; my daughters are very active musically). It's the best solution under $500 I've come across.

    My biggest complaint is that the stereo field is somewhat narrower than would be the case with other stereo micing techniques (e.g. the ORTF approach). That's not as apparent over monitors, but the headphone imaging is decidedly middle-of-the-head. However - the fact that it's impossible to screw up phase correlation and have mono cancellation effects kind of makes up for that. Not to mention that it makes setting up for a recording at least 15 minutes shorter than it used to be.....

    It's apparently used a lot for recording acoustic guitar. I've used it several times for piano (college audition CDs for teens I know), and it's worked quite well for that, as well. However, I'm not running a pro studio / treated room / highly-maintained studio piano kind of operation..... YMMV.

    - Jim
    #10
    MIDIMONSTER
    Max Output Level: -89 dBFS
    • Total Posts : 64
    • Joined: 2005/01/26 09:55:31
    • Location: Beantown
    • Status: offline
    RE: Sampled Pianos - Mono or Stereo? 2005/11/09 08:27:08 (permalink)
    ORIGINAL: René

    Rene, what about a Random Pan opcode in SFZ, how can this be acomplished???


    // Unipolar random, center to R
    pan_oncc135=100

    // Unipolar random, center to L
    pan_oncc135=-100

    // Bipolar random, L to R
    pan_oncc136=100

    // Keyboard tracked pan L to R
    pan_oncc133=100
    Combing above opcodes with the static pan= value result in a random or keyboard-tracked variation with minimum and maximum adjustments.
    -René


    Thanks! I will give this a shot.
    #11
    Jump to:
    © 2024 APG vNext Commercial Version 5.1