A secret revealed

I just read the press release for the forthcoming Garritan-Steinway piano sample collection.

Garritan-Steinway Press Release

They describe transporting the cherry-picked Steinway to, of all places, the Troy Savings Bank Music Hall in Troy, NY, to record it.

While not exactly as famous as any one of a number of NY venues they might have used, the Troy Savings Bank Music Hall is one of the greatest recording spaces on the planet. I know a number of musicians (including some Grammy-winners) who have quietly schlepped to Troy to record there. It just sounds awesome. The sound is both crystal clear, and lushly reverbed. Accoustic textures are detailed and gel perfectly.

So perfect are the acoustics, that they go out of their way not to renovate. For example, they don't dare cushion the original (and uncomfortable) wood seats lest that potentially ruin the sound. I think it's one of those things where they're not really sure why it sounds as good as it does, so they just don't mess with it.

It's really bizarre - this old music hall in the middle of an old bank in the middle of an old town. But for those who know... a great place to record.

So, I'm guessing that the Garitan-steinway sample library will be a good one (dethroning "Ivory"?)

-Peace, Love, and Brittanylips

Interesting.

I'm not so sure that that bodes well. Take a convolution impulse of the hall - sure. But there is a huge difference between playing a piano in a nice acoustice space - and playing hundreds of samples of the same piano in that space.

For example - when you play anything more than a single note, you are getting layers of that reverb, with odd phase relationships that were never in the original hall. Not good.

IMHO - most piano libraries suck because of poor mic placement, and misunderstandings of what a sampled instrument should actually be.

You think i'm just whinging? Probably - but try this: take any sampled piano and listen to it in mono. Usually, enormous phase issues that prevent them being used in a successful pop record.

And - if they use key release samples, that is a lunacy. Because ... there is no logical way to decide how loud the release samples should be. The level of a piano note goes from loud to soft as the key is held down. If you release the note quickly, maybe the key release sound will be loud enough. The longer you hold the key, the quieter the piano note becomes, so when you release the key - the release noise is inappropriately loud. You can't modulate it with key-on velocity, because that only applies to the initial note. No point using key-off velocity, because that has nothing to do with how loud the release noise should be.

In my opinion, a truely great sampled piano should be sampled with mono compatibility firmly in mind - and with no intention of layering up the reverb sound as more notes are played.

In other words - dry mono should be the ultimate. And the release is best done like a synth, with normal ADSR on the sustained sound.

The sounding board make a glorious stereo sound each time a key is hit - but that also shouldn't be layered up each time a new note is added. So I believe this should be a stereo sample, but triggered in 'mono' mode, so each new note re-triggers this stereo sound. And user adjustable, since realism isn't always desirable, in what can only be an unrealistic instrument at best.

The glorious reverb of the venue would be best applied with convolution reverb - and best chosen seperately by the user in their reverb of choice.

Many of the hardware piano makers have figured out how to make excellent sample based pianoes. They don't even need huge sample memory to get the job done (although i do hate to hear looped sounds). And if the real piano has a defective note - I would rather have a re-tuned good note than the real crap note.

Here's hoping the piano sample developers can start getting this right too. It was easier with electric pianos, because they are in mono to start with and no mic placement worries.

Rant over.

What do you think is the number of hits into the piano keys the poor pianist has to make for the library? 12'000? more?

I don't see a way to record a piano performance with a sampled grand.

.

You make a lot of very good points.

Given the comoplexity and variety of what a piano actually does, we're probably decades away from a truly accurate sample - one that can reproduce dynamic variables over time, introduce realistic sympathetic vibrations, permit realistic legato, staccato, and subtle variations in pedaling (dampers don't just do "on" and "off"), etc.

But for now, perhaps this is the latest greatest thing. Given that Steinway put their name on it (which says a lot - they are a very picky company), I wouldn't be surprised if this ups the ante.

In any case, I'm definately interested in a demo.

-PL&B