Why Bigger Isn't Better

[jjdpro]

Very interesting article about Piano Libs form Willaim Cloakley...
http://www.williamcoakley.com/articles.php?article=bigger.php

Why Bigger Isn't Better
Generally, the bigger it is, the worse it sounds… why?
Clever marketing substitutes "realism" for beauty… what's going on?
Good samples from a beautiful piano, by good techs, and developed by good programmers, doesn't make a beautiful piano… then what does?

Today, it's not unusual that synthesizers with more ROM memory sound better. But do they sound better because more memory is used or were the sounds processed by a more mature sound design team… or is it both?

When looking at the amount of memory a piano has; here are some facts to consider. First, it takes LESS time and LESS work to develop a 3 gigabyte piano than it does a beautiful thirty two megabyte piano… a fact that has contributed to the introduction of numerous MMPs (monster memory pianos) over the past few years. Second, if you can't design a beautiful thirty two megabyte piano… you won't be able to do it even with thirty two gigabytes. Surprised?

We can begin understanding these statements by taking a look at the logic behind monster memory pianos (MMPs). Previously, memory restrictions imposed what were considered undesirable limits on sampled pianos like having to loop samples and using filters instead multiple velocity layers to recreate dynamics. So the logic goes: "…wouldn't it be better if we could have a sample for every single note of the piano? And no loops. Let's get rid of the filters and have sixteen layers of samples per note from PPP to FFF to render the dynamics naturally. And let's have four layers of samples with the pedal down so we can capture the natural resonance of the piano and let's even put in some pedal thump and the sound of the dampers being lifted from the strings just for some added realism."

This logic and using large amounts of memory to achieve it is indeed "logical" …but all the memory in the world doesn't mean you'll end up with a beautiful piano. This is not to say that an MMP piano couldn't be made beautiful but, the likelihood of that happening is slim as we will see shortly. The reality is… there are sixteen megabyte pianos that sound better than most MMPs of many gigabytes and more. So why would listeners prefer the sixteen megabyte piano over a five gigabyte one? And why would producers and writers prefer to use such a small piano when MMPs are available for less money? The answer is, because the particular sixteen megabyte piano sounds better and mixes better with other instruments, qualities that are not dependent on the amount of memory used. Obviously, this doesn't mean that any sixteen megabyte piano is going to sound great but it is saying… that more memory doesn't equate to having a better sounding piano… in fact an inverse relationship exists… the bigger the piano, the more likely it is to sound poorly.

That's because it's what you do with the samples AFTER it's in memory that makes the difference. It's months of hard work and the ability to apply a delicate balance of science and art to the development. With so much memory to edit, developers don't have the multiple life times required to correctly edit gigabytes worth of data and make the piano play properly.

A refined instrument requires that every note has to be like a mastered CD; finished and perfect in every way. A great set of samples is a good start but that's all it is… a good start. Piano development is like a chain being only as strong as the weakest link… if a single step in development is overlooked or compromised; the product is less than it could be. This explains the characteristic raw (unedited) sound which marketing companies refer to as: "realism... that's the way the piano really is." So "realism" becomes the marketing focus instead of a well refined beautiful instrument. The illusion is that a listener's attention will be drawn more closely to what they think is a real piano when in fact; listeners just want to hear a gorgeous piano and a composition that keeps their attention.

Gimmicks such as samples that add pedal thump or the swishing of dampers lifted from the strings might be interesting for techies because it's "realism… it's natural" but so is an out of tune piano. And for the past 75 years, engineers dreamed of a day when they could get rid of annoying pedal thumps and other "natural" noises a piano makes. They went to great lengths to position and EQ mics so that they could eliminate these sounds.

The greatest factor for adding realism to pianos has been the increased polyphony of current products. This enables more samples to sustain simultaneously which emulates the resonance of sustaining strings in a real piano.

My first piano album was recorded using a Sample Cell card which provided a sixteen voice stereo piano. The sixteen megabytes of RAM cost $1350.00, the Sample Cell card was about $1600.00 and the ProTools system to run it about $12,000.00. The pianos that came with Sample Cell were so unusable that I had to create my own and when everyone tried to get their hands on them… I realized I had stumbled into a business.

Are the frills really adding anything?
Although MMPs come with pedal down samples, most engineers find it easier to mix the piano after disabling the pedal down samples and it's interesting that without these samples the piano generally doesn't lose any realism in the mix. Pedal down samples have proven so problematic that one new product (a software playback module) has included two very obvious buttons on the front panel for the convenient disabling of pedal down samples as well as release velocity samples.

MMP demos
A quick listen to some of the demos over the past year reveals a number of MMP developers have worked feverishly to improve their product much to their credit. But in most cases, it's impossible to determine the quality and usability of these pianos without hearing it stripped of all the effects and played in a way where you can really hear it. Whiz-bang demos make it impossible to hear defects. And typically 90% of the effects added will have to be removed if you place the piano in context with other instruments; so you need to hear the piano completely stripped of all processing. And what you're listening for is ambience IN the samples. If samples contain any detectable ambience, it can mean trouble getting the piano to mix with other instruments. Less ambience means you have more control over your production.

WHIZ BANG DEMOS
Years ago, I had been working with the Korg O1W Pro 88 for about a year at a local Church. One day, after Mass, a local high school student was hanging around raving about the new piano in the O1W. He seemed so convinced that I let him play the demo built into the unit. I'd never heard it. After a few bars, I came running over to see in which bank he had found the piano. I thought I had been through every sound in that keyboard and I was a little embarrassed that I'd somehow missed knowing about a good sounding piano right under my nose! Satisfied that he'd put one over on the head cheese, he quickly left when his mother called him. I couldn't believe my eyes; I was looking at the same piano that I'd decided not to use because it sounded so poorly. How come it sounded so good when I heard it in the demo? Well, in the demo, there was some fast playing and runs where only the real part of the piano strike was coming through, the first 1/3 second or so. As I peeled back the layers of chorusing, reverb, EQ etc and finally got to the real piano it sounded horrible. Lesson learned.

[howardv]

His comments about pedal noise made me smile. I was recording a real piano about a month ago and felt there was too much pedal noise in the delicate opening of the piece. So we adjusted the dampers to an inch of their lives and totally eliminated the noise. But then the new recording sounded like a sampled piano. Ended up readjusting the dampers to lift slightly higher and rerecording.

Howard
Ragtime (http://www.rtpress.com) Press (http://www.rtpress.com)

[tquandt]

An interesting read, however I think he is a bit biased (on the same site, you will find his article on why his pianos are better... And I wonder whether he really tried the competing pianos. When playing my PMI stuff, I cannot find it to sound 'bad' at all, although they are really big... ;)

And I think he misses the point that a realistic reproduction of a piano isn't a bad, heartless, technical reproduction, but mirrors all the qualities of the beautiful original instrument... :)

[jjdpro]

I must admit that when it's crunch time, I'm always going back to the Perfect Piano( form my Ensoniq MR ) module. That piano is amazing..

[tquandt]

I must admit that when it's crunch time, I'm always going back to the Perfect Piano( form my Ensoniq MR ) module. That piano is amazing..


:D Well, even the M1 piano had its uses for dance music... :eek:

[Gary M. Thomas]

Just a sidenote...

I have always been partial to the "EastWest Steinway B."
It may not be the biggest (or the newest), but it really does the job for me. :)

[howardv]

Coakley does seem to have a special nack for packing tight and focused pianos with quite a bit of punch into very small packages. I was looking at the Akai programing he did on his Fazioli samples about a year ago and would you believe he used a different sample rate on each of the bass notes to optimize the storage space?

I can see why he'd be daunted by the number of samples in today's pianos. Quite frankly I often wonder how developers like Worra and Michiel get such high quality products out the door so quickly. Thomas Edison sleeping habits, perhaps?

Howard
Ragtime (http://www.rtpress.com) Press (http://www.rtpress.com)

[jjdpro]

"Just a sidenote...

I have always been partial to the "EastWest Steinway B."
It may not be the biggest (or the newest), but it really does the job for me. "

Steinway B is the other goto Piano...

[Michiel Post]

:eek: :eek: sixteen megabytes of samples, for a whole instrument!
with loops and re-tuning, resulting in superior tone quality and advanced realism. :eek: :eek:

PMI uses 96 megabytes per note! What a horrible waste! :p :p

[Ouch that hurts]

Sorry, but IMO that article's just bulls***. The proof of the pudding is in the eating, and to my ear (as a trained pianist) the "MMP"s that he dismisses sound a million times more beautiful that any tiny synth piano I've ever heard.


The reality is… there are sixteen megabyte pianos that sound better than most MMPs of many gigabytes and more. So why would listeners prefer the sixteen megabyte piano over a five gigabyte one? And why would producers and writers prefer to use such a small piano when MMPs are available for less money?

Convenient lack of detail and example. Which exactly ARE these 16MB pianos that sound so much better? If he were to name names, then maybe we could do a comparison, which might at least be good for a laugh.

And who are these listeners that so prefer the 16MB pianos? Anyone done a survey? Anecdotally, I have people say that the pianos in my mixes sound "real", and are such a relief after hearing horrible artificial ones.

There is one point of course, that all choice of sound is dependent on context. Rimsky-Korsakoff said "there is no such thing as a beautiful sound", going on to explain that any concept of beauty belonging to a particular sound has to do with its appropriateness for the context. And there may well be contexts where the sound of a blatantly synth-like piano is more appropriate than the sound of a real one, or a convincing emulation.

But hear them both playing a varied piece of expressive solo music, and I know which one is more likely to get my juices flowing.

[nexus]

:eek: :eek: sixty-four megabytes of samples, for a whole instrument!
with loops and re-tuning, resulting in superior tone quality and advanced realism. :eek: :eek:

PMI uses 96 megabytes per note! What a horrible waste! :p :p

Heh, heh.http://www.northernsounds.com/forum/images/smilies/biggrin.gif

Yeah, Coakley's a good salesman, I gotta give him that. He has a way of convincing people that he's got the 'magic'.

Well he does know how to pack a 32meg ROMpler. I liked his pianos the best until the age of the Gigasampler.

But it's patently ridiculous to suggest there is something 'wrong' with the various multigig pianos becasue they are 'too large'!

I really like the Steinway included in the EWQLSO Silver and use it as well as the piano in GPO but I would NEVER suggest it could compete with these fine 'superpianos'.

I guess the age of the 'Coakleypiano' is truly over and he's got sour grapes over the new era of piano sampling.

Still his creative talents and expertise is an asset to the keyboard workstation companies like Korg and Roland, etc. I wish him success.http://www.northernsounds.com/forum/images/smilies/smile.gif

[Bruce A. Richardson]

Coakley and Bob Lentini should get together and drink beer. Perhaps they could form a less-for-more club.

He is making valid points--they are just not relevant as presented. Pianos represented by too few layers have the annoying consequence of moving back and forth on the soundstage, unless they're filtered half to death, then there's the filtering to consider. And its aftemath. One cannot reduce these shades of gray to such a simple black and white position. At least not with any hope of convincing people who know better.

Sounds like sour grapes to me. I hear what he's saying. It's just that he's not proving anything by saying it. And conversely, he's being proven wrong in many cases by people who are successfully doing exactly what he's railing against. With a well designed session and editing routine, one has absolutely no need to micro-edit every single sample. This is the art of good production design in a sampled instrument. You DESIGN the production from first setup to final distribution, exactly so that you do NOT get into needing to micro-edit everything. From instrument choice, to studio choice, to practice runs, to proof-of-concept instruments, you design your production to eliminate everything he's claiming as necessary.

[JonFairhurst]

The article makes good points: a poorly done piano won't be saved by lots of layers, and it's possible to make a great sounding piano that is efficient.

One thing about huge libs is that they are harder to test for good quality control. We've seen some posts about quality control issues on a few large pianos, and I've experienced some personally.

The good news is that with a huge instrument, you rarely run into the problem. And if you aren't in ROMpler land, you can elbow your way in and fix things, if they bother you enough. Or tell the developer and wait for an art file.

All told, I'd prefer a great piano AND lots of layers. The article takes an either/or position, which is a flawed argument.

I have noticed though that my "big" piano doesn't work all that well live. The volume of the high notes seem to drop off. I need to play around with some EQ and compression/expansion to improve things. It may not be as natural that way, but if it works better when jamming, I'm all for it!

One thing for sure, cramming the same recording into fewer bytes is the last approach I would take for prepping this piano for the stage.

-JF

[David Ferris]

There is one point of course, that all choice of sound is dependent on context. Rimsky-Korsakoff said "there is no such thing as a beautiful sound", going on to explain that any concept of beauty belonging to a particular sound has to do with its appropriateness for the context. And there may well be contexts where the sound of a blatantly synth-like piano is more appropriate than the sound of a real one, or a convincing emulation.

But hear them both playing a varied piece of expressive solo music,...


Ouch,

The "Perfect Piano" is for people who like their pianos sterile.

DavidH

[ohernie]

His comments pretty much reflect my experience with trying to use large sampled piano's in a live setting. They sound great alone at home but when I use them at the gig they don't cut it. NI's "B4" standardized the gigging Hammond softsynth. It'll be interesting to see who creates the standard for gigging piano's. The fact that I have yet to receive a definitive answer from anyone but developers regarding "what piano works for live gigs" says something in itself.

[/soupbox]

I will also say he has struck a nerve. Having spent lots of hard earned gigging money on sampled piano's I can't use I can pretty much say that, beyond basic requirements, I don't give a damn about what room was used, what piano was used, what mics were used, what preamp was used, what A/D was used, what the bit depth was, what recording software was used, whether each note was sampled, how many velocity layers each note has and whether they are looped. What I care about is whether it will fit in a live setting with the sequence that's playing through my Roland M-GS64 (Roland 88 Sound Canvas) or the group I'm jamming with. With or without sympathetic resonance, pedal noise, release samples, etc., the person that creates that piano will sell so many they won't need "group buys".

Think about it. All you developers have to do is create a piano that works better than the Yamaha "P' series, Kurzweil PC2 or Roland SRX11 and you'll have the live piano softsynth market.

Unless you prefer creating piano's you can show off to other developers on the NS forum this sounds like an economic no-brainer to me ...

[\soupbox]

Ernie

[howardv]

I have noticed though that my "big" piano doesn't work all that well live. The volume of the high notes seem to drop off. I've gotten the best results live with the White Grand and the Bardstown Bosendorfer. I've even had occasion to use the Bards for a dual piano performance where we were one short and it stood up rather nicely next to the real deal.

Howard
Ragtime (http://www.rtpress.com) Press (http://www.rtpress.com)

[ohernie]

:eek: :eek: sixty-four megabytes of samples, for a whole instrument!
with loops and re-tuning, resulting in superior tone quality and advanced realism. :eek: :eek:

PMI uses 96 megabytes per note! What a horrible waste! :p :p

Ok, so your amp goes up to 11 ...

Seriously, at what point are we point are we playing a numbers, rather than musical, game? As an engineer my gut starts kicking whenever a reality is defined by the precision and size of numbers.

I just ordered the AV piano's. What I liked about them was the *imperfections* of the piano that made them realistic. My guess is they will have too much ambience for live use, but, I liked the demo's enough to not care.

Ernie

[ohernie]

I've gotten the best results live with the White Grand and the Bardstown Bosendorfer. I've even had occasion to use the Bards for a dual piano performance where we were one short and it stood up rather nicely next to the real deal.

Howard

What did you use for a controller?
What was buffer/latency setting?

Thanks

Ernie

[howardv]

Been using a Yamaha P90 lately. I keep the us122 set for a buffer size of 512 and GS3 set to the minimum 1/2 meg buffer since I don't ever bounce to disk with it. I leave polyphony at 150 and transition voice stealing at 6. Stream to load ratio is 2 to 98. These settings plus the low latency gsif2 midi port make for nicely responsive ragtime and boogie performances.

Howard

[tquandt]

Ok, so your amp goes up to 11 ...

Seriously, at what point are we point are we playing a numbers, rather than musical, game? As an engineer my gut starts kicking whenever a reality is defined by the precision and size of numbers.

I just ordered the AV piano's. What I liked about them was the *imperfections* of the piano that made them realistic. My guess is they will have too much ambience for live use, but, I liked the demo's enough to not care.

Ernie

Ernie, in this case, numbers really count. More bit depth and more velocity layers result in a better reproduction of the real thing. That's what the big sample based pianos are aiming for - a real piano. So if the developers do their job right, their work is invisible and what you get is... well, a piano, not a synth reproduction of a piano. And as I said above: It's the character of the original that they would like to reproduce. They do not want to produce their 'own' signature sound, but mirror the qualities of the original instrument.

There are a few occasions where this is not wanted and not necessary, for example live settings with bands. Probably, there you prefer an artifical piano sound that cuts through the live mix and has a fast loading time.

However, in these settings, a lot of the criticism against big sample libraries could be used against real pianos as well. A huge Bosendorfer would be impossible to carry around and probably wouldn't fit your band mix at all. Does this mean that a cheap M1 piano sound is 'better' - because it is more appropriate to the situation...?

I leave this open to you. However, it reminds me a bit of the discussion whether a church organ is better or worse than a Hammond. Both have their uses, I guess... :|:

[spectrum]

I've been using Ivory in a live band setting recently with really good results. It is interesting that all the beautiful subtleties get erased in a live setting. But for intros and exposed playing, the sound in the live setting is pretty incredible.

Ivory works great for my live use, especially because you have really good onscreen control of the EQ, timbre, velocity curve and hammer amounts....all of which help you tailor the sound to cut through a live band better.

Definitely agree about the live thing being a very different animal and always a challenge....my MKS-20 Piano 3 setting is still one of the best things for using live with a band and I usually mix it in with whatever sampled acoustic piano I'm using too.

spectrum

[midphase]

Eric, can you tell us when and where you're playing next? It would be cool to check out your band!

[tob]

How about 1 sample only piano!!! A beautiful and very realistic 22kb Kontakt piano for all "less is more" minded out there... ;)

www.tobiasmarberger.com/stuff/Smallest_Piano_Ever.zip

/Tobias

[tquandt]

I've been using Ivory in a live band setting recently with really good results.

Hey, shouldn't you be working on the Stylus RMX updates, so that I can finally transfer my Ethno Techno stuff to RMX...? So back to work with you... :D (Just kidding, but I couldn't resist... :p )

[dpasdernick]

While I believe Mr. Coakley's opinion may be biased I believe he has a point. When hardare becomes cheaper (ram, disk space) developers and programmers get lazy. Not just sound library programmers, but OS programmers, graphics programmers, etc.

I'm still amazed at how my Roland SRJV80 expansion cards sound. Recently I bought a multi-gigabyte bass library and after playing it for a while I prefer the sound of my 32 meg Bass and Drums expansion card. PLUS.. it has a ton of decent drum sounds as well. (thank you Mr. Persing!) Listen to the sounds on the Session card by Roland. Think about the variety and quality of the sounds you get all stuffed into 32 megs of memory. Now none of them will compete with a high end library but they still sound great and are very musical.

I really wish Roland would release all of their SRJV80 expansion cards as VST's. Wrap a little player around each library (like Spetrasonics does) and then sell them for $99.00 each.

My 2 cents,

Darren

[Michiel Post]

While I believe Mr. Coakley's opinion may be biased I believe he has a point. When hardare becomes cheaper (ram, disk space) developers and programmers get lazy. Darren
My Goodness, what a nonsense!
Nor Mr. Coakley nor you obviously have any idea how much work is involved in creating a multi gigabyte sample set.
Tell me why is making 32 single 500 kbyte samples more work than recording, editing and selecting over 1.400 much larger samples?

Mr. Coakley must be an enourmously frustrated person to viscously attack his competitors with arguments that are so obviously false.

If you ever want to get back with your feet on the ground and have a good laugh listen to his demos: Perfect Pianos (http://www.williamcoakley.com/piano.php?product=V).
Ever heard such lifeless "digital" piano sounds after giga entered the market? No sustain resonance, no evolving notes, no moving harmonic content except this static william coakleyTM "perfect-loop".

[Bruce A. Richardson]

I like several of the sampled pianos for live use.

White Grand works well. Black Grand works well. GigaPiano II works very well for some settings, but I have altered the response on my version so that the very "clangiest" velocity does not sound. I do the same with White Grand, actually...that 16th velocity is a bit on the bright side for me.

I do know where you're coming from, and actually where Coakley is coming from in a sense. I just think that he's attacking the wrong aspect of the process.

[Bruce A. Richardson]

Want a great sampled piano for the stage? I will be happy to set up and run the sessions. I will need a $150,000 stipend, two beautiful interns who can cut samples and entertain me at night, a sufficient per diem to keep me sated with food and Tanqueray TEN Gimlets, a working rhythm section for one day, and my choice of engineer and studio. I will bring you a sampled piano which will do exactly what you want.

[Jake Johnson]

Coakley registered here several months ago--he's a Northern Sounds member. But he only posted a one or two messages to say hello, and then vanished. Too bad. He's been sampling pianos for a long time. Regardless of how one feels about his arguments, it's a shame he hasn't kept in touch.

[metrobot]

Michiel, I've been using the 16MB perfect piano since the mid 90's - first for my ASR-10, then transfered it with rubberchicken to my EXS24. It's a great piano, and sounds great in a mix consistently. I LOVE IT!!!!!

Well then, I decided to upgrade my piano library because the times have moved on, haven't they? So I ordered your lovely group buy and most of the pianos wouldn't install. (other than the harpsichords and Estonia, which I love). I'm just sayin, before you cast the first stone... :n: :n: :n:

Coakley is a strange guy, but to call his pianos laughable is petty considering how far he moved the art of piano sampling forward.


My Goodness, what a nonsense!
Nor Mr. Coakley nor you obviously have any idea how much work is involved in creating a multi gigabyte sample set.
Tell me why is making 32 single 500 kbyte samples more work than recording, editing and selecting over 1.400 much larger samples?

Mr. Coakley must be an enourmously frustrated person to viscously attack his competitors with arguments that are so obviously false.

If you ever want to get back with your feet on the ground and have a good laugh listen to his demos: Perfect Pianos (http://www.williamcoakley.com/piano.php?product=V).
Ever heard such lifeless "digital" piano sounds after giga entered the market? No sustain resonance, no evolving notes, no moving harmonic content except this static william coakleyTM "perfect-loop".

[howardv]

I revere and honor Coakley for how well my first 5 years of CDs stand up today. But I drop in more modern samples as they come up for remanufacture. Too bad he's not inclined to use more modern tools. Could you imagine?

Howard

[billygoat]

I revere and honor Coakley for how well my first 5 years of CDs stand up today. But I drop in more modern samples as they come up for remanufacture. Too bad he's not inclined to use more modern tools. Could you imagine?

Howard

I agree. I use the 16Mb Perfect Piano in my Proteus 2000 and after months of live inactivity, while I was recording (using Michiel's sampled pianos), I was amazed how well the 16Mb Coakley holds up in a live situation.

William Coakley is a class act. Frankly, I don't know any other developer who's as committed to the quality of his products. At least he takes the time to fix the bugs before he starts selling his products. Compare this to some of the other developers....

If he would ever move into the gigabyte territory of sampling, I am quite sure he would blow everything else away. But that doesn't seem too likely now.

[robgb]

All I've gotta say is that Kontakt 2's Steinway Lite works great for most of my stuff and it's not exactly a resource hog.

[Bruce A. Richardson]

I don't know if I came off disdainful towards Coakley, and if so, I did not mean to be. I was kind of riffing on his tone, but taken out of that context, it's a little harsh. Lentini, on the other hand....hahaha...

I wish that some of the largest pianos had that quality of intimacy and expression that I hear in his. I think the very first demo on that page illustrates this point. That is a very compelling sound to my ear.

Of course, I can hear the sound go sterile when it is sustained too long. But I think that is really not the point. How much of that example sounds really amazing? A LOT of it.

Funny, how this thread touches on some of the same themes we discussed in some other recent threads. To me, a sample which plays with expressiveness and allows me to "push" musically is always more attractive than one which may ultimately "sound better," in a final polished production.

There is also a very real phenomenon of just what "sounds better" in a live setting, or even in a pop session. There are times that what you want a piano sound to do is express its percussive content, and then to recede. In fact, this is very often the case...where you get into mixing and you find that it is difficult to get the piano into a place where it sounds rich and intimate, without bringing a negative conseqence of the mix getting clogged up with sustains and harmonic content.

Coakley knows enough as an engineer that he obviously tailored his sounds towards that end.

In this business, you don't find so many producers who have actually played and produced a lot of music...so some of what Mr. Coakley is saying has some rather uncomfortable validity. I think he is partially right. I don't think he made any friends among his competitors, and I'm not sure that some of his conclusions are really supported by the arguments he makes. But some of them are. There are some very large piano libraries I've played which are not capable of delivering a performance nearly as nuanced as that first demo on his page. They don't deliver that "explosive intimacy" that really lets you express yourself.