My very OT project...

[Ti Designs]

There are some things most people just accept. For example, bar-cons are used for shifting on just about every TT bike. They suck, you can't shift your gears without shifting your position, but that's all there is. I don't use aero bars, so I don't really care (this is where fans of Di2 should speak up). I do play piano. Everybody else accepts the fact that a piano sounds and feels like a piano, an electronic piano doesn't. Some companies put speakers into their electronic pianos, I guess if you think your ipod sounds like a piano this is OK. I've played real pianos, there's no comparison - yet...

A piano is a complex device with hammers and strings and a soundboard. You never grasp how complex it is until you try to build something that sounds like it. My first attempt was a soundboard with tactile transducers attached. The power behind a real piano is the soundboard, and a Steinway D has 16 square feet of soundboard. I used 8 square feet with three transducers covering the frequency range. It turns out that waves have nodal lines with add or cancel, so any given chord sounded worse than you can imagine. I scrapped the sound board idea for something I knew a little better, sound reproduction. My next try was just a large PA speaker built into an extension I added to my piano stand, and it sounded just like that. My third attempt (the one pictured) was my first real stab at recreating the sound from a spacial perspective. First, a piano has a single soundboard, so all the notes are in phase. When using speakers to recreate that I had to put all of the voice coils in a single plane. A real piano has the low strings on the left side and the high strings on the right, which was how I arranged my individual driver cabinets. You hear two signals with a real piano, the soundboard and the reflection of that from the piano case. My speaker cabinets have drivers mounted top and bottom, the top is driven by the signal from my Yamaha CP33, the other side is driven by a reverb processor being fed the same signal.

The piano in the picture isn't bad, sounds like a good upright piano. It's not the final project by a long ways. First, I'm gonna need a whole lot more power and a whole lot more bottom end - the next one uses an 18" sub with 500 watts to drive it. it needs to be wider too, the whole width of the piano. The real issue isn't size or power, it's the reactive dynamics of a real piano. This idea that when you press a key it makes noise, if you don't it doesn't is incorrect. Let off the damper pedal and it makes noise. Stomp your foot on the floor and a piano makes noise. I'm now playing with a pair of overhead condenser mics feeding a Lexicon LXP-1 loaded with their piano reverb to make the room sound like there's a piano there. And then there's the issue of what's called sound stage. Each section of piano strings has characteristics, it's not simply low to high. The wave form of a low note is too complex to be reproduced by a woofer, but your ears have to hear it as a single wave, and still be able to tell location between high notes and low notes. Each speaker cabinet needs to be full range single point source drivers with the lows extended on the bass units.

There are a zillion details to go over. My next attempt uses active crossovers and multiple amps. There are a number of mistakes I made with those cabinets in order to fit them into the frame, that needs to be corrected too.

All this 'cause I don't like working on bikes...

[learningtoride]

My younger sister is a composition major at the U of A and I found your project really awesome! She plays on both an electronic and standard piano at home and enjoys the gizmos at the electro studio on campus. I would love if you kept us up to date as you go with this and will be sharing this thread with her. Good luck on it's continued growth! :beer:

[rounder]

Cool. {lay a song or a sample. Promise i will not have any thing bad to say.

[Lifelover]

Very cool but at what point does it become just as easy to play/have a piano?

[Ti Designs]

The sample is going to have to wait until it's back together - I'm testing some new multi driver cabinets and the active crossover. I've been listening to pink noise all day...

It's not just the cost of the piano, it's the cost of the house I would need. I have been so close to pulling the trigger on the purchase of a piano so many times - once while I was in school and still living in a dorm. This is far more of a learning process than just an end result for me.

[Bob Ross]

While I think you may be oversimplifying the physics of piano sound (eg, "a piano has a single soundboard, so all the notes are in phase ... You hear two signals with a real piano"), I dig your pursuit of realism via acoustical means rather than electonic (physical modeling, sampling, et al). That multi-driver array behind your keyboard looks like a fun project.

[wasfast]

Very complex project to achieve but admirable. One input I'd suggest is not using 18" drivers. Surface area and power create low end. 18" drivers don't respond as quickly as smaller mass cones/voice coils and are also not as crisp on the very low end, 40-80Hz-ish

[93legendti]

iirc, Ed keeps his house temperature low, which might affect tuning of a real piano.

I'd guess the acoustics of the room are crucial, as the piano has a natural reverb (at least to my ears).

Very cool project.

[Ti Designs]

While I think you may be oversimplifying the physics of piano sound (eg, "a piano has a single soundboard, so all the notes are in phase ... You hear two signals with a real piano"), I dig your pursuit of realism via acoustical means rather than electonic (physical modeling, sampling, et al). That multi-driver array behind your keyboard looks like a fun project.


I think the in-phase in the horizontal plane idea was one of the few things I got right from that attempt. What I'm struggling with now is the position of the drivers in relationship to the player, and the frequency each one has to cover. The low notes are really low base frequencies, the wave itself is as complex as you would expect with a hammer hitting a copper wound wire under tension across a sound board. In a perfect world I would have a digital sample of the low notes sent only to a single point source speaker, and the same thing for the mids and highs. I may wind up doing that, for now I'm planning on having each section extend to the higher frequencies, but compress the high end signal. At the high end I would be expanding the signal to bring it to the front of the sound stage.

As this project grows I'm starting to realize that I often have what I need for the next step. Making the pedals feel right is a good example, the keyboard sees the sustain pedal as instruction to continue the sample output. The same switch also needs to increase soundboard harmonics, which can all be done with a reverb processor that has a pedal input.

The big subwoofer is going to be an 18" driver, it's job is to handle the signal below 25 Hz. I have a Hafler 500 powering it - that should be fun when I switch to church organ. Beyond that, driver selection has to do with voicing. I'm thinking go bright and take the tone down at the EQ, so I'm going with aluminum cones and domes as the primary drivers with soft domes filling in the tonal frequencies. Placement poses a few problems, I may have to switch to coaxial drivers.

I had to do some homework on this project so I went to see Lang Lang play in Boston. He played the first half on one piano - a Steinway D, then switched to another one was voiced brighter and seemed to have a faster action. It reminded me of something I say about bikes - most of that was the player...

[Mr. Squirrel]

dear mr. ed,
you are very creative. when i think of the sunbeams....they scatter (http://www.youtube.com/watch?v=PY1pRmfw9GQ) . please play this for me when you are finished. i will not bother you on your rides with the nice young educated women if you do.

mr. squirrel

[Bob Ross]

I think the in-phase in the horizontal plane idea was one of the few things I got right from that attempt.

Well, that's what I meant about an oversimplification...to say that the soundwaves are "in phase" because they emanate from the same horizontal plane (sic) is to neglect what happens to to those soundwaves -- and especially their harmonics -- the moment they move away from the soundboard.

...which, in the case of a piano, pretty much means any time you hear it. Because nobody listens to a piano at the soundboard, you listen to a piano at a particular distance from the soundboard. And you listen to a piano at a particular angular orientation to the soundboard. And there is a buttload of comb-filtering of harmonics due to phase cancellation when you get any distance (the X axis) from the soundboard. Plus, depending on the angle of orientation (both Y and Z axes), that comb-filtering is different; a piano absolutely does not have a uniform polar response, and so any loudspeaker system which attempted to replicate a piano's acoustic behavior would have to model this assymmetric response in three dimensions.

[Ti Designs]

Well, that's what I meant about an oversimplification...to say that the soundwaves are "in phase" because they emanate from the same horizontal plane (sic) is to neglect what happens to to those soundwaves -- and especially their harmonics -- the moment they move away from the soundboard.

I'm working on the easy parts first, a single horizontal plane seemed like the easiest of them. The active crossover has phase correction per driver which makes this point kinda moot.


Because nobody listens to a piano at the soundboard

You've just hit on another little flaw in my project. The sampled sounds weren't recorded at the sound board. Creating a sample set off the sound board of a real piano is no small task, it has little value to most people who would like their electronic piano to sound like a piano on headphones. And then there's the other little problem that I don't have a Steinway D to sample.


And there is a buttload of comb-filtering of harmonics due to phase cancellation when you get any distance (the X axis) from the soundboard. Plus, depending on the angle of orientation (both Y and Z axes), that comb-filtering is different; a piano absolutely does not have a uniform polar response, and so any loudspeaker system which attempted to replicate a piano's acoustic behavior would have to model this assymmetric response in three dimensions.

There are a few basic theories which work both for speaker drivers as well as vibrating strings. The use of line arrays to control where the signal is added or cancelled can give almost the same results as a string. The soundboard itself is where it really gets complex. The point at which the string meets the soundboard is different from string to string, the energy loss as a function of distance also changes. For right now I'm working with what I know. In testing I hope to learn some of what I don't know. This isn't a lifelong project, a year from now I hope to have something in my home that sounds and feels good. Will it sound like a really good piano in a good hall? Hell no. The saving grace here is that I'm building this thing for myself, not some hall full of people. My sonic target is about one cubic foot, maybe more if I'm doing the Ray Charles head swaying thing.

My first attempt at building a soundboard with tactile transducers was a dismal failure, but it showed me just how complex not only the sound source is, but how the tension of the strings also damps the soundboard. Think of it, 88 possible induced waves and it all works together. I build a soundboard with a single source point and any two keys sound like mud together. Worse, mud with vibrato... I honestly don't know how the piano came out before the computer, but I'm so glad it did. Think about this, the user interface hasn't changed in years. What if MicroSoft had introduced the piano way back when???

[DfCas]

I play a piano technician in real life, and I find your project interesting. If we look at pipe organs compared to electronic organs, no one has yet gotten very close to the acoustic mix that a pipe organ produces. I think its the fact that sounds originate in different places and bounce around in such a complex manner that point sourcing them is the problem.

Note that Steinway and I assume other pianos have a "diaphragmatic soundboard". It is thinned at the edges, so that different parts of the board are better at amplifying different frequencies.

I don't know if you've seen this, but a 16 year old down under built the worlds largest piano. It has some real bass.

http://www.noiseaddicts.com/2009/04/the-worlds-largest-piano/

[tannhauser]

I play a piano technician in real life, and I find your project interesting. If we look at pipe organs compared to electronic organs, no one has yet gotten very close to the acoustic mix that a pipe organ produces. I think its the fact that sounds originate in different places and bounce around in such a complex manner that point sourcing them is the problem.
\
http://www.noiseaddicts.com/2009/04/the-worlds-largest-piano/

Our entire symphony hall is the soundboard! The largest pipes have so much air flowing through them I couldn't imagine how that could be reproducible.

[Hardlyrob]

And then there's the other little problem that I don't have a Steinway D to sample.


Hey Ed - I have a nice A you are welcome to sample - not a D, but a nice piano nonetheless.

[Ti Designs]

Hey Ed - I have a nice A you are welcome to sample - not a D, but a nice piano nonetheless.

Yeh, yours was one of the reasons I started this project. The last north short ride we did I left my house at 6:00am to get to Essex on time, then we rode to your house, I got to see your piano, then back to Essex, have lunch and a few beers, then back to Arlington. All told it was about 140 miles, the last 50 were a race against the sun setting. I got home (very, very tired), turned on my piano (probably still wearing my helmet) and found that I really needed more.

[Ti Designs]

Update:

It's always good to know if you're reinventing the wheel, so I've done some checking to see who else has done this and what their conclusions were. They all came to the same conclusion, this is a speaker that doesn't reproduce sound, given the right sample it creates a piano-like sound. In other words, it's useless for listening to anything but piano. More to the point, it's not any piano. If it's a recording with room acoustics, it's all wrong. For this reason most of them thought this project wasn't worth the time or effort. I have to question this, a real piano doesn't really multitask. Nobody is using their Steinway as a kiddie pool in the summer (I hope...)

Before I start building the next generation of cabinets for this project I'm doing a lot of reading and math. Turns out that what I learned while working for Ratheon in ECM is applicable in this project. I never thought that knowing about con scan targeting would help in real life. The component rack above the keyboard is also growing. I've gone to an active crossover and added dynamic range expanders and compressors. Here's a problem that other speaker designers don't have to worry about: How do you take the same signal and make it sound like it's right there in one case (the high notes on the right side of the keyboard) but sound like it's a more distant part of a much wider signal (the high and mid components to the low notes)? The answer (I hope - I'm still testing) is a combination of power loss over distance and dynamic range. So, the subwoofer will also have a midrange and tweeter to add definition to the low notes, but they will be located on the far side and the signal will be compressed. I've set up my component rack and monitors in a spare room to do some testing. Thus far I've found that nothing good happens from delaying part of a signal...


Does anyone own a Steinway D and a tape measure?

[bobswire]

Ti, now I really understand why my son who was a pretty good musician (organ, guitar, drums) decided to remain a percussionist.
http://vimeo.com/28694952

[Ti Designs]

Update on the project:

The third generation that was originally shown had it's problems in all sorts of areas, but it was a learning experience in terms of enclosure design, crossover building and what not to do with gorilla glue. I moved onto the fourth generation piano speaker, which is almost as large as a baby grand piano and has started to take on the shape of one too.

Things that changed: First, it's now got two rows of cabinets. The first row is the simulation of where the strings are and the signal for that row is run into the BBE signal processor before being split out to individual amps by the active crossover. The back row adds both depth and definition. The woofer to the left can handle the low frequency waves of the lower octaves, but it can't produce the complexity of the notes, which is where the driver behind it comes in. I spent a week trying to get a good grasp of what makes a sound more present or distant, and how I could recreate that. It comes down to signal compression and power loss over distance, so I added a compressor to the rack and I moved those speakers to the back. Using active crossovers means having 7 amps, but it's easier than tweeking crossover networks, and ig gives me phase control over individual drivers. I gave some thought to driver selection as well. The aluminum dome midrange and the planar tweeter on the front row are accurate, but harsh. For the back row I used Morel silk dome drivers.

I haven't made stands yet, so the speakers are sitting on boxes and other monitors for position - that will change. I had tested each speaker for frequency response and I have full control over each amp, but there's still lots of tweeking to be done. I think it's safe to say it sounds like a piano. My target of a Steinway D is unrealistic because I would need a much larger room, but there's lots to explore here. I've been using the best sample I can find which is Yamaha's mono grand sample set, probably taken off a D7 concert grand. I've also added a damper program that feeds room sound into Lexicon's piano reverb (which happens to be in the Alisis NanoVerb - who knew?), so you get the feel of 88 live strings when the pedal goes down. There are a lot of other piano samples out there that I need to try.

With all these toys it's a shame I have to go to work. Who wants to work on bikes all day???

[Ti Designs]

Just a quick update. The hardware part of this is mostly done, and it didn't turn out as planned (none of my projects ever do). Most of what I learned is this: If you do to a speaker cone what a piano hammer does to a sting, you've just wrecked a speaker. Signal compression has it's place in electronic music, I combat it with overkill. Another thing I found is the size and depth of a real piano as well as the shape of the case have a lot to do with what the person sitting at the keyboard hears (Take a look at the Yamaha million dollar piano for an example of messing with a working formula). In the signal processing I needed not just settable delay for the back row of drivers, but individual control. The settings, in delay feet are 4' for the tweeter, 8' for the midrange and 16' for the woofer (cut off at 45Hz) which happens to be twice the case length for a Steinway D. I've also added a few mics, the measurement mic hanging directly above the speaker to adjust the signal processing, the side mics which feed a reverb unit and drive monitors off to the sides to control room acoustics, and finally a stereo set of condenser mics for recording as if it were an acoustic piano.

Now comes the new source. Unless anyone has any other suggestions, I'm going with the new Mac Mini (the people at Apple can't grasp the idea of never hooking a computer to the internet, it's a foreign concept to them. "you need the internet to get updates" is what I keep getting - who updates a piano???), an M-audio interface for both signal and MIDI and Native Instruments classic piano bundle with their own driver.

[christian]

So, just to be clear, this is like a Di2 upright piano?

[Ti Designs]

So, just to be clear, this is like a Di2 upright piano?

Yeh, strings still work just fine...

[Jaq]

What a magnificent obsession you have. Seriously, kudos.

[Ti Designs]

Update:

The piano speaker really works. If I play with my eyes closed I almost can't tell it's not a real piano. It's missing the thunk of the hammers, but the sound and depth are all there.

New problem: My Yamaha CP33 stage piano only has the sampled sounds of a Yamaha C7 grand piano. One advantage of the speaker over a real piano is that I can change from the Yamaha to a Steinway to a Bosendorfer at the push of a button, so it was time to add some new samples. For this I would need a sound engine. I went with the new Mac Mini - everything I needed, nothing I didn't. For a sample set I went with Native Instruments Classic piano collection which has multiple samples of three concert grands and an upright. Using an M-Audio interface I loaded the samples, ran the output to the piano speaker and found there was something very wrong. The samples don't sound as good as the internal sound of my stage piano, that can't be right! Since then I've tried just about everything, the new samples still sound pale compared to my CP33. The folks at the music store didn't buy it, they wanted me to purchase a new audio interface and a new MIDI interface. I had a better idea, I brought my Mac, loaded with the samples and drivers to their store with the agreement that if they could make the new samples sound as good as the internal voices of my stage piano, I would buy the hardware used to do it. At the end of the day the guys at the music store were looking pretty sad, the latest and greatest sample set had gotten it's butt kicked by a stage piano...

[Fixed]

what about the feel of the keyboard ?
cheers

[Ti Designs]

what about the feel of the keyboard ?


My first failed attempt was a soundboard speaker using a Clark Synthesis tactile transducer to vibrate a soundboard. That transducer is now attached to the piano stand and run off a subwoofer amp, so the keyboard has a tactile feel of the bass notes. It's still missing the hammer strike feel, bot overall it's not bad. I still play real pianos to maintain a sense of what's real, it's hard not to notice the rounded edges of the Bosendorfer or the distinctive hammer release of the Steinway. The Yamaha stage piano keyboard is very good, but it lacks anything that would set it apart from the rest.

[Ti Designs]

This is the real piano project going together. It's all together now, I've been playing it for the past few months. I honestly don't miss not having a real piano any more...

Now that the piano project is finished, I need a new project. I'm going to need a new car soon, but there's nothing out there I like. I'm thinking that a titanium tube frame EV could have astonishing performance...

[rnhood]

Nice dovetail joints on those speaker cabs.

[maxdog]

Nice dovetail joints on those speaker cabs.

I concur. Nice execution with respect to your craftsmanship.

[BumbleBeeDave]

Ed, that's beautiful!

How does it sound?

BBD

[Nooch]

it doesn't look like it sits too high off the ground -- are you playing a la schroeder?

[sc53]

Help me understand what you did: you have a electronic keyboard that you wanted to make sound like a real piano? So you built speakers, crossovers, stands and samples of real pianos that all come together to simulate the sound of a real piano whenever you strike a key of the electronic keyboard? Unbelievable! Isn't this contraption just as big as a real piano, though? Or is it smaller than a grand piano, which is what you're simulating? Also, where is the Mac mini that feeds the samples? You have another rack of amps, crossovers, and controllers somewhere?
Jeez I had no idea anybody could dream this up, much less construct it.
Congratulations!

[redir]

I don't really know what you are doing but I dig it. I build guitars so I am quite familiar with the theory of acoustics in guitars but I'd guess piano's are a different animal. One problem I think that can't be solved is the modes of the soundboard when a string is vibrated. But sampling does a great job of that. My wife plays a Roland piano into a JV1080 Roland module and the piano sounds are quite good. She just bought a Nord Electro 3 too and that has great Leslie simulations. But nothing will ever be like the real deal that's for sure.

Those cabinets look beautifully made, nice job.

[Ti Designs]

Help me understand what you did: you have a electronic keyboard that you wanted to make sound like a real piano? So you built speakers, crossovers, stands and samples of real pianos that all come together to simulate the sound of a real piano whenever you strike a key of the electronic keyboard? Unbelievable! Isn't this contraption just as big as a real piano, though? Or is it smaller than a grand piano, which is what you're simulating? Also, where is the Mac mini that feeds the samples? You have another rack of amps, crossovers, and controllers somewhere?

It got way more complicated than just a single speaker and samples. In a real piano any two notes add or cancel across nodal lines, which makes the location of the source of that note important in creating the overtones. The first prototypes were too small, even with phase correction or delay I couldn't get the same effect. The 5th prototype was built on a platform which allowed me to move the speakers around, the finished piano project's frame is aluminum channel, so while the speaker locations was determined by the previous prototype, they're still adjustable.

Another little problem is how the signal is summed. The sample is recorded directly over the sound board, with a mic on each side. My first idea was to recreate the left signal using the 1st, 2nd and 3rd drivers while the right signal is recreated by the 2nd, 3rd and 4th drivers. It turns out that when you add a signal in physical space there is comb filtering, and it sounds just like a piano with a cracked soundboard.

It will probably always have the problem of blown drivers. Speakers are just not good at recreating percussive sounds. Someone asked if I could do the same sort of thing for drums, I said no. I've done everything I can think of to reduce the damage to the drivers, three of those cabinets use tandem in-line drivers to reduce cone strain. In the end I'm simulating a string being hit by a hammer - things are going to break.

It sounds like a real piano, in fact it sounds like a number of pianos. My first attempt was to recreate the sound of a Steinway D concert grand, but change the sample set and the delay times of the rear speakers and it sounds just like a much brighter Yamaha C7. Another chance in samples and turn off the rear speakers and it can sound and feel just like an upright.

For an electronic piano it feels a little creepy at first. There is a measurement mic hanging above which listens to the room noise, when the damper pedal goes down it feeds that to a reverb and adds it to the piano signal, so it feels like there's a real piano in the room. It's also got a tactile transducer under the left side of the keyboard, so you feel the bass at your fingers.

While it's certainly not small, it is still portable. The legs and pedals come off, the frame is about 50 pounds and easy enough to move around. The signal processing equipment is housed in a flight case, the amps have their own flight case and the subwoofer sits under the piano. I can set the whole thing up in under 15 minutes.

The next step in this process is to move it to a much larger room and record something.