Piano World Home Page Forums Our Most Popular Forums Piano Tuner-Technicians Forum Possible EBVT III implementation on TuneLab

Kees: Same thing here, just got back from concerts during the weekend. I will have a chance to work on the U3 with your new numbers tomorrow, though!

Thomas: Sounds interesting. I'll compare yours to what we got earlier.

All: I like this thread, its alive, active, and people are contributing. I really think this is going to get us a great end result.

Jim (and other skilled/interested programmers) - I've never worked with the win mobile operating system. What are the programs coded in?

We spoke briefly earlier about the possibility to have a web file making the temperament files, and if that's an option I could participate, I have a pretty solid knowledge of PHP. But the ideal would be a small win mobile program that could be installed side by side with TuneLab on the handheld, and process the files fast and without the need for a network connection.

Best of all would be to get this implemented as an extra function of TuneLab itself, so if this takes off and Robert feels up to it, this would definitely be something to experiment with.

Take care all, I'll be back with the U3 aural evaluation tomorrow.

Thomas:

I like your idea, so let me explain your idea in my own words to make sure we are on the same page.

My original motivation for implementing this was to see if aural tuning instructions could be accurately simulated. This seems to be the case.

Now we can consider if the same results can also be obtained on an ETD without the somewhat cumbersome and very specific aural tuning methods which were after all designed for humans.

Starting with the temperament octave, the numbers I get by simulating the beat matching aural recipe are in practice indistinguishable from just using the offsets, as computed by Robert Scott for the average piano, or even Bill's zero IH numbers. Or we could have 3 sets of precomputed fixed offsets for concert grand, upright, and spinet.

Now if my understanding is correct the temperament extension to the midrange is designed by Bill with the goal of smoothing the beating narrow fifths in the temperament octave which would beat twice as fat in the octave above F3-F4. This must mean that the temperament moves a bit towards ET when you go up. Going down all beat rates slow down so there is really no reason to worry about the narrow fifths and any stretch scheme would suffice.

So instead of my complicated aural simulation method we can just gradually reduce the offsets from ET when we move up from F4 (your scheme is such a method). So we can just take the ET tuning produced by any ETD and apply the offsets which are now not the same in each octave.

I think this is what you have in mind Thomas, correct?

I like this approach much better than my own for practical purposes. It applies not just to EBVT but to any well-temperament if you are worried about having rapidly beating fifths in the upper midrange. And you would still have all the functionality of tunelab in hand to craft your own tuning curve, which I currently have taken over completely.

A simple general method would be to multiply each of the 12 offsets of the temperament octave by a factor t which deceases exponentially, I guess by about a factor 2 per octave up to keep the "bad" fifths from becoming too much. It can just stay 1 below F3.

Once Pat is happy with the tunings I produce now I can compute these sliding offsets and see if this method gets close enough.

Kees

Re: implementation issues.

A web based program that reads a tunelab file and returns your EBVT custom file would certainly be the easiest to implement, but I'm not sure we can assume every tuner will have internet access on location. I'm thinking of for example Jeff who tunes in rural areas it seems (looking for a smiley for "tongue in cheek).

Regarding talking Robert Scott into implementing this in tunelab, if I were him I would not be interested as this takes away a lot of the nice functionality in tunelab to customize your own stretch curve. However if Thomas' approach can be made to work, the only extension would be an optional 88 note temperament file which then can be stretched in a customized fashion with the usual tunelab method.

Finally, taking into account the small market, we should not exclude the current implementation which is that people send me an email I send them back the EBVT offsets I compute.

Kees

I think you understand my concept well, but my 88 note offset file is a little more complicated than simply multiplying the EBVT III offsets for each octave. While this produces different sized octaves, it does not take into account the offsets of the notes below you are tuning from. Octaves would be different sized, but not in the way that Patrick and Bill size them.

When I examined the offsets that you generated, I thought that when I would compare it to an ET tuning, the offsets would follow a smooth curve. To the contrary, the offsets followed a sinusoidal (roughly) pattern, which ended up close to ET by the exterior octaves, but not as close as I would have thought.

As you can see, I have roughly imitated your overall stretch and ET offsets with this method, and it follows that same sinusoidal pattern that your's does. Now we just need to see if it works on real pianos, and I'll have time to test this out on real pianos. I'm not an expert, but I should be able to tell if I am actually creating the temperament, mindless octaves, and compromises between the fourth and the fifth, as Bill Bremmer and Patrick Wingren are skilled at.

For those who want to try this, here is the EBVT III 88 note offset that I created. Simply save these numbers as a .tun file, and add your inharmonicity measurements to match the rest to the piano. The numbers will appear as custom offsets that will be added to the tuning curve. To save the tuning, save it as a new file to preserve the template.


A0 6 0 0.08
A#0 6 0 0.61
B0 6 0 -1.12
C1 6 0 1.38
C#1 6 0 -1.11
D1 6 0 0.95
D#1 6 0 -0.19
E1 6 0 -0.88
F1 6 0 0.78
F#1 6 0 -1.09
G1 6 0 1.77
G#1 6 0 -0.56
A1 6 0 -0.51
A#1 6 0 1.13
B1 6 0 -1.12
C2 6 0 1.39
C#2 6 0 -1.72
D2 6 0 0.77
D#2 6 0 0.25
E2 6 0 -1.15
F2 4 0 1.03
F#2 4 0 -1.08
G2 4 0 2.01
G#2 4 0 -1.17
A2 4 0 -0.6
A#2 4 0 0.98
B2 4 0 -1.09
C3 4 0 1.72
C#3 4 0 -1.15
D3 4 0 -0.12
D#3 4 0 0.78
E3 4 0 -1.17
F3 2 0 0.58
F#3 2 0 -1.08
G3 2 0 1.82
G#3 2 0 0.02
A3 2 0 -0.91
A#3 2 0 1.48
B3 2 0 -1.07
C4 2 0 2.2
C#4 2 0 -2.35
D4 2 0 -0.29
D#4 2 0 0.47
E4 2 0 -1.39
F4 2 0 0.58
F#4 2 0 -1.08
G4 2 0 1.82
G#4 2 0 0.02
A4 1 0 -0.91
A#4 1 0 0.9
B4 1 0 -1.2
C5 1 0 1.48
C#5 1 0 -1.37
D5 1 0 0.43
D#5 1 0 0.38
E5 1 0 -1.16
F5 1 0 0.89
F#5 1 0 -1.08
G5 1 0 2.01
G#5 1 0 -1.17
A5 1 0 -0.6
A#5 1 0 0.98
B5 1 0 -1.23
C6 1 0 1.39
C#6 1 0 -1.72
D6 1 0 0.77
D#6 1 0 0.25
E6 1 0 -1.15
F6 1 0 0.29
F#6 1 0 -0.54
G6 1 0 0.91
G#6 1 0 0.01
A6 1 0 -0.46
A#6 1 0 0.64
B6 1 0 -1.18
C7 1 0 0.86
C#7 1 0 -1.22
D7 1 0 1.22
D#7 1 0 -0.39
E7 1 0 -0.88
F7 1 0 0.93
F#7 1 0 -1.15
G7 1 0 1.7
G#7 1 0 -1.44
A7 1 0 0.08
A#7 1 0 0.61
B7 1 0 -1.19
C8 1 0 0.84


Below is a link to the excel worksheet that I used to establish these offsets.
http://www.box.net/shared/mkng4dfh3q

You could change the numbers in the leftmost column, as I manually entered those in. The rest of the numbers should fill in automatically.

The last three columns are, from left to right:
Calculated offsets
Manually typed offsets (for pasting into other programs)
An Error check (should read zero if the calculated offsets equal the ones you typed in)

On the far right, it shows what the offsets would be if you treated these numbers as different temperament numbers for each octave.

I slightly changed the formulas for the octaves directly above and below the temperament, so that they would favor (in theory), the Fifth and Octave over the Fourth.

Please feel free to critique this method, and offer any opinions you might have.

Regards,

Would I be breaking any laws if I simply shared the EBVT.tun file I created in this forum?

No.

I stand corrected. I actually tried what I proposed and it didn't work and I realized what you just wrote is the reason.

Can you formulate your method to apply to any WT?

I would suggest, if Pat is willing to put in the effort, to try your method aurally. If it works as well as my "brute force" algorithm it is much more practical, as you can still control the stretch in tunelab according to taste. You can just produce an ET tunelab file with your preferred stretch and add the offsets. I like it!

Kees

Okay! Here it is:
http://www.box.net/shared/z0dj6l0a8e

Open this tuning in Tune-lab, sample your notes like normal, and hold tight, because you are ready to try the EBVT III experience! (It probably won't be as fun as I tried to make that statement, but it should work . I'm excited by all this!).

I'll report back tomorrow when I try this on real pianos. I may try to record it, but don't cross your fingers...

If you use my Excel worksheet, just change the UT offsets in the far left column (column "C"), the rest should happen automatically. Then manually copy the calculated numbers (column "H") into column "I". Column "J" should read zero if you copied everything correctly (this is a simple error check, some may read at worst, +/- .01, depending on how Excel rounded numbers).

I would imagine it would work properly on any temperament. [Edit: As long as the offsets are fairly close to ET. I somehow remember Bill Bremmer writing that his method works as long as nothing goes beyond 5 cents of ET]

Hi Thomas,

This is great...anxious to give this a try on my friend's Young Chang 6ft grand....question...how do I get .tun file into my Iphone Tunelab?

Things are happening here! I will tune both temperaments (Kees' and Thoma's) tonight (which should be your early afternoon) on two different U3's in the same room. This will be most interesting. I'll record it on video and post it in this forum.

GPM: by bluetooth, or (the more complex, but bullet-proof way): Send an e-mail to yourself with the .tun as an attachment, open the mail on your iPhone, save the attached .tun file

Hi Patrick,

Thanks for the tips....sNot familiar with bluetooh in getting files to my Iphone. I did try the email, but my Iphone is not allowing me to download the .tun file, it shows it's there, but there is nothing to click.


Robert, is there a way to get this .tun file into the Iphone?

That would be Microsoft eMbedded Visual Studio, for versions of PocketPC that Tunelab runs on, and the language would be C/C++. The iPhone is primarily Objective-C, but with the ability to include C/C++ as well in Apple's Xcode. Android phones are Java using the Android SDK and optionally the Eclipse IDE, with C/C++ possible as a secondary add-on.

Secondary development environments are available for all platforms. Kees noted that Python might be a nice environment in which to perform a port of the code, and interpreters exist for this. The most recent iPhone dev guidelines provide a loophole that may provide a way to run Python code on iPhone. But if the desire is to run in multiple operating systems the clear choice is C/C++.

I would expect that to integrate with Tunelab in any way Robert would expect C/C++ code. (Though I have no expectation that Robert would be interested in this at all, but only he can say.)

I don't have much time to spend on this project, so if there are any other coders with interest in this and have spare cycles, please jump in! I am, unfortunately, a Matlab newbie, so that slows things down on my end. But I've written code for practically all the notable mobile platforms since the late 90's so maybe that balances out a bit.

I tuned the midrange of my Chickering with the numbers from Kees' April 06, 2011 11:29 AM post. The temperament comes out pretty good, a recognizable EBVT, though with a few half-bps discrepancies compared to when I tune it aurally. The foundation 6bps intervals were all fine. I took the iH numbers a few years ago, so perhaps I need to revisit them. I'll see if I can document the discrepancies better.

Kees is correct. I probably won't want to change TuneLab to include this custom EBVT tuning file generation algorithm. But you can do this without changing TuneLab. Just write a separate app for the Pocket PC. This app could read the TuneLab file for the IH constants, discarding everything else, and then write whatever custom version you want. Then just read the tuning file into TuneLab.

This is easy for the Pocket PC and for the laptop platforms because all the document files are available to all the apps. But in the iPhone, the documents are sandboxed so apps cannot read each other's documents. Fortunately there is a workaround (in answer to GP's question). iPhone apps can be enabled for "file sharing", and our app is so enabled. Here is the Apple web page that explains how.

Robert,

the way this thread is developing it seems there might very well be the simplest possible solution available, and it works because of the way you built the software.

I didn't realize until tonight how powerful that software really are. I absolutely love the fact that you can have "dynamic", relative custom offsets, clinging on to the tuning curve you adjust to your liking.

I also liked the combination of a phase and spectrum display. The latter comes in really handy in the treble. Nice falling time on the graph too, It gets slower higher up, right? Very natural.

Last but not least I like that you can enter iH manually, and that tuning files are stored in single text-based files.

The only thing i miss is RCT's "EQ" function, which gives a little more room for creative play.

But I'm sold, I definitely need your program. I put in an order first thing when I got home

@Kees, Thomas: I did the aural verifications tonight. The bad news is that the 6:3 approach over the break didn't work. The good news is that your (Thomas') approach is just a handful of small note tweaks away from creating an EBVT III for ETD use. The possibility of attaching the numbers as offsets in the format of an ordinary .tun file, and by that being able to stretch to your liking, is, just like you (Kees) said, definitely something to like.

I think the result was nothing short of spectacular. I will have to go straight to bed now due to an early morning meeting, but I will post the video talk-throughs tomorrow.

I have some good news too. I tried Thomas' approach but keeping the full aural simulation. I just fit a tunelab style tuning curve through the computed EBVT offsets and that is my reference ET. I then compute the offsets of EBVT with respect to this reference ET and store them as custom offsets in a tunelab file. This is exactly what Thomas did but keeping the full aural simulation.

The good part is that the offsets are less than a cent apart across the piano's I tried, from Yamaha C5 to Hellish Helsinki. So these computed offsets can be
used universally and distributed in the form of an EBVT3 template, which you can then manipulate with the full functionality of tunelab. The offsets just act as a kind of 88 note temperament.

So no coding is required, just the precomputation of a few tunelab template files for, say 3 types of pianos, either with my or with Thomas's method, whatever works best!

Remaining problem: how to deal with the part just across the break where the suggested 6:3 doesn't work. I thought the problem would be E3 is too low, as F3F4 is 4:2 and E3E4 is suddenly 6:3. If that is indeed the problem I can try a smooth transition from 4:2 to 6:3 octave across the break.

Kees

Here's my calculation of a "universal" EBVT3 template.

To use it save a a .tun file, measure your inharmonicity, probably change the partials to something you prefer and it should work on any piano, just like Thomas'. Note that I keep A4 at A=440 so comparison with Thomas' data is a bit cumbersome and I'm running out of free time at the moment.

Note the custom offsets are partial independent, and once you measure IH tunelab will compute the 3d column of offsets which are the normal smooth stretch.

Kees

A0 4 0.0 -0.10
A#0 4 0.0 1.71
B0 4 0.0 -1.15
C1 4 0.0 1.54
C#1 4 0.0 -1.30
D1 4 0.0 -1.30
D#1 4 0.0 0.42
E1 4 0.0 -1.64
F1 4 0.0 1.11
F#1 4 0.0 -1.75
G1 4 0.0 0.53
G#1 4 0.0 0.08
A1 4 0.0 -1.17
A#1 4 0.0 1.33
B1 4 0.0 -0.76
C2 4 0.0 2.48
C#2 4 0.0 -1.35
D2 4 0.0 -0.38
D#2 4 0.0 0.71
E2 4 0.0 -1.23
F2 4 0.0 0.87
F#2 4 0.0 -0.95
G2 4 0.0 2.33
G#2 4 0.0 0.22
A2 4 0.0 -0.52
A#2 4 0.0 2.21
B2 4 0.0 -0.33
C3 4 0.0 3.36
C#3 4 0.0 -1.32
D3 4 0.0 0.71
D#3 4 0.0 1.50
E3 4 0.0 -0.31
F3 4 0.0 1.81
F#3 4 0.0 -0.10
G3 4 0.0 3.09
G#3 4 0.0 0.89
A3 4 0.0 0.05
A#3 4 0.0 2.68
B3 4 0.0 0.05
C4 2 0.0 3.63
C#4 2 0.0 -1.14
D4 2 0.0 0.78
D#4 2 0.0 1.47
E4 2 0.0 -0.46
F4 2 0.0 1.65
F#4 2 0.0 -0.68
G4 2 0.0 2.07
G#4 2 0.0 0.70
A4 2 0.0 0.00
A#4 2 0.0 1.39
B4 2 0.0 -0.75
C5 1 0.0 3.49
C#5 1 0.0 1.16
D5 1 0.0 2.20
D#5 1 0.0 1.85
E5 1 0.0 0.53
F5 1 0.0 2.25
F#5 1 0.0 -0.07
G5 1 0.0 3.27
G#5 1 0.0 -0.27
A5 1 0.0 0.22
A#5 1 0.0 1.82
B5 1 0.0 -0.52
C6 1 0.0 2.27
C#6 1 0.0 -1.35
D6 1 0.0 0.92
D#6 1 0.0 0.52
E6 1 0.0 -0.86
F6 1 0.0 0.82
F#6 1 0.0 -1.47
G6 1 0.0 1.96
G#6 1 0.0 0.05
A6 1 0.0 0.16
A#6 1 0.0 0.63
B6 1 0.0 -1.15
C7 1 0.0 1.79
C#7 1 0.0 -0.57
D7 1 0.0 1.59
D#7 1 0.0 -0.37
E7 1 0.0 -0.79
F7 1 0.0 0.87
F#7 1 0.0 -1.43
G7 1 0.0 1.68
G#7 1 0.0 -1.84
A7 1 0.0 -0.37
A#7 1 0.0 0.34
B7 1 0.0 -1.37
C8 1 0.0 1.05

Sorry if the .91 cent offset is messing people up. I wanted to be able to compare overall stretch with ET, without the EBVT III graph being translated a little sharper due to the nature of EBVT III. I can change the numbers if you want, so that A4=A440.

Yes, please!

"...across the break," do you mean across the boundary between the temperament octave and the tenor section below?

Just trying to follow you guys, you're going pretty fast!