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

sounds promising!


if its 6.4 bps against A3 (which is tuned first), F3 it would, if anything be a tad too low, wouldn't it? or did I misunderstand you?



Hmm... I tried not to favor the 5th but go for equal beating instead... but of course I might have favored the 5ths anyway



Looking forward to it. This is most interesting!

Yes, sorry. My F3 is too high compared to yours.

I entered the whole mater recording and ran the comparison in tuning exam mode. Here's thescore,

There are a few odd notes at the extreme ends in the master tuning where I could not get tunelab to measure you recording.

Overall I think my program does a good job, never more off than a cent (EDIT: point, I mean, not cent!).

What do you think?

Kees

Kees, this looks promising, to say the least!!

Note that my tuning, being an actual concert tuning for the next day, had A4 at 442 Hz (hence A4 +8.0). Could you easily "transpose" my figures down to A4= 440 Hz? Would be way easier to compare to it that way. I can do it too, if I find the time tomorrow.

PS does the master tuning setting give you the partials automatically?

I don't know how to transpose yours but I can tune at your A4 and the result is here.

Notice that you actually tuned at A4=441.45Hz. You must be getting sloppy in old age . I really have your tuning at the dissecting table here!

And yes the partials are set according to the PTG exam guidelines.

Cheers,
Kees

The current "open source" solution is that people email me their tunelab files with IH measurements and I send them the tuning. A few people asked for the source code and I emailed them that.

To make this widely available the best thing is I think to have a website where you can upload your tunelab file, and it computes the tuning file for you. If I'd charge 1$ per transaction I'd make enough money to buy a (domestic) beer every 6 months.

Kees

Aah, that's right, I remember now... In placing D5, I moved both A4 and G4 from where I had them tuned originally. I do that sometimes to avoid a D4-D5 that beats wildly.

What I do is that I tune D5, then go back to G4 and A4 and check their fourths/fifths below and see if they can bear a little flattening. Mostly it works well, because often I stretch a bit too optimistically just outside the temperament octave.

This happens in my ET tuning too, by the way. If I find no place for the next note, I have to go back and listen to the 4th and 5th below and adjust them slightly.

What about a small software that could run on, say a pocket PC, iPhone, android (soon), parallel to the tuning software? Kind of those easy video converters... just put your tuning file in, then process, then save as TuneLab/RCT/VT tuning file.

It could be sold separately for a small fee. And it would open up for all kinds of customized tunings.

Apart from php which I work with quite a bit, my coding is too rusty to contribute. But I think we could find the needed skills amongst ourselves and others in this forum. You write apps, don't you? If I don't remember wrongly, Jim Moy also has a coding background.

Jim, *knock knock*

Pat: It's certainly possible with enough effort.
I'd be happy to make my MATLAB code available.

Note there are some quirks. Once the tunelab file has been converted to EBVT it should not be changed anymore in tunelab. In particular you can't tweak the stretch and you can't change the partials anymore. Of course you still can in tunelab as it doesn't know these things but the result will be nonsense.

Kees

Yes Kees, I understand,

what about if you just leave it as be - then tunelab would faithfully reproduce it all, right?

I am sitting here with your tuning sheet right now. I think we are very close. I also think we might need to get a life (friday night and everything). But I really do think it's worth it!

I'll get back with more feedback on the sheet you posted as soon as I've looked at it thoroughly. That is also a verification process that is highly needed, because 1) I could have made mistakes, 2) the piano might have wanted special solutions, acoustically.

The number 2 above is the hardest to implement logically, but I'll see what I can find.

I'm with "accordeur". The directions that Robert Scott gave me were to take samples just like you would normally do to create the tuning curve, then layer the EVBT 3 on top of that. That is how I tune ALL of my pianos now. My customers LOVE the way their pianos sound!! My only question is that the offset numbers that Robert has posted on the Tunelab website for the EBVT 3 are different from the ones that Bill has posted on his site. I have been using Robert's offset numbers. He says they came from a temperament program that he created. It would be interesting to compare the two.

Ryan Hassell
Hassell's Piano Tuning
Farmington, MO
www.hassellspianotuning.com

accordeur, Ryan: There is absolutely nothing wrong with using the temperament offsets for TuneLab.

When EBVT III is tuned by an experienced aural tuner, however, the stretch is quite different from the proportional stretch any ETD use.

This thread is about overcoming that gap, and about looking for a way to get an(y) ETD to reproduce an EBVT III tuning the way an aural tuner can.

Kees,

I checked your data and it looks very close, indeed. Now, wouldn't the smart thing here be to go backwards - for me to tune that very same piano using the examinee tuning you produced, and see what I can find by listening?

If you do it that way you copy the temperament octave offsets all over the range, whereas Bill's offsets change depending on the octave. What you do may sound good, but it's not Bill's EBVT.

Someone computed the offset numbers from Bill's website for him assuming no inharmonicity. Robert Scott computed the offset assuming a "typical" inharmonicity.

Do your customers that love it know it's not ET?

Kees

Yes, that would be the thing to do. If you want to use tunelab I can send you the file. You can find my email in my profile.

Kees

I already loaded your numbers into my RCT, which should work.

I transposed my numbers from my higher at pitch temperament just to get a hunch, and the temperament section looks great!

That said, there is definitely some work needed just above the temperament.
I noticed that I generally lie quite a bit higher in that section (F#4-B5). This is Bill's heritage: he wants to avoid beating 5ths in that area, and clearly favor the 5th. I do not particularly like that sound in ET, but it works well in EBVT III because of the different interval and octave sizes.

Have you used equal-beating here to, or proportional? (2:3) Many of Bill's interval in that 1 1/2 goes even closer to a pure 5th. That would in particular be the intervals where he states "fourth doesn't matter" and similar expressions in his written instructions.

But this is overall very cool, Kees! I'll try to tune that same C5 tomorrow, get aural verification, and get back with my findings!

Hmm, I got 6 "errors" in that range, 3 too low and 3 too high.

I used 3/2 everywhere, except in the bass 4/3 of course.
EDIT: And equal beating in the temperament octave.

Kees

At the great risk of saying, "I tried that one time, didn't like the results, so I didn't waste any time pursuing it", this is the problem I see with doing what you have said:

Any and all non-equal temperaments are always computed at zero inharmonicity. Any and all ETD's first calculate what would be a proper ET for a piano, based upon what is known as its "inharmonicity profile" (a sampling of the inharmonicity characteristics that the piano actually has). If one wishes to tune any non-equal temperament, the differences which that temperament has from ET are then applied to that calculation. This has always been seen as a correct and viable way to tune any non-equal temperament. It works, at least theoretically and on the surface.

What Robert Scott did, at least as I see it, was to find some "typical" inharmonicity characteristics and directly apply them to the published offsets for the EBVT. I discussed that with him and frankly don't understand why he did that. It does not seem logical to do it.

As an analogy, you go to an optometrist because you are no longer seeing clearly with the lenses you have been wearing for a few years. With your bare eyes, the optometrist finds the new corrective values for you. Then, for some inexplicable reason, the optometrist also applies to the corrections that are unique for you, a "typical" correction, an average of all corrections that are known. What you now would see with the new glasses is "over corrected".

When I tried these figures, I had a free download of Tunelab on a new PDA I had purchased. I was not very familiar with either one. So, bearing that in mind, the results were not at all representative of the EBVT III as I know they should sound. They were a well temperament, yes but more like an 18th Century style than late 19th Century. The results were over corrected from ET.

A well temperament of a wide diversity can and does appeal to many people. Therefore, I do not doubt that your customers have enjoyed the tunings that you have produced. However, I can see no way possible that those results match the aural description of the EBVT III as I have written them or as they are represented in the graph by Jason Kanter. You should use the published figures for the EBVT III if that is what you want to tune and not the altered figures that Robert Scott calculated.

That is my advice and opinion on that. Robert Scott is invited to reply and contribute to this discussion. I have nothing but respect and admiration for his accomplishments and contributions to the tuning profession. I just have a problem with this one, particular item.

This is where I see the principal difference in the overall sound between a piano tuned in the EBVT III (or really any WT) with the octaves as I treat them and any ET method that "pushes" the octave stretch such as Herr Stopper's, Signore Capurso's or the Cordier, Lucas Mason (or other) "ET with pure 5ths".

Any WT has the slower M3's combined with 5ths which are tempered more than in ET. The faster M3's are combined with the pure (or nearly pure) 5ths.

If ET is tuned with a wider than beatless octave such as a 6:3 type or larger, it will cause all M3's, M6's and 4ths to beat faster. That produces a certain kind of effect. The 5ths sound "cleaner" or in the case of the ET with pure 5ths, they are actually beatless. There is a remarkable "purity" and smoothness to this effect, however it is at the expense of the "warm" sound of the other more gently beating intervals.

All harmony is made more consonant in one respect but more dissonant in another. Many people describe it as "tight" or "steely". This can be a fine approach for a large concert grand that will play a piano concerto from the Romantic period or later. It will serve to make the piano project well over the orchestra and compete well with the strings who have the natural tendency to push the pitch higher in their upper registers.

On the other hand, tuning a WT with the octaves balanced with their corresponding 5ths, the "warm" harmony of slower beating RBI's is retained, yet the unpleasant sound of narrow 5ths and octave-5ths is mitigated in the 5th and 6th octaves. The warmth is retained but the beneficial effect of more highly stretched octaves is also realized.

For the pure (or nearly so) 5th and more rapidly beating RBI combinations in a WT, the octaves are far more conservatively stretched, so the tighter and tarter sound of the more rapidly beating intervals is controlled and reined in rather than exacerbated.

The result is a "best of both worlds" effect. There is key color and contrast but the adverse effects which are often anticipated with a non-equal temperament are largely suppressed.

The idea of using a moderate stretch in the central octaves for ET but "taking off", so to speak in the outer octaves can work too but at the expense of having overly wide single octaves in the 6th and 7th octaves.

The EBVT III with tempered octaves is but one solution. It is, however, as far as I am concerned, an optimum compromise for all of the above.

Actually, I did not create my EBVT offsets by applying correction to any published offsets for EBVT. Instead I attempted to derive offsets from scratch using Bill's aural tuning directions as the only basis.

Now I understand the limitations of using only 12 offsets to define an unequal temperament. Bill's aural EBVT tuning instructions cover more than just the one octave, so no set of 12 offsets can ever completely describe a temperament that is inherently a full-scale tuning. But one can at least hope to match the aural instructions in that one octave, and then further hope that the discrepancy outside of that one octave is not too bad.

If we take the aural tuning directions as the gold standard, then the goal of coming up with the 12 ETD offsets should be to try to duplicate the results of the aural tuning instructions as closely as possible. One could imagine doing this with a real piano. Suppose you had a very nice piano with a very well-designed scale and average inharmonicity profile. Imagine following the aural tuning directions for EBVT-III on that piano. Then imagine carefully measuring the offsets from that piano as compared to an ET tuning on the same piano. These could then be used as the 12 ETD offsets, at least on this same piano. And, subject to the limitations discussed in the previous paragraph, these offsets could be used to put a reasonable EBVT-III tuning on most any other piano.

The problem with doing this with a real piano is that every tuning has some error. Even on his best day, Bill's EBVT-III tuning is not going to follow his tuning instructions precisely. And even if they did, the process of measuring the pitches afterwards introduces some little error. So to avoid these two sources of error, I devised a software simulation of the "real good piano" imagined above. My software simulation is called "Temperament Designer", and it is a free download from my website. It is not an easy program to use, and to my knowledge, no one besides me has ever actually used it. Anyway, what this simulation does is calculate the consequences of any specific setting of pitches for about 2 octaves. Using the functions provided, you can "tune" any note to any other note based on beats difference or cents. It is slow and tedious, but not impossible to follow the EBVT-III tuning instructions step by step until the simulated piano is completely tuned. Then you can read out the offsets from ET for one octave and that is how I developed my 12 EBVT-III offsets. They are similar, but not identical, to the ones published by Jason Kanter.

But if I run that same simulation with inharmonicity set to zero, then I get the exact 12 offsets that Jason published. So I know that the simulation is basically correct, and I also know that Jason's 12 offsets were derived assuming a zero-inharmonicity instrument.

Now the argument has been made that it is inappropriate to take inharmonicity into account when doing this because inharmonicity has already been taken into account in constructing the underlying Equal Temperament tuning to which these offsets will apply. So it is sort of like applying them twice. But that is not what is happening. Taking inharmonicity into account when running the simulation is just being true to real piano that it models.

Let me give a very simple example. I wish to develop ETD offsets for an unequal temperament I want to call the One Perfect Third temperament. This temperament has a beatless major third, A-C#. I don't care what else it has. For my purposes I am going to focus only on this one fact. That is the aural tuning direction: Tune A-C# beatless. And just to be specific, let us say that this means the 5:4 major third, using the lowest possible coincident partials. Now if we follow the conventional wisdom of calculating the offsets based on zero inharmonicity, we would have an ETD offset for C# of -13.69 cents from ET (where A is offset 0 cents). Now suppose we were to apply this offset to a real ET piano tuning. I picked a sample Mason & Hamlin A from the TuneLab samples. In that ET, the 5th partial of A4 is 2217.135 Hz and the 4th partial of C#5 is 2234.107 Hz. After applying the -13.69 cent correction to C#5, the 4th partial now becomes 2216.510 Hz. This is close, but not exactly equal to the 2217.135 Hz from the 5th partial of A4, so this major third will not be beatless as promised.

So this illustrates that using correction figures that were based on zero-inharmonicity will not exactly achieve the aural goals that were stated. Taking inharmonicity into account during the simulated tuning reduces this error considerably. It does not eliminate the error entirely, because the assumed inharmonicity in the simulation may not match the real inharmonicity of the piano to which it is applied. But it will match better than if zero inharmonicity were assumed in the simulation.

I'm sure my code does the same as Roberts "temperament designer" but for all 88 notes and scripted.

Robert's computed EBVT3 offsets should be better than the zero IH offsets from Bill, unless the piano you have is closer to zero IH than to "typical".

Kees