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

Thanks for the responses! I don't mean to "hijack" this thread, but have noticed that my process of using the EBVT 3 with Tunelab sounds better on some pianos than others. I don't know if its just the piano. Sometimes the intervals of a 3rd seem to beat a little too fast in the upper octaves for my taste, but this is mostly on small spinets which sometimes need voicing and are not all that great to begin with. Are there certain pianos in which I should not be using EVBT 3 but rather just ET?

I hope I'm not opening a can of worms here... :-)

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

I second everything you write here, Bill - great explanation!

@Robert: Good to hear you speak about this. As we all know by now, what you percieve during aural tuning (concerning stretch a.s.o.) is not easily transformed into numbers. I think we are on to something with Kees here, bouncing the data back and fourth between calculation and aural verification. It will just take some time, and patience.

@Ryan: It's good that this can is opened, and your feedback is needed, too. There should be nothing unpleasant inside, though, on the contrary I think we are gaining new ground!

@Kees: got called on emergency tunings today, didn't have the time to tune the Y C5. Tomorrow night I will most likely get it done, though.

I think we need different inharmonicities, too. If you bear with me, and are up to it, I would suggest three different instruments. That should be enough to establish a pattern - or, if nothing else, show my inconsistency tuning aurally

One of them could be a Steinway D, which I will tune for concert soon. In between, maybe a shorter piano? What do you think?

Anyways, I'll know more tomorrow when I retune the Y C5 using your calculations.

Let me once more say that I find this research (and this discussion) very interesting, relevant, and alive

Pat: I was thinking a large upright and a spinet, but what matters most is to test the "aural simulation" on as wide a range possible.

If you send me IH data (C1C2C3C4C5 (C6 optional)) I can generate the tuning file. Instead of ih data just an mp3 of C1-C6 (i.e. just 6 notes) is easy too.

I understand you are now going to tune the Yamaha electronically with my offsets, and then critically examine it, finding what what you consider errors. I am eager to get those results.

One thing I thought of is that Bill's recipe switches to mindless octaves at F5, but in practice a smooth crossfade between the two will be implemented aurally, which I have not done (but can easily).

Kees

Robert:

I found and tried your "temperament designer". With so much manual work required it seems a bit error prone.

Are you sure your computed average EBVT3 offsets are correct? It seems to me the C is much to high.

Below your numbers compared to one I computed; they are much closer to Bill's numbers.

Robert Scott:
C 6.24
C# 0.71
D 2.07
D# 1.95
E -1.07
F 4.32
F# -1.24
G 6.31
G# 2.77
A 0.00
A# 2.15
B 1.01

Bill Bremmer (0 IH)
C 3.8
C# -1.3
D 0.9
D# 1.6
E -0.4
F 1.8
F# -0.3
G 3.1
G# 0.7
A 0.0
A# 2.9
B 0.0

Yamaha C5 (varified by Pat)

C 3.6
C# -1.0
D 0.8
D# 1.3
E -0.6
F 1.4
F# -0.3
G 3.0
G# 0.9
A 0.0
A# 2.5
B 0.2

Heintzmann 1903 large upright

C 3.5
C# -0.9
D 0.7
D# 1.2
E -0.7
F 1.2
F# -0.3
G 2.8
G# 1.0
A 0.0
A# 2.4
B 0.2

Sample Steinway D (from tunelab demo files)

C 3.9
C# -1.1
D 0.9
D# 1.5
E -0.4
F 1.8
F# -0.1
G 3.3
G# 1.0
A 0.0
A# 2.7
B 0.2

"average" from tunelab demo tun files

C 3.6
C# -1.0
D 0.7
D# 1.3
E -0.6
F 1.4
F# -0.3
G 2.9
G# 0.9
A 0.0
A# 2.5
B 0.2

Kees

Those numbers are from an older erroneous distribution. The latest EBVT-3 offsets currently distributed with TuneLab and available from our downloads page are:

C 3.31
C# -1.08
D 0.62
D# 1.38
E -0.48
F 1.49
F# -0.17
G 2.73
G# 0.93
A 0.00
A# 2.39
B -0.16

If I understand things correctly another error is introduced because these offsets are the superimposed on a tuning curve which is not horizontal. So the A and C# will be shifted slightly by different amounts and you still end up with an out of tune M3.

I agree the difference will be too small to matter in practice. On the other hand the difference between your computed EVBT3 offsets and Bill's zero IH values are also too small to matter.

The dilemma I have with my 88 note computed EBVT3 is that the offsets I compute are absolute, and they can not be used as offsets to the tunelab tuning curve, unless it is flat. So in the tunelab file that my program spits out I set the IH values to 0. You can still manually adjust the stretch, which is very useful if you want more stretch in the bass and treble. The result is very similar to that from doing an aural tuning simulation of Bill's optional large bass and treble stretch.

Kees

Yes, I will tune the Y C5 with your numbers tonight. I'll also do a couple of samples (C, octaves 1-6), maybe (like you say) a large upright and a short spinet-type piano would be a good choice.

Then you can calculate the numbers for them, send them over, and I tune the pianos according to them. You're right, this back and forth process might be the best way to refine the calculations. They are surprisingly good already, but I have a feeling that we can get them even closer.

I am very excited about this, and possibility for creating a small converter software. I think we are close to implementing EBVT III the way it deserves to be heard on an ETD tuning.

Pure 12ths are another thing that could be handled the same way. That is easier, but there it is a judgment that has to be made in the bass (3:1, 6:2 or perhaps even 9:3?)

CHAS is a third. Many possibilities.

Thanks so much for weighing in, Robert. Now I understand what you did far better than I did before and also that you had made some adjustments to it.

Apparently, Ryan Hassel used Tunelab and your EBVT III figures for a concert tuning this weekend. The artists were very high caliber, so was the piano and the music was highly complex, the kind that ET is usually thought of as being the only usable and suitable temperament. He even got the "pipe organ effect" from the tuning, so that means to me that it was an accurate representation.

Thanks again for your input.

I have to add something here (and correct myself, because there seem to be an exception that confirms the rule)... I listened to Alfredo's recording earlier today, and it does not seem to suffer from a wide temperament range nor from heavy stretch just outside it. It's actually quite a remarkable ET. For all the papers and posts by him, I still have really no idea how he does that. I know that Isaac learned directly from him, so there seems to be something that can be thought in person

All I know is this,

Tuning is not my forte. I learnt aurally first over 20 years ago. Achieving a decent temperament, was difficult, unisons were lovely. I got used to it.

Then came along my faithful PT-100 Yamaha, had different scales built-in, my doubts as a piano tuner faded away. I knew I could, but did not understand what I was doing.

Then came along the internet, pianoworld and tunelab. All lovely inventions.

Now, tunelab is like my butler, shows up first, tells me the situation, offers me options (ET, EBVT whatever), there you go. I really enjoy the trade.

Because tuning is not so much a puzzle anymore, I enjoy playing music more than ever. Amen.

Robert, I forgot to ask,

what temperament starting note are you using? If it's C and the temperament thus is C-B, one might just as well throw a game of darts... :-D

Bill's temperament is F-E(F), and an extension of the temperament have to go from that core. This dawned on me in real time earlier tonight (i just got home from having yet another date with the Yamaha C5 of this thread).

Kees: I just went through your calculations, tuning the Y C5 to your numbers, and recorded most of it (and comments) on audio. I'm uploading it as I write, but remember - I live in an arctic jungle.. It will take a while. Meanwhile, I will dwell somewhat on my findings.

There are some general things to consider:

When Bill speaks of a "pure-sounding 5th", it's not necessarily 3:2. I noticed this just now. Bill would not make it dead-on 3:2 if he had to accept a wild beating at 6:4, neither would I. The Yamaha C5 had quite a difference between the 3:2 and the 6:4. Would it be wise to calculate it as a compromise between 3:2 and 6:4?

Same thing about equal beating - there is something fishy there, because the numbers gave me unequal beatings. If anything, the fifth was beating faster.

Consistently, C4-approx G4 was too low. Could this be a result of using C-B as a basis? I don't know, just throwing a feeler.

The stretch into the upper bass (that is, on most pianos, across the break) is wider than I do it. Nothing wrong with that, just something to consider carefully. It takes some warmth away from the upper bass, and prevent the tuner from going for a really wide stretch in the lowest range.

I also made inharmonicity samples for a Yamaha U3 and a Hellas Helsinki (the closest I could get to a spinet ) I will post them as well, as soon as they are uploaded.

That you can almost already do in tunelab by selecting 3:1 in the treble and 6:2 in the bass. I believe Bernhard mentioned a while ago he goes as far as 12:4 in the low bass. You could get that effect by tweaking tunelab's tuning curve.

Similar for Cordier tuning.

Nobody seems to know what chas tuning is, my guess is it's just mindless octaves all over. To get a reasonable bass I assume you need to make a switch to equal beating 8:2/6:2 at some point or something like that.

Kees

Here is a picture of the Hellas piano so you know what your calculations are up against, Kees. The camera angle might play tricks on the beholder, but you should know this is no piano to bring home to the house...




at about 3 ft. 7 in (109 cm), it should be a nice complement to our piano sample stock collection.

These pianos are so old by now, that we are mostly using them for non-pianistic purposes - decoration, sledges, and so on :-D

Here are the links to the inharmonicity samples:

Inharmonicity samples | Hellas Helsinki

Inharmonicity samples | Yamaha U3

@Kees: Finally, here is the sound file in which I reproduce your calculated tuning on the Yamaha C5, and comment on-the-fly. Sorry about possible jumps in logic and about an unstructured file as a whole.

Yamaha C5 - aural breakdown of Kees' calculations (2011-03-27)

That said, there is very much information there. Thus you have the ball now

I am in no hurry, nor am I tired. If you want to pursue a faithful EBVT III implementation on ETD's, I'll follow you through.

The discussion gives examples of a lot of the characteristic choices in setting the EBVT III, so it might be interesting to other curious tuners to.

@Bill: if you find the time, it would of course be of immense value if you could listen through my recreation of your creation, and the way I try to explain it. This would benefit much from a solid, second opinion and there can't possibly be any person more fit to do that than you.

Let's keep the good energy flowing, folks

Thanks Pat.

I thought Bill's "pure fifths" were supposed to be pure 3:2 fifths and he mentioned you should prove them to be so.

If they are aurally pure (equal beating 3:2/6:4) it will be a different temperament, I have actually played with that in the past.

So Bill, what fifths do you really want?

Kees
EDIT:
PS Before I tuned Pat's piano remotely I actually had all "pure fifths" implemented as 3:2/6:4 equal beating in my program but I thought it was the wrong thing to do!
PPS Pat I will examine your knitpicking carefully as soon as I have time. I think the 3:2 versus 6:4 is a crucial issue.
I can't aurally tune a pure 3:2 fifth except by using its equal beating check. Aurally pure fifths to me minimize the maximum of the 3:2 and 6:4 beat rates, which is achieved at the equal beating stage.
PPPS This collaboration is a lot of fun!

Pat, I'm puzzled by the sound of the F3C4 on your recording.
To me it sounds like this, which is the sound synthesized from the IH model derived from your samples.

If the two sounds can not be reconciled the only conclusion I can see is that the tunelab IH model did not apply accurately enough to those 2 notes and either the 6th partial of F3 or the 4th partial of C4 are not where the tunelab model assumes they are.

What do you think?

Kees

Yes, slip of mind - I meant 12:4

Kees, I think they sound very much alike inharmonicity-wise! I bet it's only the amplitudes of the partials that throws us off. And that piano, by the way, has a particularly loud and nasty 9:6!

Then, about the fifths, I bet a lot will solve itself if you put your calculations back to your 3:2/6:4 What would it do for your scoring against my tuning?

Then another thing I forgot to say: the A3-A4 6:3 is more of an estimate. Bill himself starts with A3 from the fork. The placement of A4 happens when you listen to it against A3, D4 and E4. So what happens if you calculate A4 from D4 and E4 beating 2:3? I bet it will get closer!

If you have the time to try these calculations and send me the numbers, I could check the piano again just for that mid-range. It's easy and fast, and I could get back to you with the aural results in no time.

And yes, I really think this is a lot of fun, too - and a great topic to research

It now scores 100% in the temperament range. That was a **** good suggestion.


That's what I'm already doing unfortunately. Actually at no point do I tune any particular octave except F3F4 which is exactly 4:2, it's entirely by balancing beats in 4th and 5ths

A new tuning sheet and examn result with these changes is here.

Let's focus on the midrange for now as you suggested, taking things one at a time. Regarding your aural comments, I am not experienced enough in aural tuning to get everything. I do treasure your half hour recording and consider it a free lesson, but it will take me a while to digest.

When you implement this new tuning on the midrange (say C3-C5) can you write down what you perceive to be the 3 biggest problems for me to tackle? Of course I'll be happy with a bonus recording too.

I can produce a tuning file for the other 2 piano's in 5 mins. I propose to try to get things as good as possible on the C5 first. On the other hand there may be even larger issues on the other pianos. Let me know what you prefer.

Kees