Piano World Home Page Forums Our Most Popular Forums Piano Tuner-Technicians Forum Another ETD thread.

First I want to say I am open to the benefits of ETDs even though I am primarily an aural tuner.

I have discussed the pros and cons of aural vs ETDs with some of the best known ETD tuners.

We seemed to agree that ETDs produce tunings that are "good enough".

Now, I'm here to discuss the use of ETDs with little to no aural corrections, as I dare to say, is the way most tuners probably use these machines.

Having said that, my recent attempt at tuning a piano with an ETD leaves me wondering if an ETD's tuning is really good enough or not.

My reason for saying this has to do with the way these machines seem to measure Inharmonicity.

I just measured F2 10 times and got a variation of the 2nd partial of 14 cents! With people claiming precision (and some even accuracy) down to 0.001 cents, how is it possible that these machines can even come close to a reasonable tuning when they can't seem to agree with themselves on the Inharmonicity of the same string played 10 times mezzo forte?

An excellent question, and one I find endlessly fascinating, mainly because I am not a professional tuner and do not need to find the most efficient way to tune a piano in order to survive.

Measuring iH is extremely difficult. The struck piano string does not exhibit a periodic waveform, due to many factors - iH, bridge rocking, soundboard mechanical resonance motion, plate vibration, longitudinal waves, and the ever present false beats, which every string has to some extent. All these factors make precise short/discrete FFT analysis very challenging. In another thread Roberts Scott mentioned his attempts to meet this challenge in a laboratory setting - and failed.

However, a technique of convergence, where multiple measurements are made of a string and the results averaged, can yield a level of accuracy beyond that which the ear can detect. Even so, since the iH varies with tension, among other things, maintaining the piano at a constant reference pitch is essential if the iH values used are to be trusted.

IMO, the best ETD calculated temperament and stretch can only be achieved if at least one string of every note is measured and used in the calculation and then used as the reference for the remainder of the bi/tri chord. As I have shown in multiple posts over the years, the iH can vary significantly, especially at the wound/unwound break, but also when the string size changes. Each change in string size shows up on my own piano's iH measurements.

Measuring the iH is just part of a proper calculation. The amplitudes of the various partials must be taken into consideration when creating a bias in the calculation for the octave size, theoretically for every octave on the piano. Then the ETD should also consider the overall sound of the piano when arpeggios are played with the sustain pedal pressed. The interaction of the 8th and 16th partials of C1 can be distinctly heard and beat like crazy with C4 and C5 if not considered in the calculation.

The Entropy Tuner is a first attempt, not at creating an ETD replacement, but at understanding if we prefer the sound of a well tuned whole piano of the type used for Debussy, or a well tuned tempered-scale piano that works for the polyphony of Bach. Perhaps the two are compatible, perhaps not.

A number of ETDs allow you to measure the iH and partial amplitudes of every note. For the best tuning I would argue they should be used that way, but only if you have the time and the customer has the money.

Mark, you didn't mention the gear that you used to do your measuring...

You may have stumbled on to one the reasons that the lower few partials aren't used or measured while tuning in the bass!

Ron Koval

This is a very good point.

In my research into iH measurement I discovered gross anomalies in the 2nd, 3rd and 4th partials of many of the notes from A0 right up to C3. I concluded, without a shread of evidence, that bridge rocking likely distorted the measured frequencies. Nevertheless, by excluding them from the data set when conducting a regression analysis, the iH can be determined with much better accuracy. However, if my 'bridge-rocking' theory has merit, the measured values of the lower partials must be considered valid data and an offset for each excluded partial must be added to the calculation of the reference pitch for that note in order to determine the stretch. However, the offset is not required for tuning, as a higher partial would be used.

Thanks Prout,

As for averaging, these measurements were not taken in the fast real world tuning situation in which most ETDs are used. Rather, the ETD took about 5 seconds of listening to measure the string. One supposes that would produce the averaging you speak of that would reduce the variation in measurement. But 14 cents!? I still don't see how an ETD can produce an acceptable tuning with that kind of variation in precision.

Even after averaging 10+ samples of a given note, I still found, in the lower partials of the lower piano compass, iH values that diverged from the calculated iH, using the rest of the partials, by up to 10 cents. That is why the lower partials must be rejected. However, if the ETD insists on using that partial, then I agree that there is no gaurantee of an acceptable tuning.

Perhaps the EDT should also calculate the variance from several measurements and reject those over a certain limit.

That would seem logical. I tried, at Kees' suggestion, using the median of the iH measurements rather than the average and compared that to the regression analysis that rejected some of the data that seemed to me to be too far out to use. Interestingly, the difference between the median results and the modified regression results were nearly the same - certainly either set was valid beyond our ability to hear, IMO. This would indicate that an ETD that used the median of numerous tests (and it would have to be five or more, I would think) might do as well as my method.

But when we measure our samples, so the software can calculate the iH curve, the string is only measured once. Assuming the same variation I found, a questionable measurement would happen every 5 measurements. Another way to think of it is 1/5 of the notes would have measurements extremely far from the average.

I'm waiting for advocates of the ETDs to pipe in. Some of them are well known for implying that eventual improvements in ETDs will make aural tuners obsolete.

More significant is what variation in IH coefficient you got on F2?

The 2nd partial of F2 is most likely discarded as too weak to be useful/measurable. IH is calculated by fitting to a number of clear partials, which are the ones you hear.

No-one claims accuracy of 0.001 cents on piano tones.

Kees

Just a DIY guy here but I personally do not blindly follow any ETD to find a good tuning on the few pianos I work on. I start looking at what the ETD gives in terms of iH constants for 10 to 20 keys/notes taken from C1 to C6. Each note/key is sampled multiple times, then I compare and modify/reject out of range values for each group of samples, then also when graphing the iH constants. Then I verify if the ETD values give good octaves for a given group of keys (for example, all the Cs from C1 to C6) and make changes if needed. Then I can use the ETD as a reference tool for most plain strings. I tune all wound strings by ear afterwards.

IMO ETDs are not meant to measure piano's iH. Neither I know how they calculate the tuning curve or the target pitches. I do not pretend to know more than is told by the user manual. So I just follow the instructions given by the designer/seller and tune the piano. If I don't like the result I first check the piano is in fact tuned as the ETD is asking. Many times I've found the errors were on me and not on the ETD. Secondly, I look for any adjustments to get a better tuning and lastly I tweak what I don't like.

An ETD is only a tool not a scientific instrument. The results you get from an ETD depend more on your skills than on the device itself. A very good example of this is using the ETD in measuring mode to tune mindless octaves, or pure 12ths, etc. The results are outstanding.*

I have no yet beated my Verituner at setting equal temperament. I hope one day I will. Though I make a better job tuning the wound strings than VT, I still try to find better adjustments (tuning styles) to meet my aural tunings in the bass with the VT.

The ETD helps me not only to tune pianos but also to better understand and sometimes discover what really happens when I tune.

I don't think that improvements in ETDs will eventually make aural tuners obsolete. I think improvements in ETDs will eventually make us better tuners. There is a risk with ETDs, anyone with a tuning hammer and a mute could think he/she can tune a piano with an ETD, that's an illusion.



* compared with the standard tuning mode.

What device are you using?
I don't think it's very meaningful to speak generically of ETDs as if all were the same.

Doesn't say.


How does the ETD know what "clear" is? Some partials show as -- meaning not good enough to measure I guess.


I've heard people say the software is that precise. They don't clarify only on NASA grade frequency oscillators :-)

Sounds reasonable.

Rather not say.

Cybertuner requires3 acceptable measurements of each sampled note.
Sometimes it rejects a measurement entirely.
Other times it says a difference greater than 2 cents was detected, and suggests remeasuring.
After three measurements it calculates the degree of variance, ranks the result as good (green), so-so (yellow), or not good (red) and gives the opportunity to remeasure any of the samples to see if a more stable reading can be obtained.

Two other devices with which I am familiar will leave it up to the operator to select one measurement of each sampled note.

It is possible that an ETD will pick up sympathetic vibrations from poorly damped notes, or even another piano in the room. That can happen with aural tuning, too!

Mark - ETD guy here Ron had asked what gear or device you're using. That might be helpful in evaluating your experience. I have experience with TuneLab and Tunic OnlyPure. I'll take TuneLab first. Many times, in the upper section of a spinet, I get wildly different readings, so the ones that I know are off, I reject and sometimes can't even take reliable readings in that section so I skip it. The same thing can happen in the bass. (TuneLab will take as many readings as you want on each note and average them out.) After that, I look at the Tuning Curve page to see if looks like a reasonable curve. I may change the ratios to see how it affects the curve but usually go with either 6:3 or 8:4 in the bass and recently have been using 6:1 oct+12th in the treble. I found that this combination mimics Only Pure, which I like. You can set the bass section to pick the strongest partial, once you have readings, so while you're tuning that will change on the fly. As I come out of the treble section into the bass I make sure to listen to the octaves and if it doesn't sound right I go with what sounds right, not what TuneLab says. So, yes, I do aural corrections, but only for octaves and unisons and not in the temperament area. I don't know if in your scenario that disqualifies my tunings as coming strictly from the ETD. With OnlyPure it's pretty much point and shoot. I do the same thing with it in the bass section and sometimes on spinets or badly scaled pianos, disagree with it, although I would say that I think OnlyPure, on a well scaled piano, can produce a superior tuning to most aural tunings.

A couple things: Part of using an ETD is knowing how it works and all the things you can tweak and things you can't tweak. I think sometimes aural tuners think that it should be simple just to pick up an ETD and use it. It's not. There's a learning curve that's frustrating at first. For example, OnlyPure takes a while to get used to, even after using TuneLab for many years. Also, aural tuners usually talk in terms that ETD users don't understand or relate to. That's probably why more ETD tuners haven't joined in this thread. I think it's like a computer programmer talking to someone that uses their computer program. I see the aural tuners as the programmers and people like me as those who take advantage of the technology they've created. So, we need you! And I'm okay with that. Some people aren't.

By the way, I had always assumed that the accuracy claimed with ETD's related to how well it reads a note, after the inharmonicity readings are taken, not to the actual inharmonicity readings that create the tuning curve. For instance, I may read A4 before I start tuning so I can see where the piano is at, and that reading will be very accurate. Not sure that makes sense!

Once again, I really appreciate you Mark. I feel like I can join in the conversation and not just be dismissed. You're one of the people that have helped make this forum a safe place again for people like me.

Well, Mark, I know you're not using Tunic OnlyPure because you take no readings with it. So, one down...