Piano World Home Page Forums Our Most Popular Forums Piano Tuner-Technicians Forum How Did the PTG Grade Tunings before Tunelab Came Out?

Hi Chris,

Yeah, I know, but I didn't want to get into the 'discrete/short' FFT tricks used to increase accuracy. The Gabor Limit of the discrete FFT sets the minimum accuracy.

For iH measurements I use multiple techniques including the usual overlap and multiple averaged samples to converge on much higher (100x) accuracy. This takes time. In fact, I could say that my accuracy follows the basic rule. It takes me many hours to do multiple recordings of samples of varying lengths and amplitudes for a single note, and then analyze, enter the data and run regression analysis on the data. Chances are my accuracy/time ratio falls way below the theoretical limit.

However, my point is that using FFTs to tune, even with +/- 0.1Hz accuracy, is still a 2.7 cents range of error on the fourth partial of C1. Tuning using phase displacement against a reference clock can achieve a level of accuracy vastly beyond that which neither the string note the tuner is capable of reproducing.

Bill and Lucas,

You are right. My mind is simply going back to those early days. I recall now the amount of time required...no need to elaborate further. Part of the reason I had to stop being a CTE. I should have thought it through a little more.

Thanks,

Pwg

Years ago I developed a method to measure inharmonicity using continuous excitation of the string. I even wrote a PTG Journal article about it (June 1993, page 42). I used magnetic field excitation and detected the amplitude of the response of the string to that excitation optically with a photosensor and LED combo. The excitation was a precisely synthesized frequency that swept slowly across the region of interest. The process of data collection was very slow - about an hour - but it was automated. So I could start it up, leave the room, and it would finish on its own.

The idea was to overcome the limitations of the natural decay of a struck note. The excitation was continuous. At each minor frequency step, the excitation remained constant long enough for the response of the string to become steady-state, and then measure that response. I was hoping for infinitely small resolution.

Unfortunately, the imperfections of real piano strings shows up with continuous excitation too. What I found was the graph of response amplitude as a function of frequency was a broad peak with an ill-defined frequency of maximum response. And if the string has any false beats in normal struck mode, how do you think that string responds to continuous excitation at various frequencies? It has two separate peaks in the response graph! Darn. There is no free lunch after all! The measurement was about as precise as the best analyses of a normal struck string, but no better.

If you have access of the PTG Journal archives, look it up.

Thanks Robert for your comments. I will try to find the article.

prout

When I joined, there were no electronic aids. Testing outcomes varied widely.
The 1st electronic test devised by Al Sanderson used a SOT and a TI58 (I think) calculator with a small magnetic strip to store values, and an add on printer to the calculator.

The test piano was tuned aurally by three until they all agreed it was as good as they could do. The notes were measured and tediously entered into the calculator. If an entry error was made, I think you had to start over.

The piano was then detuned to specific parameters. The tuning test was done, notes measured and also entered into the TI (Texas Instruments) calculator. After a few minutes of calculating, it printed out the results.

The temperment, at that time also had to be tuned by the testee aurally and achieve 80% or better, but the tuning could be done with an SOT or strobe.

Hope my memory is working well!

Steve

Hi Robert,

I am still looking for an archival source for your 1993 article.

I did find an amusing exchange between you and Don Gilmore back in 1993 on accuracy of measurement. He claimed an acurracy of 1/6000 of a cent for his self-tuning piano, explained the methodology, then admitted that the string itself wavers more than 0.1 cent naturally, so his level of accuracy was pointless. I have a 200.0000Hz calibrated frequency source I use to check my equipment and software. Using it as a source, I too can measure the period of a single schmitt triggered square wave to seven significant figures, but as you discovered, it doesn't help in tuning a piano all that much.

Here's a thought I have rolling around upstairs...

Single partial, smooth curve calculations.
As we've heard from Bill and others that score exams, the master tuning isn't well represented by a smooth curve.

However, that is a smooth curve of a single partial that the exam uses for any particular region of the piano - what about the other partials? I believe the exam testing/partial choice was made by Al Sanderson based on the ETDs available at that time.

I did some tedious measuring and graphing when the Verituner first came to market - and found that often one of the many partials measured did come closer to a smooth curve than the others in different parts of the scale.

It would be interesting to measure a bunch of master tunings and see if there is a pattern that could be used by the ETD software to tweak the tunings even better than they already are??

Ron Koval

Ron,

The PTG Examinations and Test Standards Committee (ETSC) does not allow Master Tuning data to be used for anything except the administration of a tuning exam. I got yelled at once for quoting data for the 7th octave of a master tuning (the topic was tuning 2:1 octaves in the high treble). Even though the data was from a record that would no longer be used, they didn't want such data to be seen by anyone other than the person who would use it to administer an exam and the person who audits an exam when it is submitted to the ETSC for approval.

I understand that what you are saying is that you would like for someone to use the data for research purposes but to get the ETSC to give the approval would be very difficult. It would not be a one person decision and the likelihood that the request be approved is very remote.

What I can say is that the data looks different from calculated programs for more than one reason. The first is the partial selection. Calculated programs read octaves 1 & 2 on the 6th partial, octaves 3, 4 & 5 on the 4th partial and octaves 6 & 7 on the first partial. The exam program reads octaves 1, 2 & 3 on the 4th partial, octave 4 on the 2nd partial and octaves 5, 6 & 7 on the 1st partial. [None of that information is confidential].

Naturally, when there is a change of partial, the progression of figures takes a jump. It is always somewhat amusing, as the readings for the Examinee are being determined and announced for the person transcribing them to hear a groan or swear words from the Examinee as the note C4 suddenly is a negative number. The examinee hears the numbers slowly progressing from slightly negative at the bottom of the 3rd octave to positive in the middle and 1-2 cents positive by B3 and then suddenly, C4 is -2 or so.

I have to stop and explain that there is nothing wrong and that it is expected. Sometimes, I sometimes even play C4 and a few checks for it to show the examinee that the notes sound fine. It is not flat, Then, I go on to explain that even we examiners most often have no clue just from the numbers, whether an exam may be in passing range or not unless they are quite wildly irregular and in that case, we make no comment other than, "At this point, the numbers are meaningless. We have to wait until we determine the Pitch Correction Number and actually score the results before we will know anything at all."

The next reason that progressions are not always perfectly smooth is the actual scaling of the piano. Sometimes, there can be two consecutive pitches and then the cent value takes a whole cent or more leap. Sometimes, there is even a slight digression and then another leap after that. It seems like an error and when I see that happening, I go back to some very careful aural scrutiny. Sometimes that does reveal an overlooked progression error but other times, it is simply what is correct for that piano.

It has to do with the actual curvature of the bridge, the speaking length and changes in wire size. I have done a fair amount of plain wire scaling on older pianos for which the original scaling was unknown or even if it was known, I wanted to figure the scaling myself. I would virtually never end up with the typical kinds of wire change patterns that would be found from the manufacturer.

It seems that many of them just put on 6 13's, 4 13 1/2's, 6 14's. 6 14 1/2's, etc., on down to the tenor where they put 10 17's and 8 18's, just because every other manufacturer did that. I would usually end up with quite a different pattern. I would also sometimes find that I would have to digress to a smaller size for a couple of unisons when I crossed from the treble section to the midrange. Again, it is because of the actual bridge curve, the plate strut and then the new termination point in the midrange. The speaking length is suddenly significantly longer, so it requires a smaller wire for a couple of unisons. You see that on some Baldwin grands.

So, I'll be the first to agree that the first Sanderson attempt which took a single sample of the difference in inharmonicity between the 2nd and 4th partial of the note F4 and then calculated 3 octaves from that measurement alone was quite primitive. I remember when the SAT II came on the market and the "new and improved" FAC program came with it, it was seen as a great step forward. It still merely samples the difference between the 4th & 8th partials of F3, the 2nd and 4th partials of A4 and the 1st and 2nd partials of C6.

I have no argument whatsoever that the Verituner handles everything in an entirely different way and therefore can handle quirky and short scaled pianos much better. I, however still prefer to use the SAT IV that I have had now for 7 years for several reasons:

• Robust, dedicated device that has month long battery life and will last for many years without repair and when it does need repair, it can be totally overhauled to brand new condition in a matter of days, including shipping back & forth.
• While the initial investment is costly, it is basically a one time investment with any further expense for repairs being only a small fraction of that original cost and not generally needed for many years. The cost is comparable to some of the other platforms and does not necessitate also acquiring an expensive smartphone, laptop or tablet that can be very vulnerable to damage or breakdown. There is a comfortable warranty period and repairs also have a warranty period. The staff at the manufacturing facility are friendly and helpful. Service is quick. Tech support is available during business hours.
• Even an old, inoperable device still has value. It can be traded in or sold as is and can be sent to the manufacturer to be refurbished to brand new condition and with a warranty period. A working device in good condition can have substantial value, up to half or more the price of new and the manufacturer will also offer a good trade in value for an older device. The manufacturer sells and services all of its models and also provides SOT repair and service.
• The display is the one I started with when after more than 20 years of strictly aural tuning and is the most familiar and easily read and interpreted by me. I dislike all other kinds of displays I have seen.
• The ease of storing some 200 tunings which have been carefully worked out for use again and again on regularly serviced pianos. The tuning can be given a name and can also be off and up loaded to and from a larger library. Tuning files can also be moved around within the device as needed or desired.
  • The device offers a unique sequence program for aural tuning and storage of either equal or non-equal temperaments. Once the value for the note being tuned has been determined, the value can be stored and the device moves on to the next note of the temperament sequence to be tuned rather than chromatically.
  • Stretch curves of existing programs can be easily manipulated from wider to narrower of vice versa using the DOB function that can be changed or cancelled at will. This actually does allow for adapting a program to short or irregularly scaled pianos.
  • The ease of tuning in the Direct Interval mode. It allows me to compare, match or make a compromise between any number of coincident partials as I choose, use my own judgment that is exactly the same as using my own aural tuning judgment to determine the best value for the note in question, lock it in and store it.
  • As with all other devices, it comes with a library of non-equal temperaments and custom non-equal temperaments are easily stored and named. They can also be easily rearranged within the device.
  • The exam program is the one with which I am the most familiar. It can store 4 complete exam programs but Master Tuning files for more exams can be stored on the upper pages, named and also be off and uploaded to a larger library.
  • Can you interest me in a more "advanced" electronic tuning platform? No, you cannot. I prefer the very advanced device that I already have and will have for many years to come without the need to replace it or to have to learn how to use another platform. I like the idea that I am in control. I tell the device what to do instead of it telling me what to do.

Robert,

Got the article and read it. Very interesting . Using the computer to remove human bias and using weighted averages definitely allows for teasing out a more precise partial frequency from the noise. I understand the use of a square wave for driving the exciter. It ends up as a distorted sine wave, IMO, at the excitation point from back EMF anyway, though, today a good and cheap sine wave driver is readily available. It might reduce the need for repositioning the exciter. Not sure. I just may have to try this experiment.

Here's my take away from this thread:

1) Without modern electronics and computers, unbiased and objective judgment of a tuning would be
practically impossible.

2) It has been admitted that starting with an ETD tuning saves valuable time in creating the Master tuning, and the
aural tuners will tweak this software generated tuning for about 4 hours, EVEN THOUGH THE COMPUTER
TUNING AS-IS WOULD STILL PASS THE PTG TEST WITH HIGH MARKS! Doesn't this reek of the human
tuners getting a bit desperate? If a Tunelab tuning would pass the PTG test with high marks, I doubt if any
humans would notice the difference.

3) There are things computers can do that humans cannot, such as calculate a consistent 4:2 in the treble,
because the beats are too fast for humans to hear in the upper 1-2 octaves, using the M3/M10 test. Ditto for the
lowest 1-2 octaves, where you cannot hear the beats using m3/M6 testing for 6:3 in the bass.

4) Arguments such as being able to do math without a calculator, or do handwriting without a computer or keyboard,
do have merit. However, should humanity also keep mechanical watch making alive, even though quartz crystal
accuracy is far superior? Or should we keep audio tape and camera film around, even though the digital world has
completely eclipsed those markets?

Not at all true. It is possible to be more accurate tuning these octaves aurally than a calculated curve will produce. I not only do it in master tuning sessions but when I do routine aural tuning on ordinary pianos.

There are other tests besides M3/M10 and besides, what is discernible depends upon the skill and experience of the technician. I never use a calculated program to tune the Bass because I get more accurate results by ear. 6:3 octaves would be too sharp for octave 1, even if you could hear the tests.

Anytime I see someone say that electronic tuning is "better" than aural, I think, "There is another person who doesn't know what he doesn't know yet." And sorry to say, you proved it in what you wrote. You simply have more to learn about aural tuning if you care to.

Musicdude, you appear to be fixated on aural tuning by using certain octave checks and wanting to be able to use the same tests for the whole piano. Aural tuning does not work that way. Octave tests are a useful tool to use along the way but they are only part of the methodology. Every note that we tune aurally, apart from the stating pitch, is a compromise between more than one other interval. The final arbiter is the quality of the combined sound and not the precision of a particular octave check which may not be appropriate for a particular piano and part of the scale. Octave checks, when and if used, are only a method to help get there when and where appropriate. Experience is the key.

Ok, then how would a human aurally tune a consistent 4:2 in the top 1-2 octaves?

Your so-called "mindless octaves" appear to be just balancing double octaves and octave+5th, which is
theoretically just a balance between 4:1, and 3:1.

That is NOT the same as a consistent 4:2, which the computer is able to calculate.



For the bass, you wrote: "You may continue comparing the double octave and octave and fifth all the way to A0. However, just as with the high treble, you may begin to favor the octave and fifth or the double octave and fifth at some point in the low bass, beginning on or about C2."

Again, that appears to be simply a balance between 4:1, 3:1, and 6:1, and not a consistent 6:3 in the bass.

Whether or not that would sound "better" for octave 1 would be a matter of taste.

And it's very easy to change Tunelab to have a higher bass stretch of 8:4 in octave 1.




So the humans spend about 4 hours to "improve" a computer tuning, that would have passed the
PTG test with high marks anyways.

First off, I'm not yet convinced the humans are "improving" the computer solution at all. And even if they
did, how long would such a tuning last, over the computer tuning? The piano tuning drifts even as the tuner
is working on it, nevermind playing some hard FFF Franz Listz!

Believe it or not, I'm not against aural tuning....otherwise I wouldn't be trying to learn how to do it myself.

I DO understand that aural piano tuning is considered an art form by itself, and that is something that I respect.

But facts are facts. Computers have beaten humans in chess, and they quite clearly are dominating many other human endeavors as well......

Also if you would just use an ETD program for the master tuning, you would have people complaining about their tuning being compared to a machine.

The master tuning cannot be done using a calculated electronic program by rule. The exam program has no judgment in it. When the preliminary tuning is wrong, it's wrong.

Musicdude, if you continue to believe that electronic tuning has all the answers and is superior, you never will gain the skills to pass the exam. You will be one of those who continues to complain about it but your complaints will fall on deaf ears.

Aural tuning does not assign specific octave types to this octave or that, then abruptly change to another. It gradually shifts along the way which a calculated program cannot do and that is why they inevitably have to be corrected note by note, as small as the changes may be.

The mindless octaves approach which you ridicule is the first step I actually use to correct the deficiencies of a calculated program. That is followed and refined by rapidly beating interval checks, each of which refer back to the central octaves. Therefore, the entire piano is in tune with itself from end to end, not just dependent on octave types within a limited range. In fact, I never use the octave type tests that you mention because consistent octave types within any particular range of the piano, except a limited range in the very center of the piano is not even the goal. It is not the goal because it is not correct but yes, a computer does tune the piano very precisely incorrectly. They have been doing that every since they were first employed.

If an electronic tuning platform were allowed to be used for the entire exam, it would likely yield passing results, well in the superior range for those who actually know how to use one correctly and also have adequate hammer technique, including enough to pass the unison and stability portions of the exam, yes but that is because the tolerances of the exam allow for the fundamentally incorrect calculations that such a program inevitably makes.

No one ever completely matches the master tuning although the very best produce results that lie within small fractions of a cent of it. The RPT credential is awarded to people who show that they are competent enough to know when a piano is tuned within minimum professional standards, not those who can show that they know how to operate an electronic tuning platform reasonably well but have little or no understanding of what having a piano really be in tune actually is.

Your comments about stability show that you have no understanding about aural tuning, only your own failure to make it work. This is unfortunate because if you had approached tuning by learning aural tuning first, you would likely have an entirely different opinion about it.

You are dodging the question, which is:

How would a human aurally tune a consistent 4:2 in the top 1-2 octaves? What would be the aural test to do this?

Musicdude:

Do you actually want to tune a pure 4:2 from C6 up? Surely you don't mean that. Why would we care if a computer (mine just did) compute what no sane human would consider doing?

I just checked the data from my own piano, and tuning 4:2 from C6 yields a C7 that is 17cents sharp and a C6/C7 octave beating at 8bps on the strongest partials (2:1), and, continuing up to C8 yields a C8 that is a whopping 58 cents sharp and a C7/C8 octave beating at 60BPS!

This would, in my humble opinion, sound unbelievably terrible.

When we tune by ear we don't just tune 4:2 octaves in one place and then decide to switch octave types in another place. It all blends together. Maybe start with 6:3 in the bass and slowly change to a 4:2 or whatever. I go for 4:2 octaves in the temperament and that is it. I don't even know what I tune in the bass, but it's probably inbetween two different types of octaves.

That's the nice thing about listening. You are going for sonority. The size of the octave is irrelevant if it works with the rest of the piano.

Musicdude,

I am not dodging the question. As others have told you, tuning consistent 4:2 octaves is never the goal in the treble and consistent 6:3 octaves is never the goal in the Bass. The most appropriate size for the octave slowly changes note by note. It does take a Master Tuning Committee some time to sort that out, yes.

In an earlier post, you managed to insult every aural tuner who ever lived and is still living. It just so happens that Dr. Sanderson had the same idea you did about 4:2 octaves until he found out differently. There is also a very well known and still living technician who had the same idea that you have about the 4:2 and 6:3 but when he went to Vladimir Horowitz's technician, Franz Mohr some decades ago to try to show him how the wonderful electronic tuning device he had could do just what you say, tune consistently perfect 4:2 octaves in the treble and 6:3 octaves in the Bass, Mr. Mohr told him quite frankly and matter of factually that it was totally unacceptable.

You see, there does come a time and place when what you now believe to be the perfect tuning simply doesn't fit that description. The professional and discriminating artist would regard it as substandard. It may be fine for the work you do now and if you can manage to learn to tune the two central octaves well enough by ear to pass Part 1 of the tuning exam and use your electronic tuning platform to tune the outer octaves AND you know when to switch octave types in the highest and lowest registers, you might pass with very high scores in Part 2 of the exam. You would still have unisons and stability to pass in Part 3.

I take it that you will not be in St. Louis to attempt the exam but you still think you are ahead of the curve of all the people who have taken that exam and you want it changed so that you can use your electronic tuning platform to show everybody that aural tuning skills are unnecessary and obsolete. It isn't going to happen.