Pianotech

Expand all | Collapse all

Sensor Position Invariance. YouTube Video

  • 1.  Sensor Position Invariance. YouTube Video

    Posted 01-19-2024 04:19

    I have 'exposed' the fact to help piano tuners understand the large variance that comes from moving the microphone of the ETD, causing a large variance on the tuning measurement by a worst case of more than 1 cent, this being an intolerable amount, causing the tuner to no longer trust the ETD. I cannot tell you the number of 'private' conversations I have gotten from tuners who 'know' this to be an issue and have been distrustful of the ETD's. Well, thankfully, there is a solution.

    In prior posts, I have been stating the invariance of the sensor, and people want to know how small changes along the position of the string will affect the tuning measurement in their ETD when the sensor is used. This is a fair question and deserves an answer. 

    This video demonstrates that the sensor, when kept to an area between the hammer and the termination of the string, is invariant on measurement. The result is an astoundingly near-zero variance. Even more specifically, when I process the captures into Matlab and further analyze the variance, I see the range is from 0.03 to 0.08 cents!

    Also, yes, device is available. You can use it with confidence. It is a very high tech device employing the best design practice found in state of the art sensors used in space systems and is produced in a DoD-certified facility. 

    https://youtu.be/xUfOnQWWoD4


    ------------------------------
    Steven Norsworthy
    Cardiff By The Sea CA
    (619) 964-0101
    ------------------------------



  • 2.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-20-2024 09:21
    I don't care how much variance any ETD gives you, it should only be used as a rough guide.  Never tune from the bottom up!!!  Always set your temperament with ETD only as a guide.  I usually disagree with F3 and F#3.  Even if you set all the strings perfectly with an ETD in your temperament, it could have issues, such as a beat in one string is causing that set of unisons to sound flat.  Remember we don't listen with our eyes; listening should always be the final judge of pitch.  AND no one can tune a piano in 45 minutes!  I said tune.

    Douglas Laing
    727-539-9602






  • 3.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-20-2024 15:06

    Steve Fairchild could tune a piano in 12 minutes.  Paderewski's concert tuner could do a concert level tuning in 28 minutes.



    ------------------------------
    Parker Leigh RPT
    Winchester VA
    (540) 722-3865
    ------------------------------



  • 4.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-20-2024 15:16
    That's a touch up.  Concert tuning on a piano that gets tuned all the time.   228 strings in 12 minutes.   Mmmmm, bet that sounds great on a piano that hasn't been tuned in a year.  Kind of like trashy fast food.





  • 5.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-20-2024 17:47

    As I recall the story, when someone in the audience asked to test the piano, Fairchild replied no, that that would be unnecessary, because the piano had already been tuned by the best tuner in the room...himself.



    ------------------------------
    Ed Sutton
    ed440@me.com
    (980) 254-7413
    ------------------------------



  • 6.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-20-2024 16:02
    From my experience using the Pianosens, I suspect that the ETD’s are not as accurate as we think because they rely upon microphones in our phones. The results are mostly good, but can often be improved with aural checks, which I believe is partly due to the mic’s sensitivity to extraneous sounds. The Pianosens eliminates that problem. But still there can be anomalies which need to be addressed.
    As far as the amount of time it takes to tune a piano, obviously there are many factors involved. Using the sensor, a piano that is less than 5 cents flat or sharp I can tune in a little more than an hour and a half. But if I tuned it last week, it would be less. And it also depends how close you want to make it. I tune to within.2 cents, so that level of accuracy takes more time, and you have to do string by string if you want to want the best result possible. I’m not tuning aurally, so I can’t comment on the time it takes for an experienced aural to tune a piano. It might take less time than I do. But I know it’s not possible for me to get.2 cents by my ears alone.


    Sent from my iPhone




  • 7.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 01:43

    "...But I know it's not possible for me to get.2 cents by my ears alone."

    And I ask the question, again. Can anything but a very sensitive machine tell, or even care for that matter, if a tuning is more than 0.2¢ off? Using A440 as an example, (because the math is simple), where a 4¢ change is equivalent to about 1 Hz, a +0.25¢ inaccuracy would result in A4 being at 440.0625 Hz. Roughly 1/16th of 1 Hz. Lasting accuracy at that level of precision is way beyond real world expectations, I think. 

    FWIW, I use RCT and, including aural checks after I have completed the tuning, it takes me about 90 minutes, (plus or minus, depending), to put a stable and clean tuning on a piano that is not also requiring a pitch adjustment. Is it a concert level tuning? Perhaps, maybe at times, but I'm not sure how to even define, or apply that requirement to anything other than a great piano, kept in a supremely stable environment and serviced at least every day or so. 

    Any machine that can detect two or three decimal point accuracy in tuning might be useful in a lab but is, in my opinion, overkill in the field. Aiming for that 0.2¢ accuracy is working way too hard and I suggest that if we learned to listen carefully beyond what the ETD may be telling us we would consistently achieve close enough to that anyway.



    ------------------------------
    Geoff Sykes, RPT
    Los Angeles CA
    ------------------------------



  • 8.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 10:40

    Good point Geoff. It does feel like there's a tendency to discuss tolerances that are well beyond what's practically necessary to achieve a superb tuning. That doesn't mean Pianosens isn't useful or valid for other reasons, I would just caution against posturing over hundredths, or even single tenths, of a cent. You cannot hear that kind of difference when playing music, and even if you could no piano in a normal environment will hold that kind of tuning for long.



    ------------------------------
    Nathan Monteleone RPT
    Fort Worth TX
    (817) 675-9494
    nbmont@gmail.com
    ------------------------------



  • 9.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 11:15

    "Can anything but a very sensitive machine tell, or even care for that matter, if a tuning is more than 0.2¢ off?".  Well, unless you have the experience of tuning to this level, you might not know how it sounds.  Once you hear it, when there is no beat at all, consistently, then you have to admit that what you've been doing is not as good as it could be.  We don't wait long enough while tuning unisions, for example, to hear a slow beat.  By ear, it takes too long to hear it, so we just don't take the time to do it.  But if your app can give you that information, then you can make corrections to get within that range +- .2 cents very quickly.  You do hear it.  And when you're done with the tuning, you'll have far cleaner unisons and fewer notes to touch up.  For me, tuning the best unisons as possible has always been the goal.  If I'm inviting the client to sit down and play the piano when i'm done, I cringe when I can hear a unison that's off.  Of course, it's going to go out as time goes on.  But if you've got it dialed in, it will be in good tune longer.  I've done tunings where I did some of the unisons by ear, then compared later to the ones I did string by string.  Later, I found that the aural ones weren't as good as I thought they'd be.  I had to touch them up again.  There's a leeway, an imperfection, in tuning unisons by ear.  Much like setting an octave A3 to A4.  By ear, you can get close.  But it's not accurate enough.  You have to check by comparing beats with F3.  Where is that check with tuning unisons?  There isn't any as far as I know.  We just do the best we can, wait a few moments, and it's good enough. 

    All I can say is, for me it's worth the little extra time it takes.



    ------------------------------
    Paul McCloud, RPT
    Accutone Piano Service
    www.AccutonePianoService.com
    pavadasa@gmail.com
    ------------------------------



  • 10.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 11:47

    Geoff, I also agree with your feeling that 0.25¢ accuracy numbers are "way beyond real world expectations," if I understand you correctly. For example, I estimate that near A440, where there is roughly 25 Hz to the semi-tone, working with a test of major third (M3) progressions (during an aural tuning), the best that I can "give" myself is a one beat-per-second coincident partial resolution in the progression test. With the M3 being 5:4 that leads to about 0.25 Hz or approximately 1¢ accuracy for one of the notes of a two-note M3 interval. On a good day! Regards, Norman.



    ------------------------------
    Norman Brickman
    Potomac Piano Service
    Potomac, Maryland
    potomacpiano@verizon.net
    https://potomacpiano.com
    (301) 983.9321
    ------------------------------



  • 11.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 12:18
    Dear Norman and Geoff,
    This is the picture of what Norman has described. It was done months ago in preparation of my course which I have given to many PTG groups. It is a picture of an A4 where one of the three usisons is mistuned by one cent, or 0.255 Hz. It shows exactly what is happening as the power is immediately sucked out in the first 1/2 second, and the 180-deg crossings of the harmonics are kicking in followed by the first 1/2 cycle 180-deg crossing of the fundamental in 2 secs, and the rebound 2 secs later, the so-called 'bloom' which is not really a bloom but a re-phasing of the next 1/2 cycle where the phases realign.
    It is VERY audible when compared to the ideally tuned unisons. I play these audio files for the class. Everyone, 100% of the tuners, agree in the class. I am happy to offer my class to both of you on a zoom session.
    Respectfully yours,
    Steve
    A4 Ideal vs Mistuned by 1 cent 0.255 Hz


    ------------------------------
    Steven Norsworthy
    Cardiff By The Sea CA
    (619) 964-0101
    ------------------------------



  • 12.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 13:37

    Someone correct me if I'm wrong on this but I'm pretty sure that virtually all the good ETDs are capable of .2 cent or less accuracy. What I think Steve has been trying to say is "it's not the ETD per se, but rather the microphone". So the ETD's "accuracy" is subject to the stability and repeatability of the mic input. IOW to use computer lingo: GIGO.

    Of course I know that we're talking about micro units here, but in typical piano technician nit-pickyness over ridiculously small details, his point is about reducing input error as much as humanly (and technologically) possible. What is implied here (to me) is that a good ETD can get "close enough" because it's "hearing" is basically on par with a good trained human "ear". But a good ETD coupled with this new device can now do what no "human ear" can do. 

    The question is: "Is that important to me?" Or am I perfectly content with what I'm doing now? In which case I need not be concerned with it. 

    Again someone correct me if I'm out in left field here.

    Peter Grey Piano Doctor 



    ------------------------------
    Peter Grey
    Stratham NH
    (603) 686-2395
    pianodoctor57@gmail.com
    ------------------------------



  • 13.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 13:45

    Steven, thank you for the kind offer of a Zoom session of your class. Appreciated, but I'll hold off for now.

     

    BTW, in my post I was targeting the tuning of the piano, not tuning unisons. And as I read Geoff and Nathan's posts, I think they were doing the same.

     

    Good point on tuning unisons, and the risks with inaccuracies. I don't recall having heard the effect that you describe in my tuned unisons. The implication in what you describe, though, might be that my (and other's) accuracy in tuning unisons already is of greater accuracy (than with tuning most other non 1:1 intervals) -- considering all the coincident partials available to listen to when tuning a unison!

     

    Regards, Norman



    ------------------------------
    Norman Brickman
    Potomac Piano Service
    Potomac, Maryland
    potomacpiano@verizon.net
    https://potomacpiano.com
    (301) 983.9321
    ------------------------------



  • 14.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 14:11

    Norman and Peter,

    A couple of things:

    As stated earlier, the loss of power on the attack comes from the 180-deg phase cancellations starting with the upper partials and finally with the fundamental. You can only compare the loss of attack power by doing an A-B back-to-back comparison. I play these for the classes and people are shocked at the difference. If you just are tuning without the comparison, you would not necessarily know. I share a story with you. I have demonstrated this for more than one top tuner, concert level I might say. You need to be on a top-quality concert grand piano where everything is all lined up, and yes, it is very noticeable. The concert tuner got to 0.4 cents on the unisons, as I measured it afterwards. Then I dialed it to 0.1 and he exclaimed the he had never heard that kind of sound before!

    Look, I am just saying that the time has come for us to have a reference level way of tuning without variances, repeatable, reliable, not subjective. Why not? 

    The other thing I keep posting on is the ETD's inability to resolve acoustic smearing in the frequency analysis. Acoustic interference is most exemplified with false beats that are only a few Hz apart. The sensor cannot 'hear' the acoustic interference. It does, however, 'see' the mechanical false beats of 'vertical minus horizontal' frequencies that are partly due to poor bridge pinning or bad wire. 

    You also must remember that even with non-unison issues from note to note, if the acoustic variability that cannot get resolved in the ETD produces an error of 1 cent, the standard deviation of accumulations of all these errors across the piano is multiplied statistically by the number of notes on the piano. So let's say your standard deviation error is 0.5 cents from acoustic interference. The square root of 88 is 9.38, so if you have increased the probability of the spread of out-of-tune-ness by at least 5 cents or more. Why do you want to start out with this kind of error, knowing you are statistically increasing it as you go? On top of that, you know the piano will spread due to playing and environmental changes. You start out with unnecessary errors, and now the spread over time is even greater! It makes no logical sense if you have a method that starts out with minimal error! 

    I need customers who want to participate in the raising of the bar. That was Paul McCloud's term when he started using this new method. 

    Does it make a difference on lesser pianos? Ask Paul, and he will tell you it does, and his customers are telling him their pianos simply sound clearer and cleaner than before.

    Perhaps my device will catch on with a new generation of tuners who want to get pristine tunings from the get-go in their emerging careers. That is probably the path forward.

    My first two customers were young tuners who have not yet passed their tuning exams and are tuning in noisy environments. 

    We shall see how long it takes and how much teaching effort it takes to get on a new standard of tuning. I would love to be part of that. 

    Respectfully,

    Steve



    ------------------------------
    Steven Norsworthy
    Cardiff By The Sea CA
    (619) 964-0101
    ------------------------------



  • 15.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 17:13

    > So let's say your standard deviation error is 0.5 cents from acoustic interference. The square root of 88 is 9.38, so if you have increased the probability of the spread of out-of-tune-ness by at least 5 cents or more.

    I don't see how that's a relevant number. Tuning errors don't compound when you're using an ETD, the way they tend to aurally.



    ------------------------------
    Nathan Monteleone RPT
    Fort Worth TX
    (817) 675-9494
    nbmont@gmail.com
    ------------------------------



  • 16.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 17:18

    On the Gaussian Curve, more numbers you have, the greater the probability of finding outliers at higher sigmas. It boils down to increasing probabilities of outlier.

    One sigma is 34% on each side of the mean. Two sigma is 13.5% more on each side. Three sigma is 2.35% more on each side. 

    If you had 100 notes, and 1 cent was sigma, then 1 cents sigma would mean that 68 notes would be 1 cent off or less

    Then 27 notes would be between 1 - 2 cents off 

    Then 5 notes would be between 2 -3 cents off

    Do you want to be off that much somewhere? I don't. Why? Yes, you are increasing your probability of greater out-of-tune-ness across the range.

    It's actually far worse then this, because the lower you go in Hz, the more difficult it is for the ETD to keep the same 'cents' accuracy since the deviation in Hz is decreasing accordingly, by log(2).

    In the high register, as I have  been painfully showing, the ETD has a harder time resolving close-up acoustic interferers. 




  • 17.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 18:11

    Indeed that is the correct definition of Standard Deviation. However, Standard Deviation from all 88 notes isn't a great way to measure the quality of a tuning, for a couple reasons:

    1. We don't listen to all 88 notes at once.  What we care about is how much error there is in any given musical interval, and in practice the slower beating ones tend to stick out a lot more readily if they aren't correct.  
    2. The human ear is more tolerant of errors in the extreme ranges of the piano than it is in the middle section.  Standard Deviation assumes all the samples are of equal importance, and in this case they're not.  Taking a concrete example, let's grab the mid section of the piano from C2-C6 (49 notes), with 0.5c std dev per note as before.  You'd get a standard deviation of 3.5c.  Adding the rest of the notes in, isn't going to make the tuning objectively worse -- most of the errors will fade into insignificance -- but the standard deviation goes up to over 5c as you mentioned.

    I accept what I assume to be your basic point -- it's worth reducing the error in measurement to something well below our aural tolerance for error, because it reduces the chance that other sources of error (ex. slight tuning instability) will push the total error into the range that we can actually hear.  A dirt simple illustration of that problem (albeit one that's a little out of date) is going into the tuning exam with a pitch fork that's 0.9c sharp.  Assuming you don't know this and try to match the fork exactly, you've dramatically increased your chances of failure vs. using one that was correct.  In fact Standard Deviation would be an appropriate way of qualifying just how much you decreased your chances.

    That being acknowledged, please don't continue to imply that the standard deviation over 88 notes is meaningful.  Intended or not, it yields big scary numbers (over five cents!) that do not reflect the real-world consequences of the errors we're talking about.



    ------------------------------
    Nathan Monteleone RPT
    Fort Worth TX
    (817) 675-9494
    nbmont@gmail.com
    ------------------------------



  • 18.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 18:24
    Nathan,

    It is probably far worse than 5 cents due to my argument that the sigma increases in the lower and higher registers as I well articulated. How do I know? Measurments. Prior posts showing data.

    So, the ‘sigma is not a constant sigma across the piano. It worsens at the ends.

    Why do you NOT want to start with less error? The sensor is basically invariant and decreases the probability of starting with a large sigma.

    Steve




  • 19.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 18:33

    Nathan, 

    I am just continually baffled why some tuners want to start out with a large error, and of course the sigma increases more at the extreme ends, so that you get more outliers many cents off?

    One starts with an error, a much larger error from the acoustics-based signal processing, as I have shown with prior posts and real data.

    Then we know the piano drifts from that tuning naturally. Then we get an even larger error over time and drifting.

    Do we really want that, when we now know there is a new device that is basically invariant to begin with, that helps the ETD resolve acoustic interferers at the high end, and define more partials at better resolutions at the low end as well? 

    A lot of scientific research and engineering effort went into this. The new research I am showing basically does not exist in the literature. It is my job as a diligent researcher to know the prior art. If you think I have missed some prior art and making claims that are not true, then tell me please.

    How do I know as a practical matter that these things are relevant? Some highly respected people are testing out the device and confirming what I say is indeed the case. 

    Nathan, if you are genuinely interested in my 'class' then please have your local PTG president contact me, as I have been successfully offering the class to many monthly PTG groups over zoom.

    Respectfully submitted,

    Steve



    ------------------------------
    Steven Norsworthy
    Cardiff By The Sea CA
    (619) 964-0101
    ------------------------------



  • 20.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-21-2024 21:51

    I believe that most of us here understand what you are proselytizing with the graph and measurement results of your invention. And I don't think many of us have any argument against what you are demonstrating. In fact, bravo on your accomplishments! My point, however, is that I find that I am not alone in questioning the usefulness of these data, and the need of such accuracy as your device purports to reveal, in the real world. A 0.25¢ inaccuracy between two strings in a unison, at A4, results in a beat rate of one full cycle beat about every 16 seconds. While that may be interesting data in the lab, it's meaningless in the real world. And any tuner that can't get two strings of a unison to within less than one cent of each other is not listening to their work very carefully. Regardless of the accuracy of whatever ETD we may be using, or the method chosen to capture the vibrations of the strings, if we're not listening to the results, as the final test, in order to achieve the best acoustically audible sound we hear with our ears, we're not doing our job. And any tech that is relying on your, or any device for that matter, to give them even an acceptable tuning is, again, not listening. Regardless of our tools, we tune for our ears, not the tools. 



    ------------------------------
    Geoff Sykes, RPT
    Los Angeles CA
    ------------------------------



  • 21.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-21-2024 23:35

    Geoff,

    Do you know Bill Shull? I presented at his PTG last week. Bill was impressed when I demonstrated on his Steinway B how I tuned unisons string by string at a time and nailed them to zero cents. Call Bill and talk to him. Paul McCloud has tuned hundreds of pianos with it. He is not on the take and is not a company rep. He will honestly tell you. His customers can all hear the difference. What more can we say?

    Keep it simple: if you think this is not for you, then no need to keep writing and arguing but if you keep writing because you want to dissuade others, that is another matter left for private communication but everyone shoud be able to make up their own independent mind without someone who is uninterested trying to persuade his peers. I don't get that.

    Many are wiling to try it. Some will just imagine it is not practical and cannot hear the difference. In my class audio examples, everyone can hear the difference. You don’t know until you try. If you have not heard the difference and experienced the advantages, how can you presume?

    Can see you at Faust Harrison tomorrow night at the Orange County PTG if you want to come over.

    Steve




  • 22.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-22-2024 02:58

    Steve --

    If I wasn't interested I would not be participating in this conversation.

    I am not arguing with you, nor am I trying to persuade others either way. Along with others, I am asking questions that you choose to get defensive about instead of addressing. And you get to do that. But honestly, my questions have not specifically even been aimed at you. I was hoping for other techs to expand on the issues I was raising. Halfway expecting to be told my math, my logic and my tuning goals were wrong. 

    I think your invention is a marvel. And it's not the first time I've said so. But it's not for me as you are currently marketing it.

    In order not to start an argument, which I'm really bad at, I have responded to you privately. I will continue to monitor this thread, (who knows. I might learn something. 🙂), but I have no further need to participate.



    ------------------------------
    Geoff Sykes, RPT
    Los Angeles CA
    ------------------------------



  • 23.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-22-2024 03:09
    Geoff,

    The best way to communicate is a zoom call. We can interact. The open forum is difficult. 

    The best way of addressing your questions is to see the whole class.

    This 'back and forth' is very tedious and things get misconstrued.

    The PTG class on OC is this coming evening.

    We can still do zoom later.

    Best,
    Steve








  • 24.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-22-2024 02:35

    I can't speak for anyone else, but given the opportunity to try it out Pianosens I absolutely will (I'm thinking at a convention or something to that effect).  Dealing with the extreme high treble is pretty painful in general.  My current ETD, Tunelab, is much better than doing it by ear, but hey it could certainly stand to be improved.  I'd be curious to hear the effects elsewhere too, just saying that's where my greatest angst with existing technology lies.  So please don't feel like you need to convince me to give it a try.

    I still disagree with the claim of a per-note 0.5c standard deviation leading to a 5+ cent quantity of any sort that's meaningful for the quality of the tuning.  You mentioned that one of your testers reported notes that were off by that much; I'm not questioning the empirical evidence there (although I'm guessing it must have come from the high treble?). I think your later approach of stating probabilities on a gaussian curve is a little better, but let's run the numbers with 0.5c and 88 keys:

    The rule of thumb, as I understand it, goes that 68% of samples will fall within 1 sigma (0.5c), 95% will fall within 2 sigma (1.0c), and 99.7% within 3 sigma (1.5c). The numbers are double the ones you listed because we have to account for both positive and negative deviations. There are fancier formulas to give you the probability for any given sigma (the cumulative distribution function), but I'll probably screw them up if I try to use them.  So right off the bat, the probability of even a >1.5c deviation is only 0.3% per note, or 3/1000.

    The probability of getting at least one of those on an entire piano, if I punch that into a cumulative binomial probability function, is 0.23233113, a little more than 1 in 4 pianos.

    When you get up to 10 sigma (which is needed for 5c, with an 0.5c sigma) you're talking really tiny numbers, the odds are something like 1.5e-23 per note.

    I admit that I am really rusty on this stuff, so I might be the one making a mistake somewhere... but either I got something wrong or taking stddev * n / sqrt(n) simply isn't correct. I also get that the _actual_ sigma might be higher than that.  Sounds like it might vary a lot in different parts of the piano.  I suspect ultimately we'll come up with clearer rules of thumb based on a lot of people experimenting with it, and I look forward to seeing the results.



    ------------------------------
    Nathan Monteleone RPT
    Fort Worth TX
    (817) 675-9494
    nbmont@gmail.com
    ------------------------------



  • 25.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-22-2024 03:06
    Nathan,

    The fact is that we can easily see 1-2 cent spreading variance distributions from a single note just by moving the mic a bit. We see no change for the sensor.

    I more than clarified by clearly stating that it is harder and harder to get good std dev in the extremes of low and high registers.

    We will wrangle forever over exact statistical models, but we need not wrangle over the data that shows the invariance of the sensor. I go over all of this in my ‘class’ I make to the PTG groups.

    Best,
    Steve




  • 26.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-22-2024 09:55

    I'm arguing all of this in the interest of being fair in our evaluation of the existing technology.

    I appreciate you sharing the empirical findings. I think statical models are interesting because they help us quantity how much improvement we're getting, but it's important that we use them correctly. If you don't wish to get into it further I will leave the matter.

     I reviewed the video & graphs of the bass notes; in that case the fundamental was indeed off between 1-2 cents. The upper partials, however, were a lot closer. Tunelab does not use the fundamental for the low notes -- it has a table of partials to use for each note. I believe these were selected to get around the sorts of problems you found when measuring the fundamental. Doubtless this kind of partial selection has it's own pitfalls. But I don't think Tunelab is producing quite the magnitude of error you're claiming for the bass. I can't speak for the other ETD's because I'm not familiar with them.

    Again, I'm not out to disparage the use of a sensor, or really even question that it can provide an improvement in terms of repeatability. This is about being accurate and fair in our evaluation of prior art.

    I look forward to continued discussion with you, and hope I have not tried your patience excessively :)



    ------------------------------
    Nathan Monteleone RPT
    Fort Worth TX
    (817) 675-9494
    nbmont@gmail.com
    ------------------------------



  • 27.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-22-2024 14:09

    Paul wrote: <"Can anything but a very sensitive machine tell, or even care for that matter, if a tuning is more than 0.2¢ off?".  Well, unless you have the experience of tuning to this level, you might not know how it sounds.  Once you hear it, when there is no beat at all, consistently, then you have to admit that what you've been doing is not as good as it could be.  We don't wait long enough while tuning unisions, for example, to hear a slow beat.  By ear, it takes too long to hear it, so we just don't take the time to do it.  But if your app can give you that information, then you can make corrections to get within that range +- .2 cents very quickly.  You do hear it. ....>

    Translate simply: If you know what it sounds like, you can tell when you have (or have not) found it."

    A few years ago a Yamaha concert technician was making chapter presentations. His topic was "rising unisons." He demonstrated on a new, top line 7 foot Yamaha grand piano. He said "Listen to this unison. When I have found the right spot, the sound will seem to rise up and float above the piano. When I don't have it right, the tone will drift away downward, below the piano."

    He demonstrated this with several notes, letting each note sound for several seconds, and the technician audience began to smile in amazement. We could hear it!  He said "I don't know what it is, but I know I can hear it when I've got it and I can hear when I haven't got it."

    After that, at least for a while, on very good pianos in very good condition, I was sometimes able to tune some rising and floating unisons.

    I now wonder if what we were doing was just tuning very, very pure unisons, within the .2 cents window.

    (Years ago I learned that violinists get certain effects by imagining that the bow is weightless, just floating in the air where needed. Images like this can sometimes help us to reach higher levels of perception and performance with our "wet" operating systems.)



    ------------------------------
    Ed Sutton
    ed440@me.com
    (980) 254-7413
    ------------------------------



  • 28.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-22-2024 14:52

    Hi Ed, I assume you saw my post on this thread about the 180-deg phase shifting creating the 'bloom' effect. If you want me to describe it better, we can do a little zoom session. Let me know when you what to schedule it. Evenings are good. This evening is out because I will be doing my 'class lecture' at the Orange County PTG. I am really busy doing lots of PTG meeting sessions now. Respectfully yours, Steve



    ------------------------------
    Steven Norsworthy
    Cardiff By The Sea CA
    (619) 964-0101
    ------------------------------



  • 29.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-23-2024 09:14
    I could be wrong but I don’t think that Ed was talking about. Bloom, Steve. I attended a class by Yamaha’s Ace Ugai and I also was tutored by him in tuning unisons. He taught me to listen for the sound coming from above the soundboard. It didn’t develop over time. I didn’t have to wait to hear any bloom. That doesn’t negate the efficacy of your approach and I would sign up for your device if I wasn’t at the end of my career and trying to not spend any more money than necessary. I have heard a lot of theoretical discussion about your approach but the bottom line is whether or not it delivers results. Paul McCloud certainly thinks so and, despite all the arguments pro and contra, the only way to decide is to try it.

    Bob Anderson, RPT
    Tucson, AZ




  • 30.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-23-2024 17:07

    Thanks, Bob. Last night I was presenting to the Orange County PTG. At the end of my class slides, I went up to a Bosendorfer 185VC, put my device on C4, one string at a time individually tuning each unison, and about 10 secs per string, nailed each one to 0.0 or 0.1, and unmuted all and played the result, and it was 'perfect'. The stability and consistency and invariance allows this. It is a 'new way' of tuning.



    ------------------------------
    Steven Norsworthy
    Cardiff By The Sea CA
    (619) 964-0101
    ------------------------------



  • 31.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-23-2024 20:51

    Yes, Ace Ugai was the teacher.



    ------------------------------
    Ed Sutton
    ed440@me.com
    (980) 254-7413
    ------------------------------



  • 32.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-23-2024 09:23
    Correction: it wasn’t at a class but rather at Yamaha’s Performance Piano Seminar (which I highly recommend) where I got my tuning instruction from Ace Ugai.

    Bob Anderson, RPT
    Tucson, AZ




  • 33.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-24-2024 08:23

    I remember Ace at a NE Regional Convention in a class with David Durban. This discussion about "floating" unisons and projecting them outside the piano was entirely new to me at that time (and admittedly I was skeptical). 

    After class I tried analyzing the unisons Ace had tuned, interval checking each string to see if there was some pattern I could recognize. There was not, AND the strings were NOT tuned at perfect "zero". What's more, Ace was right there next to me and said: "you can't do that" (I'm paraphrasing) "you must listen" and then waved with his hand outside of the piano. 

    At the time I was baffled, yet intrigued at the same time. I have since come to settle on a protocol using the middle string as a fine adjustment "controller" of sorts. I know I'm not alone in this. 

    Yet, I am still eager to hear unisons tuned as "dead on perfect" as humanly (and machine assisted) as possible.  I will humbly evaluate the pros and cons. Lately I have been incorporating a bit of Tunelab into some of my work. The idea of tuning the outside strings using the machine and then tuning the middle string aurally (David Stanwood mentioned this to me recently on a phone call) seems to be an ETD synthesis of what I've been doing aurally for quite a while. I kind of like it too. 

    Looking forward to getting my Pianosens...

    Peter Grey Piano Doctor 



    ------------------------------
    Peter Grey
    Stratham NH
    (603) 686-2395
    pianodoctor57@gmail.com
    ------------------------------



  • 34.  RE: Sensor Position Invariance. YouTube Video

    Registered Piano Technician
    Posted 01-24-2024 12:06

    Indeed, you are not alone in this, Peter. I too attended one of Ace's classes in which he discussed his ideals for unison tuning. Definitely not about absolutely matching the frequencies of the strings of unisons, but about getting a certain loftier sound from the strings combined. It was a great affirmation of what many of us were already doing!

    Best,

    Alan 



    ------------------------------
    Alan Eder, RPT
    Herb Alpert School of Music
    California Institute of the Arts
    Valencia, CA
    661.904.6483
    ------------------------------



  • 35.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-24-2024 16:21

    From the video Steven provided for this thread:

     

    https://www.youtube.com/watch?v=xUfOnQWWoD4

     

    Since the PianoSens output is relatively stable across the 10 partials displayed, regardless of sensor position on the strings, I assume that the sensor has an embedded processor that is taking the actual mechanical energy amplitudes of partials, interpreting what the sensor position is, and reconstructing a fairly uniform output representation of the vibrating string.

     

    A question then: should we expect some entrepreneur, as a logical next step, to embed an ETD within the processor of a string sensor product like PianoSens? Maybe also incorporating both audio (a microphone) and mechanical energy to supplement each other? Maybe with Bluetooth coupling to a visual display on a cell phone? And of course incorporating AI! Regards, Norman



    ------------------------------
    Norman Brickman
    Potomac Piano Service
    Potomac, Maryland
    potomacpiano@verizon.net
    https://potomacpiano.com
    (301) 983.9321
    ------------------------------



  • 36.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-24-2024 17:06
    Norman, the sensor is taking raw string movement, then a preamp, followed by a 24-bit ADC. The signal processing in Matlab takes the data and does not do any normalization to the partials and doesn’t do any EQ that you imply. Call me for a further technical discussion. Steve 619-964-0101




  • 37.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-24-2024 18:36

    Steven, thanks for the offer of a phone conversation. Let me first see if I can dig in to a couple of physics papers that I have come across on vibrating strings for pianos. Not a big deal overall – I've certainly been wrong before! For those interested in a visual of what is being discussed here, try http://hyperphysics.phy-astr.gsu.edu/hbase/Waves/string.html which figuratively shows the first 6 partials.

     

    From your video, it looks to me like you are placing PianoSens at approximately 25%, 50% and 60% toward the hammer strike point from the termination point, which would be at 3%, 6%, and 7.5% of the total string length. A node for partial #10 that you show would be at 10% string length and its anti-node at 5%.

     

    You can see why I figured that there is digital signal processing going on within PianoSens – with, say, the partial #10 anti-node being right in the middle of the three sensing points, and its node close by. Yet the visual display shows very little movement for partial #10's amplitude.  And also you say in the video that the sensor output is "virtually invariant along the length of the string you want to tune to."

     

    If you get a chance to supplement the video with a few additional test points (like 10% and 80%), that would be much appreciated. Regards, Norman.



    ------------------------------
    Norman Brickman
    Potomac Piano Service
    Potomac, Maryland
    potomacpiano@verizon.net
    https://potomacpiano.com
    (301) 983.9321
    ------------------------------



  • 38.  RE: Sensor Position Invariance. YouTube Video

    Posted 01-24-2024 19:21
    Norm, 

    1/2 wave is node to node, zero to zero of the string length of the fundamental, and etc etc. I place the sensor behind the hammer in the lower register and that way I pick up more higher partials, higher than the 8th partial, etc.

    I am changing the phase points and the phases of the partials are linear phase with respect to each other. The amplitudes are changing with respect to one another according to the phase changes. It does NOT affect the frequency measurements because I am not imposing acoustic interference. My measurements are not affected on frequency, therefore. 

    I show this in the video.

    Best,
    Steve