Piano World Home Page Forums Our Most Popular Forums Piano Tuner-Technicians Forum Novel Piano Tuning App

support@pianoscope.app

Thanks!

Seems interesting solution.

Frank,

I submitted some feedback via the beta program today, but it might not have come out too clearly. With a new build, I decided to redo all the iH measurements. I was expecting, after that, that there would be enough data for a pitch raise (indeed, the old data I had entered a couple of weeks ago was still there). To my surprise, on selecting new pitch raise, I was prompted to re-enter a bunch of notes to set the initial tuning - even though I'd just sampled every single note of iH. My understanding was that the pitch raise function should have been able to use these iH measurements as a measure of the initial pitch? The manual says this, but I was never presented with the screen to do so.

Paul.

Paul, thanks for the feedback. This is most probably a bug. Carrying over the pitch from the IH measurement definitely works if you have not yet started a pitch-raise before with the same document. I will fix this. But it will take until early next week, since I am quite busy at the moment.

Thanks Frank - I've been using the same document for all my testing so far.

Paul

I worked on a Lyon & Healy spinet today and found that the solution (this time) for problematic octaves across the break was to reduce the weighting (when viewing the inharmonicity graph) down to 25%. Then the octaves and other intervals seemed to behave a bit better with a smoother tuning curve just above the break.

The bottom octave still ends up sharper than I would like - not sure if it is focusing on the 4:2 or some lower partial match than the typical 6:3 or larger for the lowest octave?

Ron Koval

Thanks Ron, I am currently working on the sharp bass issue. I currently think that the reason for the current behaviour is that my loudness model is putting too much emphasis on low frequencies which leads for example to 2:1 and 4:1 octaves being too strongly weighted in the bass making the 6:3 octaves too narrow. So I think that your change with the inharmonicity weight might only be a coincidental remedy. I will do more testing and hopefully have a new version soon for you to try.

Thanks! I felt like adjusting the Ih weight just helped smooth the octaves over the break.

Ron Koval

I have just released build 478 of pianoscope which should fix a number of issues that you all had reported.

The biggest change is a new loudness model for partial intensities that should better fit human perception. It puts significantly less emphasis on low frequencies and should fix the problem with a too sharp bass that some of you encountered. In the last days I have performed many new automated test calculations with recordings of a dozen different pianos, from a tiny spinet up to a large Fazioli. I also compared the resulting tuning curves against Verituner and PianoMeter and the resulting stretches now look very similar in all these cases. I am very interested in how the updated model works for you.

I also adjusted the partial weights of the default tuning styles to fit the new loudness model. If you have existing tuning documents that you want to re-use, I strongly recommend that you reset your styles to the new factory values. You do this by swiping left on the selected tuning style in the tuning curve view and by selecting "Discard Changes".

In the inharmonicity view, you can now choose between a fine and a coarse measurement. The fine mode requests you to play all notes up to C7, the coarse mode all As and Es up to E6. In both modes, you are free to record any notes you want as before. The fine mode is the default and is recommended if you want maximum precision. I updated the documentation to reflect the changes.

The new build also contains a couple of smaller changes and fixes which are listed in the release notes in Test Flight.

Thanks again for your great feedback. You guys are amazing!

Hakki & Ron: I plan to perform extensive measurements on this topic in the next days. The open questions for me are:

a) How big exactly is the additional error you get when you combine the partial information of every note into a single IH number provided that the inharmonicity of all notes is measured? This approach is currently used by Cybertuner, PianoMeter, PiaTune and TuneLab so the error can't be too bad, but I want to objectify the error.

b) How big is the error you get from measuring inharmonicities on detuned unisons? This is important for saving time in apps which measure many notes upfront like PianoMeter and pianoscope.

If an app measures only a sample of notes and derives the others from a model, there is an additional error, which strongly depends on the instrument. TuneLab and Cybertuner take this road, and pianoscope in the coarse mode. But I am less interested in this error as it can be avoided by simply pre-measuring more notes. And as IH measurement can be done fast, it should not be a problem. The benefits of a pre-measurement are that the tuning algorithm has more information to perform a global optimization and to allow tuning in any order.

If the answer to question a) should turn out to show a significant error – contrary to what Hakki observed – I could easily change pianoscope to use the full partial information instead of a single IH number for every note. I am curious to see the results and I will be happy to share them with you.

I would like to derive a metric and a threshold from b) with which I can prompt the user to mute the strings of a certain note when I detect an unsufficient partial / IH measurement. This would have the benefit of not having to always mute all strings for the highest precision. A question to you practiced pros: How much more time does it cost you if you have to mute all the unisons when pre-measuring?

Hello Frank!
Your program has worked very well for tuning poorly scalable pianos. But... There are a few notes regarding the interface design.
1. Allow the user to choose their brightness and contrast of squares up to the same type as in Tunelab. If this interferes with the perception of the needle, then it is possible to change the color of the needle to the opposite at the place where the needle is placed on the square.
2. Since we can now control notes and octaves by pressing and swiping, there is no need for the keyboard to always be at the bottom of the screen (except for the harmonics view). This will make room for the phase display.

Frank, thanks for the updates!

For measuring - it is so much easier and takes less dexterity to just play notes while watching the red turn to green on the keyboard display that keeps the process very quick.

To shift focus up and down while moving mutes isn't a deal-breaker if the results are better, but certainly slows down the process.

I did tune a piano yesterday that was 100-200 cents flat and couldn't get readings to begin the pre-measuring phase. I wonder if having some inharmonicity templates for average scales (something like - spinet, upright, small grand, grand)would be a help to pitch raise without measuring inharmonicity, just measuring the starting pitch of selected notes?

It may be that allowing any note to be played during the inharmonicity measuring now makes it difficult for the app to discern what note is being played when the piano is way out of pitch?


Ron Koval

Thanks, Vlad. Both your suggestions make sense. I will try them out.

Hi Frank! Today I decided to make an experiment. I chose the Gerbstadt cabinet piano with a very big non-harmonicity. I set up this piano for three programs. Two iPhone 6S and iPhone SE. Tunelab, Verituner and Pianoscope programs. At Tunelab I tried all the styles and I did not get a good result. Strongly overestimated 5, 6, and 7 octaves. Verituner made an A4-A3 octave too wide and bass too low. Pianoscope has done excellent, all settings were default and not a single error. Frank, would like a stroboscope not pale and with clear edges, this is the eye of the program. With sunlight and with high glare it is not visible at all. I liked that Vlad offers you.

D520,
Please explain: was not a good result with Tunelab and Verituner A4-A3 too wide and bass to flat. Aural interval tests?

All heard on the ear. Tunelab makes strange and illogical calculations. Sometimes an octave A3-A4 when installing 3: 1 wider than when installing 6: 3, and sometimes the opposite. High constantly sound too high. Verituner due to the fact that it measures every note, transfers errors to other octaves. There is no integrity of the system. For example, if you measure A4, then in these two programs the result is absolutely different. In Tunelab, high frequency tones sound higher than in Verituner and as a result of octaves 4: 1 sounds too high. In Pianoscope, everything sounds smoothly.

Also, please say what Verituner Style you used.

Ron, thanks for reporting this. As it turned out, this was a bug which made the inharmonicity and pitch raise views not work well with very flat instruments (>150 cents). I have just released build 483 in which this should work much better now.


In principle, I can get enough inharmonicity information out of the pitch-raise measurements. I will think about how to visually present both measurements inside the pitch-raise view, so that you can perform both measurements in a single pass.

D520, thanks for the feedback. It is good to know pianoscope has so far passed your tests.


I am already working on an update with more contast and sharpness options for the strobe.