Right. None of this was to dis Matlab or your research. I love Matlab and highly recommend it. I just wanted to push back against the implication that the cost of the tool is the important factor. I won't argue the point further.
Original Message:
Sent: 02-26-2024 12:56
From: Anthony Willey
Subject: Tune with a Single Partial or Multiple Partials? An Example...
I just want to dispel some of the mysticism about Matlab. It's just a computer program, and it doesn't cost $20k. Anybody here can buy Matlab as a home user for $150. (If you're a professional using it to earn money, it's more expensive, around $2,500.) But there's nothing uniquely special about Matlab. Like any tool, it's only as useful as the person wielding it. Matlab is really good at number crunching and it gives you easy access to a lot of powerful algorithms, but it's not the only number cruncher out there. With some extra time and effort you could get the exact same results in lots of other programs or programming languages. Math is platform independent :-)
In summary: what you do with a computer program is more important than its cost. Nobody should say "This financial analysis is accurate because I did it in a powerful spreadsheet program called Excel" or "This tuning is stable because I did it with a high end $500 tuning lever." The same goes for Matlab.
Final note: if people want to engage with this topic in a forum where Steve can respond directly, there's a discussion over on PianoWorld.
Edit: Oh, and lest anyone misinterpret, I'm not suggesting in any way that Steve isn't highly capable with Matlab.
------------------------------
Anthony Willey, RPT
http://willeypianotuning.com
http://pianometer.com
Original Message:
Sent: 02-26-2024 00:33
From: Paul McCloud
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Here is the latest video from Steve, demonstrating two examples of mic placement, showing the inaccurate readings on two pianos. It is understood that pianos have two modes of vibration, each producing a slightly different frequency. Thus actually, every string has a "false beat" in it, even if it is not heard. Hard to believe, but it's true. Our ETD's are not able to pick one or the other consistently, and so they "blend" them together. Steve gives a technical explanation, which might seem cryptic to some. Basically, using a mic with our ETD's can give erroneous results, depending upon where they are placed inside the piano. This video shows proof of this phenomenon, and is not using any of our ETD's to evaluate the results. The analysis is done by Matlab, a separate engineering program running on a computer. The cost of this program is about $20k. (I hope I got that right). What Matlab shows is that our ETD apps are not as accurate as we believe. And that mic placement problems are a source of inaccuracies. Only a sensor, which is picking up string vibration only, is able to avoid these problems.
https://youtu.be/LPG32Kr1nKM
------------------------------
Paul McCloud, RPT
Accutone Piano Service
www.AccutonePianoService.com
pavadasa@gmail.com
Original Message:
Sent: 02-20-2024 17:19
From: Paul McCloud
Subject: Tune with a Single Partial or Multiple Partials? An Example...
I did not try to use Bluetooth with the sensor, but earlier I had tried Bluetooth headphones when I was using mics and limiters. It was very disconcerting because of the latency. There are different codecs for Bluetooth, and some are better than others. But any latency throws off your coordination when you try to tune with it.
I did look into using wireless technology with the prototypes, but Steve told me that it would be noisy and have much less fidelity with Bluetooth, or any wireless technology. The real problem with all of this is the noise floor and the gain or strength of the sensor pickup. The sensor has to be able to pick up enough energy from the vibrating string, and having the iRig preamp increases the signal to a higher level that the ETD's can use. But if there is noise from the sensor, that noise gets amplified along with the signal. So, this doesn't improve things. If the wireless system has noise, that's going to be a problem too. The other thing that we have to deal with is the accuracy of the information being fed to the ETD. How would we even know if the pitch calculation is correct, given the presence of noise in the signal? Adding to that is the bulk that the electronics and a battery would add to the sensor. In the end, we agreed the best solution is to use the wired version for its small size and clean signal.
------------------------------
Paul McCloud, RPT
Accutone Piano Service
www.AccutonePianoService.com
pavadasa@gmail.com
Original Message:
Sent: 02-20-2024 14:39
From: Patrick Draine
Subject: Tune with a Single Partial or Multiple Partials? An Example...
My reading of earlier posts indicated Bluetooth significantly interfered with operating PianoSens with ETD apps. I'm sure Paul can elaborate.
------------------------------
Patrick Draine RPT
Billerica MA
(978) 663-9690
Original Message:
Sent: 02-20-2024 14:27
From: Steven Rosenthal
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Peter, watch out you don't get hog tied with all those wires. jk
Seriously, I wonder if you can put on a bluetooth dongle for the headphones. There may, or may not be a latency issue.
------------------------------
Steven Rosenthal RPT
Honolulu HI
(808) 521-7129
Original Message:
Sent: 02-20-2024 14:16
From: Peter Grey
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Paul,
I believe he has even newer information about string behavior to share as well that can answer some questions many of us have had for a long time. Looking forward to it.
Peter Grey Piano Doctor
------------------------------
Peter Grey
Stratham NH
(603) 686-2395
pianodoctor57@gmail.com
Original Message:
Sent: 02-20-2024 08:46
From: Paul McCloud
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Yes, of course this thread can continue.
The pure sound of a string coming through the sensor is so clean, it's not hard to understand how much better it is for an ETD. All the extraneous resonances are gone. My experience using it has been that it is best placed near the ends of the string, even up against the bridge, for the steadiest readings. I haven't tried using headphones to listen to this, so I'll have to try it.
------------------------------
Paul McCloud, RPT
Accutone Piano Service
www.AccutonePianoService.com
pavadasa@gmail.com
Original Message:
Sent: 02-20-2024 07:19
From: Peter Grey
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Paul,
Nonetheless we can still discuss it here among ourselves, can we not? I think it is a significant development in tuning technology. And not just for ETD users but even for aurally based tuners as the headphone jack brings the sensor pickup in info straight to the ears in a volume controlled manner. Again, it is not a panacea, but with a little getting used to gives another perspective on the matter, reducing some of the falseness on some notes, especially up in the treble.
Of course I can already hear the cry: "but we don't listen to the piano that way...", however a good unison is a good unison and thus far (again in my limited usage) I have been able to produce very good unisons with this assistance (and obviously you can control the volume). I will continue to use it and uncover it's capabilities, though I'm not in my regular tuning season here in NE.
Peter Grey Piano Doctor
------------------------------
Peter Grey
Stratham NH
(603) 686-2395
pianodoctor57@gmail.com
Original Message:
Sent: 02-20-2024 01:11
From: Paul McCloud
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Ladies and Gentlemen:
Steven Norsworthy will not be posting here or answering any more questions on the forums. He has set up a Facebook page "Pianosens" if you wish to continue to engage with him and other users.
Steve will be continuing his research on string behavior and the use of his sensor. Some of this information will be available on his FB page. He also has a website "Pianosens.com" for more information about the sensor.
------------------------------
Paul McCloud, RPT
Accutone Piano Service
www.AccutonePianoService.com
pavadasa@gmail.com
Original Message:
Sent: 02-19-2024 12:08
From: Peter Grey
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Okay, a further update:
I am now seriously impressed with this thing. I have found that moving up into the treble area is far less problematic than i.had anticipated. All it takes is a little ingenuity with wire routing and the problem is solved. It fits nicely right next to the note being tuned and reads it with no problem.
I was even able to read, tune, and "hide" some nasty false beats, and end up with surprisingly nice unisons. I'm not kidding and I even surprised myself. The sensor in conjunction with Tunelab is a surprisingly effective combination. I'm actually starting have fun.
Panacea for all piano ills? No. But it definitely opens up possibilities heretofore problematic (at least for me...others may have different solutions and that's fine with me) for many tuners.
Peter Grey Piano Doctor
------------------------------
Peter Grey
Stratham NH
(603) 686-2395
pianodoctor57@gmail.com
Original Message:
Sent: 02-18-2024 23:22
From: Peter Grey
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Okay...so the only thing I have available in the shop is this 40" SS console. Not a lot of fun, but I thought it would serve as a real world test of mic based variance vs sensor variance. Not highly scientific but I did take notes:
Using Tunelab on my Android phone I set it up on a magnetic base holding device which keeps it solidly in position. Measured IH per TL protocol abd proceeded to tune just the middle strings of the center section (strip muted) letting TL make all the decisions (no questions asked, no answers given). The range is from F3-G#5. Here are my notes:
Round 1
F3-F4...significant wide swings, sharp and flat (the display), rarely settles, hard to get it to settle solid at ZERO
F#4-E5...serious jiggling back and forth, IMPOSSIBLE to get a solid reading, can only strike a 'balance' between sharp and flat, NO ZERO
F5-G#5...reasonably solid at ZERO, close but not perfect
C5...IMPOSSIBLE to know what to do with this note, all over the place
(Since it was 7-9 cents flat to begin with I pulled in the unisons and did the whole thing again on the center strings)
Round 2 Go back and check each note tuned:
1st time...approximately 75% of notes reading slightly flat (I didn't measure, just the fact that the display showed flat now)
2nd time...now only about 30% of notes read flat (interesting)
3rd time...estimate roughly 50% of notes read flat, a few now read sharp.
After this I "corrected" everything to TL as best as I could (same difficulties as initially, nothing changed there)
Round 3 Check notes again then move ETD and recheck:
Moved ETD 1.5"...Everything OKAY until E4 which goes nuts slowly from .5 - 1.8 it's flat (this was kind of weird)
C5-G#5...showing 1.5 - 2 cts flat almost universally
Moved ETD 12" towards the treble...D4 1.7cts sharp, C5 3.6 cts flat, D5 2.5 cts flat, D#5 3cts flat, E5 3.7 cts flat, and at this point I threw in the towel.
Now I plugged in Pianosens for a comparison: Instantly about 90% more stable, notes impossible to read previously now easily read, not perfect but WAY better. No jitter, no wiggling, yes some detection of audible false beats but resolution very manageable. Had I tuned using the sensor from the beginning would easily been done far quicker. I decided at this point to tune the unisons while still using the sensor. Yes, it's a little clunky because I'm not used to it, but TL did not read any strings previously tuned as having moved sharp or flat.
Okay, likely someone is going to claim that because I had already tuned and retuned the strings several times by the time I started using the sensor that they were now much more stable than earlier. I get that. However strings that are properly set (and I do know how to do that) do not move around as randomly as the initial readings were showing due to operator error (me not setting strings properly).
My take on it is that the ETD performance was greatly enhanced with the use of the sensor. Had I been trying to actually tune the piano with the mic only I would have given up in disgust and just done it over by ear. Had I been using the sensor I would have continued throughout with it.
Again, I'm sure some will question various things here but I'm convinced that there was serious variance of pitch resolution using the mic, and virtually (repeat virtually) and greatly reduced variance with the sensor. This is just one sample so far.
I'll answer questions if there are any.
Peter Grey Piano Doctor
------------------------------
Peter Grey
Stratham NH
(603) 686-2395
pianodoctor57@gmail.com
Original Message:
Sent: 02-18-2024 21:31
From: Bill Ballard
Subject: Tune with a Single Partial or Multiple Partials? An Example...
I went: "Double dog dare…"
That was me summoning up the courage to call the issue of location variance a red herring. Like saying there's no point to comparing the fact that an apple's skin is edible and the orange's, not, because this aspect of each fruit is contained within its definition.
------------------------------
William Ballard RPT
WBPS
Saxtons River VT
802-869-9107
"Our lives contain a thousand springs
and dies if one be gone
Strange that a harp of a thousand strings
should keep in tune so long."
...........Dr. Watts, "The Continental Harmony,1774
Original Message:
Sent: 02-18-2024 10:18
From: Bill Ballard
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Nathan went: Just for grins I shot a video of C4 using TuneLab. Please pardon the crappy videography. I get that you can't really test an app against itself, I just wanted to see how much its own output changed from position to position.
You don't need to apologize for the videography. It gives a very clear picture of the situation.
I'd like to suggest to all that the thumb-on-the-scale of ETD mic positions, while real, is actually of minor consequence. My use of TuneLab is limited to one-string-per-note in the central section and 7th octave of the piano (with all unisons, aural). In the former, the power in that section is loud enough that my ETD has plenty to work with in just one spot. In the 7th octave, because greatly reduced power, I do have to move the unit twice (at the most, three times) when the EDT gets to be 6-8" away from the note being tuned.
Also worth mentioning is that for Pianosens, proximity to the note being tuned is by design not an issue, doing its work within 1/8" of the string being measured.
If an ETD mic is rarely being moved, and if its chosen position is not at the far end of the piano from where one is working, this variance is (dare I say it? yeah, I double-dog dare you!) a red herring.
But, for full disclosure, I was fully convinced of Pianosens' value even before Peter Grey's NH Chapter demo.
------------------------------
William Ballard RPT
WBPS
Saxtons River VT
802-869-9107
"Our lives contain a thousand springs
and dies if one be gone
Strange that a harp of a thousand strings
should keep in tune so long."
...........Dr. Watts, "The Continental Harmony,1774
Original Message:
Sent: 02-14-2024 21:30
From: Nathan Monteleone
Subject: Tune with a Single Partial or Multiple Partials? An Example...
Just for grins I shot a video of C4 using TuneLab. Please pardon the crappy videography. I get that you can't really test an app against itself, I just wanted to see how much its own output changed from position to position.
I'll try uploading it to the forums here too but I shared it at:
https://photos.app.goo.gl/uwvV45qSF6eavUWm6
I can largely see the effect being described here in TuneLab's display; notice that in positions #3 and #6 the pitch reading goes flat by a little more than a cent in the first second or so, and some of the others positions actually go sharp.
I think it's worth nothing, however, that they all converge on readings between -0.2 and 0.0 after a couple seconds. (I know, I moved the camera too soon after position 2 and you can't really see it, sorry.) In practice, this is what I'd actually tune to, if I was wanting to get max precision out of the ETD.
So it sounds like the data's showing us that an emf sensor stands to give us the right answer quicker, if nothing else. On the other hand, perhaps we need not be concerned that our mic-based readings are actually leading us that _too_ far astray if we're waiting for them to settle.
------------------------------
Nathan Monteleone RPT
Fort Worth TX
(817) 675-9494
nbmont@gmail.com
Original Message:
Sent: 02-14-2024 18:47
From: Steven Norsworthy
Subject: Tune with a Single Partial or Multiple Partials? An Example...
We had a lively discussion last night at the South Bay PTG where I gave a class.
One of the issues is whether we tune with a single partial or with multiple partials. If an app choses a single partial for tuning, we could be lucky to see virtually no difference where the mic was placed. But if we are unlucky we could get more than a 1 cent error range with a small mic placement movement. It is not that the movement creates a different frequency, but it creates a different resolution of that frequency.
Here is an example. Using the note C4 on a 7' grand piano, the mic was moved around to 9 different physical positions above the soundboard in reasonable locations that a tuner could place an iPhone. The .wav files were captured and analyzed in Matlab. The result was the following spread of errors with respect to placement. The data is below. Notice the sign of one relative to the other.
Partial 2: -0.88 cents
Partial 3: -0.10 cents
Partial 4: +0.93 cents
Partial 5: -0.47 cents
Partial 6: +0.79 cents
Depending on the tuning app, if it tunes on one partial, it could either be lucky (-0.1 cents) or unlucky (+0.93 cents).
Alternatively, if the app tunes by trying to fit the data statistically, it could use a mean, median, or standard deviation.
mean = 0.0540
median = -0.1000
std. dev. = 0.7874
range = 1.8100
A string sensor will not have these variances because small placement changes are inconsequential and produce one answer. Then it is up to the IH curve fitting in the app to make sure the measurements align with the IH initial calculations, thus ensuring we have extremely low variance and can trust the answer with more confidence.
Best,
Steven Norsworthy
------------------------------
Steven Norsworthy
Cardiff By The Sea CA
(619) 964-0101
------------------------------