Piano World Home Page Forums Our Most Popular Forums Piano Tuner-Technicians Forum New PC Piano Tuning Software

This one's called "Dirks piano tuner." It seems different from the others on the market, because the program starts by getting the tuner to strike and record (with the program) all 88 strings (one string only per note). It then analyzes all of the information (which apparantly can take up to a minute or 2, depending on the power of your processor), and produces a stretch. The stretch is never smooth, but jumpy, which you would expect in the real world, as each string has its own unique sonic footprint. (So it's like Verituner, in that respect). But Verituner (which I own) does continuous updates/refinements of the stretch, if you choose to allow it. This program admits only 2 recordings of the note.

Anyone heard about it, or have experience with it?

No, but playing every note would seem pretty slow, right off the bat...TLab measures all the C's or whatever you decide, and calculates quickly. I'd have to be sold on inputing ALL 88 notes...especially since much of what I personally do is fly the last top and bottom octaves by ear anyway...

Thanks for the heads up, though...

RPD

RPD, you could do every note in Tunelab...if you really wanted to

But the information would not be used as it just fits a smooth curve through your data.

This Dirk's program is interesting as it claims to use all partial of all notes to compute a "best tuning". Unfortunately Dirk doesn't specify what he considers a "best tuning" to be. And it's not really practical to play every note twice, then wait a while while the program analyzes things, then start tuning.

Not being a practical person I'd like to try it out but unfortunately the "demo version" has almost no functionality and I'm not about to pay 268 Euros to find out it does not work.

Kees

I am not a software designer, but I wonder why many programs in the market, have been designed to first calculate a tuning curve, before begining to tune.

If I were to design a piano tuning program, I would try to make calculations on the fly, while tuning, just as an aural tuner does it.

All aural tuners follow a tuning sequence and they don't have to know in advance what the target for a note could be.

I think, in that sense, Onlypure and Verituner are the unique programs that make their calculations this way (on the fly). The user has only one restriction: tune A4 first (and A3 second in the case of Verituner), from there he can tune any note in any order.

In the case of Dirk's program, in the time the user gathers info for the program another tuner can do a pitch raise or a coarse tuning!

Rafael:

The reason is the same that as an aural tuner you may have to backtrack. For example if you tune a nice temperament octave you may discover later on that the piano requires a lot of stretch and your temperament octave need to be stretched a bit too. So you have to go back and make another pass.

In other words, the best tuning for a piano depends on how its complete scale behaves. So to compute the "perfect" tuning for a piano according to some criteria you need to know how it is scaled, and in principle you need to know the partial structure of each note before you can determine that.

Practical ETD's shortcut this procedure, otherwise they would not be practical, and Dirk's program is not practical as you observed correctly. But his principle is correct, and I'm very curious what tuning would result from his approach.

It could be made practical if anyone can figure out a way to play octave clusters (C1-B1 chromatically same time), and decompose the sounds into 12 notes and do a partials analysis on each. So you'd play 8 octave clusters, then it computes the tuning and off you go. Not sure if that is possible.

Cheers,
Kees

The problem I see with this calculated tuning curves, is that the user has always to adjust them in some measure to fit each different piano.

So, why to waist time to estimate a curve that anyway will be modified by the user? Unless the programmer is assuming the tuning will be in fact the better tuning for this piano, if such a thing exists.

IMO, it is a question of personal preferences, so the better tuning does not exist!

The first time I tune a piano I have to trust my Verituner and modify the way it stretches the piano if I don't like the results, just as in aural tuning. I think this is the better approach.

I wouldn't like an unflexible program that imposes a given stretch scheme. For me it has to give me the option to stretch the piano as I want.

The user is supposed to be a competent tuner, not a beginner who doesn't know how to set the ETD. The software is a tool, not the master.

Rafael: I completely agree.

I'm just curious what Dirk did, that's all.

Kees

It may be worth the time to download the demo that just calculates for A's and E's - just to get a preview of the overall stretch. Sounds similar to the onlypure in that there isn't any user input to alter what the machine thinks is "perfect". (we all know how that sometimes works on smaller instruments!)

The needle display scale doesn't look fine enough for fine tuning, but the cents number would be a help.

Somehow it self-calibrates the soundcard?

I didn't see an overpull function in the manual.

Ron Koval

I tried it. The "self calibration" takes 5 minutes. I am a bit worried that the 5 minutes are used to hunt for credit card information on my computer as the software does not work if I don't let it past my firewall.

Apply caution!

Kees

I'll bet he is using Network Time Protocol and accessing a NTP server on the Internet. According to the Wikipedia article on NTP, it is capable of synchronizing clocks to within 1/100 of a second over the Internet. So if he uses a 5 minute sample with a possible error of 1/100 second at the start and at the end of the 5 minute period, then the maximum error is 2/100 of a second. So if he is measuring the sound card sampling rate by counting samples over a 5 minute period, his maximum error would be 1 part in 15000, which works out to about 0.12 cents, which is just about the accuracy he is claiming. Now if he had just run the calibration for an hour instead of 5 minutes, then he could have calibrated to 0.01 cent accuracy.

Hi Rafael,

I sometimes like to go back to the original "One-Touch" idea of the Verituner.

If the piano is very close to 440, I have gotten some nice results starting in Fine Mode simply starting at A4 and then tuning downward chromatically. Then, from A#4 upwards. I have only tried this with the default styles.

Edit: No recalc along the way.

Hi Joe,

Interesting. I always thought it was not possible, because the program needed to know the iH of A4 and A3 for its calculations.

I've tuned alot of pianos, when they are at pitch, starting directly in Fine mode, but I always tune A4 and A3 first, then from the break up to C8 and last from the break down to A0.

I tuned A4 then A3, and A5, A6 A2 A1, which gives a good sample for the Verituner. I start tune at the tenor break up to C8 and the work down highest bass to A0

So, 440 -> 442 Hz is barely detectable... 8 cents?

That is wacky! I think he must have gotten cents and hertz mixed up! Also, his statement appears contradictory!

BTW - here is an interesting test:
http://www.tonometric.com/adaptivepitch/
This will test your accuracy of pitch discrimination. I took it twice and did better the second time when I realized you could replay the pitches more than once. Of course, the statistics could be a little flawed if certain people decided to use an electronic device to "cheat" on the test. According to the test own ability seems to be around .6 Hz. My 8-year-old got 1.35 Hz.

Pass at 0.75Hz
Fail at 0.375

I was passing at 0.75Hz almost all of the time and at 0.375Hz about 2 out of 3 times

Paul.

Pass 0.75, fail 0.375.

Got a score of 0.45 Hz. I don't understand the Pass/Fail scores that were posted.