Hi all - I recently had the opportunity to work with a tech to help with "all things piano". Since they used Tunelab, I brushed-up on my understanding; I just have a legacy copy for my pocket pc, so it's not quite the current iphone version this tech uses.

Thought I'd do a little write-up and add to this as I think about specifics. Many of these tips/techniques applies cross platform when working with other ETDs.

Hopefully, you've had the experience of finishing a tuning and thinking - "wow, that really turned out well!" It's easy to think it was just luck, or a decent piano, or the phase of the moon... But I'd like to help you have that happen more often!

Basic rules:

1. Know your machine
2. Help your machine
3. Challenge your machine
4. Trust your musical ear (unisons/octaves)

1. Know your machine. Become familiar with the ins and outs of file management, the process of measuring and tuning, alternate temperaments, pitch raising and any specifics to that ETD. If you are going to spend any extra time making the tuning match the piano better, it's a good idea to save it for next time!

2. Help your machine. What does it need from you? How can you help it during the measuring process? How can you help it by manipulating the curve?

3. Challenge your machine. I know it is tempting to just turn it on, follow the directions and start cranking pins, but there are usually some advanced techniques that push the capability of the machine...

4. Trust your musical ear. (unisons/octaves) Hopefully you are already tuning unisons by ear. Do you sweep back and forth across "in tune" to combine the hand and ear to find the best spot? You can use that same technique to accurately find a clean and clear octave.

More to come...

Ron Koval

I am looking forward to more of your posts. Thank you.

Ah, I am impatient! What follows?

I am a Verituner user, (I don't have but the free version of Tunelab) I think VT is more flexible or adaptable to each piano and thus needs more of the human tuner to exactly match the piano.

I guess what you'll say about Tunelab will have a lot to do with VT.

Sample the piano. The technician I worked with felt there was some improvement to sample CEG's in the bass and then C's and G's above that. Setting up personal experiment and checking results should guide your note choices.

This process is the only input the ETD gets directly from the piano - make it count! Single strings only, a nice medium strength tone while Tunelab samples.

After the first note is your first chance to help the machine. Know that screen that you just tap save and go on to the next note? What were all those numbers telling you????

Look at what partials have numbers - that gives you an indication of the quality of the measuring. Try to get the most partials filled with a number. Over on the right is a single number generated, the inharmonicity constant. This is the magic number that will be used to generate the graph. How can you know if that data is actually good?

Dump the first measuring and measure again - after memorizing the iH constant. Does the machine give you the same number again? It's probably a good reflection of that string.. If the numbers differ, do it again until you can generate a consistent response to that string.

Garbage in, garbage out - help Tunelab get quality data!

Ron Koval

Being a long time TuneLab user I like what you've written Ron. Your four basic rules are right on. I follow them all the time.

Also Ron you can measure the same note again and it will compare the two for you and you can dump one and sample again. If you use several samples TuneLab will take the average of all of them.

Ok, you've got good data in the machine... or do you?

My version is old, so I'm not sure how much effect each of these situations have - it would be good for you Tunelab folks to test and report back if anything changes the iH constant. Scott, good to know on the sampling/dumping. Not sure if I'd trust an average, unless all the numbers are pretty close to each other.

I saw that Robert Scott has made some changes since my version was in use...

Part of the "know your machine" rule should probably include re-reading the manual every so often to see what you've missed, or makes more sense now.

What about false strings, have you chosen the clearest sounding string?

Is the same constant generated for each sample when the piano is 50 cents flat? 30 cents? 20 cents? Sometimes pitch raising with a saved file, or the Average tuning file might be your best first step...

Have you made sure the mic level is set properly so there isn't any clipping of the signal?

Any benefit to using an external mic?

Does it matter where your machine/mic is located in relation to the piano/area that you are sampling? Is closer better, or is down on the floor a better choice?

Plucking a string sometimes allows for measuring more partials with a clearer signal - I'm not sure of the quality of that data though when it comes to generating a curve when the piano is to be played by the hammers hitting the string.

What about noise in the room? How about fans in the area?

These are all areas for investigation - let your fellow Tunelab users know about any specific findings!

Ron Koval

Yes, if the samples are too different it may be time to sample a different note close to it instead of letting TuneLab take an average.

Hmmm...Sampling from the floor. I'll have to try that.

You will get a different set of partial readings when you pluck the string. The hammer's point of impact and relative softness emphasize certain partials, others less, as a general rule. The readings/partials are generally a tick sharper when plucked; and that 2-3 cents is enormous to the final result!

I will play with the measurement notes on every piano. C-C is the default, but I will add a note or three near the end of the tenor, start of the bass, and at C5-E5 where it crosses a plate strut, that sort of thing. I try to give the program the notes it needs to run through to get a good sound.

Looking at the last small piano; I added A2-3-4-5 and F2-3-4, for instance.

I try to avoid giving the program the last plain wire (going down), as that string is often completely different from the rest of the scale in the area. So...I'm trying to give the program notes (useful) that will allow it to create a smooth curve for the tuning.

But....I'm am always playing octaves and chords as I tune along, with double octaves and fifths along the way. The ear is the final arbiter, no matter what the program wants to do!

I also tune with Accutuner III and (gasp!) teach aural tuning!

Smiling,
I am,
Tuning for a living!
Respectfully,

Thanks Ron, good tips.

I usually have Tunelab set up to take CEG for octaves 1-6, averaging two samples for each note.

But making sure the iH constant is roughly the same for the two measurements is a good idea.

I'm usually doing 6:3 in the bass, and 4:2 in the treble.

I usually tune unisons by ear, but sometimes, I find it's easier to just tune each string with the machine.

0.02

Ron,

It was an honor to work with you last week!! I had five pianos to tune today. On the last piano (which was a Kimball La Petite) I tried the method you taught me.

Here's the process I used:

The piano was very close to pitch to start off with so I felt that overpull was not really necessary. I used the following notes to take samples (C1, F1, C2, F2, C3, F3, C4, F4, C5, F5, C6 & F6. (I previously had used all of the Cs, Es, & Gs.) I emailed Robert Scott to see if actually taking more samples would give Tunelab more information to create a better curve. He said that yes the newest version did and that when he has the time he likes to measure the C's & Fs. mentioned above. Next, I layered the "Koval Equal Well" offsets on top of the Tunlab created tuning curve. I muted off all of the Cs to one string. I set all of the C's to where Tunelab wanted them. I then tweaked each one to get the cleanest octaves possible. As I got each C to where it sounded best to my ear, I clicked on the locking mode. I then transferred that custom offset to the tuning curve, which then I could see as red dots plotted on the tuning curve. Next I toggled through different octave types trying to get the tuning curve to hit as close to possible on those red dots. I ended up with 6:3 bass and 8:2 treble. Next I changed to manual adjustment mode and tweaked the curve a little more to get even closer to the read dots. (I was a little concerned because my ear set the bass about 20 cents lower than what Tunelab wanted, but I went ahead and tuned it anyway. I also set the "auto partial" feature to go up to the highest wound string.

I have tuned this piano for about three years now and I do believe this is the best it has ever sounded. I'm excited to use this method on more pianos.

I was never trained to tune aurally, but I am a classically trained musician. I am learning that I CAN trust my musical ears to tweak and clean up octaves.

Thanks again Ron! I appreciate all of your help and knowledge!!!

Hi Ryan! A La Petite, those are tough for just about everyone! Glad to see you moving ahead...

When learning/developing/implementing any hybrid tuning technique, there is a balance each tech finds between spending extra time and how much improvement there is to the tuning.

Ryan jumped ahead to the next step - You should have decent data for Tunelab to calculate. It is set up to automatically use that data to effect the curve. You can bring up the curve to see what it looks like - only to realize there is a top and bottom curve! The top one is the main concern at this point - it is a representation of the tuning offsets that will be used for this piano.

Here's where that question always comes up - What intervals are best? Um... all, none, some? Here's the next opportunity to help and to challenge your machine.

Any EDT is set up to use the data collected to find a stretch that hopes to allow all of the same named notes to sound "in tune". I don't even want you to consider temperament at this point. We know A4 will = 440. Tunelab will calculate a location for A3 and A5.

Go ahead and tune one string of each of those. If the piano is way out, you might need to get all three strings close to minimize sound bleeding through the mutes. Those of you not used to trusting your ears, this is gonna be a bit scary...

A4 is our home base. Play it and tap the lock icon. As Tunelab 'finds' the note, the bars will stop moving - tap the stop sign to have it 'hold' that correction factor. Since you already tuned this, it should be almost exactly in tune with Tunelab! It will probably show a very small offset at the top of your screen. (I'm working from memory here...) What ever that number is, drag it down to the note name in the middle of your screen (A4)

If you toggle to the graph, the little tiny bar for A4 should be red, right on the graph.

Next, move to A3 - you already tuned it to the display, so you can always get it back.... Play A3/A4 together. Sound good? Great! Now turn away from the display, move the pitch of A3 very slightly flat and sharp and come back to where you think the A3/A4 is the best.

Look at the display while you play A3. Notate where it is in comparison to the calculation. If you want to convince yourself that your ear can be trusted, repeat the process. I found it very repeatable, especially in the middle range of the piano.

Once you are happy with A3, touch the lock icon, then the stop sign to have Tunelab find where YOU want it to be placed. Drag that offset to the note to place it on the curve.

More to come...

Ron

Simple to see, the next step is to add more A's - Let's move to A5? A4/A5, same process as before, except we now include A3, to make sure that the double octave is just as good as both singles.

Ryan used C's - the choice is yours!

Move in each direction, dragging the offset to the note to lock that note in red on the graph, also checking the double, triple and extended octave combinations across the keyboard to provide that illusion that no matter how many A's you play, they all are the best in tune that they can be.

Can someone post a screenshot of a graph with some red notes off of the curve? This is both "know your machine" and "challenge your machine" to see if you can find a better stretch for this piano.

Why can't Tunelab just "get it right" by itself? There is a general consensus that (for example)a 4:2 match between A3/A4, just a bit wide will make a decent octave for most pianos... As you can now guess, just because it is "decent" for "most" pianos , that is no guarantee that it is "best" for this piano! By using your musical ear skills to help the software, you give it a chance to do a better job.

Next you've got to practice manipulating the curve...

Ron Koval

Yeah, what is right?
Thanks Ron, this is an excellent tutorial for TuneLab.
No input really just good to see a worthwhile discussion
to help all of us maximize this software. I think it has
much more potential then we know.
Good job!

Ron,
Here are a couple of screen shots. The first is the Yamaha U1 that you helped me create. It's kind of hard to see the red custom offsets. They are lined up with the tuning curve so well.


The picture below is the Kimball La Petite I spoke of earlier. As you can see, I could not get the tuning curve to line up as well as the Yamaha U1

Hey Ryan, can you return the curve using the full automatic - using 6:3 4:2 just to show how they started out before changing the curve? Thanks!

Ron

Yamaha U1 (before tweaking tuning curve with 6:3/4:2)


Yamaha U1 (after tweaking the tuning curve)


Kimball La Petite Grand (before tweaking tuning curve with 6:3/4:2)


Kimball La Petite Grand (after tweaking the tuning curve)

...and here's the Steinway D that we did
Steinway D Tuning curve before tweaking


Steinway D after tweaking the tuning curve

I'm not exactly sure I did the above correctly. It appears that Tunelab is hitting the red custom offsets at exactly the same place on both. I wonder if in "fully automatic" mode Tunelab is making the curve hit those custom offsets on its own?

Ok, these are obviously advanced manual tweaking techniques,
which I am not familiar with as I use the automatic mode.

So I will ask a very general question: I read in the Tunelab
manual that it's ok for the higher octaves to have a higher
deviation, because the partials are not as significant. However,
on especially the Steinway D tweak, you can see the deviation
is higher for the lower octaves as well.

There are also some notes that are completely off the scale.

Also, is there any disadvantage to using C-F for the 6 octaves,
instead of C-E-G? If not, it would save me some time not to have to
measure 6 notes.

Thanks for the discussion.

Hi Paul,
As far as I can tell it is up to you which notes you choose to sample. I don't think there is a right and wrong. I started measuring more bass notes when I started getting a bass that I was not happy with. It seems to me that adding a few more notes for Tunelab to calculate give it more information to create a better curve. Another thing that I have started doing recently is to set the "Auto Partial" function to include the highest wound bass string. With "Auto Partial selection" engaged, (as I understand it) Tunelab will listen for the loudest partial and tune to that one. You can see it change which partial it is reading as you go through the bass. This seems to be a newer feature to Tunelab. I have really been liking the way my bass now comes out. As far as your question about the deviation. I'm still learning that one too. :-) I did start a thread a while back called "Understanding the Deviation Curve in Tunelab." Ron & Kees gave some good explanations.

I'll be back in a day or two. The graphs don't quite look like I remember, so I'll go see if I can play on a couple pianos to generate the graphs to show how to manipulate using manual mode. This is the really important step, so I want to show it clearly!

Tunelab seems to like the lower bass sharper than my ears do. I usually set A o to D1 10 cents flatter than Tunelab wants and get that nice deep bass, especially on 6 foot and smaller instruments. Good info above Ron and Ryan.

Have you tried a different bass interval, like 8:4? That should give you a flatter bass that the 6:3. I was almost going to make 8:4 the default.

Hello Ron,

I have been using Pocket TuneLab to tune my own Knight K10 upright. As a newbie (home) tuner I have been using the auto setting and my stretch settings have just used the user manual recommended choice of five Cs.

This so far meets my requirement although many times I wished that an alternative version of the manual had been written.

My question today is regarding unison tuning. I have been using Tunelab to tune separately all three strings. In other words not tuning the center then aurally tuning the left and right pairs. I am very careful to tune to see the opposing bars meeting in the middle but I'm wondering if Tunelab is not sufficiently accurate (repeatable) to make all three strings beatless when all are played? I note here the mention of mute bleed-through which may affect my tuning method.

All comments welcome,

Ian

Sometimes I let Tunelab tune the unisons like you have
described.

I'm not totally sure of the reason, but sometimes it's easier to do it that way. It's not that I cannot tune unisons by ear: I most certainly can, and it's usually not difficult.
But sometimes there will be a weird note, maybe the fundamental or partials are drifting in a strange way after the initial hammer blow, where it's cleaner to just let Tunelab do it.

Then when I check all three strings, it's certainly beatless, so you certainly can do it this way, although the aural method is faster, as there is less muting involved.

My 0.02

I have tried 8-4 in the bass, and it is deeper, but sometimes a bit much. It seems some pianos need a middle ground somewhere.

This mirrors my experience with many different ETDs - Although I seem to have come to a different conclusion... while the deviation graph can give a lot of information, I find that using the "musical ear" to set some octaves gives superior results. It seems that setting any single interval pure through different areas of the piano is repeatable, and precise... but not often the "right" place to tune the note for the best match to the piano.

Ron Koval

Dear Ron,

Yah, bass tuning with an ETD (any) can be problematical. The bichord strings are often slightly out-of-tune with each other. You can match ONE partial, but the two strings sounding together often create an unpalatable twang.

Choices must be made!

One thing I will suggest. When a bichord simply won't tune well I will mute the left string and favor the right for tuning first. Why? Because when the pedal is used (shift) the action slides right and strikes that string. Tuning the right string first often leads to a better sounding unison sound when the left string is added. Another thought on this is that when tuning by ear, going down into the bass, the right string is tuned FIRST, and the left added as you descend. Older pianos, particularly, seem to sound better with this approach.

I try to believe that the bass must have sounded good to someone on the factory floor, when it was first built, and approaching the bass bichord unisons by tuning the right string and adding the left (tuning up or down) seems to give a better result.

Is it mystical? Is it practical? Jeff not know. Jeff just pawn in game of Life. But....it works for me.

Practically speaking,
I am,
Sipping the Costa Rican,
One sugar,

Ok, I have a working copy that will allow me to experiment and post some screenshots here...

In the meantime, the manual for the iphone version has a good description of adjusting the curve and even a couple of screenshots of adjusting to the red bars in the treble here:

http://www.tunelab-world.com/TuneLab%20Piano%20Tuner%204.0.pdf

Pages 16,17,18 are where to look!

Ron Koval