This is a big topic and I hope to take a stab at some of the basics...

What is hybrid tuning? The partnering of human and machine to achieve higher levels of piano tuning.

In the past, most have recommended that to use a machine, one should already be an accomplished aural tuner. We all know that there are many tuners working that are not confident of their aural skills, so this topic should be of value to many working technicians.

There are listening skills that are needed, but those are the same skills used by musicians in band, orchestra and choir; matching pitches and tuning for "pure" intervals. Those of you from the world of piano performance may not have the same skill-set of listening and adjusting pitches to match another pitch - but it is a skill easier to learn than traditional aural temperament tuning.

It helps in any partnership to determine which partner is in charge of which duties. It is the same with this form of hybrid tuning. The human ear is needed to set up the framework of how the tuning will expand to the rest of the piano. The machine does just fine dividing up that framework - especially above about C4, where the inharmonicity of any piano is usually predictable. There are listening skills that can help with tuning down through the break...

All of the piano tuning gear in wide use needs to get information from the piano to use to calculate a tuning. This is the first step where the human can assist the machine. Those of you that use RCT are familiar with the varience number, because it is built in to the process. It goes something like this - the machine will measure a single string a few times and then compare to see if the measurements are consistent. Tunelab and SAT users can do this as well, jotting down the numbers or just remembering. The goal is to have repeatable numbers for the same string.

As you learn to get repeatable results, it becomes apparent that striking the key the same way while measuring helps this process along...

I'll add more later - feel free to ask questions!

Ron Koval

Unless you define what different ways the break can be tuned and how hybrid tuning can accomplish them, I can see little value for hybrid tuning on the vast majority of pianos.

Uh, haven't finished yet... and it seems that you have removed yourself from the target audience for this series by your choice of tuning methods.

Ron Koval

Actually, I am trying to keep an open mind. Are you trying to shut it?

So on to the Verituner - Filling the "I" icon is the only thing we see to determine the quality of information the machine has gathered from the piano. It follows that it is probably a good thing to fill up the "I" all the way for each note. The re-tune function gives the machine a fresh listen. Moving the unit a bit sometimes helps.

Once you have given the machine the fullest information possible, most platforms have options available for human control. It is tempting at this point to just "calculate" and then tune to the machine - what's wrong with that?

Let's take a little tuning theory detour. When a machine (and almost every modern aural tech) goes to figure out where to put the next note outside of the temperament, there are "checks" used make an exact placement of the note possible. You may run across writings using octave types (2:1, 4:2, 6:3 etc..) or a type of aural test - comparing the beat speeds of two different intervals sharing the common note to be tuned. What both of these methods do is to focus onto a single partial from each note to compare.... What is interesting is that this method of precise placement has nothing to do with the musical sense of "better" or "worse".

Think about how you may already set the second string of a unison to the first. You move the string from where ever it is to cross the line of in tune - too far the other way. Then you bring it back until you find that place that is "better". This is the musical hearing that we use in band, orchestra or choral work. There isn't any need to worry about partial matches, or theory or anything other than "better" or "worse".

wawa-wa--wa----wa------wa---------wa----wa---wa--wa-wawawa

We find that "better" place in the middle just fine without checks, because that's what humans do well.

Back to your machine. If you just calculate a tuning now, the machine will set up the octaves from end to end of the piano based on partial matches that generally work pretty well for most pianos... without any ability to HEAR where those octaves lie in the "better" or "worse" range.

Next - helping the machine part 1

Ron Koval

Thank you Mr Koval, I shall be following this thread with a lot of interest. I am sure it will turn into a very valuable thread.

Ron and list-

I will be teaching a " Fusion Tuning " class and my unison class in Bellvue next summer. It is a big untapped topic and I look forward to reading your take on it.

Best-

David Brown

Ok, here is where the paths diverge a bit...

The major tuning platforms allow for some user control. This is where you have to do your homework and get comfortable with the process.

RCT - custom Equalizer
Tunelab - adjusting the template tuning curve
SAT - double octave beat control? (I'm not really familiar with this platform)
Verituner - Custom styles

Here's a little tuning theory wonkery... if your eyes glaze over, skip to the next paragraph! Ok, you've no doubt heard about inharmonicity - the (kindof) predictable factor that makes matching a tuning to each piano such fun. Any machine that takes some measurements from a piano and then bases a tuning on those measurements makes assumptions about the unmeasured notes. If the inharmonicity follows "the rules", the machine has a good chance of coming up with a tuning that matches how aural techs approach the piano. As you may guess, the break from the plain wire to the wound strings, and about an octave above presents some challenges to the assumptions the machines make. The next assumption most machines make (and a large percentage of aural tuners)is that a smooth curve controlling one partial will result in a musical tuning....

So here is our challenge. You know how to bring strings of a unison together to find the most "peaceful" spot. Next, you have to apply that same principle to the octave. One simple way is to start with what the machine wants, then try not looking at the display and retuning the octave and see if you find a "better" location for that octave. Do it a couple of times, using the machine as a bookmark and see if you come up with the same location for a "better" octave.

Interestingly, there is a little more wiggle room in the octave width than a unison. We're gonna use that to our advantage....

Ron Koval

Personally I don't have time to play around with electronic tuners. I carried around a Sanderson Accutuner for two years, but talk about eyes glazing over. I was looking for something to do. They do give give a consistent pitch and it's always good to know how far from A440 the piano has drifted. Doing a tuning with them .......ZZZZZZZZZZZzzzzzzzzzzzzzzzzzzzzzzzzzz.

Ok fine, if you are not into electronic devices, please leave this thread alone.

I am interested, and do not wish this thread to turn into into etd vs aural.

Thanks

With tunelab you have to design your tuning curve. Despite (or perhaps because of) extensive documentation on how to do that (which is not trivial to understand) I have the feeling 90% of the users just use the default setting of 6:3 in the bass and 4:1 in the treble.

A method I like to design your tuning curve is to tune a few notes aurally all over the keyboard. For example all C's and all G's. Once you're happy with the stretch on those notes, you start tweaking the tunelab tuning curve to match what you just tuned aurally as close as possible.
Once that's done you can start listening to music on your earphones and tune by machine (except the unisons of course).

Kees

Ron,

Interesting thread... Please continue. When I do use RCT, many times the final outcome is not favorable. It seems like no matter what I try, it doesn't seem to change anything so, I complete it by ear. I'm using it less and less because of that waste of additional time for me. It's probably how or what I'm doing but, it sure is maddening. Maybe Ron or someone understands what I'm talking about.


That stuff I just don't understand at all. Simplify it for dummies like me...

The DOB stretches out (or narrows) any tuning that is programmed into the device from the note A4. It is most often used "on the fly" to boost or rein in the high treble or to stretch out or narrow the low bass.

For example, upon reaching F6 or thereabout, one can check back to the temperament octave to see what F3 and A#3 will read when F6 is in the window. If the programed tuning is too flat, the DOB can be deployed to change the program at that point to make it match what is wanted. This can be done note by note or again at C7, then F7 to build a sharper curve.

If one does not want as sharp a treble as the program offers, the reverse can be done. For example, if one wants 2:1 octaves in the 7th octave, at C7, play the previously tuned C6 and then put a negative number in the DOB until the program changes to the point where C6 played while reading C7 stops the pattern.

Similarly, if the first octave in the bass is too sharp, adjust the DOB until it tunes the low bass sufficiently flat. That can be done by playing the appropriate notes which have coincident partials with the low note being tuned and adjusting the program or by tuning a sample note aurally and then adjusting the DOB until the program adjusts itself to the amount of stretch desired.

Recently, Jeff Hickey wrote about how he accomplished the same thing by applying an offset to the program. Upon reading that, I gathered that he must be using an SAT I or II which do not have the DOB function. Only the SAT III and IV have it.

Chris Solliday RPT never uses the FAC calculation. He uses a blank program (all notes reading 0.0) but tunes some octaves aurally, then adjusts the DOB until the zero program matches his aurally tuned octaves. It is actually quicker than programming a tuning.

If one wanted to tune using non-equal temperament correction figures, have those figures programmed into a Temperament page. Tune A4 at 0.0 read on the fundamental. Tune A3 and A2 as aural octaves. Adjust the DOB until the lights stop for A3 and A3. Then enter the non-equal temperament by selecting the page. That page will then alter the DOB generated program to the desired non-equal temperament.

A DOB generated program can also be altered as desired in the high treble and low bass. This would be an example of hybrid tuning. The technician as determined how much stretch is desired by ear and then creates a program that reflects what has been determined aurally.

Ron:

Please do not think I am bashing ETDs. However, I have some valid questions about what you are writing.


I will just have to wait for you to explain how hybrid tuning handles this challenge before I can ask questions about it.


Now wait a second! How can an aural tuner use only one partial? Every interval (including unisons) use partial matches from more than one partial. Not to mention that every note can and should be checked with more than one interval.

Jeff, from what you write, you are far advanced of the majority of aural tuners in theory, math and checks.

All to often, (in classes even) I'll see a tech tune and check by making sure the thirds are pleasingly progressing... That rapidly beating sound is a simple reference of a single partial from each note being played. Historically, this is in response to the machine tuning - aural techs came up with methods to get the thirds just as evenly progressive as the early machines were able to do in the temperament. (Reading some of the early articles in the Journal on CD was educational.) Too bad that as random inharmonicity pushes single partials off from the rest, the whole tuning gets skewed by this tuning preference.

Ron Koval

Ron:

Thank you for the complement.

I am even more confused. There is a difference between ”… a smooth curve controlling one partial will result in a musical tuning.” and ”… a single partial from each note being played.”

To me, these are very different things. So I am not sure what you are talking about. I do not see a relationship between an ETD using a single partial, and an aural tuner using a single partial match (which is two different partials).

And let me add that tuning ET would be a breeze on any piano if all that was required was progressive M3s!

Oh sorry! I was trying to express controlling the placement of a note based on a single partial from that note. Like an ETD only listening to a single partial to place the note, when aural checks only reference a single partial for the to-be-tuned note...

Again, there are plenty of aural tuners that tune way above the lower rungs of aural understanding...

Ron Koval

Bill, thanks for the information about the SAT platform... I see that it IS flexible enough for a fusion (like that term!) or hybrid technique.


Ron Koval

On to the octave...

As you may know, the beating we hear between two notes can be traced to specific matches of partial at the same pitch level. Because of inharmonicity, it has been shown that there is NO such thing as a beat-less octave tuning. 2:1 4:2 6:3 8:4 12:6 are common symbols used to represent the partial number from the lower note that matches the partial number from the upper note. That's what was referred to above when the comment was made about the Tunelab tuning curve.

Here's where it starts getting a little more interesting. Traditionally, we are trained to focus on specific partials. Some tuners even focus on specific partial matches at the unison level. This particular method of hybrid tuning does just the opposite; we're going to listen to the entire blended mess and let the human wet-ware decide between "better" and "worse". Consider the following line as the representation of an octave while one note is moved from flat through "in tune" to sharp.

"way worse"_6_5_4_3_2_1_"better"_1_2_3_4_5_6_"way worse"

Here's a drill:

Fire up your handy ETD, grab some mutes and a tuning lever and go find a piano. Let's pick A4 and A3 to start. (more about this later) Tune one string of A4 to the machine. Pick a clean string and if there is bleed-through tune the other strings to get the cleanest A4 single string you can.

Now tune A3 to your machine - same drill, make it clean with one string. The machine is now your bookmark. Now while one hand plays both A3 and A4 together, move A3 slightly flat to make it "worse", then moving sharp through "better" and sharper to "worse". See if you can find "better" again without looking at the machine. Now play A3 and look at the machine. Where did it end up? Repeat. Did you end up in the same place again?

Did you notice that "better" doesn't have a specific location, but a range? Some pianos may give you a cent or two, while others balance on a knife edge.

This is what performers in band, orchestra and choirs do - almost without thinking to blend with the rest of the ensemble...

Off you go now! I'll post again after you have had a chance to give this a try.

Ron Koval

What causes some pianos to have a wide range for "better" and some to have a knife edge? I have noticed this phenomena without using an ETD. I use a tuning fork and a Peterson Virtual Strobe Tuner to set one note then go aurally from there. I have really noticed this in the mid treble. Some pianos just come out sounding really beautiful up there while others not so much because of this. I can not remember which is better knife edge or a range though...I think the ones with a bit of a range always sound better but I will have to pay more attention. Until now I could not put into words what I was experiencing so I will try to hammer down my perceptions better.

-Daniel

Welcome back - hope you all had success with tuning that first octave. (If you had problems, contact me off list- there are other things to try than what I'm planning on outlining here.)

Daniel, I've never studied the WHY of it, just noticed that different pianos behave differently.

Anyway, back to the piano. Once you've got that A3 "happy" with A4, you may do a little testing with your machine. Any of the machines allow you to control which partial it is listening to - try it to see how the 2nd, 4th, and 6th partial compare with the matches at A4...(that would force a focus on the 2:1 or 4:2 or 6:3 octave) To do that, set the machine to listen to the specific partial, lock the note switcher so it stays on A3 and then play A4. What you will most likely find in the middle of the piano is that "Yup, they all agree" -until you get really picky and notice that they all have a little drift sharp or flat. What does that mean? The "better" placement for A3 isn't an exact match with any of the partial matches with A4.

Hmmm... Tunelab folks, this just gave you a big insight into manipulating the graph. If where you want to end up isn't an exact match with any of the specific 4:2 or 6:3, or 4:1 matches, you can switch back and forth between different views to "preview" how to place the tuning between the different matches.

Verituner folks, there is an interesting thing that your machine can do that hasn't been discussed much.(I don't think...) With the note locked on A3, switch to fine mode. Realize that the machine is now "listening" to partials 1-8 to drive the motion of the spinner and needle. You should be able to see the numbers along the left side, as well as the little triangles that represent "hearing" those partials. Play A3 and you should see those triangles appear for the strongest sounding partials. Now, play A4....

Those tuning targets will now drive the spinner, based on how all of the partial matches of sounding A4 match with the partials of A3. You may even be able to see the spinner start in one direction and then switch directions with the ebb and flow of the sounding partials of A4. Cool!?

Back to a piano!

Next, repeat the tuning process with A5. Now, adding to the excitement, A5 and A4 should be "better", BUT you also want to check A3 with A5 to make sure they are also a "good" match. You will be listening for apparent "stillness". (I'm not gonna use beat-less) Here's where the wiggle room comes in - A4 is locked in stone at 440hz, but BOTH A3 and A5 can be moved/nudged/wiggled to find the "best" singles and double octave possible. The goal is to make the machine match those three notes.


Time for an aside: This goes waaaay faster than writing about it or trying it the first time! You can see that the goal is to set up a framework for the machine to then slice and dice the octave up into the appropriate atonal or tonal temperament based on your choice and input. That way, the machine has a fighting chance in coming up with an appropriate stretch for the piano. (Later on, I'll talk about over-riding the machine above the break if needed) The temptation is to only do this for the "best" instruments, but I find the biggest payoff in enjoyment and musicality comes from doing this on the smaller instruments. I still get in and out in about an hour. As you learn more about both your machine and how specific pianos react, the process becomes quicker. If you save and re-use tuning files, this only needs to be done at the first tuning.

Ron Koval

Ron:

I, and maybe others, might follow what you are driving at if you could outline the things that you think ETDs do better than ears and visa versa. This is tough reading for me.

My own solution to the "octave window" is to just use other intervals to tune with.

Ron, thanks for your input on this subject. I totally understand what you are doing here and have used a very similar process for several years now when doing hybrid tunings. When doing aural tests on intervals and octaves I often find that the best settling point "range" corresponds to the relative amplitude of the specific partial you end up trying to clean up...its amplitude in comparison to other possible target partials that are beating. It is only natural that when aurally tuning we often hear a predominant partial that is louder than the others and try to get this to sound the best. Although there is a mathamatical relationship that dictates a smaller string segment (higher partial)will produce lower amplitude, spectrum analysis/Pianalyzer will often show anomalies in this being a straight vector line, especially in the lower tenor/bass.

In the more complex mode of partial target choice in RCT one can easily pick better target partials on the expanded tuning based on this amplitude comparison and use the EQ sliders to to blend the shift. It more closely aproximates what we do aurally then any other method I know of.

When using RCT in the initial sampling mode I also take the precaution to pretune the sampled A's to a known template that aproximates the size or type of piano. (If the piano is off by more than 10-20 cents this is recommended anyways). I will then clean up the sample note unisons by ear and determine which string(s) will be excluded/muted for the best sample to take place. No sense in spending time analyzing octave relationshps ect..if the string you target is the oddball one that exhibits less than ideal readings. Even aural tuners run the risk that a strip muted piano of lesser quality forces you to assume the center/L/R string is the ideal choice for tempering or basing a tuning foundation for the rest of the piano to follow. When cleaning up the unisons you can run the risk with trichords that your initial note will be trumped by a 2:1 ratio of better matched strings. If the differences are small, coupling will mask it. On some lesser quality pianos the differences can often not be this small.

Jeff - It's not so much "ears vs. machine", though we can go there with another thread if you like...

What we'll often discover in the next few steps - Pre-tuning A1,A2,A6 and A7 is that the common heard complaint of having to let the octaves beat to make the doubles and triples sound clean can be corrected by setting the overall structure correctly FIRST before dealing with any temperament octave issues.

Ron Koval

Ugh! I would think the idea of a hybrid anything is to try to get the best of two different things. Perhaps I should just forget this Topic.

Wait! It IS what each does best/easiest... let the calculator calculate and let the ear give the calculator endpoints from which to divide...

Ron Koval

Ok, now you know where I'm going....

Go ahead and see if you can make a ladder of A's from A1(iffy on small instruments) all the way up to A7. The target is to have all of the intervals "happy". There should be a nice "ring" when you play from bottom to top. Even A1 and A7 together should sound "still".

Remember, no cheating - consider this a piano with only A's on it, no other notes....

While you go do that, I'll just ramble on here a bit.

I live in the world of the unison. While other tuners may call machine tuning so boring, I find that there is a world to be controlled trying to just listen for "stillness" and tone. Notice what we're trying to do here - tune and manipulate using intervals that are "still" or have a very slow beat. All of those "vibrato" intervals are going to be handled by our machines - they add expression to the music, but only after the overall stretch allows the instrument to "ring". In an atonal tuning, the vibrato rates range from really slow to really fast, but are stacked in a chromatic order. In tonal tunings, the vibrato rates range from really slow to really fast, but are stacked based on the tonality of Western music. Sometimes the vibrato rates of multiple intervals are exactly the same in tonal tunings...

Are you back already? Assuming you have a nice ladder of A's, figure out how to make the calculation match those A's... It's different for every platform, and sometimes, depending on the instrument you might need to be creative to force the calculation around. (Changing the iH numbers in Tunelab, for example allows for a narrower central octave...)

Go ahead and tune the whole piano and let me know if you hear a musical difference when you play. Is it huge? Probably not, but for me it makes a world of difference!

Next - no, I didn't forget about the break! Bringing in other intervals...

Ron Koval

Hmmm, I realize you are writing for 8th graders, but if you mess with the IH constants in the tunelab file all the nice interval beat/deviation monitoring that tunelab provides you with as tools to design your tuning curve will be unusable. I also doubt an 8th grader will be able to effectively edit the IH constants in the file to achieve some goal. I don't think this the way to go.

If you really need more than the 4 parameters that tunelab offers to tweak the tuning curve you should instead use custom offsets.

Kees

Basically: measure the frequencies of specific partials.

They are also good at figuring out how to use this information, but probably not good as a top notch human (yet).

Kees

Well, as far as I know. tuning wise, I could very well be an eighth grader.

I learnt tuning aurally, obviously not understanding what I was doing.

ETDs came along and helped me better understand what I was doing.

So, if I understand the point of this thread, it is basically about "teaching" your ETD to do a better job.

I am very much still interested, but do agree that tunelab calculates a nice tuning all by itself. Without needing my help.

I have been experimenting with Mr. Koval's suggestions, have now fallen back in love with aural tuning and have now started to agree with other posters that tuning aurally is faster and more efficient.

BUT, I still want to know how to teach my machine.

It is a very valuable tool, for pitch, pitch raises, noisy environments etc...

Please go on for this eight grader Mr. Koval!!!

Thanks