Official Luthiers Forum!

My experience is with steel-stringed instruments, but I'm just completing my first classical and in the process of setting it up. On SS guitars, I use 0.5 mm uniform nut compensation, a 4.0 mm saddle, and use the old bent-wire trick to move the saddle break point back and forth to determine the best intonation point. File them into the saddle and job done. I've never been called out on intonation -- although I fully expect it will happen with some picky customer.

On this new classical, I've also used 0.5 mm uniform nut compensation and the strings currently break at the rear of the 4.0 uncompensated saddle. The scale length is 660 mm. By my testing, I need to shorten various current string lengths by 3 to 9 cents.

Using Trevor Gore's eqn. 4.7-1, it seems that 1 cent is about 0.38 mm, call it 3 cents per 1 mm, so I need to pull the saddle and move the break points accordingly by 1 to 3 mm. Yes?

Is that a valid approach? I know it's not the full-meal individual-string deal, as described by "The Book", but like the age-old way of doing SS guitars, it makes sense to me.

Tim, have a read of this and get back if you still have questions.

(...but I have a vague recollection that that page might not load for you in your current location...)

_________________
Trevor Gore, Luthier. Australian hand made acoustic guitars, classical guitars; custom guitar design and build; guitar design instruction.

http://www.goreguitars.com.au

Remember the strings in terms of thickness (mass) are not a 1-6 smooth gradation but more like 1-3 plus 1-3.
Your slighter wider saddle should help, I use around 2.75mm and this crowning approach:

I copied David's system, and it works quite well for me on my guitars. My saddes are only about 2.5 mm though.

_________________
Waddy

Photobucket Build Album Library

Sound Clips of most of my guitars

Well, my brilliant idea didn't pan out so well -- the resulting saddle break points were really no closer to good compensation than the original! However, now I have a test saddle with well defined crowns, so maybe I can get this worked out for the next saddle.

One thing I'm noticing, which is confusing the hell out of me, is that the 12th fret harmonic is NOT exactly one octave higher than the open string -- it's several cents off. How can that be? Something to do with how I'm exciting the string?

Anyhow, I followed Trevor's instructions on page 20-10 of "Build" to locate my bridge with its 4 mm saddle. My nut has 0.5 mm compensation, so my distance from nut face to front face of the saddle was located at 661.5 mm. Job now is to figure out where the crowns should be located for each string. (BTW: the scale is 660, the fingerboard and saddle have a 20" uniform radius and the action is deliberately set "low" at 2.8 to 3.7 mm.)

David: your pictures and that from Waddy are not so clear on my (rather large) screen, but from your description and what I think I see, I gather that bass and treble strings are compensated in groups of 3, with the crown for the highest at the front of the saddle, and that for the lowest at the back of a 2.5 - 2.75 mm saddle. Out of curiosity (and for the sake of completeness), do you have any nut compensation? Can you work out the resulting actual compensated lengths, based on how you locate the saddle itself?

Trevor: yes, you remember correctly that I had trouble accessing some web sites from Kenya, but I've recently solved that: at least my desk top now wears a cloaking device that make its IP appear to be somewhere else. Using that, I can go anywhere, so I'm able to access the page. Gonna have to read that a couple of times!

After another coffee, I'll do some more research, go take some measurements of where my test mule saddle is really at and come up with a plan for my second attempt. Haven't thrown away a saddle for a while, but my first classical saddle will eventually end up in the bin.

No nut compensaton.
For a 650mm scale the front edge of the saddle slot is set at 652mm (or 327 from the 12th fret if you prefer).
Approximate string lengths would be:
High E- 652.25 mm
B- 653.25 mm(+ or-)
G- 654.5 mm

D-652.25 mm
A-653.25 (+ or -)
Low E-654.5 mm

These are rather approximate as I always slightly break the sharp edge on the saddle at the front and back as well as the only dimension that I actually measure is the 652 mm position of the saddle slot front edge.

Just as a comment, It probably will be most productive to concentrate on the tone of the instrument rather than "precise" intonation beyond getting it acceptable in a general sense.
Even if you do get it "perfect" for a set of, say, normal tension
strings, changing to hard or extra hard tension will put you back to square one.....

"12th fret harmonic is NOT exactly one octave higher than the open string -- it's several cents off. How can that be? "

The harmonic is flat, right? My understanding is that the reason for this is that, while an "ideal" string that had no weight or stiffness would have a perfect harmonic, the weight and stiffness makes the harmonic a little "off."

_________________
Tim Allen
"Never hurry, never rest."

Thanks David, exactly the information I was looking for and perfectly clear.


I hear you, and I don't want to get bogged down on this. I'm actually very pleased with the bass performance on this beast. Trebles could be a bit clearer, but I gather that's the usual challenge. The geometry worked out pretty close, which was a concern -- my saddle is about 1 mm lower than I was shooting for, but I do have the action set on the low side. I want to work my way through some of the ideas in Trevor's books, particularly those on body resonance and optimizing intonation across frets. Essentially, I want to gain enough understanding of the peculiarities of classical instruments, before I try to sell one to someone! They really are different from steel-string guitars.

Yep, the harmonic is flat -- enough that my ringing ears can notice it and more so than I've observed on steel string guitars, where I've found they're pretty darned close. Thanks for the explanation.

Lesson learned: don't test intonation against a harmonic on a classical.

Trevor: Have been digging in "The Book" on the general issue of correcting intonation, and in particular section 4.7.2.3 in "Design". The ability to work out the effect of a change in nut and/or saddle compensation on the tuning error across frets made intuitive sense to me, especially if you're starting with an instrument where you can measure existing tuning errors (i.e., set to target action, with a nut and saddle).

So, I built a simple Excel spreadsheet that takes an original tuning error map (Table 4.7-1), and generates a resulting tuning error map after specified nut and saddle compensation changes (Table 4.7-2), using eqn. 4.7-4. As you point out in the text, one can simply play with the compensation changes to minimize the sum of absolute errors across all frets. So, I added that "objective function" to the worksheet (eqn. 4.7-5).

Excel comes with an optimization add-in "Solver", and I assume you have probably been doing optimization with that, targeting the minimization of summed absolute tuning errors, by applying changes to the nut and saddle compensation. Easy enough to implement, but the devil must be in the details, 'cause as a rule I don't generate very good optimized error maps with Solver. Except for the low E string, my "optimized" map is much less good than yours in the book.

Now, I'm not an expert in Solver by any means, so I'm curious, is this the tool you use, and if so, what settings you employ for the optimization?

Tim, opimisation is an extensive topic in its own right, which is why I didn't specifically get into the subject in the book.

In this situation, there are a number of issues, some relating to Excel, some relating to optimisation. First, the way the equations are entered into Excel can affect results due to ill-conditioning, though this doesn't appear to have been a problem with this equation set. Second, regarding optimisation, "simple" optimisation looks for a graphical feature, a singular peak, a singular trough or maybe a point of inflection. You need to know that the expression you have and the data you feed it with produce what you're looking for and are seeking to optimise. For example, if you're looking for a trough (a minimum) you need to be sure that there is only one trough in the zone you're searching, because the function won't know if you've got the right one if there is more than one. So you may need to constrain your search. You may also need to constrain your result. For example, for a certain data set, the optimiser might suggest 12mm of nut compensation and 20mm of saddle compensation. Whilst this may well be the theoretically optimised result with the data and constraints you've set, it is unrealistic from the implementation point of view.

So the things to check are that you have the equations set up correctly and that you aren't suffering from ill-conditioning problems, that you get convergence and that you have sufficient constraints. Usual practice is to start out unconstrained, because, generally speaking, the unconstrained result may be non-intuitive but also very useful due to the insight that results. Then set constraints until you have a result that is both useful (i.e. significantly better than without optimisation) and implementable.

Now, back to normal programming...

_________________
Trevor Gore, Luthier. Australian hand made acoustic guitars, classical guitars; custom guitar design and build; guitar design instruction.

http://www.goreguitars.com.au

Trevor:
Interesting timing, as I was just about to post that I figured out why my optimization results were so miserable. I had inadvertently checked the box in Solver that constrained the compensation changes to non-negative values. Stupid error on my part! Once I unchecked the box, I was generating good results with GRG Nonlinear optimization. (No issues with rounding or "conditioning" of Excel values, as I did all necessary calculations in a single cell, rather than stepwise as you develop them in the text.)

In fact, it becomes easy to see how you sometimes accepted the Solver result as given, or reoptimized after "constraining" one of the compensation changes, either because of physical limitations to implementation, as you describe, or perhaps to shift the errors to a certain area of the finger board.

Now that I understand all that, my spreadsheet is actually pretty easy and flexible. And using it helps to understand better the limitations of any intonation approach as it affects tuning across all frets.

Now, on a more important issue: did you watch the AB-Wallabies 12-12 draw on Saturday? Good come-back by the Wallabies after a ragged first half. Still, it never seems like a proper rugby game without someone scoring a try. I know that 19 games unbeaten is not the same as 19 consecutive wins, but it's still pretty good! Wait for Eden Park!

Good to hear you got it sorted, Tim!

When I were a lad, I played a lot of RU in the UK and even though I lived pretty much in RL heartland (Lancashire) I struggled to see the point of RL, which seemed more akin to ploughing a muddy field. 45 years on, the boot is on the other foot. RL has changed into a really entertaining game (at least in Aus), whereas RU seems perpetually bogged down in technicalities.

I think the last decent game of RU I saw was the world cup final in 2003, even though Eng. beat Aus!

And now (hopefully!), back to normal programming!!

_________________
Trevor Gore, Luthier. Australian hand made acoustic guitars, classical guitars; custom guitar design and build; guitar design instruction.

http://www.goreguitars.com.au

Waddy and David, what do you use to cut those size saddle slots (2.5mm and 2.75mm)? Meant to ask this back in August when this was an active discussion and forgot.
Thanks, Wendy

I use a 10" Freud blade on my radial arm saw. It has a 2.5 mm kerf or close to it. A standard blade would give you 1/8" or close to 3 mm.

_________________
Waddy

Photobucket Build Album Library

Sound Clips of most of my guitars

Thanks Waddy. Just went and measured the Freud blade on my table saw and it is 3.24mm. Is yours a Diablo?

Just an aside... David and Waddy, really admire those tie blocks. I have trouble just getting six holes drilled.

I'm using the Tor browser right now and it works quite well for most things.

I've not noticed much of a difference in intonation when compensating the saddle over such a small thickness. In fact for the last several Classical guitars I've built I've build in the angle to the saddle much like, though to a lesser extent, a steel string guitar such that the low strings are longer then the high strings and I STILL don't have a better intonated guitar then a regular old straight one. I started doing that after reading Jose Ramirez's book but I think I will go back to a regular straight one since I can't really tell a difference, IOW both guitars are equally poorly intonated

I wonder how much sense it would make to do a split saddle. In other words 2 angled saddle slots, 1 for the treble strings and 1 for the base.

Heresy! The sky will fall on our heads!
Actually, I have seen CG with individual saddles for each string.
Don't see any problem with that.

_________________
The name catgut is confusing. There are two explanations for the mix up.

Catgut is an abbreviation of the word cattle gut. Gut strings are made from sheep or goat intestines, in the past even from horse, mule or donkey intestines.

Otherwise it could be from the word kitgut or kitstring. Kit meant fiddle, not kitten.

CallyRox wrote:"I wonder how much sense it would make to do a split saddle. In other words 2 angled saddle slots, 1 for the treble strings and 1 for the base."

Careful, there are classical builders who might want to toss you in a pond to see if you float or sink <Grin>
If you can machine this effectively I don't see why it would not work, however, filing the the double angled crowns takes me all of about five minutes.

Since I'm in the process of building my last guitar and it's for me, I'm not too worried about following any rules but I still think it would be more trouble than filing the saddle, like you said David.

Last Guitar! You're not going to leave us are you!

_________________
The name catgut is confusing. There are two explanations for the mix up.

Catgut is an abbreviation of the word cattle gut. Gut strings are made from sheep or goat intestines, in the past even from horse, mule or donkey intestines.

Otherwise it could be from the word kitgut or kitstring. Kit meant fiddle, not kitten.

There seems to be some pretty wide variations in compensation approaches, string lengths etc. in this thread. Here are the compensation measurements I got from Greg Byers. They are almost identical to measurements I once saw on Paul Jacobson's web site. I've been using them for about 12 years on more than 70 guitars with no problems or complaints. I've also retro-fitted about a dozen other guitars, some factory made and some luthier made and all of the clients have been very satisfied with the results. The set back and set forth measurements are for a 650 mm scale. The physical saddle is set back 1mm during construction and the nut is set forward by 1mm by shortening the fret board at the nut end. The saddle and nut are then compensated by the adjustment measurements.

Michael, thanks for posting that.
What kind/tension of strings? Or are these "average" measurements used to be close enough for different makes/tensions.
I just wondered, because especially with classicals the intonation on the G string tends to jump out at me, and changes with the tension of the strings.

_________________
The name catgut is confusing. There are two explanations for the mix up.

Catgut is an abbreviation of the word cattle gut. Gut strings are made from sheep or goat intestines, in the past even from horse, mule or donkey intestines.

Otherwise it could be from the word kitgut or kitstring. Kit meant fiddle, not kitten.