Monday, August 13, 2018

Update and recap

My main research is done!   It’s time to head back into the workshop and put the results to work making instruments.  

I started asking ‘can you make and design violins today using entirely methods from Old Cremona?’ and ‘how would you?’ back around 2009.  I started by exploring what others had already figured out.  I found that many people – particularly Sacconi, Hargrave, Denis, and Coates – had already paved the way, clarifying many points.  But still there were large gaps in the knowledge you would need to build this way. My questions pulled me into independent research about eight years ago. My explorations kept expanding, taking more and more of my time.  In the last three years, I laid aside making to focus my efforts on the research.  I decided not to make again until my research results essentially gave a complete picture of classical making design methods in every instrument feature.

Now that I’m closing out the research and heading on into making, some sort of recap to summarize and update seems appropriate for this blog. 

Not to say more research isn’t possible.  Far from it!  But a functionally complete enough picture to make in a historically consistent way is now in hand.  Perhaps others will pickup and continue the work?  My drive has been ‘to make in the old ways’, not to be a researcher.  The research was only a necessity along the way to my larger goals in making.   But the research reveals a way of looking at Old Italian Instrument making.  And it strongly suggests that the design processes and specific methods seen in Cremona violin family making are part of broader and older traditions of Italian instrument making.  A new kind of door is opened for exploring the broader history of instrument making.  And there is some suggestion, I believe, that these instrument making methods also link to artisan and woodworking methods more generally at that time.  Exploring the range, continuity, and diversity of such methods could be a powerful and interesting avenue for an historian of the arts.  As to violin making and my own research, time will surely bring further detail and clarity, and undoubtedly some corrections. But for me, I’m ready now to explore these old methods directly through making.

So what is this ‘functional picture of Old Cremona making’?

There are many parts to the picture; some found by my research, many found my others.  I’ll try to summarize in an accurate, but hopefully brief and not too arcane way.

Old Cremona Making:

  • Continuous with the broader arts and instrument making of the region
    • Methods, tools, and materials of a particular craft adapted as needed from the more general and shared artisan practices
      • The glues of the violin maker are just a subset of glues used by painters, woodworkers, guilders, boat builders, book binders, etc
      • Same with woods, pigments, design principles, basic tools, varnish ingredients, etc
  • Highly traditional and communal in all ways, including materials sourcing, and methods of designing and building
    • The workers within a craft follow general shared practices.
      • Thus making within Brescia has some commonality with Venice and Cremona
    • Within a region, town, or family the traditions are even more specific
      • But making within one town, or one family is even more continuous
  • Development more by evolution within tradition, rather than ‘innovation’
    • There is a combination of always being both ‘traditional’ and ‘exploring’.
      • Each feature is worked by traditional methods. But the traditional always allows a limited range of choices in the specific application of the method.
    • This combination of ‘exploring within tradition’ leads to a slow mutation and development of methods and results.
    • The choice in a build are always mostly repeats of those made in recent instruments, with only some exploration of different  choices.
    • This allows preference of repetition for choices suspected of contributing to good results in recent work. 
  • Process driven
    • You ‘prepare this’ by ‘doing that’.
    • Design is also process driven.  You ‘make this shape’ by ‘doing that’. Etc.
  • Design is more a recipe for work, rather than worked in complete detail ahead of time
    • Design results from following ‘traditional process’ during work, rather than as an ‘a priori’ creation that proceeds and then governs the work.
  • Sizing by relationship to other features is highly favored over sizing by specific measurement.
    • Today we have standards of measure to refer to.  Such standards were inconsistent or lacking in past eras.  In the past, sizing by relationship or ratios was both the tradition, and simply more practical.
      • Today, a recipe might say take 2 tablespoons of this, or 3 grams of that.  And these are standardized and practical for us.  But a renaissance or baroque era recipe will say instead ‘take a part of this and two parts of that’.
        • You do also see rough specific references. Like:
          • A ‘bean’ of
          • A ‘fingers width’
        • But these weren’t really standardized.
        • When there attempts at standards, they were given by relation to an accessible reference.  Like the ‘oncia’ carved into a public pedestal in Cremona.
        • Standards as they were often varied by region and sometimes by guild.
  •  Design shapes are governed by a combination of ‘geometry of circles and lines’, ‘simple ratios of integers’, and the boat builder and woodworker’s concept of a smooth or yar ‘faired curve’.
    • Each feature has specific traditional geometry constructions that either directly give the shape, or that govern the shape.
      • These constructions are highly consistent across generations of Cremona work
      • The Cremona constructions tend to be adaptations of constructions seen more broadly in Italian instrument making traditions
    • Each feature and geometry construction in Cremona work has traditional limited ranges of ratio choices governing it’s application
      • We see the Cremona makers freely vary their ratio choices at liberty, but only within the traditional range of choices for that feature
      • Some features present a natural ambiguity of exactly were you would measure from.   Cremona makers make these choices fairly freely.  Giving them another liberty
        • Example would be measuring from the edge or from the purfling. Etc.
      • Many of the traditional ratio ranges can be summarized in a simple rule of thumb using the concept of ‘a part’, as in ½, 1/3, ¼ etc.
        • As example, the traditional ratios for Body Length versus Body Width follow the rule of thumb that ‘the Body Length is a part less than twice the Body Width’.   So twice minus a 4th part gives us 4::7.  Or twice minus a 3rd part gives 3::5. etc.  The traditional ratios seen in Cremona work being 2::3, 3::5, 4::7, 5::9
      • Even when extreme shapes for odd instruments push outside normal ranges, we see the Cremona makers preferring to derive a solution based on the traditional ranges.  Thus for a very skinny instrument like a pochette, we see use of the traditional ratio range, but with the width divided by 2.  So for example  3/2::5
    • Both geometry constructions and ratio choices effectively ‘relate’ some features to others.
    • Some traditional methods actually give a location or shape, some only provide a ‘guide’ to position or size.  I refer to these as ‘execution’ versus ‘guide’ methods.
      • As examples, the horizontal placement of the upper soundhole eyes is ‘guide’ variously by ‘quarters across the cBout’ or/and ‘thirds of the upper bout’.  But it’s actual execution is consistently done by relating the distance across the eyes to the stop length of the strings and bridge as ‘the stop unit or a part less’.
        • As example of the variation of application, for violins and violas the distance across the eyes is used.  But a wider spacing is needed for cellos.  Instead of creating some different relationship, the old makers just a adapted the traditional relationship by applying ‘between the eyes’ instead of ‘across the eyes’.
    • While most of the outline shapes are directly given by ‘Circle and line’ geometry, some shapes – like the arching of the plates – is not directly given. In these cases, the traditional geometry and ratios given boundaries and controls to the shape.  But the boat builder’s concept of a ‘fair curve’ must complete the shape.  A ‘fair’ curve is the smooth shape you would get by bending a ‘spline’ (appropriately sized wooden slat) across the controlled points.
    • The concept of a ‘faired curve’ also was used as needed to smooth across any inconsistencies in shape.
      • This was a corrective, used more or less based on standards of the workman.
      • In a very few situations, we see Cremona makers using the underlying circle geometry in a way that does not correctly join smoothly.  In these spots we see even the best makers ‘fair’ the curve to smooth these bumpy joins.
        • The geometry requirement is that point where two arcs from different centers join must lie on the line joining their centers.   The arcs will join smoothly if and only if this condition is met.
  • The design acts as a recipe interacting with the building process, not as an ideal worked out ahead of time.
    • At each step, work is done by using the traditional geometry for a feature, and controlling the application of this geometry with guides and ratios chosen from the traditional ranges of options for the feature.
    • The work on the current feature keys of actual size and location of the features it relates to in the instrument as being built. 
      • In other words, there is no ‘correction’ back to a preplanned ideal.
      • Any inconsistencies or asymmetries are incorporated as the work unfolds, rather than corrected out.
      • This is a major factor in the ‘organic’ character of classical Cremona work.
  • The traditional collection of geometry construction and ratios choices for the ‘guide’ and ‘execution’ methods of a particular type of instrument effectively related every feature to some of the other features in the instrument, and not to others.  Thus part of the ‘tradition’ had to do with which features are related to which.
    • The same geometry, combined with different ratio choices and initial sizing can give very different final results.
      • Some features traditional gave some choice of geometry constructions.
        • The soundholes and scroll are examples were there are some traditional variations of choice in the actual geometry
        • We also see variation in the geometry around the corners and the risers coming from the main bouts into the corners.
        • But most features had just one consistent geometry.  And where there were choices, they were very limited by tradition.
      • Almost every feature gave a range of traditional ratio choices to use.
      • Most features gave some additional liberty through the choices of including or excluding things like margins in making the ratio calculations.
        • Do we measure from the outer edge or the purfling.  Or from the top or eye, the center, or the bottom. Etc.
      • Combined, these ‘choices’ meant that very different instruments like a tenor viola, piccolo violin, a pochette, and a cello can share almost all the same geometry constructions
  • Some of the development and evolution of the methods came through slow changes  in the roles of ‘guide’ and ‘execution’ methods.  And the matter is complicated by the occasional appearance of ‘incidental relationships’ that arise only accidentally from a combination of choices.  Each maker repeats most choices from instrument to instruments, so sometimes these incidental relationships can persist across a whole group of instruments. Several processes can be seen related to the long term interaction of these things.
    • Sometimes what begins as an ‘incidental’ relationship can become a ‘guide’ relationship.
      • In the first Cremona generations, we see ‘quarters across the cBout’ as a ‘guide method’ for distance across the upper soundhole eyes.   But some of these same makers also very much favored making the cBout and upper bout width related in a 2 to 3 ratio.  ‘Incidentally’ this caused the distance across the eyes to also be related to ‘thirds across the upper bout’.   Two generations later, we find some makers using the ‘thirds across the upper bout’ as the ‘guide’ now.   These makers weren’t always making the cBout and upper bout have a 2 to 3 relation, and the ‘quarters across the cBout’ guide wasn’t actually followed with consistency any more.
      • Similarly, we sometimes see what was a ‘guide’ becoming the ‘execution’. 


I won’t go into more detail about the particular constructions and traditional ratio ranges here.   But anyone with questions, or wanting further detail is invited to contact me.

What’s important in all this is that we see the Old Cremona makers followed flexible but highly structured traditions.   And we see this in the instruments.   We see something individual and organically loose in a sense, but with a great deal of orderliness and sense shining through.

Their traditions supported them.   By making within the tradition, certain features were guaranteed to be related to other certain other features.   But they weren’t straight jacketed by their tradition either.  It allowed them great latitude, just in a structured way.

They could launch off into a highly uniquely sized project, yet have confidence that the important functional relationships would all be well preserved. 

Very importantly, their structured methods allowed them to repeat what was known to work, and still allowed them to tinker at the edges seeking improved results.

By looking at their work in this way, we can see the choice within the tradition embodied in each instrument.  It’s sort of a conversation across generations of Cremona work.   Now we can listen in on that conversation.   We can see precisely when Stradivari decides to revisit a choice from Andrea Amati.  Or when Del Gesu final pushes a step to far and has to add an extra arc to the soundhole design, even though it doesn’t join smoothly.   Beyond eaves dropping to follow their conversation, we can join in.   We also can step into the bounds and freedoms offered by their traditions.   We can make an instrument picking up from the conversation.  We can repeat the choices we see in some instruments we like, and see what the method gives us.  Then repeat and change as many choices as we like in the next, exactly as they did.

We should not underestimate the power of evolving ‘trial and error’ learning processes.   That is part of the nature of ‘practice’ for a musician.  It is part of the deeper nature of the arts and the sciences.  Lone invention and genius have a place in human history.  But probably even more than we commonly recognize or admit, human accomplishment tends to build from our collective accomplishments.  And is something of many trials and attempts, and evolutionary in nature about human practices and even our ideas.

Recent news gives us an extreme example.   Google recently turned its neural network A.I., Alphazero, to the problem of playing chess.   It actually learns, in an neural network ‘evolution’ process.  It tries many things, and learns what works and what doesn’t.   It has massive computing power, so it can run an extraordinary number of learn trials in a short time.  But the process is essentially adaptive, favoring what works and discounting what doesn’t.  Basic trial and error with structured favoritism and countless repetitions.   That’s the same as evolution.   Alphazero was given the rules of chess.   They turned it on.  Nine hours later, it played a tournament of 100 games against Stockfish.  The Stockfish program was preciously the strongest chess player every.  Stronger than the best human masters.  Stronger than any other computer chess program.  But Stockfish isn’t like Alphazero.   Humans feed Stockfish a compendium of human theory of chess, the openings, and endings, the ways of evaluating pieces and situations.   Then Stockfish adds brute calculation to find the best moves.   Alphazero was given none of that.  Just the rules.  But nine hours later, it decisively beat Stockfish.   28 wins, no losses, 72 ties.  In nine hours, Alphazero had become the most powerful chess entity ever to exist – through massive trial and error learning. We should never underestimate trail and error learning – repetition and limited variations, with reinforcing feedback reflecting the success of the results.

That is what this Old Cremona system amounts to.  What my research found is an approach to making based on following a long slate of traditional choices, with only slight variations from past choices, combined with evaluation of results, and lots of repetition across generations of makers.  The slate of choices made is like a DNA encoding, if you will allow.  With each instrument they got to review the results.   They could then follow their suspicions and make a few changes to the choices, perhaps hoping to improve an appearance feature or something about performance.  If the hunch paid out, they could repeat that choice.  If not, they could try other variations. This is the formula for evolution, and the formula for neural network learning.  Evolution processes of this kind are probably the most powerful design engines every to exist.


So how do we put this research to work?  And why bother?

One way to implement the research is simply to start building and making choices following the Cremona example.   This is what I’m doing.  

But there are other ways to learn from this approach to looking at classical making.  Besides directly showing us the ‘guide’ and ‘execution’ methods of the Old Cremona making, these also reveal which features were traditionally related to which features.   And, indirectly, they reveal a way of looking at the features of the violin and understand the ‘knobs and levers’ of variation in the classical work.   So for example, we see that ‘channel width’ is one of the levers we can move to vary arching.   So a maker who perhaps doesn’t want to go all the way to ‘doing as they did’, might instead decide to still use basically the same structures and relationships, and to make adjustments by moving the same ‘knobs and levers’.


Thank you for exploring ‘revival making’ with me.

No comments:

Post a Comment