Documentation GigaHarp
The GigaHarp
DECO3100 InfoStudio2009
Assignment 1 - Data Sculpture
Kuan-Yu Kevin Chen
Idea Development:
Initially
During our first activity we were to find three datasets as an excecise to familiarise ourselves with reading datasets and finding a trend within them. Two of the datasets I listed consisted of data from benchmarking Graphics Processing Units and their respective Cards. This instantly also became the basis of my idea for the first assignment.
Initially the idea was to involve using the power draw and efficiency of graphics cards dataset as well as the benchmark results of graphics cards dataset to create a trend. These two datasets in conjunction with the price of each Graphics Card, a trend is therefore to be shown using using the "Performance per watt" and "Performance per dollar" comparisons. However this did not push the boundary of visual design far enough, and a dataset showing exactly just that trend is already available in simple graphs. So then the main question came up, how do I make this remotely geeky idea better? The answer was, make it more geeky.
Therefore I decided to dive a lil deeper to information just for GPUs, which means i have restricted myself into using only specifications from GPUs rather then the cards. From this I started to concentrate on transistor counts per GPU die and the amount of texture data they can process, and plot them all against the amount of power draw of each card.
This was not geeky enough still.
Final Idea
Finally I have decided to dive even deeper into just the fundamentals of Processing Units. Which means, at the basis of all computing power be it grapics, sound, data functions, is a processing unit. And this unit consists of purely transistors.
The development of fabrication methods and size of transistors has leaped exponentially through out the years of computing. We are now making more powerful processors by shrinking each transistor size, which allows us to fit more transistors per set die size. What this means is that for a set processing unit die size, as the transistor size decreases, and therefore the transistor count increases, the core die is able to pass more raw data in the forms of zeros and ones which translates to increased performance.
Therefore the trend was set, the development and evolution of transistors and processing units. The attributes used will be transistor count, raw bytes of data the core can process, and transistor size. The main part of this was that the information is not to just be visually informative, but also to do it using sound.
Visualising the Model:
Initially the basic idea of the model is for it to be an instrument. After gathering some inspiration, a music box in a certain movie captured my eyes immediately.
An idea therefore was considered to use something similar to a music box, but not visually like one.
However the musicbox idea was not able to be realised. To produce sound, and to have it continuously loop, the design of visualisation keep looking remotely like a music box itself. Maybe another method of sound creation might be more flexible visually.
There are only a few fundamental ways to produce sound:
- Physical Contact Type (music box, xylophone, glass harp)
- String Type (concert harps, violin, cello)
- Pressured Air Type (flute, organ)
All instruments falls in or between these categories. They can be either solely only one type, or can be hybrids of multiple types.
The decision was then to try out string type to see what can come out of it.
Immediately a realisation was made on how flexible and similarity some string type instruments have on the graph plotted on ManyEyes. Conceptually the model requires two backings and strings inbetween. And the placement, sound pitch, strings etc. are all to be data driven. So the decision was to use the graph from ManyEyes, plot it and make it into a visualisation model:
- Transistor numbers will be where the dots are situated.
- Data processed will be number of strings
- Transistor size will be the pitch of the strings
With these in mind, the visual design of the model was then underway. Using the shape of the graph, sketches were made simulating the side and top views of the model. Mix and match was done to see the most visually appealing and probable model to produce.
From these views the 3D mix and match was sketched.
From the skeches, the most visually compelling and accurate to the idea was S:1 T:2. However this design would require a lot more work then possible. Large ply wood is required to be cut to shape, then vacuum bent, over 120 strings are required, and each string would require its own tensioner. Due to time constraints, S:2 T:1 was selected instead to be the basis of the model due to its simplistic design and the ability to exploit using only one model to produce two of the backs.
From here on, ergonomic and structural strength was considred. How is the harp to be held? How can I keep the harp from buckling under tension?. The challenge was to combine these two together and have it integrated into the model without disrupting the data.
In the end, the simple U curve (second sketch from top) was selected due to its simplistic and ability to blend in with the model and compliment it.
The Model and Further Development:
Work was then underway in Rhino to plot out the graph on the XY plane. The data needs to be sorted and made into a format that rhino can read. The curve was then made and work in Grasshopper was then underway.
Curves were then further added within Rhino to enhance the body of the harp. Some modifications were done here and there to make the harp visually more appealing.
The Complete Digital Model in Rhino
The model was then exploded, and the face extracted to form a template for the laser cutting machine. Pieces are rotated and fitted onto the 800x400mm template. Text and labels were also added onto the template as etching.
Fabrication:
Link
The above links to the full article of the construction of the model.
Pieces were cut using the laser cutter.
The pieces were then glued together accordingly to make three curves.
Weights were used to compress the three curves for a stronger bond and left overnight.
They three pieces were then glued and screwed together.
Two of the similar curves are the back of the harp while the third curve is for support and holding.
Sanding was then done to fix up the edges and give the model an overall smoother feel.
Estapol was sprayed to protect the wood from moisture.
Strings were then threaded through the holes and tied at both ends.
The strings are then tensioned using the single tensioner.
Problems and Difficulties:
Most of the difficulties and challenges were with the design of the model. The fabrication process was reasonably straight forward. The only major concern and issue was when the harp joint at the front started to split when tensioning the first and second strings. Because of this these two strings do not play a note properly.
Another issue with the harp is the use of a single tensioner for all the lines. This created issues of non-equal tensions between all the strings. However this was the single most efficient method so the problem was within acceptable specs.
The harp was able to produce sound, however it is not as rich as a real concert harp. This is due to the low tension, lack of sound chamber and usage of standard nylon fishing wire instead of proper harp steel core, steel casing strings. The decision to use nylon strings only was due to the price of actual harp strings, which was around $30 per string. With this model which had 20 strings, that could easily reach $600 on strings alone (Even though some strings can be cut in half and reused).
In the end the model went smoothly with only a few problems and issues besides the slight splitting of the joints.
Completed Sculpture:
The complete model.
The annotated photograph
Files:
Dataset
Rhino Model
Grasshopper File
AutoCad File
DECO3100 InfoStudio2009
Assignment 1 - Data Sculpture
Kuan-Yu Kevin Chen
Idea Development:
Initially
During our first activity we were to find three datasets as an excecise to familiarise ourselves with reading datasets and finding a trend within them. Two of the datasets I listed consisted of data from benchmarking Graphics Processing Units and their respective Cards. This instantly also became the basis of my idea for the first assignment.
Initially the idea was to involve using the power draw and efficiency of graphics cards dataset as well as the benchmark results of graphics cards dataset to create a trend. These two datasets in conjunction with the price of each Graphics Card, a trend is therefore to be shown using using the "Performance per watt" and "Performance per dollar" comparisons. However this did not push the boundary of visual design far enough, and a dataset showing exactly just that trend is already available in simple graphs. So then the main question came up, how do I make this remotely geeky idea better? The answer was, make it more geeky.
Therefore I decided to dive a lil deeper to information just for GPUs, which means i have restricted myself into using only specifications from GPUs rather then the cards. From this I started to concentrate on transistor counts per GPU die and the amount of texture data they can process, and plot them all against the amount of power draw of each card.
This was not geeky enough still.
Final Idea
Finally I have decided to dive even deeper into just the fundamentals of Processing Units. Which means, at the basis of all computing power be it grapics, sound, data functions, is a processing unit. And this unit consists of purely transistors.
The development of fabrication methods and size of transistors has leaped exponentially through out the years of computing. We are now making more powerful processors by shrinking each transistor size, which allows us to fit more transistors per set die size. What this means is that for a set processing unit die size, as the transistor size decreases, and therefore the transistor count increases, the core die is able to pass more raw data in the forms of zeros and ones which translates to increased performance.
Therefore the trend was set, the development and evolution of transistors and processing units. The attributes used will be transistor count, raw bytes of data the core can process, and transistor size. The main part of this was that the information is not to just be visually informative, but also to do it using sound.
Visualising the Model:
Initially the basic idea of the model is for it to be an instrument. After gathering some inspiration, a music box in a certain movie captured my eyes immediately.
An idea therefore was considered to use something similar to a music box, but not visually like one.
However the musicbox idea was not able to be realised. To produce sound, and to have it continuously loop, the design of visualisation keep looking remotely like a music box itself. Maybe another method of sound creation might be more flexible visually.
There are only a few fundamental ways to produce sound:
- Physical Contact Type (music box, xylophone, glass harp)
- String Type (concert harps, violin, cello)
- Pressured Air Type (flute, organ)
All instruments falls in or between these categories. They can be either solely only one type, or can be hybrids of multiple types.
The decision was then to try out string type to see what can come out of it.
Immediately a realisation was made on how flexible and similarity some string type instruments have on the graph plotted on ManyEyes. Conceptually the model requires two backings and strings inbetween. And the placement, sound pitch, strings etc. are all to be data driven. So the decision was to use the graph from ManyEyes, plot it and make it into a visualisation model:
- Transistor numbers will be where the dots are situated.
- Data processed will be number of strings
- Transistor size will be the pitch of the strings
With these in mind, the visual design of the model was then underway. Using the shape of the graph, sketches were made simulating the side and top views of the model. Mix and match was done to see the most visually appealing and probable model to produce.
From these views the 3D mix and match was sketched.
From the skeches, the most visually compelling and accurate to the idea was S:1 T:2. However this design would require a lot more work then possible. Large ply wood is required to be cut to shape, then vacuum bent, over 120 strings are required, and each string would require its own tensioner. Due to time constraints, S:2 T:1 was selected instead to be the basis of the model due to its simplistic design and the ability to exploit using only one model to produce two of the backs.
From here on, ergonomic and structural strength was considred. How is the harp to be held? How can I keep the harp from buckling under tension?. The challenge was to combine these two together and have it integrated into the model without disrupting the data.
In the end, the simple U curve (second sketch from top) was selected due to its simplistic and ability to blend in with the model and compliment it.
The Model and Further Development:
Work was then underway in Rhino to plot out the graph on the XY plane. The data needs to be sorted and made into a format that rhino can read. The curve was then made and work in Grasshopper was then underway.
Curves were then further added within Rhino to enhance the body of the harp. Some modifications were done here and there to make the harp visually more appealing.
The Complete Digital Model in Rhino
The model was then exploded, and the face extracted to form a template for the laser cutting machine. Pieces are rotated and fitted onto the 800x400mm template. Text and labels were also added onto the template as etching.
Fabrication:
Link
The above links to the full article of the construction of the model.
Pieces were cut using the laser cutter.
The pieces were then glued together accordingly to make three curves.
Weights were used to compress the three curves for a stronger bond and left overnight.
They three pieces were then glued and screwed together.
Two of the similar curves are the back of the harp while the third curve is for support and holding.
Sanding was then done to fix up the edges and give the model an overall smoother feel.
Estapol was sprayed to protect the wood from moisture.
Strings were then threaded through the holes and tied at both ends.
The strings are then tensioned using the single tensioner.
Problems and Difficulties:
Most of the difficulties and challenges were with the design of the model. The fabrication process was reasonably straight forward. The only major concern and issue was when the harp joint at the front started to split when tensioning the first and second strings. Because of this these two strings do not play a note properly.
Another issue with the harp is the use of a single tensioner for all the lines. This created issues of non-equal tensions between all the strings. However this was the single most efficient method so the problem was within acceptable specs.
The harp was able to produce sound, however it is not as rich as a real concert harp. This is due to the low tension, lack of sound chamber and usage of standard nylon fishing wire instead of proper harp steel core, steel casing strings. The decision to use nylon strings only was due to the price of actual harp strings, which was around $30 per string. With this model which had 20 strings, that could easily reach $600 on strings alone (Even though some strings can be cut in half and reused).
In the end the model went smoothly with only a few problems and issues besides the slight splitting of the joints.
Completed Sculpture:
The complete model.
The annotated photograph
Files:
Dataset
Rhino Model
Grasshopper File
AutoCad File
Add a Comment
About
© 2009 Created by Andrew on Ning. Create a Ning Network!
You need to be a member of infostudio 2009 to add comments!
Join this Ning Network