I'm not too worried about these issues. The underlying issue with why we want an on-off response in the motor is this: we want the participant to have just ONE bit of data given by the environment; we want to show that with just this one bit, PLUS their feeling of their own movement, they can get a sense of space. This is what lets us show that the feeling of a space need not depend on more than that one data point. And this lets us make a strong argument that the sense of space is constituted via movement. When we have analogue response, the participant is gleaning multiple bits of data; and the data maps out vectors, from weaker to stronger responses, and so inherently has a kind of oriented spatiality to it. Experientially, I should also say that the feeling of being connected with something beyond me in the case of the digital glove, clearly arose from the way *I moved*; I have to do something to get a bearing. Whereas, with the analogue glove, it was more like being beholden to a vector field beyond me which ALREADY laid out its directions; I need to move only into order to reveal already given vectors. I.e., my thought is that the digital apparatus gives you more of a sense of being participant in space constitution.
No doubt the cut off implementation is going to give artifacts and interesting stuff. But I think these are artefacts as to where QUANTATIVELY, in *an already given space*, the cutoff is to happen. But our interest is in our moving relation to things and boundaries as QUALITATIVE, in a 'space of constituion'. So I'm not worried. On the other hand, coincidentally, I was thinking this morning that hysteresis, metastability, transitions, etc., are the point where machines become interestingly living. This is because I must have plugged my USB hard drive at *just* the wrong microsecond this morning as my computer was booting up, because it failed to recognize it properly and then wouldn't recognized it properly on a second, third, etc. insert. I thought my drive was fried, but it turned out to be some weird glitch that: didn't recognize the drive; yet left the driver resident in memory; and couldn't fix up that problem. When I shut down and restarted the drive worked fine. I take it this because there is something about the process of a system getting ready to function that nearly always leaves an unclean transition. I.e., we want computer systems to either be off or on, started up (and ready to recognize hardware) or not (and not ready to recognize), but the infinitely fast transitions that we envision always take time. And that's where machine get interestingly ornery, where we want to say: it's died on me, it's failed to live. -----Original Message-----
From:
owner-memory-place@concordia.ca [mailto:
owner-memory-place@concordia.ca] On Behalf Of
adrian@adrianfreed.com Sent: March-05-11 5:18 PM
To:
morgan@msutherl.net Cc:
memory-place@concordia.ca;
post@memoryplace.posterous.com Subject: RE: Hysteresis and Metastability
Thanks for the pointer although I should emphasize that the point of my
note
was to warn the memory-place group of the methodological rabbit hole:
not send them down into it.
You don't need to avoid comparators. There is one in every gate of every
digital computer...
You just need to know that digital is an illusion and that maintaining
the illusion is hard work and has some obvious
and some less obvious results. I nearly got caught by this myself
recently scaling up multitouch pressure fabric arrays we are working on.
Bigger than about 12x12 skew and jitter-induced distortion from
distributing the ADC clocks around the arrays is enough to impact
performance. Without a lot of hard work we will never get near the pitch
precision a cellist has available to them for example.
One doesn't usually add a nonlinearity as you suggest in the analog
world because the circuitry to do the nonlinearity
adds even more of its own noise (and a temperature dependence, requiring
noisy diodes to compensate for). Take a look at the LM13700 ecosystem to see what I am talking about.
> > Here's a suggestion: rather than using a comparator, think about using the
> venerable sigmoid or the hyperbolic tangent functions so that the
> "amplification" around the threshold can be tuned to be more gentle:
>
http://mathworld.wolfram.com/SigmoidFunction.html +
>
https://secure.wikimedia.org/wikipedia/en/wiki/Sigmoid_function > > You can get a feel for the functions here:
>
http://www.wolframalpha.com/input/?i=Plot%5BTanh%5Bx%5D%2C+%7Bx%2C+-6%2C+6%7D... >
http://www.wolframalpha.com/input/?i=Plot%5B1%2F%281+%2B+E%5E-x%29%2C+%7Bx%2C... > > I'm not sure how to implement this with analog electronics, but you could do
> it with a little Atmega ATtiny chip:
http://www.sparkfun.com/products/211 or
> a spare Arduino (overkill).
> > Morgan
> > On Sat, Mar 5, 2011 at 1:41 PM, wrote:
> > > Xin Wei asked me to forward these thoughts to the group in general. I
> > sent them to him originally because I didn't went to be known as the
> > chief source of rain on your parade…
> >
> >
> > After I heard you wanted to change the sensor device to a switching
> > instead of a continuous mode I reflexively told Patricia to look into
> > hysteresis (a basis function for state space modeling) for the glove
> > sensor.
> >
> > Hysteresis is what engineers learn about to analyze the usual solutions
> > to a thorny and deep problem that arises when moving from "continuous"
> > signals to bivalued signals. The solutions have various names, e.g.,
> > positive feedback comparator, Schmitt trigger
> > (
http://en.wikipedia.org/wiki/Schmitt_trigger).
> > The idea is to eliminate what some engineers called "chatter" (others
> > jitter) a switching back and forth between the two states when you are
> > near the thresshold. At this threshhold very small modulations in light
> > are "amplified" into gross switching changes by the apparatus. These may
> > be caused by ambient interference from other light sources, noise from
> > the light source or sensors themselves or more interestingly your own
> > shakiness.
> >
> > I am concerned that I have just taught enough to the implementation team
> > to essentially ruin the experiment. The signal conditioning I mentioned
> > is inherent in the construction of this device and may deeply distort
> > the results. The idea of a "clean" edge/experience is loaded. You are
> > looking for emergence of walls and rooms but you may have constructed
> > them with the comparator.
> > What if fuzzy, chattery edges were fundamental and primary to
> > experiencing edge
> > and memory.....etc. What if random modulations are fundamental
> > to the perceptual apparatus (which there is evidence of in the case of
> > the eye)?
> >
> > Oops.
> >
> > Remember the coin-in-the-water demonstration? I claim the water is a
> > window. It doesn't have the familar spatial relationship or material
> > qualities that windows have in regular rooms. Why can't part of the
> > roomness be defined in the electronics?
> >
> > At the very least we have to acknowledge that what we experience using
> > the apparatus is an entanglement of us, the light source and the whole
> > apparatus.
> >
> > p.s. It is not the comparator by itself that is the problem. I hear you
> > all talking of the apparatus as if going to two states or a "single"
> > sense modality somehow simplified or reduced something. In fact the
> > implementation of a comparator like this is an amplifier of the noise at
> > the thresshold - a recoding of barely perceptible signals into
> > perceptible ones - a trade of space and time. ().
> >
> > If you are tempted to sweep this concern under the rug remember:
> >
> > 1) that 100pS jitter in audio signal reproduction is audible
> > 2) the related problem of metastability:
> > "metastability is an inevitable result of any attempt to map a
> > continuous domain to a discrete one"
> >
> > "Serious computer and digital hardware bugs caused by metastability have
> > a fascinating social history. Many engineers have refused to believe
> > that a bistable device can enter into a state that is neither true nor
> > false and has a positive probability that it will remain indefinite for
> > any given period of time, albeit with exponentially decreasing
> > probability over time. However, metastability is an inevitable result of
> > any attempt to map a continuous domain to a discrete one. There will
> > always be points in the continuous domain which are equidistant (or
> > nearly so) from the points of the discrete domain, making a decision as
> > to which discrete point to select a difficult and potentially lengthy
> > process.[8] If the inputs to an arbiter or flip-flop arrive almost
> > simultaneously, the circuit most likely will traverse a point of
> > metastability. Metastability remains poorly understood in some circles,
> > and various engineers have proposed their own circuits said to solve or
> > filter out the metastability; typically these circuits simply shift the
> > occurrence of metastability from one place to another."
> >
http://en.wikipedia.org/wiki/Metastability_in_electronics > > 3) These sorts of things are probably at play in biological signal
> > mechanisms at the cellular level e.g., gradients in cells.
> >
> >
