HNHacker News
TopNewBestAskShowJobs

jason_s

474 karma · joined February 18, 2015

Embedded systems engineer in motor control and signal processing.

https://twitter.com/jms_embedded

submissionscomments
jason_s··on The dedoublifier, part one
I'm assuming he's going to use continued fractions to find the best rational approximation to a given floating-point number.
jason_s··on On Reviews
Microchip Technology decouples yearly reviews and pay raises. (And pay raises aren't necessarily once a year, it ebbs and flows with company earnings.) I like this a lot; it removes a lot of the stress from the review process.
jason_s··on The Tower of Babel and the Fighter Plane (2013) [pdf]
Yeah, it'll probably be a car crash or a hospital equipment failure :/
jason_s··on The Tower of Babel and the Fighter Plane (2013) [pdf]
Great read!

Complexity is evil evil evil (my take on it: http://www.embeddedrelated.com/showarticle/132.php)

<quote from Conway's article>

Incidentally, today’s automobiles share this problem of a massive amount of on-board code and the bugginess that accompanies it. A few years ago, some Mercedes owners found that when they pushed a certain button on the navigation system the driver’s seat moved.

</quote>

I have this sinking feeling that my death is going to occur because of some interaction between software/firmware design and an unexpected situation (e.g. stuck in an elevator between floors because the firmware is expecting internet packets from Twitter or something). Hope I'm wrong, but the complexity of electronics seems to be increasing fast enough for system reliability to decrease.

jason_s··on Atlassian files for IPO
You can use either Atlassian markup or Markdown to edit Confluence pages, it just takes a few extra keystrokes and I can't remember how to do it right at the moment. Not a piece of cake like Github/Stackoverflow, but at least it's possible.
jason_s··on Computer Sciences Courses That Don't Exist, but Should
Robot Odyssey!!!!!

https://en.wikipedia.org/wiki/Robot_Odyssey

jason_s··on What RSA key size to use for an SSL certificate
the answers you linked to cite key sizes of c * 2^k where c = 1,3,or 5 -- like 3072 = 3 * 1024, 2560 = 5 * 512.
jason_s··on Really Random Number Generator
just because binary data is "random" doesn't mean it yields independent, identically-distributed random variables --- whatever the source.
jason_s··on What RSA key size to use for an SSL certificate
Is there a reason 3072-bit or 2560-bit RSA keys can't be used?
jason_s··on Really Random Number Generator
OK but where's the statistical testing to verify it?

For microelectronics there are better ways for true digital random number generators, e.g. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.99....

jason_s··on Chebyshev Approximation
Ironically, that's one of the few useless parts of mathematics, and the least favorite part of my elementary school education.

Suggest group theory or linear algebra instead. ;-)

jason_s··on LSD – Line Square Dot Game
...except for the bug where the ball gets trapped in a line and pops out of an enclosed polygon.
jason_s··on Chebyshev Approximation
Microchip does have a standard license for maybe 99% of its firmware libraries, that includes this proviso, that just forbids you from using the software on microcontrollers made by other manufacturers:

  Microchip licenses to you the right to use, 
  modify, copy and distribute Software only
  when embedded on a Microchip microcontroller 
  or digital signal controller that is integrated 
  into your product or third party product
  (pursuant to the sublicense terms in the
  accompanying license agreement).
If there are any other restrictions, I'm not aware of them.
jason_s··on Chebyshev Approximation
I'm not sure I would believe the comment by the person who you are responding to.
jason_s··on Chebyshev Approximation
PIC is not the name of the company. It's Microchip Technology, Inc. You seem to be upset about something, since I don't think you have the facts straight about the XC series of compilers; the source code for XC16 and XC32 is posted here under "Source Archive" http://www.microchip.com/pagehandler/en-us/devtools/dev-tool...

Microchip has a number of open-source initiatives. You can use the internal guts of MPLAB X from Java via MDBcore, for example, as a scriptable debugger/simulator.

jason_s··on Chebyshev Approximation
Чебышёв!
jason_s··on Chebyshev Approximation
Heh. Two immediate comments:

- It's not as bad as that. Chebyshev approximation + other techniques for evaluating mathematical functions are pretty easy to learn.

- You're absolutely right. Welcome to mathematics! The mountain goes infinitely high. There are tons of technical papers I'll never be able to digest simply because they've compressed lots of thoughts into cryptic-looking but standard notation.

jason_s··on Chebyshev Approximation
is there a way to have Pade approximation but with continued fractions, so you just truncate the series as necessary? I tried looking into this a few months ago, but my head just started to hurt.
jason_s··on Chebyshev Approximation
>Well then go lobby for creating dsPIC support in LLVM!

Um. Been there, done that. Well, you see, [: discussion of internal politics deleted :]

jason_s··on Chebyshev Approximation
well, I work for Microchip, so I won't be writing code for other manufacturers' processors, at least not for my day job :-)
jason_s··on Chebyshev Approximation
I've heard of it. Rust is probably the most promising to me.

Unfortunately just because a language exists, doesn't mean it is usable on a particular embedded system architecture. The processor core and system libraries need support.

jason_s··on Chebyshev Approximation
I agree with Jacobolus --- even with say, 40 points and a 5th order Clenshaw/Horner rule, you have a 40x5 matrix; parallelize along the long dimension.
jason_s··on Chebyshev Approximation
>The author of this blog post might as well have used Clenshaw’s algorithm

So is that essentially sort of like Horner's Rule but for Chebyshev polynomials? e.g. instead of a0 + (a1 * x) + (a2 * x * x) + (a3 * x * x * x) + ... , use a0 + x * (a1 + x * (a2 + x * (a3 + ... so that the coefficients are used in the reverse order, and instead of having to maintain a sum and a value of x^n, we just have a single accumulator; in Clenshaw's algorithm we have the two accumulators bk1 and bk2, but they look similar.

jason_s··on Chebyshev Approximation
>This blog post’s statement that “higher degree polynomials can have problems with numerical accuracy” turns out to be a myth

(author here) I should have qualified or corrected that statement in my article. There are a few reasons I rarely use n>5:

- when polynomial approximation is appropriate, high-degree polynomials are usually unnecessary on embedded systems (0.1% accuracy usually sufficient)

- it's usually a sign that polynomial approximation is being misapplied and there are better ways (via range reduction, transformation like 1/f(x), etc.)

- with least-squares, computation of the coefficients can have problems with numerical accuracy. When I've used least-squares fitting polynomials with n>10, I often get a complaint from my numerical tool (MATLAB or Python) that the matrix in question is ill-conditioned. That may be because I used the Vandermonde matrix rather than Chebyshev polynomials, or because I used a large number of points. Chebyshev approximation doesn't have the same problem, I don't think, and I should not have mapped my experience with least-squares fitting onto Chebyshev approximation.

Evaluation of the polynomials themselves should be fine (after all, we can easily compute a 20th-degree Chebyshev polynomial).

jason_s··on Chebyshev Approximation
The author (me) frankly hates C and C++. Haven't touched it for use on a PC since 2009. However I use it every day for my job working on dsPIC motor control, because I have no better options.

When I am trying to teach someone about algorithms, I turn to Python because it is more readable, quicker for me to write, it's platform-independent, and I can incorporate it into an IPython (now Jupyter) Notebook and postprocess that into my blog articles with a minimum of hiccups.

jason_s··on Continuum Analytics Raises $24M Series A for Anaconda Python and PyData
We would be absolutely stuck w/o conda's environment manager -- I can make sure we use exactly the same versions of the packages across our whole team.
jason_s··on What’s New in Python 3.5
I hate the new @ operator for matrix multiplication, it just looks ugly. (And I use numpy matrix math all the time.) But whatever.
jason_s··on Fast Approximate Logarithms, Part III: The Formulas
interesting -- never heard of Estrin's method; thanks!
jason_s··on The Property Jungle: I'm a security researcher, get me out of here
DM = ?
jason_s··on Fast Approximate Logarithms, Part III: The Formulas
No mention of Padé approximants? I'd just use a Chebyshev approximation transformed into a Padé approximation, then perturb the coefficients to optimize maximum error.
← PreviousPage 6 of 7Next →