[sudo-discuss] Spectral Response Curves

[StCredZero]

On Tue, Nov 12, 2013 at 12:10 PM, Patrik D'haeseleer <patrikd at gmail.com>wrote:

> I'm not sure why people are so obsessed by 470nm specifically. If you look
> back at the original literature, you'll see that the response is quite
> broad, and is actually a fairly good match for the overall blue channel
> response you posted earlier.
>

I have had this in mind while designing this app. If I thought I needed
real precision, I wouldn't have tried at all, because I have some idea of
the limits of the setup. However, 470nm is supposed to be the frequency
that people react most strongly to, and people who have read popular
interest articles on the subject will be familiar with this.


> Compare:
> http://www.jneurosci.org/content/21/16/6405/F5.expansion.html
>
> http://www.maxmax.com/images/Cameras/Technical/NikonD200_SpectralResponse.jpg
>

Unfortunately, the Nikon D200 is not the target platform. It's the iPhone 4
and newer models and I'm having a hard time finding spectral response data
on the newer models. (AFAIK, they are variants of the IMX145, and I don't
know if the IMX145 datasheet would apply.) Looking at the spectral response
of other hardware is good enough to validate this *might* work. However, I
really should validate the resulting instrument by accurately measuring
what its response actually is. "Yeah, it's probably good enough" is ok for
my personal use, but it's really a different story if I'm going to
distribute it.


> So yeah, for marketing purposes, it's probably more effective if you can
> tap into the hype around 470nm. But if you really want an app that's
> effective against insomnia, you might as well measure the blue channel all
> by itself...
>

As I said before, if I thought I really needed much more accuracy than
that, then I would not have proceeded with this. Even so, I really need to
be able to tell people precisely what they are getting. What if the peak
turned out to be more like 450nm and dropped off unexpectedly quickly? A
user would want to know. If I know exactly what the curve is and can
provide that information, knowledgeable people like you can make an
informed decision about how well this works.


> PS: Also, simply subtracting the red channel from the blue gives you such
> an incredibly crude measure of the light incoming at any one particular
> frequency,
>

The intent has always been to create a relatively crude instrument.


> that using highly precise measurement equipment to validate the result is
> rather nonsensical.
>

Uh, no. Home audio equipment is comparatively crude, and most consumers
know this, but equipment of decent precision is still used to measure its
performance. This isn't nonsensical. It's providing good information to
users. Really, they have enough uncertainty to deal with when making
decisions and unlike yourself, they don't necessarily have a lot of
experience gauging how inaccuracies can stack in a system. The more of that
kind of judgement you can remove from their process, the easier it is for
them.


> The DIY spectroscope Public Lab really gives you more than enough
> precision for what you need (although perhaps not enough for what your
> customers *want*, but what do they know...)
>

It's what they can *know* not what they can make an educated guess at in a
field they know relatively little about. I don't think, you're really
putting yourself in the shoes of the target demographic.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sudoroom.org/pipermail/sudo-discuss/attachments/20131112/7df8edc9/attachment.html>