25 comments

[ 3.3 ms ] story [ 63.1 ms ] thread
Neat. Can one resolve the arnumber parameter stamp.jsp takes for articles in general or did that require being logged in?
Hmm, looking into it myself, it looks not generalizable, short of searching for the document title and hoping you get lucky. This wasn’t a direct paywall bypass.

It looks like this is actually just a version of the document published a different way.

Posted to HN from IEEE GLOBECOM 2008 - 2008 IEEE Global Telecommunications Conference:

https://ieeexplore.ieee.org/abstract/document/4698465

Same thing but free from IEEE Communications Magazine, vol. 47, no. 11:

https://ieeexplore.ieee.org/abstract/document/5307460

Person I replied to clicked the PDF button on the freely-available version.

I think I remember IEEE making everything available a few years ago…there’s probably an HN discussion if it happened (but it might have been ACM (or both)). If so, this would be an ordinary dark pattern of extracting payment where it is not absolutely necessary.

Anyway, I googled the title based on my possibly accurate memory.

In fairness my impression is that very few HN members ’give back’ by joining such organizations. It’s not part of today’s startup mindset.

ACM is the one pushing towards open access for everything. IEEE is not as far along with that. It gets really annoying, I'm a member of a couple IEEE societies and find that I have to use my institutional login (GA Tech in my case) to reliably get access to their publications. With ACM, I join a SIG and I pretty much get the entirety of the SIG's digital library (even before the bigger open access push).
Orthogonality is a beautiful concept.

In physics, different frequencies don’t interfere with each other—they're always independent. However, in digital operations and mathematics, this isn’t automatically true, so we need to carefully select the right frequencies.

OFDM leverages this by ensuring that, in the frequency domain, certain frequencies remain completely separate, meaning changes to one frequency (in amplitude or phase) won't affect the others.

This allows us to capture the maximum information with limited samples.

> In physics, different frequencies don’t interfere with each other—they're always independent.

Are you overlooking harmonics? Are you overlooking the same frequency case of noise cancellation, via phase shift?

Phase shift isn’t different frequencies?
Please check out the 3rd paragraph of this article.

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

I’m aware of how ANC works. It uses the same frequencies with reverse amplitude. The comment you objected to however was talking about different frequencies.
> reverse amplitude

Right? One of my dumb noobie questions has always been why textbooks explain ANC as "180 phase shift" instead of "reverse amplitude". Do they really do an entire DSP pass to break apart a signal and phase-shift it just to get the output that a minus sign could get? There must be something missing from the plum-pudding model.

It's only a sign flip when the detector and emitter are perfectly in phase. Adding delay causes the required phase shift will drift slightly. Also note that the phase offset of a delay is frequency dependent, so this drift will be as well.

Another reason is that devices and materials have different frequency responses, so an external signal needs to be filtered to match the levels in the device.

I wasn't clear : phase shift has a clear example: think of two sine waves, one whose value at time t is sin(t) and the other sin(t+pi/2). they'd have the same period, and frequency, but would be phase shifted by that addition in the time parameter.

Here's a graphed example.

https://www.wolframalpha.com/input?i=Plot%5BSin%5Bt%5D%2CSin...

For another fun example, here's the residential power in current American homes: two 120v A/C feeds, out of phase so that you can combine them in the breaker box to get 240v A/C.

https://www.wolframalpha.com/input?i=Plot%5B120*Sin%5Bt%5D%2...

You're correct - I oversimplified assuming no doppler effect and a linear time invariant system.

The key is that we can still analyze and manipulate frequency components independently in REAL PHYSICAL WORLD, which enables techniques like OFDM.

Sounds a little bit like CDMA. My professor in college didn't make it obvious why CDMA is any better than just assigning one carrier frequency to each transmitter. [1]

The Wikipedia page for OFDM [2] says something like, if there's interference or attenuation affecting one frequency, it's better to spread that out instead of completely losing one channel. Which kinda makes sense, and would make more sense if I'd ever done RF work. I'd love to do SDR as a hobby some day.

[1] The explanation in the book was something like, "See with 100 frequencies, you can create 100 chips and each user gets 1 chip, so you can support 100 users" which sounded pointless to me.

[2] https://en.wikipedia.org/wiki/Orthogonal_frequency-division_...

Any spread-spectrum technology has advantages when there is fading only some frequencies (a normal thing to happen in cities with these large surfaces that reflect RF called "buildings").

CDMA has an added advantage that the bit-error-rate is inversely proportional to the number of people communicating. With just giving everybody a single channel (or timeslot for TDMA), you either have enough or you don't. When you give everybody their own code, then you end up with either a small number of very high quality channels (few people communicating) or a large number of much lower quality channels (many people communicating). This was seen as advantageous when cell networks were rapidly growing.

I don't think it's used much for terrestrial anymore, but it is still used in satellites.