Hacker News Reader

Comment by dexwiz | Hacker News Reader

Can someone explain why it takes so long to form an image? The article reads like they still don't have enough data to see anything. From my understand of physical measurement, you start with a `blurry` image (or a graph with a wide peak), and it slowly comes into focus with more data. If the image is too faint because of the signal to instrument precision (noise) ratio is too low, you will never get an image, no matter how much data you have. I was bad at analytical chemistry, so my understanding may be completely wrong.

View on HN

> "If the image is too faint because of the signal to instrument precision (noise) ratio is too low, you will never get an image..."

I'm curious why this makes sense to you? Does it make sense that you could use an ordinary film camera and take a picture with an arbitrarily short exposure, say a picosecond?

With very, very low SNR it can take a long time to build up data. The reason why this is possible is because the real data, the signal, is correlated with itself, it builds up consistently with each addition of measurements. Noise, however, is uncorrelated, it's random, so it tends to evenly cancel (or even) itself out with additional measurements.

If you're talking about making a map of the astrophysical or cosmological backgrounds of gravitational waves -- you're right. As you integrate (i.e. capture more data), the noise will tend to cancel out while the signal will reinforce itself. But the problem is that the signal-to-noise-ratio is so small -- essentially zero -- that, with current detector sensitivity, it would take literally forever to get an image with any features at all.

> The article reads like they still don't have enough data to see anything.

That's because whatever signal is out there is quite a bit fainter than the initial estimates predicted it would be. If the signal were as strong as the initial estimates predicted, it would have been seen by now.

Reply on news.ycombinator.com