Talk:Anti-aliasing filter

Jargon cleanup

The article contains an excess of terms most people would not understand if they were to come across this article.

Take the anti-aliasing article for comparison. It explains the terms and concepts in detail. This article on the other hand expects you to know all of these terms and concepts, rather than explain them.

Help and feedback would be appreciated.

-Omnicloud (talk) 17:24, 30 June 2008 (UTC)

I agree. I sure don't understand this! Uber-Awesomeness (talk) 23:01, 24 January 2009 (UTC)

I agree with both sides on this, however, this is a technical area - and naturally has to contain jargon. Personally I think the article needs very little remedial work, but does require expansion. Bloodholds (talk) 20:25, 24 May 2009 (UTC)

Incomprehensible sentence: "A toggle can on-off anti-aliasing filter as the world's first camera which has its capability." 89.217.21.82 (talk) 17:41, 17 March 2014 (UTC)

Is it misleading to say "sharply cut off aliasing"? wouldn't reducing aliasing be the opposite of sharpening?Goatonastik (talk) 23:28, 23 October 2014 (UTC)

Sharply cutting off is not related to sharpening. Dicklyon (talk) 06:10, 24 October 2014 (UTC)

Optical Anti-Aliasing

In the section on Optical anti-aliasing filters, it says that "fill factor alone ... can provide a significant anti-aliasing effect" Isn't that false? Don't you need an AA filter of some kind to eliminate signal that's higher frequency than the pixel pitch, regardless of whether the pixels are 100% coverage? —Preceding unsigned comment added by 152.2.202.114 (talk) 20:50, 15 May 2009 (UTC)

No, it's quite true; in fact, the filtering effect of a 100% fill factor is better than what is achieved by a standard optical anti-aliasing filter, which is why the cited source shows a higher quality image depends more on a large fill factor than on the optical anti-aliasing filter. You can understand this by working out the transfer functions. With a 100% fill sensor, the sinc function transfer function has its first zero at the frequency that would alias to DC, so it does a good job of attenuating large-scale (low-frequency) aliasing. With the standard optical anti-aliasing filter that splits a point into two points, the transfer function is a cosine; it can be arranged to put the first null at whatever frequency you want, but then it quickly comes back up to a big peak again, as opposed to the sinc whose first sidelobe is much further down. I added some more sources. Dicklyon (talk) 03:21, 27 July 2009 (UTC)

The real problem with image sensors, though, is the Bayer filter array, such that aliasing gives color fringing. The article mentions the problem with Bayer filter arrays, but doesn't explain why. Gah4 (talk) 06:50, 4 August 2022 (UTC)

Article Clean up

The grammar in this article was really poor. I went through and fixed a lot of it as well as consolidated some of the language. Still needs a lot of work though. 18.62.28.79 (talk) 00:11, 10 January 2015 (UTC)

Other possible ideas for improvement:

1) Combine audio and optical into an examples section. 2) Section on systems where an AA filter is not used (for example diffraction limited optical, some RF)

Good idea or too specific? 18.62.28.79 (talk) 00:29, 10 January 2015 (UTC)

Still not clear

I had trouble understanding the article. Optical filters are made of glass, and I think this filter is electronic. It would be helpful if someone showed exactly what this type filter does in some non-technical manner. Input and output comparisons, not how it does the task, but what it does, maybe some images. The lead paragraph is almost unintelligible to someone not in the field. It should be be prefaced with a general, non-technical paragraph or two before it gets into the technical material. I hope someone can do this. — Preceding unsigned comment added by Nantucketbob (talk • contribs) 12:49, 6 February 2015 (UTC)

It is an optical filter, though not made of glass. It uses birefringent materials, which have a different index of refraction for different polarizations. You might remember calcite, which has been a favorite for many years, though it seems that lithium niobate is more usual. One filter will generate two spots, one shifted horizontally. You need another one to shift vertically, but the result from the first one is polarized. Usual is a quarter wave plate to convert to circular polarization, so the second one will work. A point source object will make four points on the image, spaced depending on the pixel size of the sensor. Gah4 (talk) 07:03, 4 August 2022 (UTC)

Optical low pass

Spelt as two words by:

• Canon (with a hyphen)
• Nikon
• Red (again, hyphen)
• AbelCine
• Sony
• Arri

--Shivertimbers433 (talk) 05:53, 9 April 2015 (UTC)

Optical low-pass filter is grammatically correct so that's what we go with here. ~Kvng (talk) 16:56, 6 May 2022 (UTC)

Sampling rate?

In the "Bandpass signals" section there's an example of sampling FM radio, stating that "and the sampling rate would be no less than 400 kHz". I'm no expert, but shouldn't that be 200 kHz with IQ sampling? https://dsp.stackexchange.com/questions/672/complex-sampling-can-break-nyquist states that to sample complex signals from -100 kHz to 100 kHz it's enough to sample at 200 ksps. 178.164.216.62 (talk) 05:15, 24 January 2024 (UTC)

The section doesn't mention IQ sampling. It depends some on how much filtering you put in, though. As well as I know, they put a mixer before the ADC, as it is easy enough to do in CMOS. There is the GS1299, FM radio in one CMOS IC from antenna input to headphone output. I believe it uses a low (compared to 10.7MHz) IF before the ADC. Then the stereo decoder is completely digital. I tried to find the patent for it, but didn't find it yet. But yes, if you do IQ sampling, then you have a lower sampling frequency. But you get two samples each time, so the actual rate (samples/s) is the same. Gah4 (talk) 17:06, 24 January 2024 (UTC)

OK, I found this one: RDA5807FP which has a better description of the circuit. And it might not be the exact one, but this patent is by the same company. As you note, it does IQ sampling. The local oscillator outputs + and - for I and Q, and the LNA outputs + and -, so all it needs is some analog switching. Then there are I and Q ADCs, and input to the digital FM demodulator and stereo demultiplexer. Decoded audio to two DAC, and output amplifiers to drive 32 ohm headphones. Also, the headphone cable is the antenna! Do note, though, that the mixers are before the ADC. I don't know if this qualifies as a WP:RS, though. Gah4 (talk) 19:21, 24 January 2024 (UTC)