Wikipedia:Reference desk/Archives/Computing/2011 March 6

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

March 6

Question about Dell's early days

I read somewhere that Michael Dell at one point was making $80 grand a month while he was still a pre med student making computers in his dorm room. I have serious difficulty seeing how that's physically possible, but that's not why I'm posting here. I read that he got his materials to start with from an IBM supply house that was obligated to order in large quantities and he got them to unload their looming inventory to him at cost. Getting them to do that for him must have been a result of his persuasion skills. But what I want to know is a little more about whatever extremely fortuitous circumstances he found to get so many buyers. I mean I know the lower prices he was able to give gave him a huge advantage, but getting enough people to make that kind of business? Maybe it's just because I'm a super pessimist, but I'm pretty sure if no-name me put out ads advertising a computer for 20 or even 30 percent less than the well-known vendors, a large proportion of people would ignore me, assuming I was a shyster, and go with the known vendor. Thanks.76.27.175.80 (talk) 01:50, 6 March 2011 (UTC)

As someone responsible for buying PCs in sizeable numbers back in the late 1980s and early 1990s, I can assure you that when I told my boss I could buy straight from IBM at $6,000 a unit, or "IBM compatible clones" at $3,000 per unit, it was clones all the way. But it did depend on the person responsible for the money. When a more conservative accountant took over we did have a sudden switch to the ("nobody ever got fired for buying...") IBMs. HiLo48 (talk) 02:34, 6 March 2011 (UTC)

Metasearch blogs for topic x

I want to keep updated about what the blogophere says about topic x, is there a program that would go through blogs and show me new entries for this topic? Eventually, would it aggregate all results in one page?

Technorati[1] or Google Blog Search[2] might help. --Colapeninsula (talk) 14:54, 7 March 2011 (UTC)

Mixcraft Robotic Voice

I have mixcraft on my computer, and am interested in knowing how to make a voice sound like a robot without having to download more effects because I have limited internet access. The best example of what I'm trying to do is the effect Chris Qualls from Electric Valentine (the male) sometimes adds to his voice. Another example would be like the voice in the song 26 Basslines by Benga. If you are unfamiliar with any of these examples (you probably have to wait till around the end of most of these songs to see what i mean): http://www.youtube.com/watch?v=WD4yRSUn7uc —Preceding unsigned comment added by 75.8.107.139 (talk) 05:01, 6 March 2011 (UTC) http://www.youtube.com/watch?v=16Kx13S0-0g&feature=related http://www.youtube.com/watch?v=1wSVL8Hqqv8 http://www.youtube.com/watch?v=2_5ufrHgtrI&feature=related http://www.youtube.com/watch?v=OQZ3NPtzyqY http://www.youtube.com/watch?v=Ri_5HdHGGCg http://www.youtube.com/watch?v=p4D-hCO8Y2k

P.S. - Please no unnecassary comments on your opinion of the music (good or bad), the songs are mearly meant as an example of the effect I'm looking for. —Preceding unsigned comment added by 75.8.107.139 (talk) 04:03, 6 March 2011 (UTC)

A Google search for mixcraft robotic produces many results. Robotic voice effects may be of interest. 92.15.18.16 (talk) 12:03, 6 March 2011 (UTC)

I think the "robotic voice" that OP is referring to is autotune (or "pitch correction") effect, applied with high intensity to a human voice. Auto-tune is often used this way, but it can make other sounds that are less "robotic" - it is a general-purpose and highly-configurable audio processing step. From a signal-processing standpoint, all of these are special cases of the phase vocoder (the core effect that non-linearly modulates phase and frequency of a voice audio signal). Nimur (talk) 19:53, 6 March 2011 (UTC)

web cache

I would like to use a web cache to reduce bandwidth on Windows 7, but I am having difficulty finding one which is easy to use. I tried Polipo but the Windows version lacks some of the features and kept crashing after ~10 mins. I tried Squid (software) but it's extremely confusing, requires extensive setting up, and according to their site the Windows version lacks many features. I tried nginx as a forward proxy but the Windows version doesn't support cache features. Are there any other free web cache software which works on Windows which I can try? It would be especially good if it was portable and could be used on any computer via usb drive without spending ages setting up. Thank you 82.43.92.41 (talk) 11:19, 6 March 2011 (UTC)

Squid really is the best choice - because it's free, stable, configurable, and widespread, it has the best documentation and support-tools. Maybe you need help finding better install instructions? The official Windows distribution has a one-click binary installation tool; and documentation is available. Setting up sophisticated caching behavior is hard, but that's because the process is inherently complicated and there are thousands of options. Pre-configured Squid examples may fit your needs, and you can customize the configuration options as you need.

If you want, you can run a local instance of the proxy server, by putting an installed version of Squid (or any other proxy server) on the portable media. But this is not even necessary. Instead, set up the cache/proxy server once; and make it accessible on the network; then simply set each client to use the pre-configured proxy/cache that is running remotely on the network. For example, configuring Firefox to use a proxy-server takes a few seconds following the official instructions. Even if you manage to run a proxy-server on a thumb-drive, you can't change the local machine's network client settings just by plugging in a USB stick. Each client must be configured individually (and you need administrative access to the client in order to set up proxy and network settings). Nimur (talk) 15:51, 6 March 2011 (UTC)

Name of an old Windows game

I remember back in the 90's, when I used Windows 98, there was a game included with the OS (at least I played it on that OS and I conjecture it was part of the OS or part of some game package for that OS). In that game, you played a little spider. You were on a field and on that field a snake was creeping around. Now as you moved the spider across the field, it would leave a string behind. When the string enclosed a rectangle with the outer boundaries, that rectangle would become grey and the snakes would no longer be able to move into that area. The goal of the game was to capture a specific percentage of the field of play from the snakes. Does anyone know the name of this game and if it was part of the OS or part of any game package and which? Would appreciate any help. Toshio Yamaguchi (talk) 15:03, 6 March 2011 (UTC)

I gather from your description that this is some kind of clone of the classic arcade game Qix, but I can't identify the exact game. JIP | Talk 18:29, 6 March 2011 (UTC)

A quick search finds JezzBall but it doesn't appear to involve snakes or spiders. Nil Einne (talk) 21:21, 6 March 2011 (UTC)

Volfied looks pretty close... --Wirbelwindヴィルヴェルヴィント (talk) 02:45, 7 March 2011 (UTC)

Is it one of these ganmes on this website? http://www.anetforums.com/posts.aspx?ThreadIndex=29061 General Rommel (talk) 10:20, 7 March 2011 (UTC)

Ok, first of all thanks for your help. The game I have in mind seems to be a remake of Qix. The principle of play is exactly the same. The difference is (as mentioned) the graphics looked more modern, the 'marker' controlled by the player was a small spider and the enemy was a greenish / yellowish snake. The field of play was pink. Toshio Yamaguchi (talk) 11:25, 7 March 2011 (UTC)

There's a game a lot like your description on FreeArcade, but I'm at work and it's blocked, so I can't tell you the name of it. Kingsfold (Quack quack!) 19:31, 7 March 2011 (UTC)

I'm fairly sure it's not Volfied. The "marker" in Volfied is not a spider but a small spaceship, and the enemy is not a snake but various cybernetic alien monsters, depending on the level, and the field of play is a 1990s-style sci-fi landscape, not pink. Unfortunately, I still don't know what the exact game is. JIP | Talk 19:41, 7 March 2011 (UTC)

Actual pixel contents of RAW image files

What does a digital camera's RAW image file actually contain in the actual pixel data? RAW file format#Sensor image data says that there is 12- or 14-bit RGB intensity data for each pixel. Is this in essence the same thing as the 8-bit RGB intensity data in a normal bitmap image file, only 16 to 64 times more accurate? Are there separate intensity values for the individual R, G and B subpixels instead of an overall RGB intensity for the whole pixel? Is there any other pixel-by-pixel data? JIP | Talk 18:28, 6 March 2011 (UTC)

The term RAW does not refer to a particular format, just to the fact that many cameras store data internally in idiosyncratic ways that do not correspond to any formally specified file type. So the answer is, any of these things are possible, and many others as well. Specific answers can only be given for a specific type of camera. Looie496 (talk) 19:12, 6 March 2011 (UTC)

Well, I am mostly interested in the raw image formats of my own DLSR camera (Olympus E-520) and my mother's DSLR camera (Canon EOS 5D Mark II, costing over four times as much). JIP | Talk 19:28, 6 March 2011 (UTC)

Raw data should also include detailed parameters for the camera acquisition details - ugly low-level details such as timing variations in the sensor analog-gain for hardware auto-exposure; field readout order, if the data came from a CCD sensor; between one and four optical black data areas, lens control parameters; vibration compensation parameters; defective pixel masks; long exposure pattern noise image data; light metering; auto-focus sensor details, for SLR cameras that have a physically separate AF optical device; accelerometer, gyro, lens stabilization, and other motor parameters, and so on. Details of each camera manufacturer's RAW format are proprietary. Some manufacturers compress the data (losslessly, usually). Some manufacturers encrypt some or all of the raw optical- or meta-data. dcraw, a free and open-source software implementation that decodes most major brands and models' RAW formats, has detailed technical descriptions - probably the best documentation is dcraw.c. This is a program source-file: in lieu of a 400-page manual, the inline comments explain the specific bit-orders for some 400 variants of RAW file format.

For example, your Olympus E-520 packs the entire file in a huffman coded container; it uses a 29-bit-shifted storage format for what appears to be unsigned-integer pixel levels; it uses some magic-number gamma correction lookup tables - for example:

{ "OLYMPUS E-520", 0, 0xfd2,	{ 8344,-2322,-1020,-7596,15635,2048,-1748,2269,7287 } },

Those numbers aren't in the RAW file, but you need them to convert a RAW image into a proper digital photograph.

On the other hand, the Canon 5-D can support more sophisticated matrix metering, and stores that AE matrix in the RAW file. (Even still, it also needs some magic numbers for its gamma correction and other stages of "raw" to "JPG" conversion).

The moral of the story: the RAW container is more than just uncompressed pixel-data in bayer sensor format. It contains everything that the camera's digital electronics "know" about the photography operation - everything that would get stuffed into an EXIF tag, plus a lot of extra information about the camera's hardware and software configuration, the scene/environmental conditions, everything about the sensor, lens, optical-, and electronic data paths that the camera is programmed to know and compensate for. Nimur (talk) 19:43, 6 March 2011 (UTC)

Yes, I know that a raw image file contains more than the pixel-by-pixel intensity data. I know that it contains things like exposure time, aperture value, ISO speed rating, focal length, colour temperature, date and time of exposure, internal details about the camera's firmware, and much, much more. My question was specifically about the pixel-by-pixel data. After all, the EXIF data itself cannot describe the actual image, only the conditions under which it was made. JIP | Talk 19:57, 6 March 2011 (UTC)

Okay. So then the answer is, "depending on your sensor", the exact RAW pixel-by-pixel data will be a bayer mask of data (roughly, "two greens and one red and one blue value" for each pixel. Each of these element values are n-bit unsigned integers (n usually ranging from 8 to 16 for modern camera sensors). These numbers represent a nonlinearly-scaled, position-dependent light-intensity value; a lens model must be applied to account for optical geometry; and a sensor gamma must be applied to account for sensor characteristics; and an exposure algorithm must be applied to account for time-variation of pixel amplitudes during read-out. Each pixel column- and row- will also contain a set of optical black pixels, who are unexposed to light, with the intent to meter sensor analog electronic noise levels for black level correction; this must be combined with the bayer-mask before gamma-correction to shift the intensity-scale. Following those steps, you will have "clean" bayer data, which can be demosaiced into RGB or to the color-space of your choosing; following this stage, you can use meta-data or user-selected parameters to apply contrast correction, white-balance, and any other image manipulation you like. The most trivial algorithm would assume everything is "linear" and simply read the Bayer data, discard one green pixel, and pretend that the 8 most significant bits of each pixel are "RGB" data - which will work, but the image will look very crude. At the end of the day, a camera sensor is nothing but an array of analog to digital converters; the hard part is that physical light intensity does not cleanly correspond to voltage, and therefore the raw bits representing voltage data need a lot of pre-conditioning to make a good image. Nimur (talk) 20:24, 6 March 2011 (UTC)

The photon intensity to voltage is the more linear, it is instead the F stops, shutter and ISO that doubles at each step and hence the need to convert the values to match that progression. --Aspro (talk) 21:49, 6 March 2011 (UTC)

Putting it another way. The RAW file does not contain the actual pixel data. Rather it contains a unit value representing 3 pixels worth (one for each colour) recorded by the image sensor. A 12 bit image has a 'depth' of 2^12 = 4096 discrete levels of information compared to the 2^8 = 256 discrete levels of information in an 8 bit image. However, (and this is where it can be confusing) as the 8 bit represents 3 pixel worth of information, it can reproduce 256*256*256 = 16,777,216 values. Also, to be pedantic but necessary to answer your question fully, the brightness values recorded by the camera's image sensor are converted into a logarithmic value so that more information can be stored in lighter tones and less in the darker parts of the image. This is why the RAW files always looks a bit darker – in other words, what you see is a compressed image file. So it isn't RGB like a jpg, its RAW.--Aspro (talk) 21:24, 6 March 2011 (UTC)

JPEG is not RGB; it is invariably a luma-chroma colorspace - YUV, and typically with chroma subsampling. That means that a JPEG does not contain color-information for every pixel - unlike RGB data or Bayer RAW data. If you open a "JPG" file in a tool like Photoshop, you can extract an "effective" RGB value for each pixel, but when you use a tool like Photoshop to edit a JPG file, you are actually working with a rasterized, decoded image layer that is no longer a JPG. See color space in JPEG for more details. There are a lot of critical concepts here that should not be mixed up: color-space transformation (Bayer to Y-Cb-Cr to eventually RGB for screen rasterization); chroma subsampling, usually during the pre-processing for JPEG compression; gamma correction and dynamic-range reduction (as the 12-bits get shoved into an 8-bit format); and lossy quantization, which is actually part of the JPEG algorithm. Every one of these steps can be either "lossy" or "lossless," depending on the image signal processing flow. When you dump RAW pixel data, you have control over each of those independently. If you use a professional-grade image processing software, you can micromanage that data-flow. Most people do not want to go into this depth with their RAWs - usually, people just want to use RAW because they feel that they can extract all 12 bits of data for customizing their photographic contrasts and color outputs. Nimur (talk) 22:12, 6 March 2011 (UTC)

It was in an attempt to draw a comparison to the "effective" RGB value that I phrased it like that. It went through my mind that he may be less familiar with tiff etc - I was looking form something he could relate to. The way the OP's states his question suggest to me that he would appreciate a simple explanation that directly answers his question (although I could be wrong about that).--Aspro (talk) 22:37, 6 March 2011 (UTC)

Yes, but raw data is not simple - so a simple explanation would be incomplete! Fundamentally, raw data is more complicated than an RGB bitmap. A simplified view would be to treat the Bayer image as an GRGB bitmap, but that would be insufficient, and directly displaying the file as such would look crummy. Sorry if my explanations have been too technical; I work with camera raw-data on a very low level, and I consider these details to be important. Nimur (talk) 00:43, 7 March 2011 (UTC)

Doing mirror writing in Microsoft Word?

Is it possible? I did it once before, ages ago, but I forget how; and I think now I've got Windows XP it's become impossible, but I might be wrong - I asked this once before, but as far as I can recall, the only way to do it is to make a picture of the writing and flip it, then paste it back into the document. I remember this made for a smeary distracting mess of the words, didn't work at all. Has anything changed in recent times - can anyone help? Thanks Adambrowne666 (talk) 19:55, 6 March 2011 (UTC)

Does Wordart still exist in your version? Graeme Bartlett (talk) 20:47, 6 March 2011 (UTC)

Wordart still exists in Office 2010 General Rommel (talk) 10:13, 7 March 2011 (UTC)

Yes, I do have Wordart - I've opened it in my document ... I've got a feeling I used Wordart the second time I mentioned, ended up with the smeary distracting mess - but maybe I was doing it wrong? Adambrowne666 (talk) 10:58, 7 March 2011 (UTC)

When you say "make a picture of it", are you referring to taking a screenshot? If so, it is not complicated. In Windows, you can press the print screen key to copy the desktop. Then, you can paste that back into Word. I suggest pasting it into Paint so you can crop out the stuff around your text. In either Paint or Word, flip the image. The problem is that the image will not scale. It will be blurry or pixelized. With Word Art (as suggested), you type your text in word art. Then, you can select Flip Horizontal (which I believe is under Rotate in the Word Art menu). Finally, if this is just a matter of printing it reversed, most printer dialogs have a "mirror" option in the print layout section of the printer properties when you are printing. You don't reverse the text in your document, it just prints reversed. -- kainaw™ 19:58, 7 March 2011 (UTC)

Beauty, thanks, Kainaw and others - will check it out. Adambrowne666 (talk) 00:58, 10 March 2011 (UTC)