Slashdot Mirror

Sort of, but not entirely. He's redrawn the London Underground in a "circular" style (scroll to the bottom) which *isn't* the same as the well-known one.

Well, human hearing, not including intensity levels beyond threshold of pain, covers 120 dB. Moreover, narrowband signals are perceptible when several dB _below_ broadband noise. This makes me say that audio is not quite a solved problem even on the electronics side (let alone on the speakers side). Various things seem to pop up, and not just from crazed audiophiles. Recently I was reading how it turned out that some amplifiers who were specced at very low THD only made that measurement at moderate signal levels, and had very large distortion for low signals (one ironic example was the hugely overpriced Halcro amplifiers). The thermal memory issue may be another thing.
I did an informal ABC/HR test a year and a half ago with four people, which was blind but not doubleblind as the switching was done by telling another person to switch. It was with the DAC I mentioned. I had two IV's which I could switch in after the DAC using a small relay. One was using the AD797. The other one was derived from Hawksford's discrete current feedback IV (figure 4-4 in http://www.essex.ac.uk/csee/research/audio_lab/malcolmspubdocs/C111%20Current%20steering%20transimpedance%20amplifier.pdf ). To get the distortion of the Hawksford IV near -120 dB (checked with rented distortion analyzer), we added extra gain so we could use more feedback. For testing, we tested each person one at a time using Stax electrostatic headphones fed by this hybrid transistor/tube amp, which is about 5 ppm THD: http://gilmore2.chem.northwestern.edu/images5/gilmore4_1.png
We did eight trials for each person. One couldn't tell the reference (5/8). The other three could tell it (averaging 7/8).
Did another test. There were two of the Hawksford IV per channel, set up as differential since the DAC and the amp were both balanced. Connecting the emitters of the current mirror transistors between the two separate IVs dropped the differential output 3rd and 4th harmonics further (slightly raised 2nd for single ended operation) (I'm pretty sure this infringes on a Nelson Pass patent...). I used a relay for that as well, and did the test switching that connection on and off. Two people could tell the difference (averaging 6.5/8).
Small sample size, not double blind, and ultimately anecdotal. Nonetheless it was sufficient to convince me that the audio hobby has more mysteries to reveal than seems at first.

For those interested, the IEEE Conference on Computational Intelligence and Games is being hosted by the IT University of Copenhagen, and Mario AI isn't the only competition. There's also simulated car racing, Ms. Pac-man, 2k Bot Prize and the Starcraft RTS AI Competition (!!!).

I think you should read this:

http://www.iser.essex.ac.uk/files/conferences/epunet/2005/docs/pdf/papers/brunello.pdf

One of the things stated in the paper is that when the labour market is very easy (everyone can pack up and move, and everyone will always accept a new job) then general training is very expensive to provide for an employer. The consequence being that only company-specific training is given, but regular training rarely if at all. See China for a good example.

However, when people change jobs a bit less, there is less risk for the employer to lose his investment. Apparently, the Dutch labour market is more flexible than the UK one, so employers demanded a bit more assurance on this item. This seems to be supported by other observations on the labour market as well.

I guess that comeuppance never really materialized? You thought you were "smarter" than your more popular tormenters, and that one day, you'd show them- you'd study hard for that computer science degree and make a million-billion dollars, while Suzy the homecoming queen wound up working at Denny's and those mean jocks wound up in prison or working as auto mechanics. Guess it didn't work out that way- instead you've been working at Best Buy and languishing in some junior college for years, while those popular kids you loved to hate have already graduated college and actually started their lives, leaving you in the dust. It's unfair, but there's nothing you can do about it.

http://www.iser.essex.ac.uk/news/2009/03/04/popular-kids-earn-more

Popular kids earn more

Being popular with your peers at school could mean you earn more as an adult. That's according to new research by a team at ISER.

The research, which used American data from the Wisconsin Longitudinal Study, found a clear link between a student's popularity and their level of earnings later in life. As part of a wider long-term or longitudinal survey, High School students in the area were asked about their friendships, and then, 35 years later, how much they were earning.

The study asked students to nominate up to three best friends from their class. When analysing the data, the ISER team deemed those students who received high numbers of nominations as most popular. Those students who gave higher numbers of nominations were deemed more gregarious or out going.

The research showed that being gregarious had no effect on the students' earnings later on, while being popular did. Every extra friend nomination was associated with a two per cent higher wage, and there was a 10 per cent earnings difference between the bottom fifth and top fifth of the popularity range.

Commenting on the research findings, Professor Steve Pudney said the work emphasized the critical importance of the early development of social skills alongside learning as a basis for economic success in adult life.

The research also found evidence that the early family environment and the type and size of school play a significant role in shaping friendship networks.

I've done research and writing at federal institutions, private and state universities and commercial concerns, collaborated with people and labs in a dozen or so countries, and submitted to journals in several different fields. Never once did I hear LaTex mentioned as something available to write with or as a format acceptable for manuscript submission. I happen to be familiar with LaTex due to years of Linux tinkering, and from working with people who also happened to be at least modestly capable with it. Even so I'd use something that didn't require concern with command/control syntax. My brain is better used on the science and language syntax.

Microsoft Word can track changes according to collaborator. A particular format need only be created once, then saved as a template, many of which are available for download. There are various referencing packages that merge well with Word. I have run across other researchers who preferred something else for writing, but never have I run across one who did not have Word available or was not adequately familiar with it.

Perhaps there are fields I've not worked in that allow use of LaTex for writing and submission. I'll bet there are none that require it, and Word is acceptable to most if not all.

http://www.essex.ac.uk/linguistics/clmt/latex4ling/journals/ is a short article listing LaTex friendly journals. I disagree with the assessments about Springer and Elsevier, as every one of their journals I've written for did not list LaTex as acceptable. That leaves a very short list of journals that do accept it (and two major publishers that do not accept it). The list is a lot shorter than just the list of >35,000 journals referenced by NIH/National Library of Medicine's PubMed, the database I'm most familiar with.

Mod me down if you must for dropping the *nix flag and waving the enemy's, but these are the observations of a trained observer.

As a former High School InfoTech student and current College Programming Student, I really don't find textbooks that useful at all. Truthfully, the only use I ever get out of textbooks (other than reading the questions the teacher's assign) is reading the examples and the using the reference section.

Not only do examples and references exist on the web, but it is SO much easier to use a reference with hyperlinks than to have to jump between pages of a book

If you really need some good ideas I have a list of resourses:
- CodeSyntax - Basis syntax for Java,C,Python,etc
- JavaBat - different levels of Java puzzles (ajax handles compiling/etc, no software required)
- Eddie's Basic Guide to C Programming
- ANSI Dictionary - unbelievably nice ANSI dictionary, fully cross-referenced.

Consider setting up a wiki-book full of information, labs, excersies and tutorials. This is a computer class after all and information should be easy to find without needing to pack yet ANOTHER heavy book around. To make your job easier, you could allow the students to add stuff to the wiki (log activity of course), even setting up a page where they can add useful websites they've found.

The choice of font makes a huge difference to readability. Professor Arnold Wilkins (U. Essex,UK) has conducted extensive studies on the 'hygiene of reading'. his recommended font is Verdana with Times and Courier among the worst. (You want to avoid fonts with high autocorrelations.. where all the letters are very similiar looking and so it's difficult to keep track on the right place in the text). He also recommends using lower contrast and avoiding flicker as best you can:

"We read under light that is too bright and that flickers continually but imperceptibly, disturbing eye movements and visual search. We read text that is too small with fonts that are too striped, impairing reading speed. We read from displays that direct glare into the eyes, causing headaches. There are simple steps we can take to make our reading more hygenic."

http://www.essex.ac.uk/psychology/psy/people/wilkins/wilkins.html

Actually what you said is "you're not going to get harmonic distortion" which is the opposite of reality. Here's the peer-reviewed watershed work on jitter and harmonic distortion, published in the Journal of the AES. Educate yourself. http://www.essex.ac.uk/ESE/research/audio_lab/malcolmspubdocs/C44%20DAC%20Transisiton%20distortion,%20jitter,%20slew%20rate.pdf

> However, the inkjets (which, by the way, are very far from perfect) are quite uncorrelated from the matter at hand.

Heh. Actually they're very relevant to the matter at hand. I've tried to explain why briefly below.

> It isn't as if they just sat around a few decades and researched

Except that *is* what happened. (link, link) IBM did a lot of the early work in the early 1960s but never came up with something that'd appeal to mid- and small-business and consumers. Siemens made an early breakthrough here, and a thermal inkjet prototype was developed by HP in (IIRC) 1979. It was too primitive to bring to market, so HP continued to polish it until they introduced their first commercial inkjet in (again, IIRC) 1984. This wasn't a perfect printer, in fact according to Canon (which actually had more patents than HP on inkjet tech at this time) their engineers would "lose face" (warning: .doc) if they shipped such a lousy printer. The first inkjets didn't sell very well but HP continued to invest in them until they had 300 dpi resolution in about 1988, at which point they first tasted commercial success.

> No, it came in iterations, every few months or years a new model.

Iterative improvements came only around 1989-90, when sales started booming. Until then, imagine yourself an HP R&D manager and wonder if you would have greenlighted research that took almost a decade to get significant sales (on top of almost 15 years of prior research).

I've often said that /.-ers have a very warped view of innovation, coming no doubt from a world where all you need to build something new is a freely-available library here and some sample code there, and hey presto! new app! Innovation outside software (in hardware, in pharmaceuticals) is a good deal more complicated than that.

Jitter is actually a bigger problem than the filtering. Plus, 96 kHz... that means even higher clock frequency and more jitter sensitivity. It's a big issue, and the distortions produced when jitter (phase noise) gets in the DAC chip are complex and audible in the ppm, unlike the low order even harmonics of class A tube or MOSFET amps. A good technical paper on the nature of jitter-induced distortions in digital audio: http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C134%20Paper%20121st%20convention%20(c orrected).pdf

Class D amps have a long way to go. THD is meaningless, as the blind test studies by GedLee that were presented at the Audio Engineering Society convention a few years ago show, THD doesn't correlate with the distortion detectability, since the type of distortion is far more important. Crossover distortion from class B and AB stages, and effects of jitter, are audible in the parts per million. There's another type of distortion that doesn't affect THD measures at all, but is perceptually significant: thermal memory distortion. There's a good description of it and ways to decrease it here: http://peufeu.free.fr/audio/memory/ (there's also an AES paper linked there that describes how to measure it in real amps). Of course, tubes don't exhibit such distortion, and is my guess as to one of the reasons some people prefer them despite higher THD than typical solid state amps (however, this higher THD is simply due to most tube amps being simple; a tube with constant current load is more linear than any single solid state device; you can easily make a tube amp as linear as a solid state one if you use as many tubes as you would transistors).

In regards to vinyl, I do agree that overall CDs win. But your general argument on quality fails, because you forget that the human ear weighs different distortions differently (blind testing by GedLee presented at the AES a few years ago shows perception of distortion doesn't correlate with an unweighted metric such as THD). Low order even harmonics are undetectable until several percent. Crossover distortion from class B or AB output stages is detectable in the parts per million (and the presence of one distortion doesn't necessarily mask the detectability of another). The problem with digital is jitter (phase noise). Jitter in the I2S signal going into the DAC chip is converted to amplitude errors of a very complex nature, and are easily audible. There's a fairly technical but fascinating paper from the journal of the AES here http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C134%20Paper%20121st%20convention%20(c orrected).pdf
Worse, most DAC chips use high oversampling (usually 8x), which further increases sensitivity to jitter, as to avoid the need for a steep anti-imaging filter in the subsequent analog stages. There are other problems in a digital system, such as the digital filters in the resampling hardware, which are only partially alleviated by 24 bit 96 kHz systems.

Besides mixing and recording, most consumer digital gear has very poor jitter rejection. Jitter (phase noise) getting in the DAC creates amplitude errors of a very complex nature, and unlike some types of distortion (say the common low order even harmonics of a speaker or class A MOSFET or tube amp), is audible in the parts per million. A good paper from the journal of the Audio Engineering Society is http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C134%20Paper%20121st%20convention%20(c orrected).pdf On the same website, there's another AES paper describing the jitter problems added by the ubiquitous S/PDIF interface.

It's not just about high frequencies. Jitter (phase noise) getting into the DAC chip creates amplitude distortions that affect a wide range of frequencies. There have been enough papers in the journal of the Audio Engineering Society on jitter so I won't go into it any further; here's a good example: http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C134%20Paper%20121st%20convention%20(c orrected).pdf

I especially hate how lots of high end manufacturers use toroidal power transformers. These are convenient for they have smaller size, but they also have larger bandwidth, letting in all the noise from the house wiring get right through. EI cores = win!

There truly are, however, lots of problems with PC audio, though there are easy solutions. The PC is an extremely RF noisy environment. Though you can't hear high frequencies from interference getting into the audio circuitry, such frequency is often modulated, multiple frequencies form beat frequencies which can fall in the audio band, and most importantly, affect the electronics in two specific ways: they modulate device parameters and thus intermodulate with the audio signal, and by exceeding slew limits of the gain devices, they create effects akin to the well studied transient intermodulation.

The most common solution is to just use the digital out and an external DAC/receiver, but that has the problem of jitter added by the interface and poor jitter rejection in most modern DACs. If you check jitter specs in the datasheets from TI and Analog Devices, which produce the most commonly used DAC chips, this is immediately obvious. Even using an asynchronous resampling with say a CS8421, you only get some jitter attenuation. The real problem here is S/PDIF, because it is a synchronous interface and thus you're not just sending data but timing information which is analog in nature. Hawksford's paper years ago in the journal of the audio engineering society has more details on the severe problems with S/PDIF: http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C41%20SPDIF%20interface%20flawed.pdf

Toslink is not as good as a good coax (assuming you avoid ground loops) because the optical transmitter and receiver devices (they're all variations of the good old TORX/TOTX pair) have really high intrinsic jitter. The problem with coax is that, while the cable has characteristic impedance of 75 ohm and the source has 75 ohm output impedance and the receiver has 75 ohm output impedance, the RCA connectors do not (it's physically impossible for an RCA connector to have that). Since S/PDIF is serial, you're transmitting a bit at a time, and the total frequency is about a dozen or so MHz. So you get reflections at the RCA, and increased jitter. Jitter remains a problem even if your receiver has an asynchronous resampler before the DAC, sine those only have limited jitter attenuation. I've just scratched the surface about the problems of S/PDIF. Hawksford had a great paper way back in the journal of the AES about how flawed it is. http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C41%20SPDIF%20interface%20flawed.pdf

Pretty much. Here's the paper I mentioned; it's from 1992 and the standard has not changed since: http://www.essex.ac.uk/ese/research/audio_lab/malc olmspubdocs/C41%20SPDIF%20interface%20flawed.pdf

Thats being activly worked on

There is some reaserch going on using gumstix (tiny pc's running linux) to control flocks of flying things (small helicopters in this case)

Grid swarms

they are also trying to do this with larger outdoor RC planes "just need to find a safe place to do this" is there comment :-)

CDR? I'm concerned with the D/A process, specifically in systems with S/PDIF from transport to external DAC, where any jitter in the signal reaching the DAC, including that created by reflections due to impedance mismatches along the transmission line, is translated into distortion of the analog audio. Two journal papers containing a good overview are this one and this one.

I am aware of several instances where the glass mastering of CDs produced discs with so much jitter that the artists objected to the sound quality and held release until the issue was fixed. One is here; another artist with the problem was Steely Dan (Google it). Sony Music has gone through extensive attempts to reduce such jitter issues in their glass mastering. And none of these are concerning the effect of jitter whereby a block is missed, but the one where a far smaller amount that does not create data loss causes distortion in the audio by the mechanisms noted in the two papers I linked to (not to mention tons of others, as this is a very old issue).

Your mistake in replying was that you only considered the digital parts of the chain, and completely ignored the analog portion of any D/A conversion -- you forgot you're producing an analog signal in the end -- and any jitter reaching the DAC will create analog distortion.

CDR? I'm concerned with the D/A process, specifically in systems with S/PDIF from transport to external DAC, where any jitter in the signal reaching the DAC, including that created by reflections due to impedance mismatches along the transmission line, is translated into distortion of the analog audio. Two journal papers containing a good overview are this one and this one.

I am aware of several instances where the glass mastering of CDs produced discs with so much jitter that the artists objected to the sound quality and held release until the issue was fixed. One is here; another artist with the problem was Steely Dan (Google it). Sony Music has gone through extensive attempts to reduce such jitter issues in their glass mastering. And none of these are concerning the effect of jitter whereby a block is missed, but the one where a far smaller amount that does not create data loss causes distortion in the audio by the mechanisms noted in the two papers I linked to (not to mention tons of others, as this is a very old issue).

Your mistake in replying was that you only considered the digital parts of the chain, and completely ignored the analog portion of any D/A conversion -- you forgot you're producing an analog signal in the end -- and any jitter reaching the DAC will create analog distortion.

If you think swarms of little cute robot dogs are scary, have a look at what's going on in our lab over here:

http://cswww.essex.ac.uk/staff/owen/research.htm#T he%20Flying%20Gridswarm,%20and%20the%20UltraSwarm

The deactivation button might be there... if those who pay for the research want it there.

Philosophy has several roles wrt computer science (CS):

• CS going meta: To reflect on what is computer science (for example, is it a science, an engineering, or a branch of mathematics) and what is the scope of the discipline.
  
• Metaphysics & ontology: To reflect on the subject matter of computer science, specifically computer programs and computing machines, and to invesitgate their metaphysics (what are programs 'made of'?)
  
• Epistemology: To investigate the validity, applicability, and methodology of experiments using computer programs as simulations (astronomy, biology, chemistry, etc.)
  
• To pose questions arising from the discipline.

Papers, books, and conferences on the philosophy of computer science: http://pcs.essex.ac.uk/

Philosophy has several roles wrt computer science (CS):

• CS going meta: To reflect on what is computer science (for example, is it a science, an engineering, or a branch of mathematics) and what is the scope of the discipline.
  
• Metaphysics & ontology: To reflect on the subject matter of computer science, specifically computer programs and computing machines, and to invesitgate their metaphysics (what are programs 'made of'?)
  
• Epistemology: To investigate the validity, applicability, and methodology of experiments using computer programs as simulations (astronomy, biology, chemistry, etc.)
  
• To pose questions arising from the discipline.

Papers, books, and conferences on the philosophy of computer science: http://pcs.essex.ac.uk/

...the flesh is weak" comes out as "The meat is rotten, but the wine's great".

Seems like I have to repeat myself over and over again, since this is an urban legend. According to MACHINE TRANSLATION: An Introductory Guide:

The `spirit is willing' story is amusing, and it really is a pity that it is not true. However, like most MT `howlers' it is a fabrication. In fact, for the most part, they were in circulation long before any MT system could have produced them (variants of the `spirit is willing' example can be found in the American press as early as 1956, but sadly, there does not seem to have been an MT system in America which could translate from English into Russian until much more recently --- for sound strategic reasons, work in the USA had concentrated on the translation of Russian into English, not the other way round). Of course, there are real MT howlers. Two of the nicest are the translation of French avocat (`advocate', `lawyer' or `barrister') as avocado, and the translation of Les soldats sont dans le café as The soldiers are in the coffee. However, they are not as easy to find as the reader might think, and they certainly do not show that MT is useless.

BTW, since this book is no longer available in the stores, the whole contents is placed online. I recommend reading this book to anyone who is interested into the subject of MT. It really is a nice introduction into the subject.

Since when has Essex been in the US? Here is the guy doing the research and here is where he is located. Spot the blue stuff between the East coast of the US and Colchester? Not sure Google has mapped out Europe yet mind you ;-) O.

My alma mater had an 'Intelligent Autonomous Vehicles Lab', which consisted of M680x0 VME machines mounted to a motorized chassis with a bunch of sensors (collision, light, ultrasound, video). I'm pretty sure they were powered by rechargeable lead-acid cells when they were 'off the leash'. I dread to think how much all that must have cost back then. Probably a few grand per robot. :-/

More details can be found in this introductory paper.

I'd suggest going with a similar design, but using slightly less esoteric hardware - a soekris device, perhaps. Unfortunately, that'll probably blow most of your budget alone. :-(

--

"No man but a blockhead ever wrote, except for money."
-- Dr Johnson

Put that in your pipe and smoke it, Joi Ito...

(My blog; my blogging application.)

Damn lameness filter! You asked for it.

Just find a church with services at different times. I have M.E. and find it difficult to get to church in the morning - so I go to the chaplaincy at my local university which has services at 12:10. When I lived in Amsterdam I went to Christ Church which had a service at 16:00.

Also look here for a quick introduction to morphology. Pay special attention to bullet six (zero morphology).

Please smack yourself several times in the head with a large crowbar until you understand linguistics.

The `spirit is willing' story is amusing, and it really is a pity that it is not true. However, like most MT `howlers' it is a fabrication. In fact, for the most part, they were in circulation long before any MT system could have produced them (variants of the `spirit is willing' example can be found in the American press as early as 1956, but sadly, there does not seem to have been an MT system in America which could translate from English into Russian until much more recently --- for sound strategic reasons, work in the USA had concentrated on the translation of Russian into English, not the other way round). Of course, there are real MT howlers. Two of the nicest are the translation of French avocat (`advocate', `lawyer' or `barrister') as avocado, and the translation of Les soldats sont dans le café as The soldiers are in the coffee. However, they are not as easy to find as the reader might think, and they certainly do not show that MT is useless.

BTW, since this book is no longer available in the stores, the whole contents is placed online, though the server appears down right now. I recommend reading this book to anyone who is interested into the subject of MT. It really is a nice introduction into the subject.

Seize is another weird one.

The improved i-before-e rule.

(Now, a rule with too many exceptions is useless, but hey, it's English.)

If you're convinced you're a champion speller, try this and this -- then post your results here. :)

The University of Essex
Cosmic Pussycat Designs (okay, maybe this one should be banned)
you get the idea...

Since 2000 residents of the BRD with non-German parents can get citizenship if one parent has lived in Germany for eight years and has a job or something.

The pressure to change the law was largely internal, not from the EU. An informative working draft detailing how the SPD-Green coalition changed German citizenship requirements.

This is not quite along the lines of violently hot or explosive materials, but it is quite surprising if you've never seen it: http://www.essex.ac.uk/psychology/visual/thatcher. html

I would have to say that a good starting place for you would be The Greengrocer's Apostrophe

(CD's should be CDs)

The proposition of this happening at any time in the future may be considered optimistic to say the least. All you need to do is look at the current methods using text, such as babelfish, and you'll realize how far there is to go. Many times people who make comments about so called Machine Translation have no clue what is actually involved in the process. It involves resolving between many ambiguities. This is heightened in spoken translation, where homophones must also be accounted for.

Take for instance the following sentence: "It's in the pen." Do you know what it was talking about? Would you even if I told you what "it" was? A pen could mean a pig pen, or it could mean a writing instrument. Granted, if you knew I was talking about a spring, you'd know it was a writing instrument, but teaching a computer to know that is no trivial task.

For more info, visit http://www.essex.ac.uk/linguistics/clmt/MTbook/

The audio fidelity on the whole is as good as can be expected. We all know what MP3 encoding does to your tunes if you have a reasonable speakers

All mp3 decoders are not created equal and I sure wish reviewers would dig a little deeper. If they go beyond navigation and I/O features and cover audio quality, it's typically only to mention background hum or hiss or a wimpy headphone amplifier.

mp3 at high bitrates created with a competent encoder (LAME is one) can sound pretty good. Decoding with cheap 16 bit DSPs such as the TMS320C54x used in the Rio One is hard to do-- you have to watch out for error accumulation (e.g. please round to zero instead of simply truncating).

If you don't believe me that decoders differ, consider these tests of PC decoders. Unfortunately, no one seems to do such detailed testing on embedded decoders.

I'm giving my Rio One to my nephew who will be so thrilled to have it that if he hears the high-frequency errors on playback of middlin' bitrates (192kbps), he probably won't mind.

I don't know what I'm going to replace it with, though. I know I shouldn't expect much out of a $80 player powered from on AA cell, but there's no guarantee that a $400 Rio Riot or iPod will be glitch-free: they might have spent the whole power and cost budget on LCDs, hard drives and amortizing development.

A StrongARM-based PocketPC might be the answer- plenty of horsepower to run less compromised 16 bit decoder or even a 24 bit or floating point one. It should also be able to decode ogg vorbis....

By the way, instead of reasonable speakers, I'm using a good pair of headphones. Much more bang for the buck when it comes to revealing audio defects, though the Sonys tend to be a bit shrill (well the older V6s that I have) for long-term listening. These are the same model we used when I was at E-mu for all normal testing. The only thing more revealing was the elements from a good pair of Sennheisers in a set of noise protection muffs to cut background noise by 23dB. Also, some of the ATC guys have Grado electrostatics.

that we've seen similar threads before... I pop this big question.

Most people who want a wearable, at the current time, are complete geeks. Usually these types of people want stuff THEIR way, and they try to be economical when doing it, so why buy something like this?

I too want a wearable, and one that attracts the GOOD kind of attention, not the bad kind.. I decided to set about making one, and learned a LOT from various websites.

Of course, going the PC-104 route will still cost you a pretty penny, you can get a unit with a LOT more storage and custom input and display options. There are several pages on the net that go into detail on how to modify visor type displays and conceal them behind a normal pair of sunglasses.

For anyone serious about making a wearable, I say grow your own. It is a learning experience you can be proud of, instead of a product that you blew a pay check on, and then complain.. "I wish it did this, and that"

-fc

Read the linkked article. Your exact questions are answered in the FAQ.

To whit:

3 My mp3s are made from 16-bit CDs - why use a 24-bit decoder?

As you probably know, when you encode a CD to mp3 format, you don't store an exact copy of the original signal. When an mp3 is decoded, you don't get those original 16-bits back, but an approximation that should sound similar. When the decoder puts together all the elements held in the mp3 file, the arithmetical result can be very accurate in numerical terms, even if it's not exactly what was on the original CD. If you round it to 16-bits, you add a small amount of extra distortion to this reconstructed signal, getting even further away from what was on the original CD. If you round it to 24-bits, you're still adding distortion, but it's 256 times quieter than that added by rounding to 16-bits.

and,

4 I only have a 16-bit sound card - what use could a 24-bit decoder be?

If you calculate the result to 24-bit accuracy, and then round it to 16-bits, you gain nothing - the result will match all the standard 16-bit decoders. However, if you dither the result from 24-bits down to 16-bits, you can avoid all the distortion generated by rounding to 16-bits, and the result may sound better. Please read this article about dither for a fuller explanation of this.

First of all, mp3 decoding is done through basically synthesising waves according to the (lossy) information being used, and outputting them as 16 bit 44.1K audio in the vast majority of cases.

mp3 audio is inherently lower quality than 16/44 (some other formats like WMA or 'mp3pro' are even more blatantly inaccurate, making up data out of the blue) and so, to my knowledge, the most common approach has been to just cast the resulting sample values to an int or something, which is the same as truncation of the value. This results in quantization distortion, and since it's just lousy mp3, who cares?

However, it is possible to decode mp3 to 24-bit resolution and up- and this is where it appears Apple's approach to these things gets interesting. My own experience with this started when I got iTunes running on a MacOS 8.6 system- before then, there was no chance of running iTunes, and I'd been using other means of playing mp3s, like SoundApp, which remains a nifty program but didn't prepare me for what I was going to hear from iTunes.

Briefly, I do audio mastering work, and have very high resolution audio gear coming off my Mac- and here's the deal- I started playing tunes off iTunes, and was very startled to hear them playing with a depth and dimensionality that I was totally unaccustomed to. The sound was more 3D than 2D, despite the mp3 sources. Why?

For a possible answer you might look here, at some tests done with 24-bit mp3 decoding libraries, and consider Apple's background in pro audio. Put simply, it's my suspicion that iTunes is decoding to 24 bit or 32-bit floating point, and dithering the result to 16 bits for the sound output DACs. This is a substantially more sophisticated approach than the usual 'mp3 is cheesy anyway' direct truncation, and it yields considerably better sound. I can't get direct confirmation of this by citing iTunes source, as iTunes is closed source- so I linked to a 24 bit decoder review page to drive home the point that this technology is out there and in use.

What does this have to do with iPod? Easy- what iTunes can do in software, iPod can do in embedded hardware. I think it would be a good idea to analyze the performance of iPod compared to other mp3 playing portables- and see whether iPod is pioneering high-resolution mp3 decoding and dithering in a portable. This can be measured: the noise floor will be different and up to 20 db lower compared to simple truncation! It is also likely to sound distinctly different as well- high-quality headphones might make this equally obvious.

Just thought I'd raise the issue, since the Register has apparently commented that it sounds nice, and I've had similar observations about iTunes... the _character_ of the improvement in sound is very much resolution domain stuff, and Winamp users can apparently get an example of this type of sound through a 24-bit MAD mp3 library plugin. If my hunch is correct, Apple are already routinely doing this in their products to get a more 'high-end' sound, including iPod- and it may be a first in mp3 portables. More research (by someone who _can_ just run out and buy a Nomad and an iPod and start measuring them ;) ) is indicated :)

here

and here

These links all are by other users who installed linux on the TP701 laptops.

• Skips samples during the first second of file, resulting in audible click.
• Audible low frequency glitches in many files.
• Faults in the decoding engine itself include audible mistakes below 15 kHz and a few mistakes above too.
• The right channel is decoded correctly (only occasional 1-bit difference from l3dec), but the left channel is destroyed...
• Sonique HQ decode mode also inverts the signal.

Actually, the Nitrane and Sonique decoding engines are quite error-prone. mpg123 and the fraunhofer decoders are the only ones proven to be the best. Check out this study which points out the differences in decoder quality. His methodology seems sound.

zsazsafrazs

While contextual knowledge can increase the qualitiy of a translation; the amount of world knowledge necessary to translate a typical web page is simply astounding. Most users of a translation system simply do not want to wait for hours to translate a simple sentence.

And, there is the problem of linguistic knowledge. Most web pages are not written in "proper" English, but in some Web-speak-lingo. This requires the system to be very robust.

The most successful use of MT in corporations today are situations where a very simple grammar and lexicon is used, and very little world knowledge ois required. For instance, the Xerox corporation has its own translation system that translates component manuals. The technical writers that write the original version of the manual are required to use very simple English only, without any ambiguities and with very simple constructions.

For translation software that has multiple language targets, i would have expected it to first resolve the meaning of the English sentence into an internal semantic representation before using it to emit Spanish or French.

This "internal semantic representation" is called an Interlingua. It has been used in various MT systems, with varied amounts of succes.

The most important advantage of an Interlingua-based MT system is that is does not require a translation engine for each language pair. For instance, if you create a system for English, French, Dutch and German texts, you only need to create four analysis engines:

1. English -> interlingua
2. French -> interlingua
3. German -> interlingua
4. Dutch -> interlingua

1. interlingua -> English
2. interlingua -> French
3. interlingua -> German
4. interlingua -> Dutch

1. English -> French
2. English -> German
3. English -> Dutch

1. French -> English
2. French -> German
3. French -> Dutch

Clearly, it is easier to integrate new languages into a interlingua system than into a transfer system.

While contextual knowledge can increase the qualitiy of a translation; the amount of world knowledge necessary to translate a typical web page is simply astounding. Most users of a translation system simply do not want to wait for hours to translate a simple sentence.

And, there is the problem of linguistic knowledge. Most web pages are not written in "proper" English, but in some Web-speak-lingo. This requires the system to be very robust.

The most successful use of MT in corporations today are situations where a very simple grammar and lexicon is used, and very little world knowledge ois required. For instance, the Xerox corporation has its own translation system that translates component manuals. The technical writers that write the original version of the manual are required to use very simple English only, without any ambiguities and with very simple constructions.

For translation software that has multiple language targets, i would have expected it to first resolve the meaning of the English sentence into an internal semantic representation before using it to emit Spanish or French.

This "internal semantic representation" is called an Interlingua. It has been used in various MT systems, with varied amounts of succes.

The most important advantage of an Interlingua-based MT system is that is does not require a translation engine for each language pair. For instance, if you create a system for English, French, Dutch and German texts, you only need to create four analysis engines:

1. English -> interlingua
2. French -> interlingua
3. German -> interlingua
4. Dutch -> interlingua

1. interlingua -> English
2. interlingua -> French
3. interlingua -> German
4. interlingua -> Dutch

1. English -> French
2. English -> German
3. English -> Dutch

1. French -> English
2. French -> German
3. French -> Dutch

Clearly, it is easier to integrate new languages into a interlingua system than into a transfer system.

While contextual knowledge can increase the qualitiy of a translation; the amount of world knowledge necessary to translate a typical web page is simply astounding. Most users of a translation system simply do not want to wait for hours to translate a simple sentence.

And, there is the problem of linguistic knowledge. Most web pages are not written in "proper" English, but in some Web-speak-lingo. This requires the system to be very robust.

The most successful use of MT in corporations today are situations where a very simple grammar and lexicon is used, and very little world knowledge ois required. For instance, the Xerox corporation has its own translation system that translates component manuals. The technical writers that write the original version of the manual are required to use very simple English only, without any ambiguities and with very simple constructions.

For translation software that has multiple language targets, i would have expected it to first resolve the meaning of the English sentence into an internal semantic representation before using it to emit Spanish or French.

This "internal semantic representation" is called an Interlingua. It has been used in various MT systems, with varied amounts of succes.

The most important advantage of an Interlingua-based MT system is that is does not require a translation engine for each language pair. For instance, if you create a system for English, French, Dutch and German texts, you only need to create four analysis engines:

1. English -> interlingua
2. French -> interlingua
3. German -> interlingua
4. Dutch -> interlingua

1. interlingua -> English
2. interlingua -> French
3. interlingua -> German
4. interlingua -> Dutch

1. English -> French
2. English -> German
3. English -> Dutch

1. French -> English
2. French -> German
3. French -> Dutch

Clearly, it is easier to integrate new languages into a interlingua system than into a transfer system.

The `spirit is willing' story is amusing, and it really is a pity that it is not true. However, like most MT `howlers' it is a fabrication. In fact, for the most part, they were in circulation long before any MT system could have produced them (variants of the `spirit is willing' example can be found in the American press as early as 1956, but sadly, there does not seem to have been an MT system in America which could translate from English into Russian until much more recently --- for sound strategic reasons, work in the USA had concentrated on the translation of Russian into English, not the other way round). Of course, there are real MT howlers. Two of the nicest are the translation of French avocat (`advocate', `lawyer' or `barrister') as avocado, and the translation of Les soldats sont dans le café as The soldiers are in the coffee. However, they are not as easy to find as the reader might think, and they certainly do not show that MT is useless.

BTW, since this book is no longer available in the stores, the whole contents is placed online. I recommend reading this book to anyone who is interested into the subject of MT. It really is a nice introduction into the subject.

The `spirit is willing' story is amusing, and it really is a pity that it is not true. However, like most MT `howlers' it is a fabrication. In fact, for the most part, they were in circulation long before any MT system could have produced them (variants of the `spirit is willing' example can be found in the American press as early as 1956, but sadly, there does not seem to have been an MT system in America which could translate from English into Russian until much more recently --- for sound strategic reasons, work in the USA had concentrated on the translation of Russian into English, not the other way round). Of course, there are real MT howlers. Two of the nicest are the translation of French avocat (`advocate', `lawyer' or `barrister') as avocado, and the translation of Les soldats sont dans le café as The soldiers are in the coffee. However, they are not as easy to find as the reader might think, and they certainly do not show that MT is useless.

BTW, since this book is no longer available in the stores, the whole contents is placed online. I recommend reading this book to anyone who is interested into the subject of MT. It really is a nice introduction into the subject.