I think the lettering added to the dial (along with the GIMP logo) should NOT have been pure black, but a shade of gray matching the numbers on the dial. The words "GIMP Years" aren't quite centered either.
I also would've chosen a gauge that doesn't skip numbers at the begining! Since the chosen gauge is not symetrical, it just looks crooked. The 5 should be horizontal from the 25, and the 15 should be at the top, but since there's a tick missing between 0 and 5, it's out of wack. Speaking of symmetry, the location of the bezel bolts also makes this image look crooked.
How about a monthly news letter? There's a local game store (board games, but still similar people) that sends out an email once a month. It talks about that latest this and the greatest that. It's got a trvia question, and the first responder gets a $10 gift certificate. If a person reads the news letter and goes off in search of the answer to the question, they've got your store on their mind. People have been conditioned to think of the EBs and Gamestops when they want a game. You've got to get them to think of you too. The only way to get them to do that is to remind them you're there.
Could I buy board games cheaper online? Of course. Why do I buy them from my local store (25minutes away too, and I don't like in the middle of nowhere!)? Because the staff is made up of knowlegable, nice people. They clearly care about what they're doing/selling. It's pleasant experience to go to the store and buy from them. If you can make your customers walk out of the store smiling, they'll come back!
As some one above alluded to, you can use phase modulation, or QAM (quadrature amplitude modulation) to acheive this. You could also use AM. Let's say that you and I (TX and RX) agree that 1V is 0, 2V is 1, 3V is 2, 4V is 4. Each time you send a cycle of the sinusoid, you send one of those amplitudes, and I decode it to a 0,1,2, or 3 (the equivalent of 2 bits). That same can be done with phase modulation or a combination of the two.
You're right, and if you don't beleive him, do a Monte Carlo simulation (or just work out all of the possibilites, there's only 9 combiniations of winning doors and iniial guesses). You can then see that 2/3 of the time switching will make you a winner. It's counter intuitive, but it's true.. The subtlty comes in because he opened the door AFTER you made your initial choice, therefore the original probabilities are not affeted.
More mathematicall: A is the event that you've chosen the correct door. B is the even where Monty Hall opens the door. P(A)=1/3. P(A|B)=1/2. Had Monty opened the door BEFORE you choose your door, then P(A|B) would be correct, but since it's after, P(A) is the correct one.
We have both cubicles and offices where I work. As on of my cubicle dwelling co-workers said to someone who had just upgraded to an offce, "Nice office! I'll took the extra money."
But what if between one sample and the next sample, the amplitude of the signal goes up by *half* of a digital "step"? Do you get no change in the sampled value for that time interval, or do you get a full step change for that sample interval? It has to be one or the other.
When you go to reconstruct the original analog signal, these digital compromises are introduced as noise. It's usually imperceptible, but maybe there's an audiophile out there who can actually hear it.
This is quantization noise. Quantization noise can be modeled as a uniformly distributed random proccess, provided you have enough bits (16 is plenty).
The signal to quantization noise ratio is approximately 6*NumerOfBits. For a CD that 16*6 = 96dB. This means that the power of the quantization noise will be 2.5e-10 times the signal power when the signal is at full scale.
Me thinks this is a bad idea. Having all of the motor type stuff on 12V and the digital stuff on 5V or 3.3V keeps the supplies sepperate. A motor will produce some weird noise on the supply 5V supply rails if used as the web site suggests.
Wood: Actually a non-perishable resource, if the right species are used. Maple is good.
Then shaped by machines power with electricity generated from Oil, or Polluting Coal, or *heavens* NuCuLar energy, or River Blocking Dams and stained/finished with checmicals Made from Oil.
If you're just going to zoom in on voice, you might as well just record the voice and look at a plot of it. That would be far more repeatable. However, I don't think a human will be able to look at a time-domain representation and match it to the speaker. Looking at a spectrogram (ie a decimated short-time fourier transform plotter as an image map) would be better, but still isn't quite there. You can see formants and such, but each speaker will essentially look the same still.
Some more info on dynamic time warping as it applies to speech recognition; it's not too different from speaker recognition
Think about this for a second. If the recording is different by one sample inserted (a.125 ms delay will do this at 8kHz sample rate), then this whole plan blows up. Humans aren't that accurate. You need some method for coping with the fact that when you speak a sentence twice there are bound to be temporal differences. For this, one ususally used dynamic time warping (DTM) or some such thing.
Displaying it on an o-scope is slightly better, but for one it will fall victim to the same short-coming described above.
Brett
For a graduate class at Goergia Tech. Voice identification/verification is not a travial problem. I don't ever recall hearing our professor talk about "voice prints."
Most modern voice identification systems use linear predictive coded (LPC) ceptra and either hidden Markov models (HMM) to evaluate how close a a given speaker is to a known user.
Having said that, I don't think it makes a very cool demo as the result is simply a number. In the case of speaker verification this number represents the probability that the speaker is who he claims to be.
Good idea, but I don't think this is what you want.
Let's say the air craft is sitting still, rotor rotating clock-wise. The center of rotation will be at the axel. As the craft begins to move forward, the center of rotation will move to the right. At some certain airspeed, the mu-1 condition will be met, and as you have pointed out, all points (except the tip on the right side of the craft, which is now the center of rotation) will have forward velocities.
What I think the poster was getting at is that the airfoils are moving the wrong direction to produce lift. To the right of the axel, the front-to back component of the velocity is possitive in the forward direction, but since this blade rotates clockwise, on the right side of the axel, the blade needs a possitive reverse component to produce lift. I think that's what they meant by "backwards."
That's true when the helicopter is sitting still, but not when it is moving.
Think of a wheel rolling on the ground. The center of the rotation is actually the point where the wheel touches the ground, NOT the center of the wheel. This is only true when the wheel does not slip of course.
I'm guessing these drives would have a sort of "wear leveling" just like they have in most compact flash cards.
The wear leveling works by keeping a table of what physical flash is mapped to what address. The trouble comes when power is yanked whilst the table is in the middle of an update.
Slashdot Mirror
I like the idea, but not the implmentation.
I think the lettering added to the dial (along with the GIMP logo) should NOT have been pure black, but a shade of gray matching the numbers on the dial. The words "GIMP Years" aren't quite centered either.
I also would've chosen a gauge that doesn't skip numbers at the begining! Since the chosen gauge is not symetrical, it just looks crooked. The 5 should be horizontal from the 25, and the 15 should be at the top, but since there's a tick missing between 0 and 5, it's out of wack. Speaking of symmetry, the location of the bezel bolts also makes this image look crooked.
Brett
How about a monthly news letter? There's a local game store (board games, but still similar people) that sends out an email once a month. It talks about that latest this and the greatest that. It's got a trvia question, and the first responder gets a $10 gift certificate. If a person reads the news letter and goes off in search of the answer to the question, they've got your store on their mind. People have been conditioned to think of the EBs and Gamestops when they want a game. You've got to get them to think of you too. The only way to get them to do that is to remind them you're there.
Could I buy board games cheaper online? Of course. Why do I buy them from my local store (25minutes away too, and I don't like in the middle of nowhere!)? Because the staff is made up of knowlegable, nice people. They clearly care about what they're doing/selling. It's pleasant experience to go to the store and buy from them. If you can make your customers walk out of the store smiling, they'll come back!
A better comforter cannot heat the bed before I get in it.
Brett
I think you're confusing bandwidth (measured in Hz) with channel capacity (measured in bits/second).
Brett
ps I am a radio engineer
As some one above alluded to, you can use phase modulation, or QAM (quadrature amplitude modulation) to acheive this. You could also use AM. Let's say that you and I (TX and RX) agree that 1V is 0, 2V is 1, 3V is 2, 4V is 4. Each time you send a cycle of the sinusoid, you send one of those amplitudes, and I decode it to a 0,1,2, or 3 (the equivalent of 2 bits). That same can be done with phase modulation or a combination of the two.
Brett
You're right, and if you don't beleive him, do a Monte Carlo simulation (or just work out all of the possibilites, there's only 9 combiniations of winning doors and iniial guesses). You can then see that 2/3 of the time switching will make you a winner. It's counter intuitive, but it's true.. The subtlty comes in because he opened the door AFTER you made your initial choice, therefore the original probabilities are not affeted.
More mathematicall: A is the event that you've chosen the correct door. B is the even where Monty Hall opens the door. P(A)=1/3. P(A|B)=1/2. Had Monty opened the door BEFORE you choose your door, then P(A|B) would be correct, but since it's after, P(A) is the correct one.
Brett
We have both cubicles and offices where I work. As on of my cubicle dwelling co-workers said to someone who had just upgraded to an offce, "Nice office! I'll took the extra money."
Offices are nice, but they cost you money.
Brett
Doesn't the guy at the end of the video look like Bong Soo Han? He played Dr. Klahn in Kentucky Fried Movie.
Brett
But I never really cared for the video slots.
There's something about a real reel reeling that sends me reeling...
Brett
--"You keep using that word...I do not think it means what you think it means."
Sure enough, Belkin does make some "pre-n" stuff:
Wireless Pre-N Router
Brett
"Does your cell phone have some kind of magic keyboard that attaches to it?"
Predective text input: T9. Most modern phone seem to have it.
Brett
But what if between one sample and the next sample, the amplitude of the signal goes up by *half* of a digital "step"? Do you get no change in the sampled value for that time interval, or do you get a full step change for that sample interval? It has to be one or the other.
When you go to reconstruct the original analog signal, these digital compromises are introduced as noise. It's usually imperceptible, but maybe there's an audiophile out there who can actually hear it.
This is quantization noise. Quantization noise can be modeled as a uniformly distributed random proccess, provided you have enough bits (16 is plenty).
The signal to quantization noise ratio is approximately 6*NumerOfBits. For a CD that 16*6 = 96dB. This means that the power of the quantization noise will be 2.5e-10 times the signal power when the signal is at full scale.
No one will hear that.
Brett
Me thinks this is a bad idea. Having all of the motor type stuff on 12V and the digital stuff on 5V or 3.3V keeps the supplies sepperate. A motor will produce some weird noise on the supply 5V supply rails if used as the web site suggests.
Check out:
Skotrat
Beer Town(home of the Asscoiation of Brewers)
Tasty Brew
Stout Billy's
Brett
He said longer not faster.
That govenor is there to keep you uner the rated speed of the stock tires. This is a good thing.
Brett
Acutally, Tivo offers this service on Series 2 machines with the home media option.
Brett
Like Tivo Web?
Brett
Wood: Actually a non-perishable resource, if the right species are used. Maple is good.
Then shaped by machines power with electricity generated from Oil, or Polluting Coal, or *heavens* NuCuLar energy, or River Blocking Dams and stained/finished with checmicals Made from Oil.
Brett
If you're just going to zoom in on voice, you might as well just record the voice and look at a plot of it. That would be far more repeatable. However, I don't think a human will be able to look at a time-domain representation and match it to the speaker. Looking at a spectrogram (ie a decimated short-time fourier transform plotter as an image map) would be better, but still isn't quite there. You can see formants and such, but each speaker will essentially look the same still.
Some more info on dynamic time warping as it applies to speech recognition; it's not too different from speaker recognition
Brett
Think about this for a second. If the recording is different by one sample inserted (a .125 ms delay will do this at 8kHz sample rate), then this whole plan blows up. Humans aren't that accurate. You need some method for coping with the fact that when you speak a sentence twice there are bound to be temporal differences. For this, one ususally used dynamic time warping (DTM) or some such thing.
Displaying it on an o-scope is slightly better, but for one it will fall victim to the same short-coming described above.
Brett
For a graduate class at Goergia Tech. Voice identification/verification is not a travial problem. I don't ever recall hearing our professor talk about "voice prints."
Most modern voice identification systems use linear predictive coded (LPC) ceptra and either hidden Markov models (HMM) to evaluate how close a a given speaker is to a known user.
Having said that, I don't think it makes a very cool demo as the result is simply a number. In the case of speaker verification this number represents the probability that the speaker is who he claims to be.
Good idea, but I don't think this is what you want.
Brett
I agree with what you say. Let's expand a bit.
Let's say the air craft is sitting still, rotor rotating clock-wise. The center of rotation will be at the axel. As the craft begins to move forward, the center of rotation will move to the right. At some certain airspeed, the mu-1 condition will be met, and as you have pointed out, all points (except the tip on the right side of the craft, which is now the center of rotation) will have forward velocities.
What I think the poster was getting at is that the airfoils are moving the wrong direction to produce lift. To the right of the axel, the front-to back component of the velocity is possitive in the forward direction, but since this blade rotates clockwise, on the right side of the axel, the blade needs a possitive reverse component to produce lift. I think that's what they meant by "backwards."
Brett
That's true when the helicopter is sitting still, but not when it is moving.
Think of a wheel rolling on the ground. The center of the rotation is actually the point where the wheel touches the ground, NOT the center of the wheel. This is only true when the wheel does not slip of course.
Here's a link with some information about rolling.
Brett
In fact, here's a white paper from Sandisk.
Brett
I'm guessing these drives would have a sort of "wear leveling" just like they have in most compact flash cards.
The wear leveling works by keeping a table of what physical flash is mapped to what address. The trouble comes when power is yanked whilst the table is in the middle of an update.
Brett