Jack Crossfire's blog View Details
Archive for July, 2013
Posted by Jack Crossfire | Jul 31, 2013 @ 02:40 AM | 8,149 Views

Wood was sanded down to form the curve.

Drawing the model outline on the mold ended up very important.

...Continue Reading
Posted by Jack Crossfire | Jul 30, 2013 @ 02:16 AM | 7,394 Views

3.3" diameter cans worked much better than 2.94" diameter cans.

The mane difference from the MIT radar was the microcontroller replacing the soundcard, the virtual ground being 1.6V instead of 5V & omission of the lowpass filter. The LM324 was used for the op-amp instead of a MAX414.

Fixed point 16 bit FFT on the microcontroller worked reasonably fast. Hand coded assembly language for the STM32 would probably be 25% faster. 8 bits would be no faster.

256 window size = 1ms
1024 window size = 4ms
4096 window size = 20ms
16384 window size = 90ms

There is no practical way to implement navigation based on frequency modulated continuous wave in software, but it was useful to see how it's going to be done when it becomes practical. The microcontroller sampled its 16384 sample FFT window at 1.2Mhz. The frequency was modulated in a 15ms ramp.

A piece of metal was placed 3 meters in front of it to test it indoors.

with metal:

...Continue Reading
Posted by Jack Crossfire | Jul 26, 2013 @ 05:05 AM | 8,828 Views
Making your own microwave frequency circuits is still as expensive as it ever was.

That's the heart of a 2.4Ghz heterodyne receiver. It's $47.

That would be the 2nd most expensive part, at $43.

...Continue Reading
Posted by Jack Crossfire | Jul 24, 2013 @ 01:32 AM | 7,667 Views
Posted by Jack Crossfire | Jul 23, 2013 @ 12:13 AM | 7,545 Views
The mane concern for an embedded radar is the magic filtering which turns the harmonic frequencies in the IF output to a distance & velocity. That's still an area of research & it takes a lot of CPU power. It all but eliminates anything besides measuring velocity.

The method of choice is the fast fourier transform. A frequency counter doesn't seem to do the job.

There is the proprietary assembly language FFT for ARM. Then, there's the standard fixed point one, written in 1989. When the Arduino age began, someone ported it to 8 bits & discarded the 16 bit version so all Goog gives you is the 8 bit version. Fortunately, the 16 bit version is still available in the gstreamer source code.

It needs 3 * 2 * window_size of RAM, limiting you to 16384 samples.

MIT provides sample IF output from the radar project. You can load it in an audio editor & see the ranging in a spectrogram. The mane problem is an audio editor can't perform the spectrogram just during the pings, so there's a lot of noise where the pinging is idle.

The IF output is not a simple sine wave corresponding to the nearest object, but many harmonics from many objects. You can play the IF output & hear the harmonics.

For detecting velocity, the MIT radar does a 11025 sample fourier transform. The sample rate is 44100.

For detecting distance, it does an 882 sample fourier transform. It has a lot of noise which is eliminated by taking only the difference between 2...Continue Reading
Posted by Jack Crossfire | Jul 19, 2013 @ 07:38 PM | 6,644 Views
Radar is probably going to be the dominant navigation method for personal drones, whether indoor or outdoor. It was how the skycrane hovered over Mars without GPS. Automotive demands are greatly lowering the cost, but the famous MIT course made it famous.

You still can't buy a solid state module which spits out a 3D radar image. Those are still reserved for the rich & famous. You have to get RF components & build your own front end.

Modern radar uses a frequency modulated continuous wave. It tunes the transmitted carrier wave up & down in frequency. Then it multiplies the transmitted waveform at any instant by the reflected waveform (mixing) at the same instant. The product is a sine wave at the difference in frequencies (the IF output).

The reflected waveform's frequency is always a delayed copy of the transmitted frequency, allowing the distance & doppler shifted velocity of the target to be deduced from the difference in frequencies. A 2nd ping is required without modulating the carrier frequency to separate the distance from velocity.

2 receivers can be used to determine the direction of motion. The mixed output frequency (IF output) can be deduced by a microcontroller ADC.

Accuracy depends on a very rapidly sweeping carrier frequency with a very smooth slope. The MIT radar uses an XR-2206 function generator to modulate the carrier frequency. The transmitter is a voltage controlled oscillator. A crusty old microcontroller PWM won't...Continue Reading
Posted by Jack Crossfire | Jul 19, 2013 @ 02:17 AM | 7,028 Views
The next step with Marcy 1 is fiberglass. The last venture into fiberglass was 10 years ago. There were no goo tube videos or web pages on how to do it in 2003.

Now, there are some long winded goo tube videos but still no web pages. Most of them use a 1 sided mold & allow excess resin on 1 side. Aerospace parts quite clearly involve layering 2 resin coated cloths, compressing it in a 2 sided mold to squeeze out excess resin.

The canopies for RC copters seem to be made of 1 fiberglass layer in a 1 sided mold. The mold forms a smooth outside & the cloth has barely the minimum required amount of resin.

Creating a 2 sided mold would be an ordeal. It would require carving the shape out of a big piece of foam, gluing some wood together, or shaping clay. Then, the foam could form a negative mold out of plaster or be used on its own.

Mold making is 1 area where 3D printing could pay off.

In other news, there was the Roving Networks RN-41. It's expensive, big, much slower than wifi, but so much easier to get working than a directly connected wifi dongle. It's the easiest way to give your phone or tablet a wireless connection to a robot when the speed doesn't matter.
Posted by Jack Crossfire | Jul 17, 2013 @ 08:47 PM | 6,849 Views
So a guy your age has 2 masters degrees in aerospace. He continued going to school until age 31, under scholarships, while you were dicking around with video editing software in your bedroom.

He had a lot of management jobs in charge of lots of money. He survived multiple RIFs & got hired as a contractor in the place everyone wants to live after the sequester eliminated most contractor jobs there.

That represents the level of caliber of who's getting jobs in Calif*, nowadays, especially in the space program. They have lots of formal education through their entire lives, not Google search experience. They have lots of experience in a structured environment, not building things in their garages. They have lots of responsibility outside of paying their own rent.

That's a big change from 15 years ago. The open source & maker movements may have originally benefited some people, but in the fullness of time seem to have diluted the value of independently gained knowledge. It doesn't matter if you made a fusion reactor in your garage anymore or wrote a program in your bedroom anymore, because now, everyone does that.

What's differentiating the successful guys from the masses of unemployed makers is now the rarity of a structured environment & formal education.

Well, I'm a few days from signing off on moving to Dallas. Enough time has passed, avenues attempted, & dead ends reached to be satisfied that staying in Calif*, let alone having the good jobs, requires a much higher caliber resume than what worked 6 years ago. Reading about the people who are able to stay here is good evidence of that. No unturned stone is going to magically appear, after the move.

The cost of living isn't a factor, anymore. Most people rent because real estate is too unstable & rent is the same, everywhere except NY.
Posted by Jack Crossfire | Jul 17, 2013 @ 02:34 AM | 6,462 Views
Mini Marcy 1 (1 min 10 sec)

After much testing, it quite clearly doesn't have enough lateral control. It eventually creates an airflow which sucks it into a wall. It also creates an airflow which sucks itself to the ceiling. In a stable hover, it eventually starts lifting faster & faster until it requires a throttle cut to get it to come down again.

The computer can keep it in the air until it runs out of horizontal room. It can supposedly entertain a human pilot until it eventually runs out of room.

It seems to be a matter of excessive efficiency. Complete flights with full pitch showed it tilting but never going anywhere because of its own vortex. The shape of the room seemed to allow movement only when far from any wall.

...Continue Reading
Posted by Jack Crossfire | Jul 16, 2013 @ 10:52 PM | 6,257 Views

Well, they don't look like paper, but as in all things with the new economy, it's more about symbolism than reality.

More intriguing than the "paper" aspect was that they're the simplest things which can fly. They're unpowered. There's just enough resolution in the photos to see the airplane uses 2 stock Plantraco actuators. The monocopter doesn't seem to have control.

Obviously, they're not autonomous. They could be useful for plotting temperature & humidity vs. altitude. They couldn't plot air pressure vs. altitude because air pressure is needed to determine altitude. There's a chance of using them to plot radio reception for a different radio than the one used for flying. Cameras aren't quite small enough to fit on there.
Posted by Jack Crossfire | Jul 16, 2013 @ 03:44 AM | 7,012 Views

NASA has 2 global hawks which they can only afford to fly 10 days per year. That's 1/5 as many hours as they could afford to fly the space shuttle.

They sometimes drop something out of their rear end.

...Continue Reading
Posted by Jack Crossfire | Jul 13, 2013 @ 04:37 AM | 6,454 Views

When Petman came out, they insisted it was purely to test clothing. In the end, Petman was the precursor to a humanoid purely intended for search & rescue, but you can imagine what Atlas is really heading towards. There's no other way to fight a war against IED's.

Surprising that they already bought the robot from Boston Dynamics instead of having the competitors build it or that DARPA is even buying robots. They previously funded development but never took possession of a product.

Posted by Jack Crossfire | Jul 12, 2013 @ 02:01 AM | 6,763 Views
Technical Details (3 min 57 sec)

It's sort of a backwards business model, resulting from the mantra that charging money for data is evil, so they launch a satellite with no software & charge money for the user to write software after it's in orbit. It's not the most efficient way of doing it, but it's the only funding option left when you take out the data.

It does bring to mind a practical application. How much money would a cubesat make as a private communications relay that anyone could rent & bounce their own signal off of? There's certainly a huge win for RC flying. There's also private internet access from anywhere, without a bandwidth cap.

Cubesats have now been around long enough & matured long enough that there's just about a fairly standard satellite bus & fairly standard way of launching them that's just within reach of what a small business can afford.



They're all standardized on the 8 bit Arduino platform & you don't want to reinvent everything the satellite bus does, so your software has to conform to that world. That's the compromise required to get the cost down, the same as adopting Intel so the world could afford PC's.

They quote $40,000 for the launch & $10,000 for the satellite. They don't mention the orbit lifetime, though anecdotally, they're lasting 1 year whether they're transmitting or not. It would have to make $4200/month to recover the cost of the...Continue Reading
Posted by Jack Crossfire | Jul 11, 2013 @ 05:41 PM | 6,191 Views
The oracle of MIT had a nugget of information about starting your own business:


If all you've done is make steady paychecks from someone else since 2000, you're probably worth absolutely zero now. That's why you're the inflation generation.

Unless you're Mark Zuckerberg, you'll normally get a better return on investment from the stock market than starting a business. A rich dad cannot be underestimated. I've never seen any business succeed without a rich dad, regardless of knowing how to do something no-one else can. Nowadays, anyone can know anything, so your only leverage becomes a rich dad.

People on the street generally think they come up with an idea, pitch the idea to investors, & the investors provide the seed money to start generating revenue. Unfortunately, it doesn't work that way except in the movies.

No venture capitalist will ever fund a business until after it's already generating revenue. It doesn't have to be profitable, but it has to already be selling something. You can pound the pavement all you want & have the key to anti gravity, but it won't happen until you make money on your own. They're stodgy, unimaginative bankers who wouldn't recognize Jesus if he landed on their Shenzou, but that's why China is #1 & the rest of the world cooks their fries.

The dilemma is as soon as you sell something, it's copied & becomes worthless. The initial product is basically a freebee to...Continue Reading
Posted by Jack Crossfire | Jul 11, 2013 @ 12:29 AM | 6,518 Views
The best motors have been 90% efficient for many years, a gain made possible by Neodymium magnets & software controllers being cheap enough to replace brushes. Most of the losses are now in the winding resistance & eddy currents in the core.

Since then, halbach arrays & coreless windings have been attempted, but never caught on. Coreless windings have no eddy currents, but produce a lot less torque. Halbach arrays can theoretically create a lot more flux in the armature, but seem to be impossible to manufacture in the maximum density circular array without destroying the magnets.

People increase the spacing of the magnets or use linear arrays. They end up with the same or less flux than the standard arrangement. There can be some weight savings in getting the same amount of flux with less magnets, but it's not improving efficiency.

Refrigerator magnets are halbach arrays created by magnetizing ferrite paste in a circular magnetic field. Not sure Neodymium magnets can be magnetized the same way.

The default Goog search shows some kind of halbach array & coreless winding getting 95% efficiency. The motor has to be started by hand & doesn't move anything.

The only real improvement not in the realm of free energy is the superconducting motor. All you need is some liquid nitrogen & the winding resistance becomes 0.
Posted by Jack Crossfire | Jul 10, 2013 @ 11:22 PM | 6,284 Views
X-47B Completes Second Carrier-based Arrested Landing (0 min 28 sec)

After 13 years, it finally did a carrier landing. For all the talk of it being the future of warfare, these programs are always on the verge of cancellation. The UCAS program was a privately funded idea in 2000, taken up by DARPA in 2001, canceled in 2006, then taken up by the Navy with another 7 years before it would ever make a carrier landing.

Now that it's done everything required for a carrier based weapon, its future is uncertain. There's no follow on program. There's no plan to mass produce it. US can't afford to expand the military.
Posted by Jack Crossfire | Jul 06, 2013 @ 06:12 AM | 6,256 Views
July 4 in Pleasant Hill (3 min 28 sec)

Pleasant Hill on day 1. Traffic getting out was snarled. The trick with parking was to park as far away as possible.

Fireworks run (4 min 2 sec)
...Continue Reading
Posted by Jack Crossfire | Jul 05, 2013 @ 09:08 PM | 5,944 Views

The Chinese are whacking out a lot of flying toys for XMas. They're trying so many ideas, it's hard to imagine any idea not already being attempted. The mane features are gesture control, caged copters, folding copters, tilt rotors, IR obstacle detection, IR altitude detection, recording video, replaying flights, shooting things, blowing bubbles.

The guy recommends everything he reviews, but obviously not all the features really work & most of the stuff he reviews obviously either never makes it to mass production or never sells significantly. It shows just how hard it is for an idea to succeed.

He did reveal Sphero executing a hand drawn route with a certain amount of accuracy. Getting a smooth arc like that from a hand drawing is immensely hard with a wheeled vehicle, but Sphero was able to do it.

Then, there was this. So far, staying upright on it is taking a huge amount of exertion. The skateboard operation is busted. It would take buying a commercial board & many hours of practice to get proficient. It's not the intuitive solution envisioned.

Astrology predicts Saturn's direction reversing on July 8 at the same time the moon is new is going to result in a massive financial victory & falling out of love with everyone from the last 30 years.
Posted by Jack Crossfire | Jul 04, 2013 @ 07:20 PM | 5,758 Views
So there were improvements to Marcy 1, achieving smaller size, slower RPM, more efficiency, simpler manufacturing. The problem is she's now probably too good to open source. The latest airframe is unlike anything else. It's a lot more docile than the bladestar, more controllable than the vectron.

There isn't a day job providing income. If someone offered a day job, it could be open sourced, making it work ware. The days of charity employment in exchange for donations are over. There are so many open source projects, they're as good as high school diplomas.

The design is good enough that it's worth a try to make money on it. The design is so simple that there can only be 1 sale of a massive number. It would take 1 million units to finish with a net profit, but there's no way a peasant can afford 1 million units.

Kickstarter could fund maybe 1000, but only after the design was copied & mass produced by someone else with the long green. The kickstarter project would only fund the cost of manufacturing, leaving an outcome of 0.

If Air Hogs gave a royalty for it, the royalty would amount to maybe 1 cent per unit.

Naturally, many others are racing to achieve the same improvements, so if commercializing Marcy 1 doesn't happen in a certain amount of time, she'll be open sourced.
Posted by Jack Crossfire | Jul 03, 2013 @ 03:15 AM | 6,494 Views
Proton rocket failure compilation (2 min 20 sec)

The latest in the string of recent Proton rocket failures as seen from some heavily firewalled Russian TV footage & some phone cams. Clearly an engine failure in the beginning, causing a sideways tilt, followed by the routine underdamped oscillations of a PID controller trying to recover, fully gimbaling the engines. Then it finally gives up, ejects the payload, shuts down the engines & explodes. Russia doesn't have a range safety device.

It's a spectacular setback for Russia & US, since US can't afford to launch the final space station module without the Proton. US can afford better video equipment, however. Why can't US have more failures, so they can be captured from multiple camera angles, in 3D, slow motion, HD?

Подготовка к пуску ракеты-носителя Прото&#108 (2 min 23 sec)
...Continue Reading