The lunch box finally started moving, after 2 weeks of baby steps between commutes. An extra 2 days were burned in a last ditch effort to get it going on a dspic33. The pickit ended up completely bricked when it tried to update its firmware in a virtual machine. Virtualbox can't emulate USB ports accurately, which probably causes any USB firmware update to fail. Fortunately, the ARM solution took a lot less time to implement.
The servo steering was bang on, with no D term required. Didn't have any stripped gears, despite using the cheapest plastic Tower Hobbies servo. Making it talk to a phone app took an immense amount of code, but proved vital.
It's a really important launch, because it's the 1st time a prototype of a vehicle which could someday carry humans is being launched since 2011.
The prototype can't carry humans, has no seats, no instrument panel, & has never been photographed on the inside, but we're told it simulates the requirements of a crewed vehicle, if it's ever funded.
It took 10 years to build this one. There isn't enough money to launch another one for 4 more years. Humans are going to ride Dragon long before Orion ever takes them anywhere.
It's being launched on your favorite rocket, the one you saw in person twice, carrying She Jehovah's payloads. The fact that it takes the largest rocket in the world to launch a tiny capsule shows how much more energy it takes to put a spaceship around the moon than the nominal mission to low Earth orbit. It's been 40 years since such a large rocket for such a small capsule was part of the human experience.
The manned version will be partly made in Europe, by trading investments in the space station from the 1980's for services today for which there was no money. The space station has proven an invaluable bank account for services NASA couldn't possibly afford today.
The rounded cover looked like something from the old west, when it was 1st unveiled 10 years ago. It turned out to not be a stylized artist rendition, but the actual shape.
The 3DR current sensor outputs 0.3V for 5A, 0.1V for 2A, 0.05V for 1A. It was designed for much higher current. It's nonlinear, under 5A. After tearing it down while thinking it over, the decision was made to try a home made shunt resistor.
Based on the internet, the target current is 0-5A. A simple power supply exercise reveals a small piece of wire drops 0.027V at 5A. It would need 122x gain to reach full range. Bench power supplies are quite useful for measuring extremely low resistances.
Then, it was decided a full power regulated system would need to sense current, voltage, gyro, & a voltage representing target power. Then it would need a fixed point multiply. None of the G-buggy circuit could be reused. It would need a higher end microcontroller with the full MP lab build system, finally using the pickit that arrived years ago. The era of the home made programmer with gpasm was over.
Then it was decided making it through hole wasn't worth it. A tried & true surface mount ARM would support bluetooth for phone configuration, 400Mhz for the hand controller, & have real floating point power calculations.
The ancient dspic33fj128mc802 may never be used. It's not worth building a through hole board & bringing up yet another build environment.
It was decided the trail was too rutted to get any training effect with the home made rover. After months of passively researching the problem, finally did a sort by price on Horizon, yielding the Tamiya Lunch Box, a 30 year old design. It was the largest thing for its price, big enough to navigate the rutted trail.
Tamiya Lunch Box $105
- working motor included
- plenty of room for gear
- endurance over 1 hour with 5Ah 12V battery
- obsolete 1987 design
- needs bearings
The cheapest hobbyking option was the Quanum Vandal.
Quanum Vandal $64
- modern design
- no motor included There is a 3700kV & 2200kV brushless motor in the apartment.
- probably too low to get up curbs
- unknown endurance
- 142x46x24mm battery
$64 for proportional steering was a lot better than 6 months ago, but the deciding factor was of course
Horizon - free shipping + $10 tax
Hobbyking - $52 shipping, cash only + no tax
The cheapest for the largest size was the lunch box, for all its horror stories. The mane horror stories are the inferior plastic bushings & the chassis bending. There wasn't enough money for any more than the very cheapest thing big enough to get over the curbs.
The funny thing about day jobs is when they're gone, you think you'll spend like no tomorrow when they come back. When they come back, you realize why you never spent like no tomorrow before they were gone. Inherit it, steal it, get lucky on kickstarter, or qualify for bigger loans, but no-one makes money from working.
Step 1 is to determine the current draw of the box & PWM range of the stock ESC.
Step 2 is convert the G-Buggy electronics to a power sensing throttle, outputting real PWM signals for the throttle & steering. The voltage regulator needs to be replaced with a shunt resistor calibrated for the current draw.
With the full headlights, it should be a rolling city on the trail.
Capturing HDMI video from an iPad 2 is like going back to the world of 1997. The capture cards are hundreds of dollars. They use PCI. When was the last time something used PCI? They use ancient custom FPGA compression chips to output ancient, horrible quality MPEG-2 compression. The cables are a fortune. The 30 pin connector is 2 generations old. There is no Linux support. There are cheaper grey market items on ebay, but the grey market has never actually shipped anything.
The thing is $50 camcorders have been compressing HD video from image sensors for years. They all use the same $1 Ambarella chip to generate high quality x264 compression. It takes parallel data in the same format from an image sensor as an HDMI decoder would generate. You could hack something out of an FPGA to forward data from an HDMI cable to a stock camcorder. The trick would be the soldering.
The leaders of the cheap market are the Timeleak HD72A & Roxio Game Capture HD. Only the Timeleak can capture HDCP protected content but it has no compression. An intriguing device is the Grabbee HD, which compresses to USB. If the grey market item actually arrived, it would have to be used in Windows. Video capture in Linux is something which has risen & fallen, but it's something that would be rarely used. It would only capture 1 game on the iPad.
Ultramap finally compiled, after a week. The phone no longer showed a static menu button, so it required a layout change, which required recompiling the program. Adb no longer worked with the new phone & required an upgrade, but Eclipse wouldn't upgrade, so a new version of eclipse was required.
Eclipse wasn't catching all the errors in the version change & running it with constant missing class errors. The xmlns:android="http://schemas.android.com/apk/res/android" line wasn't compiling anymore but wasn't necessary. The phone version was API 19 but Eclipse was compiling 20. Eclipse needed to install API 19 from Window->Android SDK manager. The API number isn't the same as the 4.4 number & there's no easy way to find out what API correlates to the 4.4 number.
All the project.properties files needed to be hand tweeked to say android-19. The AndroidManifest.xml files then needed to say android:targetSdkVersion="19"
Installing Google maps API is a multi step process. You have to download an Eclipse project called Google play services in Window->Android SDK manager, import it as an existing project from the android-sdks/extras directory, & compile it as a library. It has to be linked into your app project, but there are multiple properties dialogs with multiple fields for libraries.
Adding the jar file to Package explorer->Ultramap->properties->project references doesn't work. The entire google play services...Continue Reading
The decision was made to convert contact lens agitator #2 to a traditional levitator, since so much work went into it. Would find a cheap thing to hang from it. Instead of reusing the existing electronics, would make a simpler circuit using knowledge gained over 9 years to do a better job. The original circuit had a full H bridge. The new circuit would just pull.
Opening it after 9 years revealed a few critters. It was indeed a steaming pile of dog turd. It was a miracle the original circuit ever worked at all, since it didn't use star grounding, relied on a dog slow LM324, had a lot of unnecessary parts for what seemed to be lowpass filtering. It relied on extreme capacitance to work around the grounding. It was the product of many experiments without an oscilloscope.
The 1st attempt used a MOSFET with full belt & braces snubber diodes. It quickly overheated & destroyed itself. It went back to a tried & true BJT with no snubber diodes.
It was finally time for a 3rd stab at a contact lens agitator. After years of levitation & home made linear motors, this one would be a tried & true laboratory shaker, using a stepper motor. The linear motor couldn't make enough thrust, was too noisy, & shed metal filings. The levitator had proven completely ineffective but looked neat.
There was hope a computer fan could do the job. Those turned out to use a dead simple half bridge. 1 pin is always 12V. The other 2 pins alternate going to ground. They spin in only 1 direction, no matter the polarity.
The next step was a traditional brushless gimbal motor. A 3 phase motor controller would have been ideal, but completely unaffordable. A pair of BJT's from a burned out lightbulb would do the job. They didn't need spudger diodes like MOSFETs. The same half bridge arrangement of a computer fan actually provided enough of a shaking motion when applied to a brushless gimbal motor.
After much effort, a C program for driving it with PWM wouldn't compile properly. The compiler choked on a counter equality comparison. After redoing it in assembly, PWM was once again a noisy failure. Even at 22khz, it was too noisy because the PIC at 8Mhz couldn't get the timing close enough. A linear regulator would lower the 20V input to control speed, the heat from which would heat the payload.
How to mount a regulator on a CPU heatsink to heat the payload....Continue Reading
That was disappointing. They lost 3 in 2007, when an oxidizer tank exploded. Like last time, they'll never release the cause of the accident. This was probably another oxidizer tank failure. The pilots would have been ejected by the blast. The guy in the left seat would have been the unlucky one, knocked by just the right piece of carbon fiber to knock him out.
There comes a point in a space program when enough people have been lost by the same cause that there is a definite safety issue in the management or the system. Hybrid rocket engines just may not be safe enough because they require too much gas under too much pressure for current materials. People have struggled with carbon fiber tanks for decades. They haven't been consistently able to contain a rated pressure.
It's hard to believe they'll be able to make another vehicle without any money coming in from customers. The plan was always to mass produce them, but only after the money was coming in from the 1st one. Dick Branson had slowed funding to a crawl. It's hard to slow it any further without stopping it.
It's yet another story of another guy giving his life in the quest to reach space. As long as space is just 62 miles away, they'll keep reaching for it & some won't come back. Maybe we'll lose 1 per decade. No-one died in the 1990's, when the number of people going up was at its peak. There hasn't been enough money since then. The 2010's are still benign compared to the 2000's.
Virgin Galactic/Scaled was the last to continue the legend of people coming to the high desert of Calif* to ride a vehicle into space, test pilots whose names no-one knew, who rode horses & drank at a run down bar. The legend is really over. Kazakstan ended up being the place for going into space. The new space cowboys drink Vodka & drive trucks with dashcams on the wrong side of the road.
Working for the Modern San Francisco Startup is a new experience. The commute is 90 minutes. Everyone was born after you went to your last baseball game. All the jobs are now in the city, where 15 years ago they were all in the valley. The current cycle back to the city began in 2007 when real estate plummeted. Now the city is ferociously expensive while the valley is the wasteland.
All the assets are stored on web application cloud servers: asana for project management, bitbucket for code repository, gmail for email, gliffy.com & chart.io for documents. Even all the lunches are ordered on obscure cloud services like eat24hours.com. It's surprising how much cloud still isn't Google, if you look hard enough.
There's no more full time IT guy & server room. It's all web services. There are but 2 areas this cloud web generation will absolutely not use touch cloud services for, however: DATING. They all met their partners in BARS & NIGHTCLUBS. NO ONLINE DATING! NEVER!
The other thing is working from home. For all the webification, there is no working from home. The topic of physically being in an office is still as sacred as where the romantic relationship starts.
It's surprisingly easy for them to find the best people for the job. Programming is no longer a black box known only by a few savants, but manestream material, nowadays. It's a significant change from 2001, when it was very hard to find someone who knew what they were doing.
You wouldn't believe how hard it is to make a fan. The mane problem was noise, which RC parts aren't designed to avoid, but which is a critical design feature for appliances. The other problem was safety. A prop saver did absolutely nothing. The prop still shattered in an obstruction test. This is 1 area where 3D printing a shroud would be useful. For now, a wire provides some minimal protection from accidentally coming in from the side.
Anothor 9.0V rover run at rock bottom 9min/mile. Hit 8m16s in the fastest mile. A 90 minute commute doesn't leave any time for hacking, so the balky steering lives on.
The very last rover run was after canned soup, frozen chicken, & a long session of diahrea. At 9.2V, with reduced D gain, the steering was still impossible. The fastest mile was 8m4s. The entire flat section was at 8m49s/mile.
More of the .25 mile graph poked above the 8m mark. The uphill part went faster than the downhill part, so trail condition affected speed more than voltage. It was definitely harder, partly because of lack of exercise & because it was 2 days after the last rover run. The brushless rover never did match the G-buggy in consistency or steering, as bad as the g-buggy was.
It was the last of the intense speed runs. The commute has put an end to the illustrious 4 year running career & the salads. Speed never equaled its peak in Feb. It's interesting to look at the training log & ponder what worked.
For the 1st time in 4 years, was much less hungry in the last week. Felt much less appetite from lack of exercise. 15 hours a week of sitting in a train is pure hell on a body. Looks like the fitness of the last 4 years has come to an end, but these are the sacrifices required to make 1/2 what you did 10 years ago, in the new economy.
Always amusing to see photos that were impossible just 3 years ago. It's taken for granted now, but only a tick of the clock ago, camera sensors weren't fast enough to get razor sharp, handheld images from an airplane in near total darkness.
We all had the experience of trying to get night shots from airliners. It was completely hopeless for a camera made in 2008.
That took working the proverbial ass off.
Forget about live video of auroras. The only way to see what an aurora looked like in real time before 2012 was to drive 3000 miles. Most of the world could only imagine from timelapses & artist renditions. Now everyone has seen an aurora in real time.
Never build 1 when you can build 2 for twice the price. Everyone knows about Stennis space center, the A-1 test stand where the SSME & F-1 were tested. Lesser known is the even bigger B-1/B-2 test stand where the complete S-IC & shuttle core were test fired.
Little did you know an identical set of test stands was built in Huntsville, in an age long before Stennis. The mighty S1C test stand was built from 1960 - 1964 for $30 million & would test the 1st 4 S-IC stages. You can't even buy a house for $30 million, anymore.
Had granola cerial with no pepto bismol, since it might interfere with nutrient absorption. At 9V, it did the actual 13.1 segment in 1h58m2s including red lights. Red lights took 30s. The fastest mile was 8m7s on the uphill. Speed was highly erratic, with some quarter mile segments going extremely fast. Any benefit from PWM regulation was lost to flips & spinouts. Steering was non functional. Ate a caffeine free powergel at 11.5mi, which went everywhere. Couldn't drive while eating. The 1Ah battery had no problem. Got a bowel movement at 13.72mi.
The answer is still no. You can't make your own private router with 2 mile range out of a pair of LTE phones. As far as what can be deciphered from the hindi accents, it's not a point to point version of LTE but a way for devices to discover nearby devices. They broadcast a very brief 128 bit ID on the LTE physical layer which is directly received by devices within 500m, to allow device discovery, but the data communication is still over a paid, bandwidth capped, tower based LTE plan.
Of course, the hype is just as insane as the Intel Edison 2 weeks ago & the Google Tango before that. Does anyone remember what Google Tango was?